Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Social media and health: social support and social capital on pregnancy-related social networking sites
(USC Thesis Other)
Social media and health: social support and social capital on pregnancy-related social networking sites
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
i
SOCIAL MEDIA AND HEALTH:
SOCIAL SUPPORT AND SOCIAL CAPITAL ON
PREGNANCY-RELATED SOCIAL NETWORKING SITES
by
Heather Jane Hether
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(COMMUNICATION)
December 2009
Copyright 2009 Heather Jane Hether
ii
DEDICATION
To my children, Easton and Emma,
who fill my life with such joy, love, and sweetness.
iii
ACKNOWLEDGEMENTS
I could not have completed this dissertation without the support of so many
wonderful people. First and foremost, I would not be here today if it weren’t for the
support and encouragement of Sheila Murphy. One of my first professors at USC when I
was pursuing a Master’s degree, Sheila has been a steadfast supporter during the past
eight years and her encouragement in my decision to return to school full-time to pursue
my PhD was a turning point in my life. I have learned so much from Sheila as both an
educator and researcher and I am so thankful for all the wonderful research opportunities
on which we collaborated.
There have been other inspiring and influential professors at Annenberg, such as
Peggy McLaughlin, Michael Cody, Sandra Ball-Rokeach, Alison Trope, and Doe Mayer
from whom I have learned much and who have also made my time at Annenberg so
interesting. I also owe a special thanks to Professor Tom Valente, from USC’s
Department of Preventive Medicine, who has been so generous with his time and
expertise. Tom’s work has been a central influence on my own and I can’t thank him
enough for his mentorship.
I feel a special affection for the members of my cohort and I have fond memories
of our first year together. Chris Chavez, in particular, has been a good friend, always
willing to listen and commiserate or cheerlead as required. There are too many other
friends and peers to name at Annenberg from whom I have learned much and who have
made the past several years a wonderful experience.
iv
I also want to thank my parents, Ross Burkett and Carole Barton, for their interest
and support in this project. My friends, including Stacy Lieberman, Julie Miller, Sue
Lambe, Andrea Mayberry, and Leslie Gaudineer in particular, provided much needed
social breaks. Kelly Reinig also stepped in to help with childcare at a critical point in my
dissertation writing and I am thankful that she provided such excellent care for my
children.
Most importantly, I have to thank my husband, Michael Hether, who gave me a
room of my own and the quiet time to use it. Michael has been my most ardent supporter,
pushing me forward whenever I was too tired or doubtful. He never stopped believing in
me and I would not have finished this project without his support. My children, Easton
and Emma, bring so much joy into my life and have kept me grounded. Their sweetness
and laughter make everyday a blessing, and it is for them that I would so hard. This
dissertation required sacrifice from Michael, Easton, and Emma, and I will be forever
appreciative of their support.
v
TABLE OF CONTENTS
DEDICATION .............................................................................................................. ii
ACKNOWLEDGEMENTS ......................................................................................... iii
LIST OF TABLES ....................................................................................................... ix
LIST OF FIGURES .................................................................................................... xii
ABSTRACT ............................................................................................................... xiii
PREFACE ....................................................................................................................xv
CHAPTER 1: THE INTERNET AND HEALTH .........................................................1
The Internet as a Popular Health Information Resource ..........................................1
Social Media and Health ..........................................................................................5
Social Networks and Health.....................................................................................9
Diffusion of Innovations Theory ...........................................................................14
Pregnancy as a Unique Health Issue ......................................................................17
Conclusion .............................................................................................................21
CHAPTER 2: SOCIAL SUPPORT AND SOCIAL CAPTIAL ..................................23
Social Support ........................................................................................................23
Online Social Support ............................................................................................29
Social Capital .........................................................................................................31
Measuring Social Capital .......................................................................................35
Social Capital and Health.......................................................................................38
Online Social Capital .............................................................................................38
The Convergence and Divergence of Social Support and Social Capital ..............39
Conceptualization ............................................................................................39
Measurement ....................................................................................................41
Application .......................................................................................................42
Research Questions ................................................................................................42
CHAPTER 3: METHODOLOGY ...............................................................................47
Methodology Organization ....................................................................................47
Content Analysis Methodology .............................................................................47
The Sample ......................................................................................................48
Coding Procedures ...........................................................................................50
Pilot Testing ...............................................................................................50
Interrater Reliability ...................................................................................50
Protocol ......................................................................................................53
Instrument ..................................................................................................53
vi
Units of Analysis..............................................................................................54
Content Analysis Measures....................................................................................54
Social Support ..................................................................................................54
Health-related Measures ..................................................................................58
CHAPTER 4: CONTENT ANALYSIS RESULTS ....................................................60
RQ1: Demographic Characteristics of Social Networking Members ....................60
RQ2: Health Information in Social Networking Sites ..........................................62
RQ3: Specific Health Issues Discussed on Social Networking Sites ...................63
RQ4: Dimensions of Social Support Exchanged on Social
Networking Sites ..........................................................................................64
CHAPTER 5: SURVEY METHODOLOGY ..............................................................66
Survey Measures ....................................................................................................67
Social Support ..................................................................................................67
Social Capital ...................................................................................................68
Frequency and Intensity of Use .....................................................................70
Participation .....................................................................................................72
Health Outcomes ..............................................................................................73
Knowledge .................................................................................................73
Attitudes .....................................................................................................74
Behavior .....................................................................................................77
Analysis..................................................................................................................81
CHAPTER 6: SURVEY RESULTS ............................................................................83
Sample Characteristics ..........................................................................................83
RQ5: Frequency and Intensity of Use and Health Outcomes ................................86
Knowledge ......................................................................................................86
Attitudes ..........................................................................................................87
Behavior ..........................................................................................................88
Reasons for Support Seeking and Health Outcomes ............................................90
Knowledge ......................................................................................................90
Attitudes ..........................................................................................................90
Behavior ..........................................................................................................92
Perceived Usefulness of the Site and Health Outcomes .......................................95
Knowledge ......................................................................................................95
Attitudes ..........................................................................................................95
Behavior ..........................................................................................................96
RQ6: Participation and Health Outcomes ..............................................................98
Knowledge ......................................................................................................98
Attitudes ..........................................................................................................98
Behavior ..........................................................................................................99
RQ7a: Perception of Social Support and Health Outcomes ................................100
Knowledge ....................................................................................................100
Attitudes ........................................................................................................100
vii
Behavior ........................................................................................................102
RQ7b: Informational and Emotional Social Support and Health Outcomes .......104
RQ8a: Social Capital and Health Outcomes ........................................................105
Knowledge ....................................................................................................106
Attitudes ........................................................................................................106
Behavior ........................................................................................................107
RQ8b: Bonding and Bridging Social Capital and Health Outcomes ...................109
Knowledge ....................................................................................................109
Attitudes ........................................................................................................109
Behavior ........................................................................................................110
CHAPTER 7: SOCIAL NETWORK ANALYSIS METHODOLOGY ...................113
Measures .............................................................................................................114
Survey Data ...................................................................................................114
Degree .....................................................................................................115
Content Analysis Data ..................................................................................118
Centrality and Centralization ...................................................................119
Closeness ...........................................................................................120
Betweenness ......................................................................................121
Density ....................................................................................................121
Tie Strength .............................................................................................121
Heterogeneity ..........................................................................................122
Analysis ...............................................................................................................122
CHAPTER 8: SOCIAL NETWORK ANALYSIS RESULTS .................................123
RQ9: Social Network Measures and Health Outcomes .......................................123
Incoming Support Network .................................................................................123
Indegree .........................................................................................................123
Outdegree ......................................................................................................124
Outgoing Support Network .................................................................................125
Indegree .........................................................................................................125
Outdegree ......................................................................................................126
RQ10: Comparison of Social Network Measures and Survey Scales of
Social Support and Capital ...............................................................................128
RQ11: Structure of the Network and Dimension of Social Support ...................131
RQ12: Bonding and Bridging Social Capital and
Health-related Social Networking Sites ............................................................138
Site A .............................................................................................................139
Site B ..............................................................................................................140
RQ13: Tie Strength and Social Support ..............................................................142
RQ14: Diversity of Ties ......................................................................................144
Age .................................................................................................................144
Race................................................................................................................145
Employment Status ........................................................................................145
viii
Location .........................................................................................................146
Pregnancy Experience ....................................................................................147
CHAPTER 9: DISCUSSION AND IMPLICATIONS ..............................................149
Major Findings and Implications .........................................................................151
Content Analysis ............................................................................................151
Social Networking Member Survey...............................................................155
Frequency and Intensity of Participation and Health Outcomes .............155
Social Support and Social Capital and Health Outcomes ........................163
Social Network Measures of Degree and Health Outcomes ....................166
Social Network Analysis................................................................................169
Social Network Indicators: Valid Measures of Social Support
Or Social Capital? ....................................................................................169
Network Structure and Social Support ....................................................171
Bridging and Bonding Social Capital, Weak and Strong Ties.................172
Methodological Limitations and Implications .....................................................175
Content Analysis ............................................................................................176
Survey ............................................................................................................176
Future Research ...................................................................................................179
Social Support Substitution............................................................................179
Hierarchy of Social Support...........................................................................180
Health Impact of Social Networking Sites.....................................................181
Social Network Analysis, Social Support, and Social Capital.......................182
Conclusion ...........................................................................................................183
REFERENCES ..........................................................................................................184
APPENDICES ...........................................................................................................198
Appendix A: Support-seeker Code Sheet ............................................................198
Appendix B: Support-provider Code Sheet .........................................................207
Appendix C: Member Survey ..............................................................................208
ix
LIST OF TABLES
Table 3.1: Interrater Reliability ................................................................................52
Table 3.2: Typology of Social Support used in Content Analysis ...........................56
Table 4.1: Characteristics of Social Networking Members .....................................62
Table 4.2: Most Common Health Topics for Advice-seekers..................................64
Table 5.1: Reliability and Descriptive Statistics for the dimensions of the
Social Provisions Scale ...........................................................................69
Table 5.2: Reliability and Descriptive Statistics for the
Social Capital Scales ...............................................................................70
Table 5.3: Descriptive Statistics for Frequency and Intensity Variables ................72
Table 5.4: Descriptive Statistics for Participation Variables ...................................73
Table 5.5: Descriptive Statistics of Attitude Outcomes ...........................................75
Table 5.6: Descriptive Statistics for Overall Attitude Toward Pregnancy...............77
Table 5.7: Descriptive Statistics for Behavior Outcomes ........................................79
Table 5.8: Behavior Change Distributions ...............................................................80
Table 6.1: Characteristics of the Survey Sample .....................................................84
Table 6.2: Standardized Betas Indicating Significant Associations Between
Frequency of Site Use and a Positive Attitude Toward Getting
Pregnant Again........................................................................................87
Table 6.3: Standardized Betas Indicating Significant Associations Between
Seeking Esteem Support and Pregnancy-related Attitudes .....................91
Table 6.4: Standardized Betas Indicating Significant Associations Between
Seeking Emotional Support and Attitudes Toward the Pregnancy .........92
Table 6.5: Adjusted Odds Ratios Indicating Significant Associations Between
Seeking an Opportunity for Nurturance and Behavior Change ..............94
x
Table 6.6: Adjusted Odds Ratios Indicating Significant Associations Between
Perceived Site Usefulness and Behavior Change ..................................97
Table 6.7: Standardized Betas Indicating Significant Associations Between
Social Support and Prenatal Attitudes .................................................101
Table 6.8: Adjusted Odds Ratios Indicating Significant Associations Between
Social Support and Behavioral Change as a Result of Participating
on the SNS ............................................................................................103
Table 6.9: Summary of Significant Associations Between Perception of
Social Support and Outcomes ..............................................................105
Table 6.10: Adjusted Odds Ratios Indicating Associations Between
Social Capital and Behavioral Change as a Result of Participating
on the SNS ............................................................................................108
Table 6.11: Standardized Betas Indicating Significant Associations Between
Bonding and Bridging Social Capital and a Positive Attitude
Toward Being Healthy .........................................................................110
Table 6.12: Standardized Betas Indicating Significant Associations Between
Bonding and Bridging Social Capital and Overall
Behavior Change ..................................................................................112
Table 7.1: Social Network Measures used to Predict Health Outcomes................117
Table 8.1: Standardized Betas Indicating Significant Associations Between
Indegree and Health Behaviors .............................................................124
Table 8.2: Significant Associations Between Social Support, Social Capital,
Social Network Measures and Health Outcomes ..................................129
Table 8.3: Pearson’s Correlations Between Social Support, Social Capital,
and Social Network Measures ..............................................................131
Table 8.4: Comparison of Five Support Networks on Social
Networking Sites A and B ....................................................................133
Table 9.1: Significant Associations Between Participation
Variables and Outcomes .......................................................................156
Table 9.2: Significant Associations Between Dimension of Support
Sought and Outcomes ...........................................................................159
xi
Table 9.3: Significant Associations Between Support Seeking and
Providing and Outcomes .......................................................................162
Table 9.4: Significant Associations Between the Perception of Social
Support, Social Capital and Outcomes .................................................164
Table 9.5: Significant Associations Between Social Network Measures
of Degree and Outcomes .......................................................................168
xii
LIST OF FIGURES
Figure 2.1: Example of a Sociogram used to Depict a Social Network ....................26
Figure 4.1: Distribution of Responses per Support Solicitation ................................61
Figure 4.2: The Different Kinds of Health Information
Sought by Advice-Seekers ......................................................................63
Figure 4.3: The Exchange of Five Different Dimensions of Social Support ............65
Figure 5.1: Percentage of Respondents Who Sought Each Dimension
of Support from the Social Networking Site ...........................................71
Figure 8.1: All Support on Site A ............................................................................135
Figure 8.2: All Support on Site B ............................................................................135
Figure 8.3: Informational Support on Site A ...........................................................135
Figure 8.4: Informational Support on Site B ...........................................................135
Figure 8.5: Emotional Support on Site A ................................................................135
Figure 8.6: Emotional Support on Site B ................................................................135
Figure 8.7: Esteem Support on Site A .....................................................................136
Figure 8.8: Esteem Support on Site B .....................................................................136
Figure 8.9: Relationship Support on Site A.............................................................136
Figure 8.10: Relationship Support on Site B .............................................................136
Figure 8.11: Frequency of Support Provisions to Same Member on Site A .............140
Figure 8.12: Frequency of Support Provisions to Same Member on Site B .............141
xiii
ABSTRACT
A large number of Americans search the Internet for health information.
Moreover, Internet users are no longer solely passive consumers of online health content,
but they are also active producers as well. Social media, such as social networking sites,
are online communities where patients can pool their knowledge and experience to
manage their health care more confidently. However, little is known about the impact of
participation on these kinds of sites. This dissertation used three methodologies to
explore the health-related impact of these sites. First, a content analysis was conducted
with a sample of 572 support-seeking messages and 1,965 support-providing messages
from two sites during one month. Members were largely American, Caucasian, and
married. They sought information on a range of health topics, but especially issues
related to the beginning and end of pregnancy. Further, informational support and
emotional support were most in demand. Second, using survey data collected from 114
pregnant members of these sites, this dissertation tested whether participation was
associated with health outcomes. The survey also compared the impact that perceptions
of social support and social capital had on outcomes such as prenatal knowledge,
attitudes, and health behaviors. Results show that members who provided more support
tended to be the most influenced by their participation on the site. Further, social support
was associated with more attitudinal outcomes than was social capital, suggesting some
experiential differences in these constructs. Third, a social network analysis compared
social network metrics with survey scales of social support and social capital. The results
showed that social network metrics were correlated more with social support, suggesting
xiv
further differences between these constructs. Social network analysis also suggested that
there was a hierarchy of social support wherein the network disintegrated (it became
more decentralized and less dense with fewer members) as the demands of the support
provision increased. Social network analysis also supported the findings of the content
analysis in illustrating that this was a weak tie network that largely exchanged
informational support. The implications of these and other findings and directions for
future research are discussed.
xv
PREFACE
Increasingly, the Internet is being relied upon as a source of health information by
a large number of Americans. According to a 2009 study by the Pew Internet &
American Life Project, 80% of Internet users had searched online for health information
(Fox & Jones, 2009). Studies have shown that the information individuals obtain from the
Internet informs their medical decision-making, influences their health-related attitudes
and behaviors, and impacts their relationship with their physician (Anderson, Rainey, &
Eysenbach, 2003; Bass, Ruzek, Gordon, Fleisher, McKeown-Conn, & Moore, 2006; Fox
& Rainie, 2000; Gerber & Eiser, 2001). A population that is increasingly relying on the
Internet as a source of health information (Broom, 2005) suggests that scholars should
investigate the content and impact of this health information resource.
While the Internet is a cornucopia of health information, this dissertation focuses
on social networking sites that enable individuals to exchange social support about a
particular health issue, in this case, pregnancy. A 30-year history of social support
scholarship has shown that social support is associated with a variety of positive health
outcomes, such as improving resistance to and recovery from disease (Cohen, 2001;
Seeman, 2001; Spiegel & Kimerling, 2001); reducing mortality (Berkman, 1995; House,
Landis, & Umberson, 1988); and improving prenatal health behaviors (Schaffer & Lia-
Hoagberg, 1997). In fact, Burleson and MacGeorge (2002) explain that because there is
such a large number of studies that have associated social support with a variety of
physical and psychological outcomes the reviews of the relevant literature tend to be
either vague and impressionistic or very specific to particular stressors, populations,
xvi
outcomes, or some combination of these variables. They further explain that despite the
variations in the studies and their specific outcomes, the literature overwhelmingly
supports a positive association of perceived social support with physical and mental
health.
Social support flows through social networks (Hall & Wellman, 1985); hence,
much research has been conducted on the health benefits of membership in well-
connected networks as opposed to the potential adverse effects of isolation or even
membership in poorly connected networks. The emergence of online social communities
has created a new paradigm of social networks, one in which members may never meet
face-to-face. Therefore, despite the fairly substantial body of research examining the
influence of social networks on health, most of the extant research has focused on offline
social networks. Furthermore, while there is a fair amount of scholarship about social
support in online contexts, this work has largely focused on the use of list-server bulletin
boards or email threads as the vehicles for social support. Thus, the exchange of social
support in online social networks represents a new arena of scholarship.
While the association of social support with positive health outcomes has been
well documented, another related construct, social capital, has recently garnered attention
with researchers interested in public health. Similar to the social support literature, the
burgeoning field of research on this construct has found some associations between social
capital and positive health outcomes. In Kim, Subramanian, & Kawachi (2008)’s review
of the literature examining the association of social capital with health, they found several
studies showing that indicators of social capital had a protective effect on mortality
xvii
outcomes at the state, regional, and/or neighborhood level in the U.S. Further, at the
individual level, they found that studies show social capital was associated with better
rated self-health.
Researchers recognize that social support and social capital are related constructs
(Lakon, Godette, & Hipp, 2008; Lin, 1999; Wellman & Frank, 2001); however, little, if
any, research has sought to empirically investigate the relationship between them. This
dissertation attempts to resolve this gap. Furthermore, this dissertation also contributes to
the literature by: 1) examining what kind of health information is circulating on social
networking sites related to pregnancy; 2) assessing the impacts of participation on these
health-related social networking sites; 3) examining the associations between social
support, social capital, and health outcomes in an online environment; and 4) using
multiple methodologies, specifically a participant survey and social network analysis, to
test the validity of different measurements of the same, or similar, constructs.
This dissertation consists of nine chapters. In Chapter 1, the Internet and its role
as a health information resource is discussed. The role of social media in health care is
also introduced, including new ideas that reflect the convergence of technology and
health care, such as those reflected in the concept, “Medicine 2.0.” The research on social
networks and health is elaborated, and the health topic at the center of this dissertation,
pregnancy, is discussed as a unique health issue that is especially relevant for study of
social networking sites and health outcomes.
In Chapter 2, two of the primary theoretical constructs of this study, social
support and social capital, are discussed in detail, including their research histories,
xviii
current tensions in their literatures, and how they have been associated with health. Social
network analysis as a methodology to measure social support and social capital is also
introduced. Chapter 2 concludes with the research questions that guide the current study.
Chapter 3 briefly introduces the three methodologies used in the current research,
a content analysis, a survey, a social network analysis, and then describes in detail the
content analysis methodology. Chapter 4 follows with a presentation of the findings from
the content analysis. Chapter 5 describes the survey methodology and Chapter 6 presents
the findings from the survey. Chapter 7 describes the social network analysis
methodology and Chapter 8 presents the results from the social network analysis.
Chapter 9 includes a discussion of the findings presented in the previous chapters.
The theoretical, methodological, and practical implications of the current study are
described, as well as suggestions for future research.
1
CHAPTER O
E
THE I
TER
ET A
D HEALTH
The Internet as a Popular Health Information Resource
Numerous studies have indicated that the Internet has become a preeminent health
information resource for many Americans. In 2009, the Pew Internet and American Life
Project reported that 80% of Internet users had searched online for health information
(Fox & Jones, 2009). Studies suggest that the Internet has become a valued tool for
patient self-education that is changing how patients manage their health care (Anderson,
Rainey, & Eysenbach, 2003; Bass et al., 2006; Broom, 2005; Diaz et al., 2002). Further,
as will be described later in this chapter, there are other changes underfoot in the health-
care industry that suggest a climate in which technology, and the Internet specifically, is
only going to assume a more central role for patients and health-care consumers.
Several studies portray a similar profile of Internet users who are most likely to be
searching for health information online. In general, the characteristics of this group
include being female, Caucasian, younger than age 65, having a higher education level,
and more Internet experience (Atkinson, Saperstein, & Pleis, 2009; Baker, et al., 2003;
Fox, 2005; Fox, 2006; Fox & Jones, 2009; Houston & Alison, 2002; Ybarra & Suman,
2006; Ybarra & Suman, 2008). Some studies have also found poorer health status was
associated with searching online for health information (e.g. Baker et al., 2003; Houston
& Alison, 2002). Meanwhile, an association with income remains inconclusive. Some
researchers have found an association between higher income and searching online for
2
health information (e.g. Fox & Jones, 2009) while others have not (e.g. Baker, et al.,
2003).
Studies have also identified several kinds of information that online health seekers
are pursuing. A recent study by the Pew Internet and American Life Project found that
the two most frequent health topics of interest to online health information seekers were
information related to a specific disease or medical problem (66%) followed by
information related to certain treatments or procedures (55%) (Fox & Jones, 2009). Other
researchers found that 43% of Internet health information seekers tried to diagnose a
health problem with online health information and 33% tried to treat a health problem
with such information (Ybarra & Suman, 2006; 2008).
Research suggests, however, that the Internet supplements, rather than substitutes
for, health information from other sources (Fox & Jones, 2009; Ybarra & Suman, 2008).
Fox & Jones (2009) found that health professionals, such as doctors, remain the number
one source of health information for American adults, with 86% of American adults
seeking their advice when they need medical information or assistance. Baker, Wagner,
Singer, et al. (2003) found that online healthcare information had very little impact on
how often Internet users contacted their healthcare provider. In their study, more than
90% of survey respondents said that the Internet had no effect on how frequently they
made such contacts. Ybarra & Suman (2008) suggest that Internet users are using online
health information to compliment information from their healthcare professional.
Furthermore, unlike Baker et al.’s (2003) study, they found that online health information
both motivated health information seekers to contact a healthcare provider and it made
3
them feel more comfortable with the information provided by their healthcare
professional.
In terms of user effects of online health information, there seems to be some
consensus that user knowledge is increased by online health information. For example,
Houston and Allison (2002) report that 81% of Internet users who search online for
health information reported “learning something new” the last time they were online.
Similarly, Baker, Wagner, Singer, et al. (2003) found that among survey respondents
with a chronic healthcare condition, 48% said health information on the Internet
improved their understanding of their condition. Beyond the association of Internet health
information and knowledge, there seems to be little consensus in the literature about more
substantive impacts of Internet health information. Baker, Wagner, Singer, et al. (2003)
found that only 16% of respondents indicated that Internet health information affected the
treatment they were undergoing and only 7% said that it led them to seek care from
different doctors. Fox & Jones (2009), however, report some more substantive impacts of
searching for health information online. They found that 57% of online health
information seekers say their most recent search had an impact on them. Of these
individuals, 60% indicated that the information they found online affected a treatment
decision related to their illness or health condition, 56% said it changed their overall
approach toward their health or how they take care of someone else’ health, and 53% said
the information they found online led them to ask their doctor new questions or to get a
second opinion from another doctor.
4
While the evidence is still mounting in terms of how online health information
affects its users, several scholars have described the potential risks and benefits related to
the availability and use of online health information. Cline and Haynes (2001) present a
comprehensive review of the relevant literature and identify the following benefits and
risks of online health information. The benefits include widespread access to health
information; the interactivity of the medium that promotes tailoring of information and
interpersonal interaction; and the potential for anonymity. The risks of online health
information include disparities in access (e.g. “the digital divide”); difficulties in
accessing and processing information due to “information overload,” disorganization of
the Internet, searching difficulties or inaccessible language; and concerns about the
quality of the information. Other scholars describe potential intangible and tangible
effects of using the Internet for health information. Some intangible effects include the
benefit of making patients feel more comfortable and confident about their care, and
some additional tangible effects include how this information might affect the patient-
physician interaction — either positively because more educated patients might require
less time being educated by their physician on the basics of their condition, and they can
therefore focus on more sophisticated issues. Or, contrarily, online misinformation could
require physicians to spend more time correcting erroneous information that their patients
learned from the Internet (Bass et al., 2006; Baker, Wagner, Singer, & Bundorf, 2003;
Broom, 2005; Bylund et al., 2007; Gerber & Eiser, 2001).
As the previous discussion has illustrated, there is no straightforward answer to
the question of how online health information affects the health of its users. Since the
5
Internet is home to a vast amount of health information, presented in a variety of different
formats, establishing a conclusive relationship between this information and health
outcomes is a complicated endeavor. This dissertation, however, attempts to provide a
response to the above-mentioned question by examining one particular area of the
Internet health information landscape, social networking sites, specifically those focusing
on the discussion of pregnancy and prenatal care.
Social networking sites have emerged in the era of Web 2.0 as the second
incarnation of online support communities. While online support communities have
flourished on the Internet since the early days of the medium in the form of list-servs or
email threads, more recently a new generation of online support communities has
emerged. Web 2.0 represents a conceptual and practical shift wherein all Internet users
are encouraged to post content to the Internet. Now, Internet end-users are positioned as
producers, as well as consumers, of online content. Social media, such as social
networking sites, are emblematic of Web 2.0 by providing communities within which
members can form social connections and share information. Moreover, participation in
online communities represents a distinct way that Internet users can access health
information and they are forums that are increasingly being recognized as spaces that
facilitate patient-empowerment (Sarasohn-Kahn, 2008; Shirky, 2008).
Social Media and Health
In Web 2.0, health information on the Internet is not simply a one way flow of
statistics and information from a traditionally recognized healthcare expert — such as a
physician or healthcare organization like the Centers for Disease Control and Prevention
6
— to the public. Instead, health information on the Internet is found on a variety of web
sites where content is user-generated by members of the public, such as on blogs and
social networking sites. This notion of user-generated content is at the core of the concept
of “Web 2.0,” a concept that suggests a more interactive environment than “Web 1.0” in
which users were primarily situated as readers (Hughes, Joshi, & Wareham, 2008;
Sarasohn-Kahn, 2008). In this latest evolution of the Internet, everyone is recognized as
having something to contribute to the conversation, whether the conversation is about
politics, entertainment, or health.
The idea of harnessing the power of collective knowledge has been popularized in
Surowiecki’s (2004) catchphrase (and title of his book), “the wisdom of crowds,” in
which his thesis argues that large groups of people are often smarter than their smartest
members. Further, “even if most of the people within a group are not especially well-
informed or rational, it can still reach a collectively wise decision” (Surowiecki, 2004, p.
xiii). Similarly, in the context of health, the Internet, and social networking sites in
particular, is being imagined as a space where “the collective wisdom can and should be
harnessed” (Eysenbach, 2008, n.p.). Sarasohn-Kahn (2008) elaborates, “when patients
managing the same chronic condition share observations with each other, their collective
wisdom can yield clinical insights well beyond the understanding of any single patient or
physician” (p. 5).
Data from a 2009 study conducted by the Pew Internet and American Life Project
reports that 60% of Internet users, or 37% of U.S. adults, who searched for health
information online (a demographic they describe as “e-patients”) have accessed social
7
media related to health. Social media refers to user-generated content like blogs (web
logs or online personal commentary); rankings and reviews of health care providers and
medical facilities; online discussions; social networking sites; etc. According to the
report, 59% of e-patients have either read someone else’s narrative about a health issue
posted on a website such as a blog, or consulted online reviews of healthcare providers. A
much smaller number of e-patients have actively participated in social media. Fox &
Jones (2009) report that 6% of e-patients have posted comments or queries in online
group forums and 5% of e-patients have posted online reviews of doctors.
While the Pew Internet and American Life Project suggests that social networking
sites are not often used for health queries (Fox & Jones, 2009), their data do not
investigate how participation on health-related SNSs is associated with health status. For
example, Fox & Jones (2009) indicate that 80% of Americans describe their health as
good or excellent; therefore, one presumes that a population of healthy adults has fewer
reasons to participate in online health discussions than those who are managing a health
condition. However, their data indicate that 39% of e-patients use a social networking site
like FaceBook or MySpace. The popularity of these sites suggests that as Internet users
age and / or develop a health issue, they may be more inclined to participate in social
networking sites that have a more explicit health focus due to their familiarity with the
format.
User-generated health content on the Internet is one manifestation of what some
researchers and industry experts cite as a shift to a more “patient-centered” model of
health care (Shirky, 2008). New terms such as “Medicine 2.0” and “Health 2.0” have
8
been coined to reflect a new model of decentralized healthcare that relies heavily on
interactive Internet technology (Hughes, Joshi, & Wareham, 2008). This model is based
on core values of empowerment, participation, and collaboration in the pursuit of health-
care. The Internet assumes a key role in empowering patients to manage their own health-
care more confidently. Sarasohn-Kahn (2008) describes Health 2.0 as a movement that
uses “…social software and its ability to promote collaboration between patients, their
caregivers, medical professionals, and other stakeholders in health” (p. 2).
Medicine 2.0 has been similarly defined (Hughes, Joshi, & Wareham, 2008);
however, Eysenbach (2008) suggests that Medicine 2.0 is a broader term under which
Health 2.0 falls. Eysenbach (2008) elaborates, “…Medicine 2.0 also stands for a new,
better health system, which emphasizes collaboration, participation, apomediation, and
openness, as opposed to the traditional, hierarchical, closed structures within health care
and medicine” (n.p.). Further, researchers intimate that Medicine 2.0 is a term with an
academic focus, while Health 2.0 is a term used more often by those in industry
(Eysenbach, 2008).
Few academic studies, however, have been published that examine the health
impacts of participation on health-related social networking sites. A search of the
literature only revealed one related study that examined participation on a Japanese social
networking site for people coping with depression (Takahashi, Uchida, Miyaki, Sakai,
Shimbo, & Nakayama, 2009). This study used mixed methods that collected quantitative,
qualitative, and social network data to examine the health benefits and risks of
participation on this social networking site. The authors found that there were both
9
positive and negative aspects of participation in this community. Foremost, members
benefitted from providing and receiving social support. Further, the enhanced
opportunities for privacy and anonymity within this community were seen as assets that
could facilitate the exchange of support without the stigma of a face-to-face encounter.
There was, however, at least one major risk of participation. The authors suggest that
participation on the site could result in a psychological burden, wherein members could
become over-involved in someone else’s emotional problems, resulting in — or
increasing — their own depression.
This study is notable because it focuses on a social networking site related to a
health issue; however, it has at least two significant limitations. First, while the study
collected survey data, the sample size was quite small, N = 37. In addition, the authors do
not report statistical results of any associations between participation on the site and
health-related outcomes, such as knowledge, attitudes, or behaviors. In addition, it should
be noted that most diseases are not at risk of spreading online, as this study suggested of
depression. Therefore, the generalization of some of these findings to other health
conditions may be limited.
Social
etworks and Health
Online social networks are a provocative area of research for scholars interested
in health because they represent a new arena within which to extend knowledge about
how social relationships are associated with health outcomes. Social network analysts
working in public health have already established the importance of social ties to health
outcomes. Research has shown that certain health behaviors like smoking, drug use, and
10
even obesity are impacted by an individual’s social relationships (e.g., Christakis &
Fowler, 2007; 2008; Neaigus et al., 1994; Valente, Unger, & Johnson, 2005). As
Christakis (2007) aptly summarizes “people are connected and so their health is
connected” (p. 373).
Social networks have been described as “the web of social ties that surround an
individual” (Berkman, 1984, p. 414). Valente, Gallaher, & Mouttapa (2004) describe
social network analysis as “a set of theories, methods, and techniques used to understand
social relationships and how these relationships might influence individual and group
behavior” (p. 1686). Social network analysis, as a methodology, can be distinguished
from other social scientific methods because rather than analyzing associations between
variables or characteristics at the individual level, social network analysis is concerned
with the tie, or relation between people, and therefore uses relational data (Scott, 2007;
Valente, in press). Ties can be examined on a dyadic, “local” level — the pairs of people
that are connected to each other and their characteristics — in addition, ties can be
examined at the level of the entire network. For example, there are measures that indicate
the overall structure of the network, such as density (the ratio of how many people are
actually connected compared to the overall number of possible connections) and
centralization (whether there are a few key people at the center of the network around
whom a lot of activity revolves). On the local level, ties can be examined in terms of their
connectivity using a measure like degree, which measures the quantity of a network
member’s direct contacts. Additional network measures will be discussed in Chapters 2
and 7.
11
Online social networks are particularly interesting because they have the potential
to operate on a completely different scale than offline networks. Online networks are not
constrained by geography, scheduling, or any other logistical constraints that may limit
their offline counterparts. The Internet can also enlarge social networks by erasing or
minimizing socio-demographic differences, like gender, ethnicity, or age that may
impede the formation or size of offline networks. These structural differences provide
new opportunities to learn more about the dynamics and processes of networks.
Meanwhile, there already exists much research about offline networks that can facilitate
the study of online networks.
Social network analysis is conducive to the study of public health because many
health behaviors are social behaviors that may be susceptible to peer influence. For
example, a recent social network analysis followed more than 12,000 participants and
their contacts for 32 years (1971 to 2003) and examined smoking behavior (Christakis &
Fowler, 2008). The results show that there is a tendency for smokers to be associated
with other smokers and nonsmokers to be associated with other nonsmokers, with
relatively few connections between these clusters. The study found that the risks of being
a smoker were higher if one was connected to a smoker. Moreover, physical distance did
not dissipate the intensity of effects between contacts. That is, regardless of whether the
contact lived in the immediate geographic area or not, smoking behavior was still related.
Christakis & Fowler (2008) also found that over time smokers moved from the
center of the network to the periphery. Therefore, when smokers were younger they were
popular and as they grew older they moved to the fringe of the network. Valente, Unger,
12
& Johnson’s (2005) study on adolescent smoking found a similar relationship between
youth popularity and smoking. This network analysis examined smoking behavior of
more than 1,400 grade six students across 16 middle schools in Southern California.
Their results show that adolescent popularity was associated with greater risk for
smoking.
Other health issues that have been examined with social network analysis include
sexually transmitted diseases, substance abuse, and obesity. Bettinger, Adler, Curriero, &
Ellen (2004) found that risk perceptions of sexually transmitted disease varied by
network position. Individuals in central or bridge (linking two smaller sub-groups in the
network) positions perceived lower levels of risk for a sexually transmitted disease than
did isolates and, consequently, they were less likely to use a condom. Neaigus et al.
(1994) examined the social networks of intravenous drug users (IDUs) and found that
IDUs had ties that extended beyond drug use and often included marriage, kinship, or
friendship and that reciprocity between network members was common. The authors
argue that findings such as these can be used to design more effective public health
interventions. Finally, Christakis & Fowler (2007) examined whether weight gain in one
person was associated with weight gain among his or her social ties. The authors found
clusters of overweight people and analyses showed that an individual was more likely to
be obese if he or she had an obese friend. In addition, persons of the same sex were more
influential on each other than those of the opposite sex.
In sum, the research presented on social networks thus far suggests that people are
more likely to be connected if they are similar. The studies previously described show
13
that smokers are more likely to be clustered together, as are overweight people, as are
intravenous drug users (Christakis & Fowler, 2007; 2008; Neaigus et al., 1994). While
members of the same network are likely to exhibit similar behaviors, other research has
shown that one’s position in the network is also an important predictor of behavior. Some
network positions are associated with greater risk for certain negative health behaviors.
Other research has also shown that a dearth of social connections is associated with
adverse health effects. Berkman (1984) reviews a series of key studies about social
networks and health and shows that there is a consistent pattern for increased mortality
and morbidity rates associated with decreases in social connection. For example, several
studies show that coronary heart disease is associated with a lack of social connections.
Therefore, the evidence indicates that networks are associated with various health
outcomes.
A survey of the network literature, however, reveals that there is no “one size fits
all” theory to explain how networks influence behavior. Instead, the causal explanations
will vary by the health outcome under study, the socio-demographic characteristics of the
network members, and the network structure. However, scholars have suggested the
following theoretical explanations as to why networks might influence member
behaviors. These include group norms (Christakis & Fowler, 2007, 2008; Valente, in
press; Valente, et al. 2004), social learning theory (Valente, in press; Valente, et al.
2004), social influence (Christakis & Fowler, 2007; Valente, in press; Valente, et al.
2004) , social support (Berkman, 1984, 2000), access to resources and materials goods
(Berkman, 1984, 2000), homophily and selection (Valente, in press; Valente, et al. 2004),
14
and diffusion of innovation (Luke & Harris, 2007; Valente, in press; Valente, et al. 2004).
Social support and social capital (access to resources) are two of the major theoretical
foci of this dissertation, and they will be addressed in the next chapter, along with
homophily, the tendency of similar people to associate with each other. However,
diffusion of innovations is another theory which informs the current study; therefore, it is
discussed briefly below.
Diffusion of Innovations Theory
Diffusion of innovations (DOI) theory (Rogers, 2003) is relevant to this
dissertation because it is a theory that explains how new ideas, or innovations, spread in a
population. Rogers (2003) defines diffusion as “the process in which an innovation is
communicated through certain channels over time among the members of a social
system” (p. 5). Rogers conceives of communication as not merely a one-way linear
transfer of information but “a process in which participants create and share information
with one another in order to reach a mutual understanding” (Rogers, 2003, p. 5). This
notion of communication reflects the same spirit of communication in Web 2.0 and social
media in particular. More importantly, Rogers’ (2003) and subsequent scholars’ (e.g.
Valente, 1995) discussion of diffusion is important to the current study because DOI is a
theory that lends itself to using structural models, like social network analysis, to assess
how new ideas or innovations spread through a community or network of people. More
specifically, there are two central ideas from diffusion of innovations that are particularly
relevant to this dissertation. The first concerns how media and interpersonal influence are
15
addressed in DOI, and the second is DOI’s identification and discussion of opinion
leaders as central in the diffusion of new ideas.
Diffusion of innovations theory distinguishes between the effects of interpersonal,
face-to-face communication and traditional mass media communication. Rogers (2003)
argues that interpersonal communication is more effective at persuading an individual to
accept an innovation than are mass media. Other scholars have also recognized that
interpersonal communication is more persuasive than mass media. For example in Katz
& Lazarsfeld’s (1955, 2006) seminal work on the two-step flow of communication they
found that opinion leaders were more exposed to media and that these opinion leaders, in
turn, were stronger persuasive influences in the overall network than the media at large.
In other words, opinion leaders became exposed to new ideas from media and they, in
turn, intercepted, interpreted, and diffused the information to the network. Katz &
Lazarsfeld (2006) argued “the traditional image of the mass persuasion process must
make room for ‘people’ as intervening factors between the stimuli of the media and
resultant opinions, decisions and actions” (p. 33). Furthermore, opinion leaders are not an
elite group of people in the community, in terms of socio-demographic characteristics,
instead the authors clarify that an opinion leader “can best be thought of as a group
member playing a key communications role” (Katz & Lazarsfeld, 2006, p. 33).
More recently, health communication scholars have also found that interpersonal
influence is more strongly associated with behavior change than are media (Cassell,
Jackson & Cheavront, 1998; Piotrow, Kincaid, Rimon, & Rinehart, 1997; Valente &
Saba, 2001; Valente, 2002). Valente (2002) argues that there is a trade-off between
16
program reach and outcomes. He suggests that interpersonal communication is more
effective at changing behaviors, but lacks the broad reach of a media intervention,
whereas mass media reach more people, but are less effective at changing behaviors.
Rogers (2003) only briefly addresses the potential role of the Internet in diffusion.
He notes that the Internet is a combination of a mass medium and interpersonal
communication with the potential to accelerate the diffusion of an innovation more
rapidly. For this dissertation, the research site is a particularly poignant example of how
the Internet can be a hybrid of communication channels (Cassell, Jackson, & Cheavront,
1998). Social networking sites provide one-to-many communication, as do mass media;
however, they also are the vehicles for one-to-one communication between a social
support provider and a social support receiver. Therefore, they have the potential to be
one of the most influential kinds of health information web sites on the Internet because
they have some of the benefits of interpersonal communication (e.g., personal
contact),with the benefits of mass media (e.g., broad reach) (Valente, 2002). Therefore,
these sites have the potential for strong interpersonal influence and rapid diffusion of
ideas.
As previously mentioned, diffusion of innovations theory recognizes that there are
key individuals, called opinion leaders, within a community or network that are more
important than others for facilitating the spread of a new idea or innovation. Rogers
(2003) defines opinion leadership as “the degree to which an individual is able to
influence other individuals’ attitudes or overt behavior informally in a desired way with
relative frequency” (p. 27). Opinion leaders are not necessarily the earliest adopters of an
17
innovation, although they do usually adopt innovations before the majority does (Valente
& Pumpuang, 2007). However, these individuals are important because they are role
models for the community and, as such, have a strong interpersonal influence on the
adoption of new ideas or behaviors. Public health program planners have embraced the
concept of opinion leaders as a means of facilitating health behavior change programs.
These programs are sometimes called peer influence, peer education, peer networks,
interpersonal counseling or outreach programs and involve the selection and use of
opinion leaders chosen from within the community to act as role models for health
behavior change (Valente & Davis, 1999; Valente & Pumpuang, 2007).
Opinion leaders are typically identified in a network as the individual with the
most nominations from other network members. This means that other network members
identify them as someone they go to for advice or information, or they might be someone
that others consider as one of their best friends. Someone who is identified as an opinion
leader is someone who has a lot of power in the community (Valente, 1995). The notion
of an opinion leader is important to this dissertation because opinion leadership is a type
of network model that will be applied to this particular network to further understand its
structure and potential influence. This will be further elaborated in Chapter 7, the
methodological chapter that pertains to the social network analysis.
Pregnancy as a Unique Health Issue
This dissertation examines social networking sites that focus on the health issue of
pregnancy and pre-natal care. This health topic was chosen for methodological,
theoretical, and practical reasons. Females were selected as research participants because
18
research has established different patterns of Internet use among women, as compared to
men. For example, research has shown that there are significantly more women than men
who use the Internet to search for health information (Fox & Fallows, 2003; Kommer &
Rainie, 2002) and that women are more likely than men to use the Internet to maintain
larger networks of distant contacts (Boneva, Kraut, & Frohlich, 2001). Women are also
more likely than men to participate in social networking sites (Hargittai, 2007). In
addition, a single sex study was designed in part because research has also shown that
peer-to-peer influence is stronger among same-sex dyads (Christakis & Fowler, 2007;
2008).
Theoretically, this health topic was selected because pregnancy is a health issue
unlike any other. Pregnancy is often considered a joyful time in a woman’s life, with an
anticipated happy outcome. It also has an established trajectory and defined timeline.
However, while pregnancy may generally be a happy time, it may also be accompanied
by ambiguity and stress as pregnant women adjust to their changing bodies, face diverse
pregnancy symptoms and risks, become a patient in the health care system, modify their
lifestyle to accommodate their pregnancy, and prepare for motherhood. Therefore,
pregnant women may find themselves relying on information from diverse sources to
help them navigate the myriad of issues confronting them. In particular, the conditions
facing pregnant women may potentially position the media as a source of expert
information as dependency theories would suggest (e.g. Ball-Rokeach, 1985; 1998).
Therefore, this health topic has the potential to contribute to the theoretical literature
19
related to media/new technology and health impacts, as well as to social support and
social capital especially as they pertain to women and online communities.
Research suggests the Internet is among the sources pregnant women are likely to
reference for health information. In 2008, on a list of the top 10 most popular social
networking sites for women, seven of them had groups or ongoing discussions related to
pregnancy (Ningthoujam, 2008). Further, the data suggest that social networking sites
that focus on the issues of motherhood, parenting and social support have a large volume
of traffic. For example, one of the most popular web sites for women,
www.cafemom.com, was estimated to have had more than 1,000,000 visits in August
2009 as did another site for parents, www.parentsconnect.com. Another site dedicated to
the issue of social support for a myriad of topics, www.dailystrength.org, was estimated
to have had more than 400,000 visits during the same month (retrieved from
www.trafficestimate.com, September, 10, 2009). While these figures are not limited to
women who only participate in the pregnancy communities, several of the pregnancy
groups on these sites report memberships in excess of 1000 members (e.g. dailystrength
and cafemom). While it is difficult to assess how many members actively participate in
these communities at any one time, these data highlight the popularity of these forums.
Finally, pregnancy is a health issue of practical relevance to millions of American
women and newborns. In 2008, the preliminary estimate of US births was 4,247,000,
representing a birth rate of 13.9 per 1,000 (Tejada-Vera & Sutton, 2009). In addition,
according to the National Center for Health Statistics, annually there are approximately
20
1,000,000 pregnancies that end in miscarriage or stillbirth (Ventura, Mosher, Curtin,
Abma, & Henshaw, 2000).
Furthermore, in 2004 the US maternal mortality rate was 13.1 deaths, per 100,000
live births. While this number decreased dramatically during the 20
th
century, the CDC
reports that since 1982 there has been no further decrease in the United States. Moreover,
the World Health Organization reports that there are 40 countries with a lower maternal
mortality rate than the U.S. Maternal mortality, however, is only one complication of
pregnancy. Other, more prevalent, complications include low birthweight babies (8.2%),
depression (10%), and gestational diabetes (2-10%) (www.cdc.gov). The costs associated
with pregnancy complications extend beyond personal losses and include a societal
financial cost. For example, the March of Dimes reports that in 2005 the annual societal
economic costs (includes medical, educational, and lost productivity) associated with
preterm birth in the US was $26.2 billion
(http://www.marchofdimes.com/prematurity/21198_10734.asp).
Therefore, the data show that pregnancy is a health issue that affects a
considerable number of Americans each year. More than 80% of women will deliver an
infant in their lifetime (http://www.cdc.gov/reproductivehealth/maternalinfanthealth/
PregComplications.htm). Furthermore, while most pregnancies result in healthy, live
births, there is a considerable cost associated with complications that do arise (Edlin,
2003).
Studies have found that prenatal health and well-being are associated with
newborn and postpartum health. It is well known, for example, that a woman’s prenatal
21
health impacts the health of her newborn. Inadequate weight gain, improper nutritional
intake, consumption of illegal drugs or alcohol, and smoking have all been identified as
risk factors for low birth weight babies and other complications (e.g., Chomitz, Cheung,
& Lieberman, 1995). Beyond these explicit health behaviors, other psychosocial
variables, such as social support and stress, have also been associated with pregnancy
complications and postpartum outcomes. Several studies have found that a lack of social
support and prenatal stress are associated with lower birthweight babies, preterm delivery
and postpartum depression (Conway & Kennedy, 2004; Cronenwett, 1984; Dejin-
Karlsson, et al., 2000; Feldman, et al., 2000; Liese, Snowden, & Ford, 1989; Raufuss,
2003; Sable & Wilkinson, 2000). The studies suggest that interventions with pregnant
women have the potential to improve health outcomes, especially if they address the
psychosocial variables of stress and social support.
Conclusion
The Internet is now a primary source of health information for many Americans,
and has proven to be particularly popular among women. Among the health topics that
are likely to be important to women are pregnancy and prenatal care. Further, it is well
established that women’s physical and mental health during their pregnancy can have
important health implications for both themselves and their unborn child. In particular,
research has shown an association between a lack of social support and pregnancy
complications and poor postpartum outcomes.
Despite the more than 4,000,000 live births that occur annually in the United
States, pregnancy is a health condition that many women might find themselves
22
experiencing alone, that is, with no one else in their network pregnant at the same time.
This suggests that women might go online in search of a community of other women also
experiencing pregnancy. Furthermore, there has been little research about online social
networks and health outcomes. However, the extensive literature on offline social
networks has shown that networks can influence an individual’s behavior. Social
networking sites offer the dual persuasive edge of relaying interpersonal communication
in a mass media environment, thereby combining the persuasive appeal of personal
contact with the reach of media.
This dissertation attempts to fill the gaps in the literature related to participation in
health-related social networking sites and health outcomes, specifically related to
pregnancy. This dissertation focuses on extending the literature in three ways. First, on a
practical level, it assesses the health impacts of participation on health-related social
networking sites. Second, theoretically, it extends the literature related to social support
and social capital in online environments. Third, methodologically, it extends the
literature by testing several measurements of the same and similar constructs to facilitate
a better understanding of their validity.
The next chapter discusses social support and social capital, including the current
issues in their respective literatures, how they have been examined in the online
environment, and how they are associated with health outcomes. In addition, the chapter
concludes with the presentation of the research questions that guide this dissertation.
23
CHAPTER 2
SOCIAL SUPPORT A
D SOCIAL CAPITAL
Social support and social capital are among the theoretical cornerstones of this
dissertation. Both constructs have substantial research histories that were initiated in
offline environments and have since been applied to online contexts. Furthermore,
because research has found positive associations between these constructs and health they
have both caught the attention of epidemiological and other researchers interested in
public health outcomes. This chapter begins with a brief history and discussion of some
of the key issues concerning these constructs, including the central issues of how to
define and measure them. Next, the chapter examines how these constructs converge and
diverge. Finally, the chapter concludes with a presentation of the research questions that
guide this dissertation.
Social Support
Research on social support has spanned more than three decades and is
voluminous. Scholars from a variety of disciplines including psychology, sociology,
anthropology, epidemiology, gerontology, and communication have studied this
construct, resulting in a multidisciplinary approach that has extended understanding of
social support while also sometimes making it more complicated. The study of social
support is concerned with the process by which social relationships might promote health
and well-being (Cohen, Gottleib, & Underwood, 2000). Two mechanisms have been
identified as the potential causal pathways that enable social support to positively
24
influence health: the first is typically referred to as the main or direct effect model, and
the second is the buffering modeling (Cohen & Wills, 1985).
The main effect model argues that social support has an ongoing positive
influence regardless of whether there is a specific stressful event or not. This model
describes how having regular positive experiences with supportive people creates positive
affect, helps individual avoid negative experiences, and may be associated with healthier
behaviors. In this model, social support impacts physical health through physiological
mechanisms such as neuroendocrine or immune system functioning (Cohen, 2000; Cohen
& Wills, 1985).
In contrast, the buffering hypothesis states that it is primarily in times of stress
that social support is beneficial. Stress is conceptualized as a potential threat for which
individuals feel they lack the requisite coping skills (Cohen & Wills, 1985). Social
support is hypothesized to attenuate the stressful event by either buffering an individual
from negative physiological outcomes through reappraising the event as less stressful,
affecting the neuroendocrine systems, or by encouraging healthy behaviors (House,
1981).
Throughout its research history, social support is a term that has been plagued by
a lack of agreement among researchers as to how it is best operationalized and measured.
In fact, one of the few things researchers have agreed upon is that there is no agreement
on how to best define and measure social support (Barrera, 1986; Cohen, 1988; Dunkel-
Schetter, et al., 1996; House, 1987; Sarason, Sarason, & Gurung, 1997). As a result,
within the social support literature there are at least two distinct research perspectives —
25
one grounded in sociology, the other, psychology — that have guided the investigation of
this construct. The two perspectives are described below.
Within a sociological perspective, social support is conceptualized and measured
as one’s integration, or embeddedness, in a social network. Within this perspective, some
scholars further differentiate between a paradigm that examines the existence or quantity
of an individual’s social relationships as compared to a social network analysis that
examines the overall structure of the network of relationships in which an individual is
situated. The former approach, which examines the existence or quantity of social ties,
focuses primarily on the presence or absence of specific ties (e.g. marital, participation in
community organizations, or contact with friends) and how those relationships may be
associated with particular outcomes (Berkman & Syme, 1979; House & Kahn, 1985;
House, Kahn, McLeod, & Williams, 1985). For example, this approach may be best
exemplified by studies that examine how marital status is associated with health
outcomes. The focus is simply on the presence or absence of that specific social
relationship, without a consideration of its quality. Analyses in this tradition have
shown how the presence of a marital tie is positively associated with health status –
especially for men, less so for women (Berkman, 1984; Berkman & Syme, 1979; House
& Kahn, 1985).
Also within a sociological orientation, social support has been examined with
social network analysis. As briefly discussed in Chapter 1, social network analysis uses a
distinct theory, methodology, set of mathematical models and computer software to
produce network metrics, such as density and centralization, to better understand how the
26
structure and characteristics of a social network are associated with the social support that
flows through it (Berkman, Glass, Brissette, Seeman, 2000; House & Kahn, 1985,
Valente, in press). Simply, a social network can be described as “the web of social
relationships that surrounds individuals” (Heaney & Israel, 2002, p. 185). More
specifically, Hall & Wellman (1985) describe a social network as “a set of nodes that are
tied by one or more specific types of relations between them” (original italics) (p. 25).
The authors further explain that nodes, in social support research, are typically individual
persons and the ties represent the flow of resources from one person to another. The
resources (i.e., support) that move through a network can vary in terms of quality (the
dimension of social support, such as informational or emotional), quantity, (how many
times support flows between these two specific members), multiplexity (how many
different dimensions of support flow through the link), and reciprocity (whether support
flows in both directions or just one). Social network analysis also uses sociograms, as
illustrated in Figure 2.1, to visually depict the structure of networks and the flow of
resources from one entity to another.
Figure 2.1: Example of a sociogram used to depict a social network
The circles are nodes and
represent individuals.
The arrows indicate a
tie and the flow of
support. It can flow
one way or it can be
reciprocal as indicated
by two arrowheads.
27
While the literature sometimes conflates the terms social support and social
networks, it should be noted that not all social networks provide support (Berkman,
1984). Furthermore, in social network analysis, social support is seen as a resource that
flows through the social ties that comprise a social network. While some researchers have
approached the study of social ties strictly through measuring their presence or absence
(e.g. see Berkman, 1984 and Barrera, 1986 for reviews), others have suggested that a
richer analysis is possible by examining the quality of support as well as the social
network (Barrera, 1986; Stansfeld, 2006).
Focusing on the quality of social support is precisely what a psychological
orientation toward this construct does. More specifically, a psychological orientation
typically examines an individual’s cognitive and emotional perceptions of which specific
dimensions of social support are available to them. These different dimensions of social
support are usually referred to as the functional content of relationships. Research in this
tradition may also examine actual support received or perceived adequacy or satisfaction
of support (Barrera, 1986; Burleson & MacGeorge, 2002; House, Kahn, McLeod, &
Williams, 1985). In this paradigm, social support can be defined as “the social resources
that persons perceive to be available or that are actually provided to them by
nonprofessionals in the context of both formal support groups and informal helping
relationships” (Cohen, Gottlieb, & Underwood, 2000, p. 4).
Within the psychological perspective, many typologies of support functions have
been developed. These typologies have often been used as the basis to create
measurement scales that typically use self-reported data to analyze respondents’
28
perceptions of the support available to them. Herein lies an additional weakness within
the study of social support. A large number of support typologies and scales have been
developed, leading some scholars to comment that “most investigators develop their own
scales, so there are almost as many different measures as there are studies” (House &
Kahn, 1985, p. 94). However, among the many typologies there are some similar types of
support identified, although sometimes with different labels. Typical of social support
scales is Cutrona and Suhr’s (1992) typology of social support, in which they distinguish
among five different dimensions of social support: 1) informational; 2) emotional; 3)
esteem; 4) tangible; and 5) social network. Informational support is best represented by
factual advice and feedback. These also might be considered task oriented messages that
provide information or problem solving (Walther & Boyd, 2002; White & Dorman,
2001). Examples of informational support include information on symptoms or
treatments related to a health issue. Emotional support is provided through expressions of
concern, caring, empathy and sympathy (Cutrona & Suhr, 1992; Walther & Boyd, 2002).
Esteem support is provided through recognizing another’s worth or through the
expression of complimentary statements or admiration (Walther & Boyd, 2002). Tangible
aid refers to support in the form of actual physical help. In Miyata’s (2002) study of
online support groups for Japanese mothers, she measured this kind of support by asking
respondents about the provision of childcare services. Finally, social network support
involves referring someone to another person “or group of people who share a common
set of experiences or expertise” (Walther & Boyd, 2002, p.159).
29
Across the different conceptualizations and measures of social support, research
has demonstrated a positive relationship between social support and physical and/or
mental health (Aneshensel & Stone; 1982; Berkman, 1984; Cohen, 1988; Burleson &
MacGeorge, 2002; House, Landis & Umberson, 1988; Krause, 1990). For example,
scholars have found that social support is significantly associated with resistance to and
recovery from disease (Cohen, 2001; Seeman, 2001; Spiegel & Kimerling, 2001);
reducing mortality (Berkman, 1995; House, Landis, & Umberson, 1988); and improving
prenatal health behaviors (Schaffer & Lia-Hoagberg, 1997). Furthermore, positive
associations between social support and health outcomes are not restricted to face-to-face
encounters. Instead, the exchange of social support in online platforms is also associated
with positive effects as described below.
Online Social Support
Several studies have documented the effectiveness of online social support. For
example, Miyata’s (2002) study of online support groups for Japanese mothers found that
mothers who received more Internet support had greater well-being than those who
received less Internet support. Similarly, Dunham et al. (1998) found that young mothers
in a Canadian city who participated consistently in an online social support group
reported greater decreases in stress as opposed to those who participated less often. In a
study of women with breast cancer, Rodgers and Chen (2005) likewise found that
participation in an online support group resulted in greater optimism towards breast
cancer, improved mood, decreased psychological distress, and strategies to manage
stress.
30
Online social support has been popular since the early days of the Internet. It
initially became popular through the use of list-servs and bulletin boards where
participants typically communicated to each other through an email thread that was sent
to all members of a support group. In the era of “Web 2.0” where information sharing and
collaboration are celebrated, social networking sites have emerged as the latest
incarnation of online support communities. Indeed, there are a growing number of social
networking sites that have developed around the exchange of social support for specific
health issues like cancer, (e.g. mycancerplace.com), HIV (e.g. hivpassions.com), and
diabetes (e.g. tudiabetes.com). In addition, several other “umbrella-style” sites have been
created that house many different forums for the exchange of social support for a variety
of health issues (e.g. dailystrength.org and patientslikeme.com).
The Internet remains popular as a channel for the exchange of social support
because it erases the constraints of fixed schedules, geographic location, and
embarrassment that may limit the seeking of face-to-face social support. Another benefit
of online social support sites is their potential to create large, heterogeneous support
communities by flattening out social hierarchies related to gender, social class, ethnicity,
age, and lifestyle (Boase & Wellman, 2006). Subsequently, online forums have been
lauded for their ability to connect diverse individuals in a network of weak ties (Walther
& Boyd, 2002; Wellman & Gulia, 1997; Wright & Bell, 2003). This argument, now
famously coined in the phrase (and article title) “the strength of weak ties” (Granovetter,
1973, p.1360) refers to the notion that weak ties, while lacking the emotional intensity of
strong ties, have the potential to be rich sources of information. Granovetter argued that
31
strong ties link individuals who are homogenous and therefore most likely have access to
the same information. Thus, it is through weak ties that individuals can access new,
diverse information. This, and other research, suggests that relationships with similar
others might be preferable for the provision of emotional support while relationships with
dissimilar others might be preferable for information support (Granovetter, 1973; Preece;
1999; Wright, 2000).
Social Capital
Like social support, social capital is a construct with a substantial research
history. Sociologists Pierre Bourdieu (1986) and James Coleman (1990), economist Glen
Loury (1977) and, later, political scientist Robert Putnam (1993) are often identified as
central figures in this construct’s conceptualization and development. Social capital is the
idea that our social relationships have value. It refers to the resources that are embedded
in social networks that provide benefits to group members (Lin, 1999). Social capital
describes how participation in social relationships yields benefits to social actors that may
be economical, physical, political, emotional or social, among others. Beyond this general
description, the literature on social capital, like that of social support, is plagued with
inconsistencies and disagreements among scholars, thus making a coherent explanation
of the construct vulnerable to even further disagreement.
The conceptual origin of this construct — the idea that group membership can
have beneficial consequences — goes back even further to the classical roots of
sociology and the work of Durkheim and Marx (Portes, 1998). However, beginning in the
early to mid 1990s this concept had a resurgence of popularity among researchers
interested in notions of community. It was at this time that Putnam (1993, 1995) ignited
32
the collective imagination when he wrote about an American decline in social capital and
used the metaphor of the lone bowler to illustrate this concept. Also around that time, the
construct of social capital became popular with researchers interested in the social
determinants of health. Since then, many researchers studying health outcomes have
considered the potential impact of social capital on health. Evidence of this heightened
interest is visible in the rapid increase in publications about this topic. Kawachi,
Subramanian, & Kim (2008) report that from 1996 to 2006 published research papers on
social capital and health increased more than ten-fold.
Some of the tensions within the social capital literature result from disagreement
about the unit of analysis with which to define and measure it. Some scholars see social
capital as an attribute of a group or community, while others see it as an attribute of an
individual. Putnam is among the scholars who regard social capital as a community level
resource. He defines it as the “features of social organization such as networks, norms
and social trust that facilitate coordination and cooperation for mutual benefit” (Putnam,
1995, p. 67). Indicators of social capital in this perspective include civic engagement,
such as participation in community organizations, and mutual trust among community
members (Kawachi, Kenney, Lochner, & Prothrow-Stith, 1997; Putnam, 1995; 2000). In
this perspective, social capital is a property of the group, not of the individuals who
comprise the group. For example, there could be high levels of trust and cooperation
within a group, resulting in high levels of social capital. However, there could be one
individual group member who is not trustworthy, nor cooperative, but that person still has
access to the high social capital of the group (Kawachi, Subramanian, & Kim, 2008).
33
Further complicating matters, in this conceptualization of social capital, it is typically
measured at the individual level through variables such as participation in community
organizations, voter turnout, newspaper readership, etc. These data are then aggregated to
the group level, so that neighborhoods, states or other groups can be compared, rather
than individuals.
Portes (1998), on the other hand, is among the scholars that conceptualize social
capital as an individual-level attribute citing the intellectual legacy of Bourdieu,
Coleman, and Loury. Portes (1998) argues that social capital is best understood as an
attribute of individuals that describes “the ability of actors to secure benefits by virtue of
membership in social networks or other social structures” (p. 6). At the individual level,
social capital is typically measured by examining an individual’s social relationships.
Analyses of this kind will typically measure the quantity and strength of an individual’s
social relationships and identify the resources to which an individual’s relationships
provide him access (van der Gaag & Webber, 2008).
Portes disagrees with conceptions of social capital as an attribute of groups or
communities by characterizing Putnam’s argument as tautological. Portes argues that
social capital at the community-level becomes both a cause and an effect of itself. In
this conception, social capital “leads to positive outcomes, such as economic
development and less crime, and its existence is inferred from the same outcomes”
(Portes, 1998, p. 19).
Scholars also note other dualisms within the study of social capital. For example,
several scholars suggest that there are distinct structural and cognitive components of
34
social capital (Baum & Ziersch, 2003; Hawe & Schiell, 2000; Portes, 1998). This dualism
recognizes that one element of social capital is the networks that are required to form
social capital. People need to be connected and relationships need to be cultivated to
create social capital. The other element of social capital refers to the content, which is the
actual resources to which social capital provides access, which may be intangible goods
such as trust and reciprocity or other more tangible goods, such as money or products.
Another distinction in social capital research concerns the difference between
bonding and bridging social capital (Adler & Kwon, 2002; Kavanaugh, Reese, Carroll, &
Rosson, 2005; Gittell & Vidal, 1998; Putnam, 2000; Yuan & Gay, 2006). Putnam (2000)
describes bonding social capital as the strong ties that develop between people of similar
backgrounds. It is inward looking and reinforces exclusive identities and homogenous
groups, while providing social and psychological support to members of a particular
community. Bridging social capital, in contrast, typically is manifested in weak ties that
link people across disparate social backgrounds. Bridging social capital is beneficial for
providing links to external assets and information diffusion (Putnam, 2000). Putnam also
notes, however, that bonding and bridging are not either-or dichotomies, but that groups
typically manifest more or less of each dimension. Studies have found both bonding and
bridging can be found in online groups, as illustrated in the work of Norris (2004) and
Ellison, Steinfeld, and Lampe (2007).
Norris (2004) examined self-reported data from the Pew Internet and American
Life Project and found that members of most online groups participated in them to meet
both bonding and bridging needs, although bonding was “slightly stronger” (p.36) than
35
bridging. This finding was reflected not only by participants of social support groups, but
also by members of political, religious, and professionally oriented online communities.
Further, although Norris did find some bridging across age groups, her research showed
that participating in online groups did little to facilitate bridging across racial,
socioeconomic or class divides. In another study of how Facebook, an online social
networking site, use was associated with social capital, Ellison, Steinfeld, and Lampe
(2007) found significant associations between Facebook use and both bonding and
bridging social capital. Unlike Norris’s (2004) findings, however, this study found the
association was slightly stronger for bridging social capital. This discrepancy may be due
to the way in which social capital was measured. A discussion of approaches to
measuring social capital follows below.
Measuring Social Capital
Most measurements of social capital are quantitative and use indicators such as
voter turnout, newspaper readership, and membership in voluntary organizations (Hawe
& Shiell, 2000). Several studies have examined social capital with a secondary data
analysis of individual-level survey data that was collected for another purpose, and then
aggregated to the community, state, or national level (Baum & Ziersch, 2003). For
example, Kawachi, Kennedy, and Glass (1999) used data from the General Social
Surveys, aggregated to the state level, as indicators of social capital. From the survey
data, they constructed measures of social capital that included variables such as
community trust, reciprocity, and membership in voluntary associations. Similarly,
Norris (2004) used data collected from a Pew Internet and American Life Project about
36
their experiences on the Internet, such as how much the Internet has helped them find
people who share their interests, and how much the Internet has helped them connect
with people from different racial or ethnic backgrounds, and used those measures to infer
bonding and bridging social capital.
Williams (2006) extended the tools available to measure social capital in both
online and offline contexts by developing a survey instrument to assess this construct.
The Internet Social Capital Scales (ISCS) are two scales for measuring bonding and
bridging social capital in online and offline contexts. The scales are theoretically
grounded in Putnam’s (2000) work on bonding and bridging social capital, as well as
Granovetter’s (1973) work on strong and weak network ties. Consequently, the bridging
scales focus on the externally directed ties that are representative of this kind of capital
while the bonding scales reflect more exclusive, inwardly directed social ties. Ellison et
al. (2007) used these scales in their study of Facebook use and found them to be reliable
(Cronbach’s alpha = .87 for bridging social capital and Cronbach’s alpha = .75 for
bonding social capital).
Another approach to measuring social capital is with social network analysis. As
previously discussed in Chapter 1, social network analysis is a theory and methodology
that looks at the relations among people, organizations, or states (Garton,
Haythornthwaite, & Wellman, 1997; Scott, 2007; Valente, in press). Researchers using
this approach conceptualize social capital as a relational construct (Lakon et al., 2008) in
which the unit of analysis is the tie, or the relation, between entities. Social network
analysis can measure social capital as both an individual and group level attribute
37
(Borgatti, et al., 1998; Kawachi et al., 2008). At the individual level, it can examine
social capital in the ties between an individual and a group (egocentric measures) as well
as between all members of a particular group (network measures) (Lakon et al., 2008).
Network analysts avoid measuring social capital with cultural indicators (e.g. norms,
trust, etc.), and instead examine social capital strictly through network measurements
(Borgatti et al., 1998). For example, some egocentric measures of social capital might
include network size and degree. The more people that are in an individual’s network,
and the more people he or she is connected to (the measure of degree), the more social
capital that can be cultivated, as represented by presumably more resources to which an
individual would have access.
Within social network analysis bonding and bridging social capital can be
measured in several ways. For example, by examining the characteristics of the nodes
that are connected together (e.g., the ties) social network analysts can identify whether
the ties are more representative of either bonding or bridging social capital. Ties of
similar socio-demographic characteristics suggest the presence of bonding social capital,
while connections between dissimilar ties suggest the presence of bridging social capital
(Lakon et al., 2008). Furthermore, tie strength is also an indicator of bonding and
bridging social capital. Tie strength is a “combination of the amount of time, the
emotional intensity, the intimacy (mutual confiding) and reciprocal services which
characterize the tie” (Granovetter, 1973, p. 1361). Some researchers also suggest that the
exchange of emotional support is an indicator of tie-strength (Wellman, 1982; Wellman
38
& Wortley, 1990). Stronger ties are an indication of bonding social capital, and weaker
ties suggest bridging social capital (Granovetter, 1973; 1974; Williams, 2006).
Social Capital and Health
Several studies of social capital have examined how indicators of social capital,
such as trust and participation in community organizations, are related to mortality rates
(Pearce & Davey Smith, 2003). For example, Kawachi, Kennedy, Lochner, and
Prothrow-Stith (1997) analyzed data from 39 US states and found mortality rates were
strongly related to per capita density of membership in voluntary groups and by levels of
social trust, both indicators of social capital. In a follow-up study, Kawachi, Kennedy,
and Glass (1999) also found an association between low social capital and self-rated poor
health, even after controlling for individual-level factors such as low income, low
education, and smoking. In 2005, a study conducted in Australia found that regular
involvement in informal social networks (an indicator of social capital) was positively
associated with better mental health, but not with better physical health (Ziersch, 2005).
Online Social Capital
In the context of the Internet, social capital has been applied primarily as a means
of investigating online communities. Specifically, there has been an active debate in the
literature about the meaning of community on the Internet, especially as it compares to
offline communities and social capital has been a part of this discussion. In a search of
the literature, no studies were found that examine social capital and health outcomes in
online contexts. Moreover, Viswanath (2008) wrote “it is too early to predict the impact
of the Internet…on interpersonal communications, social capital and health” (p. 266).
39
This dissertation attempts to fill this gap by examining how social capital in an online
community about a health issue is related to health outcomes.
The Convergence and Divergence of
Social Support and Social Capital
The literatures that have developed around the concepts of social support and
social capital, while anchored in very different theoretical frameworks, have many
similarities. The similarities extend from how each construct is conceptualized and
measured to its application. Many scholars who theorize about one of these constructs
typically give a salutatory nod of recognition to the other construct, but the link between
these two constructs has not been examined in depth. Below, some of the significant
similarities and divergences of these constructs are addressed, as well as their
implications.
Conceptualization
As previously noted, the social support literature is fractured between a
sociological and psychological perspective, which contrasts an emphasis on the structure
of an individual’s relationships as opposed to the supportive content of them.
Interestingly, the tensions within the social capital literature parallel those in the social
support literature. The split in the social support literature between a sociological
approach and psychological approach is thematically similar to the split in the social
capital debate about whether social capital is a property of the group or the individual.
The tensions in both literatures reflect a struggle between an external focus on the overall
40
structure of a network of relationships versus an inward focus on individual perceptions
and access.
A close reading of the scholarship related to social support and social capital
reveals further similarities. Social capital is the notion that individuals can benefit or
profit from their social contacts. In essence, people can use their relationships to get
ahead or achieve a personal goal. The classic example is an individual using his or her
contacts to find a new job, as seen in Granovetter’s (1973) study. Coleman (1990), a key
social capital theorist, describes social capital as “resources that can be used by actors to
realize their interests” (p. 305). At the individual attribute level, the resources that one’s
social relationships provide can assume a variety of forms, such as information, access to
new contacts, and even tangible benefits, such as money or goods.
Many theorists working with social support also describe social support as
consisting of resources from one’s personal network. For example, Cohen and Hoberman
(1983) describe social support as “the various resources provided by one’s interpersonal
ties” (p. 100). In Burleson and MacGeorge’s (2002) review of the construct, they cite
Caplan’s (1974) work on enacted support and his definition of it as “supplies of money,
materials, tools, skills, and cognitive guidance” (p. 380). Similarly, Cohen et al. (2000)
describe social support as “social resources that persons perceive to be available or that
are actually provided to them” (p. 4). Moreover, within the psychological perspective all
of the various scales reflecting the dimensions of social support, such as informational,
emotional, esteem, social network, and tangible (Cutrona & Suhr, 1992) reflect many of
the same resources that are available through an individual’s social capital. Therefore,
41
where social capital theorists focus on “resources” and social support theorists focus on
“support,” functionally these two elements are very similar.
Furthermore, there are even more direct parallels between bonding social capital
and bridging social capital. Much of the currency of bridging social capital is focused on
the informational benefit of connections with diverse others. Moreover, weak ties have
been shown to be better resources (e.g. Granovetter, 1973) for bridging social capital. In
contrast, bonding social capital is typically more focused on the exchange of emotional
support, of which strong ties are better resources. Therefore, bonding social capital can be
shorthand for the exchange of emotional support, while bridging social capital can be a
proxy for informational support.
Measurement
Most importantly for present purposes, the overriding similarity in how both
social support and social capital can be measured is through the use of social network
analysis. As described above, there are theorists working with each construct who
approach their work from a sociological perspective that focuses on social integration. In
this approach, these researchers situate the flow of resources — be them social capital or
social support — within a social network. With both constructs, the social network is
seen as central in facilitating the flow of these resources to network members. Therefore,
the structure of the network — who is connected to whom — is important for both.
Furthermore, researchers have suggested using the same network metrics as evidence of
both social support and social capital (e.g. Borgatti et al., 1998; Lakon, Godette, & Hipp,
2008; Wellman, 1982).
42
Application
As previously described, both social support and social capital have been
associated with positive health outcomes. Furthermore, the suggested causal pathways
linking these constructs to health outcomes have also been proposed to be the same.
As this section has suggested, there are many similarities in how social support
and social capital are theorized, measured, and applied. However, a fundamental
difference of these two constructs is the larger theoretical framework within which each
construct resides. Social capital is anchored in an economic model and, following the
capital metaphor, it has often been theorized as a means to an end, rather than an end in
itself. In contrast, social support is typically theorized as an end in itself. Moreover, social
support theory tends to lacks the broad, societal, implications inherent in social capital.
However, while the theoretical frameworks are different for these constructs, what
remains unresolved is whether social support and social capital are experienced
differently by individuals in their day-to-day lives. This dissertation seeks to provide
insight into this relatively unknown relationship as reflected in some of the research
questions, which are discussed below.
Research Questions
The first set of research questions examine the health topics that are addressed on
social networking sites related to pregnancy. Since the Internet is becoming a premier
source of health information that is especially popular with women, it is important to map
the informational landscape of these online communities and identify the health topics for
which women are seeking support. This analysis has practical implications for health care
professionals by providing information related to the concerns, health beliefs, and needs
43
of this particular population. Thus, the dissertation addresses the following research
questions:
RQ1: What are the demographic characteristics of this population of social
networking members?
RQ2: What type of health information (symptoms, complications, treatment,
diagnosis, risk factors, prevention) is sought in these forums?
RQ3: What are the specific health issues that are discussed in these forums?
RQ4: Which dimensions of social support are being exchanged in these
forums?
The second set of research questions examines the health impacts of participation
on a pregnancy-related social networking site. Currently there is little or no research that
examines the association of participation on a health-related social networking site with
health outcomes, such as knowledge, attitudes, and behaviors — commonly used
outcome variables in the public health field (Valente, 2002). This dissertation will begin
to fill the void in the literature by examining whether the frequency and intensity of use
as well as participation are associated with knowledge of prenatal care, attitudes towards
prenatal care, and prenatal behaviors. Thus, the dissertation addresses the following
research questions:
RQ5: How is frequency and intensity of use of a health-related social
networking site associated with health outcomes? Specifically, is participation
significantly associated with: (a) knowledge of prenatal health, (b) attitudes
toward pregnancy and prenatal health, and (c) and health behavior?
44
RQ6: How is participation on a health-related social networking site
associated with health outcomes? Specifically, is participation significantly
associated with: (a) knowledge of prenatal health, (b) attitudes toward
pregnancy and prenatal health, and (c) and health behavior?
Furthermore, both social support and social capital have been associated with
positive health outcomes (Aneshensel & Stone; 1982; Berkman, 1984; Cohen, 1988;
House, Landis & Umberson, 1988; Kawachi, Kennedy, & Glass, 1999; Kawachi,
Kennedy, Lochner, & Prothrow-Stith, 1997; Krause, 1990; Pearce & Davey Smith, 2003;
Ziersch, 2005). Social capital is also a construct that theorists suggest has two distinct
dimensions of bonding and bridging. These two distinct dimensions will be tested
individually to see if they are associated with health behaviors.
Moreover, since bridging social capital facilitates informational exchange and
bonding social capital facilitates emotional support a separate research question will
examine how the social support measures that explicitly assess these dimensions of
support (informational and emotional) are, themselves, associated with health outcomes.
Results from these two sets of analyses will help assess how these two constructs
converge or diverge. Finally, social network measures of social support and social capital
will be used to predict outcomes to better understand the influence of network relations
on health. Thus, the dissertation will address the following research questions:
RQ7: How is the perception of social support on a health-related social
networking site associated with health outcomes? Specifically,
45
a) Is the perception of global social support significantly associated
with: (i) knowledge of prenatal health; (ii) attitudes toward pregnancy
and prenatal health; and (iii) health behavior?
b) Are there differences in how the specific dimensions of informational
and emotional support are associated with: (i) knowledge of prenatal
health; (ii) attitudes toward pregnancy and prenatal health; and (iii)
and health behavior?
RQ8: How is social capital on a health-related social networking site
associated with health outcomes? Specifically,
a) Is the perception of global social capital significantly associated with:
(i) knowledge of prenatal health; (ii) attitudes toward pregnancy and
prenatal health; and (iii) and health behavior?
b) Are there differences in how the specific dimensions of bonding and
bridging social capital are associated with: (i) knowledge of prenatal
health; (ii) attitudes toward pregnancy and prenatal health; and (iii)
and health behavior?
RQ9: How are social network measures of social support and social capital on
a health-related social networking site associated with health outcomes?
Specifically, are these measures associated with: (a) knowledge of prenatal
health, (b) attitudes toward pregnancy and prenatal health, and (c) and health
behavior?
46
The final series of research questions focus on the theoretical and methodological
issues related to social support and social capital. Most of the analyses rely on the social
network analysis either completely or in part. In a broader sense, most of the questions
seek to untangle the theoretical and methodological overlap between social support and
social capital. Furthermore, these research questions seek to better understand which
dimensions of social capital (bonding or bridging) and social support (with an emphasis
on emotional and informational) are most strongly cultivated or provided on the site.
Finally, this dissertation seeks to further understand how social support and social capital
flow through social networking sites, as well as the relation between tie strength and
these resources. Thus, the dissertation will address the following research questions:
RQ10: Are social network measures of social support and social capital more
strongly related to survey scales of social support or social capital?
RQ11: How does the structure of the network change based on the type of
social support being provided?
RQ12: Is bonding or bridging social capital more strongly cultivated on
health-related social networking sites?
RQ13: How is the strength of a tie associated with the exchange of social
support?
a) Are weak ties more likely to exchange informational support?
b) Are strong ties more likely to exchange emotional support?
RQ14: Across which, if any, socio-demographic variables do the ties exhibit
diversity?
47
CHAPTER 3
METHODOLOGY
Methodology Organization
This dissertation uses three methodologies to address the research questions
identified in the previous chapter. First, a content analysis is used to identify the health
topics and social support exchanged on the social networking sites. Second, a survey is
used to assess the health-related impacts of participation on social networking sites on
site members. Finally, a social network analysis is used to explore theoretical and
methodological issues related to measuring and comparing social support and social
capital constructs. The protocol and instruments for each methodology were reviewed
and approved by the university’s Institutional Review Board (IRB).
The dissertation is structured such that each method and its associated measures
are presented in their own chapters, immediately followed by their findings in the
subsequent chapter. The dissertation is structured this way so that the reader will be able
to more easily follow the various measures within each methodology and their associated
findings. In this chapter, the content analysis is described, including its measures and
instruments. In the next chapter, the results from the content analysis will be presented.
Content Analysis Methodology
A content analysis was conducted to systematically identify the contents of the
messages exchanged in these communities, the socio-demographic characteristics of
community members, and to gather network data for the social network analysis.
48
The Sample
The sample consisted of social networking sites that were created around the topic
of pregnancy. A social networking site (SNS), in this project, is defined as an online
community in which group members participate in discussions related to the topic-of-
interest. Typically, for an individual to participate on these sites they have to become a
member. Becoming a member to most sites is free. The process is usually quick and
involves picking a screenname (the name that the member will be identified as in her
postings), a password to log on to the site, and filling out an online form with some basic
information.
These sites are distinguished from list servers or other online chat rooms in the
following ways. First, unlike a list server where information is exchanged via an email
message that is sent to all members, SNSs are web sites that users log on to participate.
The transcript of ongoing conversations (the message thread) remains posted on the web
site for an indeterminate amount of time, and members can reference these threads,
and/or contribute to them as long as they remain uploaded on the site. Second, social
networking sites are sites in which members are encouraged to create a personal profile
on their own web page hosted on the site. These profiles typically describe the member’s
interests, what state they live in, their age, and other demographic and personality
variables. Often the profiles include a personal photograph (a head shot), and/or
photographs of their family and friends. Members are rarely identified by their full name,
more often the members’ screenname is a pseudonym, such as Momof2, Janesmom,
MommytoaPrincess, etc. In addition, social networking sites also allow users to construct
49
social networks — a network of other users with whom they have a social connection —
and make these networks visible to other participants (boyd & Ellison, 2007).
The sample was identified through Google search engine. Key word combinations
such as “pregnancy, support,” “pregnancy, forum,” “pregnancy, social network,” “mom,
network” were entered into the search algorithm. The results that were returned from this
search were then clicked through and individually examined to see if the site met the
criteria, as described above. The principal investigator read through approximately three
pages of results or until the results appeared to be less and less related to the search terms.
In all, there were 10 social networking sites discovered around the topic of pregnancy
that met the criteria previously stated.
Of these 10 social networking sites several of them were sub-sections of larger
web sites that were created for a broader purpose. For example, there were web sites
about motherhood, social support for a variety of health issues, and parenting. Within
these web sites, there were smaller communities of participants interested in discussing
pregnancy and prenatal care. It is these smaller communities and web pages that are the
focus of this analysis.
Of these 10 pregnancy-related social networking sites, the two sites that had the
most messages, without exceeding 300 messages in the month previous to the one
selected for analysis, were chosen for the content analysis and social network analysis.
Three hundred messages was selected as the cut-off point for two reasons. First, 300
original postings provide a sizeable sample to conduct a comprehensive analysis.
Secondly, the costs of coding more messages per site became prohibitive.
50
From these two sites, all of the “original” messages (the initial posting in a thread
which is typically support-seeking), along with the first 10 responses were analyzed for
one month during the fall of 2008. Messages posted from the site moderator(s) and
solicitations were excluded from the sample. In addition to coding the message content,
the socio-demographic information of the message posters (seekers and respondents) was
also coded. The coding categories are elaborated further below and the full text of the
code sheet is available in Appendix A and B.
Coding Procedures
Pilot testing. Prior to content analyzing the messages, three coders were trained
through an in-depth tutorial on the coding sheets and coding manual. The coders
conducted the pilot testing of the code sheets by coding a sub-sample of 25 messages in
batches of five to seven, and then comparing their coding. Any differences in coding
were discussed, and refinements were made to the code sheets and coding protocol.
Coders continued coding all 25 messages and by the end of the pilot testing the coders
were trained in the methodology.
Interrater reliability. After pilot testing the instrument, only two coders (the
principal investigator plus an assistant) participated in the remainder of the study.
Interrater reliability was assessed on approximately 15% of both the advice seeking
messages and the responses. Reliability was computed with simple percent agreement,
one of the most popular coefficients of reliability, because of its ease of understanding
(Neuendorf, 2002; Rourke, Anderson, Garrison, & Archer, 2001). Simple percent
51
agreement was computed for each variable by dividing the number of times the the two
coders agreed on each variable by the total number of judgments of that variable.
While Cohen’s kappa and Scott’s pi are also frequently-used reliability
coefficients, they have been criticized as being too conservative (Neuendorf, 2002).
Furthermore, these coefficients are problematic in the case of extreme distributions where
a high percent agreement may equate to a low Cohen’s kappa or Scott’s pi (Perrault &
Leigh, 1989; Potter & Levine-Donnerstein, 1999 as cited in Neuendorf, 2002). In these
analyses, Cohen’s kappa was particularly problematic and misleading for several
variables, wherein low variance resulted in a very low kappa. For example, while percent
agreement for seeking informational support was 93%, the kappa value was only .25.
This variable is a good illustration of the paradox that Feinstein & Cicchetti (1990)
discuss wherein a variable with high percent agreement results in a low Cohen’s kappa
due to the fact that kappa is affected by the prevalence of the variable under examination.
In this example, most messages sought informational support; therefore, in calculating
Cohen’s kappa the marginal totals were often unbalanced, but fairly symmetrical – a
situation which reduces kappa. As Viera & Garrett (2005) explain, “kappa may not be
reliable for rare observations” (p. 362). Therefore, only percent agreements will be
presented to avoid misleading kappa values.
On the support seeking messages, reliability ranged from 78% to 100%, and on
the support providing messages, reliability ranged from 80% to 100%. For the seeking or
providing of each dimension of support, reliability was as follows: (1) informational
support, 93% and 83%; (2) emotional support, 81% and 80%; (3) Esteem support, 97%
52
and 93%; (4) relationship support, 89% and 92%; and (5) tangible support, 100% in each
reliability sample. See Table 3.1 for complete reliability data.
Table 3.1: Interrater reliability
Variable Percent agreement
Advice seeking messages
Explicit request for support 98
Seek informational support 93
Seek emotional support 81
Seek esteem support 97
Seek relationship support 89
Seek tangible support 100
Seek health information 80
Kinds of health information (prevention,
symptoms, diagnosis, treatment, complications)
91 (average)
Health topic 99 average
Other informational support topic 97 average
Socio-demographic characteristics of
advice-seeker
Geographic location of advice-seeker 95
Gender 100
Race 96 average
Marital status 84 average
Employment status 90 average
Pregnancy status 89
Previous pregnancies 79
Age (Pearson’s r) .999, p < .001
Advice providing messages
Provide informational support 83
Provide emotional support 80
Provide esteem support 93
Provide relationship support 92
Provide tangible support 100
Negative comments 99
Socio-demographic characteristics of
advice-provider
Geographic location of advice-provider 93
Gender 100
Race 80 average
Marital status 76 average
Employment status 89 average
Pregnancy status 82
53
Table 3.1, Continued
Variable Percent agreement
Socio-demographic characteristics of advice-
provider
Previous pregnancies 77
Age (Pearson’s r) .999, p < .001
Protocol. The coders independently coded the sample of supportive messages.
After reliability data were collected each coder began coding a sub-sample of messages.
Each coder coded approximately one half of the entire sample, which included two weeks
of messages from each site. Coders first coded the messages manually on paper code
sheets. After the sample of messages was coded, the data were entered into an electronic
code sheet using Qualtrics online survey software. This online code sheet allowed for
easy conversion of the data into an SPSS data file for analysis.
Instrument. This content analysis used two code sheets to analyze the sample of
support messages. The first codesheet dealt with the contents of the support seeking
message and the socio-demographic characteristics of its poster, including age, race,
geographic location, and whether this was a first pregnancy. The second code sheet dealt
with the contents of the support-provision messages and the socio-demographic
information of the support-provider. Social support was coded using a typology based on
the work of Cutrona and Suhr (1992) that identified five different dimensions of social
support: informational, emotional, esteem, social network, and tangible. This support
typology identifies support categories commonly recognized in the support literature and
it has also been applied in another study of online support messages (Braithwaite,
Waldron, & Finn, 1999).
54
Units of Analysis
The sample included all messages that were posted on the two selected sites
during the month-long study period. There were 704 members who participated during
the month across both sites, all of them women except one. All of the members’ screen
names were converted to numeric codes to protect their confidentiality. The collection of
socio-demographic data was limited to the information that was provided by members,
either through their head shot or their personal profile. For example, the ethnicity of
participants was primarily coded based on the headshot that members may have posted of
themselves. Without such an image, it was virtually impossible to collect this
information. Likewise, all other socio-demographic variables were coded either from
information posted in the member’s profile, or information posted in their message.
These variables included the members’ geographic location (U.S., international, unable to
tell) along with the specific U.S. state or international country; their gender; race; marital
status; employment status; age; pregnancy status (whether they were currently pregnant
or had been pregnant before); and their number of biological children.
Content Analysis Measures
Social Support
Cutrona and Suhr’s (1992) typology of social support provided the framework for
the content analysis coding categories. These coding categories reflected the different
dimensions of social support that were asked for and provided on the social networking
sites. There were five dimensions of social support that were coded: 1) informational;
2) emotional; 3) esteem; 4) social network / relationship; and 5) tangible. The content
55
analysis coded what kind of support was being asked for and what kind of support was
offered in response. In the current study, the application of this typology of support was
also further informed by Braithwaite, Waldron, & Finn’s (1999) study of social support
in online groups for people with disabilities. In this study, the researchers specifically
defined social network / relationship support as an attempt to create structural
connections, links between people, as opposed to emotional connections. This was a key
distinction not explicitly described in earlier work by the original researchers.
Informational support included the request for knowledge or understanding about
a specific issue or situation. Informational support included requests for suggestions /
advice; referrals to other sources of knowledge/expertise; situation appraisal; and
teaching. Emotional support referred to the request for emotional feedback or
understanding, where the problem was framed in terms of emotions and the focus of the
message was typically about someone’s feelings. This kind of support included affection;
sympathy; understanding/empathy; encouragement; and prayer. Esteem support typically
referred to one’s skills, abilities, and intrinsic value. It is distinguished from emotional
support because it focused on a person’s self-concept and self-perceptions rather than
their feelings. This kind of support included validation and relief of blame. Social
network support, as previously stated, was understood in the context of Braithwaite,
Waldron, & Finn’s (1999) study where they defined social network support as “attempts
to create structural connections” (p. 135) among people with similar interests and
concerns. In this study, we renamed this dimension relationship support because one
might argue that any posting to an online community reflects the seeking of social
56
network support. Therefore, relationship support included providing access to new
companions; solicitations for friendship; physical presence; and companionship. Finally,
tangible support referred to requests and provisions of physical aid, such as objects like
money or maternity clothes or services like babysitting. These exchanges can either be
loans or gifts. The dimensions of support are described in the context of support-seeking;
however, a similar understanding of these dimensions was also used to identify what kind
of support was being provided. See Table 3.2 for a complete description and examples of
the different dimensions of social support that were identified in this study (Braithwaite,
Waldron, & Finn, 1999; Cutrona & Suhr, 1992; Cutrona, Suhr, & MacFarlane, 1990).
Table 3.2: Typology of social support used in content analysis
Dimension Definition Examples
Informational
support
Requests for
knowledge or
understanding
about a specific
issue or
situation.
Suggestions/advice: Someone to offer ideas
and suggest actions, (e.g. “When should I tell
my boss I’m pregnant?”)
Referrals: Referral to some other source of
help or information, (e.g. “Where can I find
out about licensed daycare providers in my
area?”)
Situation appraisal: Someone to reassess or
redefine the situation, (e.g. “I’m only 30
years-old and my doctor has suggested an
amniocentesis. I thought these were only
recommended after age 35. Why would he
recommend one now?”)
Teaching: Someone to explain the facts, or
news about a situation or about the skills
needed to deal with the situation, (e.g. “What
are the risks associated with an
amniocentesis?”)
57
Table 3.2, Continued
Dimension Definition Examples
Emotional
support
Problem is
framed in terms
of emotion and
participant is
seeking some
kind of
emotional
feedback or
understanding.
Affection: Includes “virtual” physical contact
such as sending hugs, or using emoticons.
Generally this kind of support is provided,
without being asked for (e.g.: ☺)
Sympathy: Someone to feel sorrow or regret
for the support-seeker’s situation, (e.g. “I
wanted to tell everyone that I miscarried my
baby at 9 weeks.”)
Understanding/empathy: Someone to
understand the situation, often through
personal experience, (e.g. “Am I the only one
whose mother-in-law is making them
crazy??!!”)
Encouragement: Someone to provide the
recipient with hope and confidence, (e.g. “I
am really scared of going through labor. I
don’t think I can push this baby out.”)
Prayers: Someone to pray with/for another
member, (e.g. “Please pray for me and my
baby, that everything will be okay.”)
Esteem support Focuses on a
person’s self-
concept and
self-
perceptions.
Validation: Someone to express agreement
with the support-seeker’s perspective on the
situation (e.g. “Did I do the right thing by
telling my mother-in-law that I don’t want
her in the delivery room?”)
Relief of blame: Someone to alleviate the
support-seeker’s feelings of guilt about the
situation (e.g. “I am feeling so guilty that
something I did caused my miscarriage.”)
Compliment (coded in the provision of
support only): Saying positive things about a
support-seeker, (e.g. “You are so strong for
handling everything the way you did.”)
58
Table 3.2, Continued
Dimension Definition Examples
Relationship
support
Attempts to
create structural
connections
with other
individuals or
groups.
Access: Someone to provide access to new
companions, including access to another
support group (e.g. “My baby tested positive
for Down’s Syndrome, do you know anyone
else who’s gone through this that I could talk
to?”)
Friendship: An explicit request to make
friends and participate in the group. Also
includes any exchange of personal email
addresses or requests/offers to talk at a later
date or in a private conversation, (e.g. “I just
joined this group because I’d like to make
some new friends with other moms-to-be.”)
Physical presence: Someone looking to meet
other women in same geographic location.
(e.g. “Is anyone else living in the Boston area
who would like to meet for coffee?”)
Companions: Someone who needs to be
reminded of the availability of existing
companions with whom they can talk to
about a problem, (e.g. “I don’t think there’s
anyone in my life who I can really talk to
about how I’m feeling right now.”)
Tangible
support
Requests for
physical aid.
Loan: A request to lend the support-seeker
something including money.
Gift: A request to give the support-seeker
something, including money.
Direct task: A request to perform a task
directly for the support-seeker, (e.g.
babysitting).
Health-related Measures
If the coders identified a message as seeking information, a second question asked
whether health information was sought. Health information was defined very broadly and
59
was used to identify “any topic or issue that dealt with the human body.” If health
information was sought coders identified the general kind of health information that was
sought as among the following: prevention/screening, information about how an injury or
complication could be prevented, such as amniocentesis and other pre-natal tests; risk
factors, information about the risk factors associated with a particular complication, such
as age, family history, pre-existing medical condition, etc.; symptoms, pregnancy-related
symptoms, such as morning sickness; diagnosis, the act of identifying whether one is
pregnant; treatment, a medical course of action, such as bed-rest or dietary restrictions;
and complications, serious and unanticipated complications of pregnancy, such as
gestational diabetes. The other category captured health issues that were not subsumed
under the other categories.
In addition, the specific health topic of the support-seeking message was captured
through a variable that listed 27 pregnancy-related health issues, including different
specific symptoms, morning sickness and weight gain, pregnancy complications, such as
gestational diabetes and breech baby, and issues having to do with the baby’s health, such
as circumcision and baby’s daily health, e.g. breastfeeding, etc. Health issues that did not
have a pre-determined category established were coded under an open-ended “other”
category and during analysis new categories were created for health issues that appeared
multiple times, such as stomach/digestion complaints and general aches and pains.
60
CHAPTER 4
CO
TE
T A
ALYSIS RESULTS
During the one month examined in this study September 2008, there were 572
solicitations for support (original requests) posted across both target sites. Of those
messages, 92% (525) contained explicit requests for support. In response, there were
2,380 messages posted, with a range of 0 – 24 responses per original request.
Interestingly, 12% (64) of messages that explicitly asked for support did not receive any
response. Thirty percent of the support solicitations also included follow-ups from the
advice-seeker (the woman who started the message thread). The number of these follow-
ups ranged from 1 to 7, with a mean of less than 1 follow-up per message (mean = .45).
The final sample of responses excludes these follow-ups and only examines support
provided from other members. Furthermore, the sample is limited to a maximum of 10
responses per support-seeking message. Therefore, the final sample consists of 1965
responses, with a range from one to 10 responses per support-seeking message. As
illustrated in Figure 4.1, there was a mean of 3.44 and a mode of 1 response per original
message.
RQ1. What are the demographic characteristics of this population of social networking
members?
There were 704 members across both research sites. All of the members were
women, except for one. Forty-six percent of the members were Caucasian, three percent
were African American, and less than one percent were Asian American or Hispanic. The
mean age of members was 27, with a range of 16 – 47 years, and a standard deviation of
61
Figure 4.1: Distribution of responses per support solicitation
5.74. Forty-five percent of the members were married and 19 percent had an unmarried
partner. Eighty percent of the members were pregnant at the time of the study and 28%
had had multiple pregnancies. As for location, 56% of the members were living across 49
US states, plus the territories of Guam and Puerto Rico; 11% were living in other
countries; and for 34% of the members, their location was unknown. In the US, the five
states with the most members were: 1) Texas (4% of members); 2) Pennsylvania (3% of
members); 3) Florida (3% of members); 4) Ohio (3% of members); and 5) New York
(3% of members). Table 4.1 reports the complete demographic description of the sample.
The majority of members provided advice only (50%, 353 members), while 30%
(209) sought advice only, and 20% (142) of members assumed both roles during the
course of the month. There were also some demographic differences between the
members who sought advice only and those who provided advice only. The advice
providers were older than the advice seekers, (28 versus 26, t (182) = -2.06, p<.05) and
more likely to be Caucasian (55% versus 32%, χ
2
(1) = 6.39, p<.05).
14%
19%
16%
12%
8%
7%
6%
4%
3%
4%
7%
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
20%
0 1 2 3 4 5 6 7 8 9 10
Percentage of Overall Support
Solicitations
Number of Responses Per Support Solicitation
62
Table 4.1: Characteristics of social networking members
= 704
Support role
Advice-seeker only 30%
Advice-provider only 50%
Both advice-seeker and advice-provider 20%
Gender
Women 99%
Ethnicity
Caucasian 46%
African American 3%
Hispanic >1%
Asian American >1%
Unable to tell 49%
Marital status
Married 45%
Unmarried partner 19%
Single 3%
Divorce or separated >1%
Unable to tell 32%
Pregnancy status
Currently pregnant 80%
Multiple pregnancies 28%
Location
US 55%
International 11%
Unable to tell 34%
Age
Mean age 27
Unable to tell 63%
RQ2.What type of health information (symptoms, complications, treatment, diagnosis,
risk factors, and prevention) is sought in these forums?
Sixty-six percent (n = 379) of messages sought health-related information. As
Figure 4.2 illustrates, of those messages seeking health information, 37% (140) sought
information about pregnancy symptoms. Thirty percent (112) of messages sought
63
information related to complications; 17% (66) sought information about treatment; 14%
(53) sought information related to a diagnosis; 11% (43) sought information related to
risk factors; 5% (17) sought prevention information; and 7% (27) sought information
related not easily categorized into one of the above categories. It is important to note that
these categories are not exclusive—any one message might have been asking for more
than one kind of health information.
Figure 4.2: The different kinds of health information sought by advice-seekers
RQ3. What are the specific health issues that are discussed in these forums?
A large variety of health issues were addressed in these forums. The health issues
that evoked the largest amount of advice-seeking were: 1) determining a pregnancy
diagnosis (11%); 2) labor and delivery (9%); 3) prenatal testing (9%); 4) miscarriage
(6%); 5) food safety and nutrition (6%); 6) abdominal pain (as a potential complication)
(6%); 7a) baby movements (5%); 7b) morning sickness (5%); 7c) other miscellaneous
symptoms (5%); 8) weight gain (5%); 9) general aches and pains (4%); and 10) stomach /
digestion complaints (4%). Together these issues accounted for 75% (279) of all the
health-information seeking on these sites. There were many more health issues that were
7%
5%
11%
14%
17%
30%
37%
0% 10% 20% 30% 40%
Other
Prevention
Risk Factors
Diagnosis
Treatment
Complications
Symptoms
64
addressed, albeit with much less frequency. The health issues addressed included not only
issues relating to the mother’s health and the status of the pregnancy, but issues that
addressed the baby’s health and well-being after birth, such as breast and formula
feeding. However, as Table 4.2 suggests, a central concern of advice-seekers was on
managing various pregnancy-related symptoms.
Table 4.2: Most common health topics for advice-seekers
Health Issue (n = 379) Rank Frequency Percent
Determining
pregnancy
1 42 11%
Labor and delivery 2 36 9%
Prenatal tests 3 34 9%
Miscarriage 4 24 6%
Food safety and
nutrition
5 22 6%
Abdominal pain
(complications)
6 21 6%
Baby movements 7 18 5%
Morning sickness 7 18 5%
Miscellaneous
symptoms
7 18 5%
Weight gain 8 17 5%
Aches and pains 9 15 4%
Stomach/digestion
issues
10 14 4%
RQ4. Which dimensions of social support are being exchanged in these forums?
An analysis of the support-seeking messages shows that informational support
was most in demand, followed by emotional support. Relationship and esteem support
were far less in demand. There were no requests made for tangible support. Interestingly,
as Figure 4.3 illustrates, the kind of support asked for was not always provided. For
example, while 87% of messages sought informational support, only 77% of responses
65
provided this kind of support. Further, 49% of messages sought emotional support while
58% of messages provided this kind of support. There was, however, an even exchange
of esteem and relationship support, with 4% of messages asking for and receiving these
kinds of support. Finally there were no requests for tangible support, and none was
provided.
Figure 4.3: The exchange of five different dimensions of social support
0%
4%
4%
58%
77%
0%
4%
4%
49%
87%
0% 50% 100%
Support Sought
Support Provided
66
CHAPTER 5
SURVEY METHODOLOGY
A survey of social networking web site members was conducted to examine how
participation on such a web site is associated with health outcomes. In addition, the
survey collected social support and social capital data to examine how these constructs
are associated with health outcomes. Finally, social network data were also collected for
the social network analysis.
Survey respondents were recruited from the same two social networking sites
used in the content analysis. After receiving permission from the site moderators,
recruitment messages were posted in the online forums inviting members to participate in
an online survey related to their social networking experiences. To increase the potential
number of survey respondents, the recruitment notice was also posted on six additional
pregnancy-related social networking sites, half of which had a significantly greater
volume of traffic than the two sites used in the content analysis. Most of the sites that
were contacted gave permission to post the recruitment message. However, one large site
denied access on the basis that it wanted to “protect” its members and several others
asked that the recruitment message be posted in a forum specifically for solicitations.
Further, while the sites were told that their names would not be published for the privacy
of their members, two of the sites (americanpregnancy.org and babyfit.com) specifically
asked that their support in this study be reported. The recruitment message was posted on
eight social networking sites in all.
67
Survey respondents were restricted to adult (age 18+) females. Qualtrics online
survey software was used to create the survey. The relevant portions of the survey can be
found in Appendix C. The recruitment notice was posted on each site early in the month
(October 2008) immediately following the month in which the content analysis was
conducted (September 2008). As an incentive, there was a lottery of five electronic gift
certificates of varying amounts (ranging from $25 to $100) to Target stores. Respondents
could complete the survey any time between early October and early December 2008. At
least two follow-up messages were posted on the sites reminding members of the survey
and their invitation to participate. Unfortunately a response rate cannot be calculated
because it is unknown how many site members actually viewed the survey invitation.
After two months, the survey link was disabled and all the data were downloaded and
prepared for analysis. SPSS 17.0 was used to analyze the data. Incentives were disbursed
to survey respondents by random selection.
Survey Measures
Social Support
On the participant survey, social support was measured with the Social Provisions
Scale (Cutrona & Russell, 1987; 1990), which parallels the content analysis coding
categories. The 24-item scale was reproduced in whole and is included in Appendix C.
The scale measures six dimensions of social support. Five of these dimensions parallel
the dimensions measured on the content analysis: 1) informational support; 2) emotional
support; 3) esteem support; 4) social network / relationship support; and 5) tangible
support. In addition, the Social Provision Scale also measures a dimension labeled
68
nurturance, which is the general perception that “others rely upon oneself for their well-
being” (Cutrona & Russell, 1987, p. 38). While nurturance is not a standard measure of
social support, it was retained in the current survey to assess how important the
opportunity to help others might be to participants.
Reliability analyses revealed the Social Provisions Scale was highly reliable.
Cronbach’s alpha for all 24-items was .96, and reliability ranged from .77 to .86 for each
of the six dimensions, as reported in Table 5.1. Scales were created for each dimension of
social support, in addition to a global social support scale (the average of all 24 items).
Because of the parallel between informational and emotional social support and bridging
and bonding social capital, only informational and emotional social support as well as the
global indicator of this construct, were used to predict health-related outcomes, thus
enabling a comparison between these four dimensions.
Social Capital
On the member survey, social capital was measured using a modified version of
Williams’ (2006) Internet Social Capital Scales (ISCS). These scales included seven
items measuring bonding social capital and 10 items measuring bridging social capital in
an online community. The scales omitted three items from the original bonding subscale
because they did not apply well in this particular health-related context. These items
were: “The people I interact with online would put their reputation on the line for me,”
“The people I interact with online would be good job references for me,” and “The
people I interact online with would share their last dollar with me.” In addition, in its
69
Table 5.1: Reliability and descriptive statistics for the dimensions of the Social
Provisions Scale (4-pt Likert)
Dimension
Cronbach’s
Alpha (α)
Mean Median Mode Standard
deviation
Skewness
Guidance
(informational
support)
.85 2.96 3.00 3 .736 -.597
Attachment
(emotional
support)
.86 2.51 2.25 2.25 .808 .045
Reassurance of
worth (esteem
support)
.78 2.94 3.00 3 .537 -.715
Social
integration
(social
network
support)
.77 3.24 3.25 3 .490 -.894
Nurturance .78 2.29 2.25 2.00 .676 -.079
Reliable
alliance
(tangible
support)
.86 2.89 3.00 2.75 .679 -.304
Entire scale .96 2.79 2.83 2.58 .598 -.650
current use the scale changed the dollar amount of a hypothetical loan from $500 to $100:
“If I needed an emergency loan of $100, I know someone in this online community
that I can turn to.” See Appendix C for the complete scale.
Similar to other studies (Elison et al., 2007; Williams, 2006), reliability analyses
revealed the Internet Social Capital Scales were reliable. As Table 5.2 indicates,
Cronbach’s alpha for the 10-item bonding subscale was .93, and for the seven-item
bridging subscale it was .92. Cronbach’s alpha for the entire 17-item scale was .93. Two
scales were created for each dimension of social capital and were used to predict health
outcomes, along with a global social capital scale (the average of all 17 items).
70
Table 5.2: Reliability and descriptive statistics for the social capital scales
Dimension Cronbach’s
Alpha (α)
Mean Median Mode Standard
deviation
Skewness
Bonding SC .90 3.12 3.29 3.29 .967 -.19
Bridging SC .94 4.00 4.00 3.80 .695 -.63
Global SC .94 3.62 3.65 4.00 .773 -.438
Frequency and Intensity of Use
Several variables assessed how frequency and intensity of use were associated
with health-related outcomes. First, two variables assessed how time spent using the site
was associated with outcomes. The first of these variables measured how long
respondents had been members of the SNS. There were six response options ranging
from less than one week to more than 9 months. The mean score for this measure was
4.72, indicating that the average length of membership was more than six months. Next,
respondents indicated how many minutes they spend on the site at one time. There were
five response options that spanned 15 minute increments and ranged from 15 minutes or
less to more than one hour. The mean score of 2.95 for this question indicates that the
members spent an average of 30 minutes on the site in one session.
Additional questions assessed how intensely members used the site. Respondents
were asked how many different kinds of social support they sought from the site. The
question stem said “Please check all the reasons that you go to this web site” and there
were seven response options. These response options reflected the following dimensions
of social support: informational; emotional; esteem; social network / relationship (two
response options were available, one reflecting the desire to make a connection with a
new friend and another reflecting the desire to connect with other support groups); and
tangible support. It addition there was a response option that reflected the “opportunity
for nurturance” (Cutrona & Russell, 1987) and one that said “to relax and have fun.”
As Figure 5.1 indicates, several of these response options had highly uneven
distributions (> 90% selected one value); therefore, only the variables that were more
evenly distributed than a 90:10 split were used to predict health outcomes. These
included seeking informational support (80% said yes), emotional support (80% said
yes), esteem support (61% yes), the opportunity to nurture (75% said yes), and to relax
and have fun (75% said yes). In addition, all of the nine response options were added
together (including an “other” category) to create an index variable that reflected all the
support that women sought on the site.
Figure 5.1: Percentage of respondents who sought
social networking site
Finally, the self-perceived perceived usefulness of the site for respondents was
also measured. This was a scale variable created by averaging the responses to three
0
6%
4%
Opportunity for nurturance
To relax and have fun
Esteem support
Relationship support
Emotional support
Informational support
new friend and another reflecting the desire to connect with other support groups); and
tangible support. It addition there was a response option that reflected the “opportunity
” (Cutrona & Russell, 1987) and one that said “to relax and have fun.”
As Figure 5.1 indicates, several of these response options had highly uneven
distributions (> 90% selected one value); therefore, only the variables that were more
han a 90:10 split were used to predict health outcomes. These
included seeking informational support (80% said yes), emotional support (80% said
yes), esteem support (61% yes), the opportunity to nurture (75% said yes), and to relax
yes). In addition, all of the nine response options were added
together (including an “other” category) to create an index variable that reflected all the
support that women sought on the site.
Figure 5.1: Percentage of respondents who sought each dimension of support from the
perceived perceived usefulness of the site for respondents was
also measured. This was a scale variable created by averaging the responses to three
20 40 60 80
80%
80%
61%
75%
75%
Opportunity for nurturance
To relax and have fun
Esteem support
Relationship support – New friend
Relationship support – Group referral
Tangible support
Emotional support
Informational support
71
new friend and another reflecting the desire to connect with other support groups); and
tangible support. It addition there was a response option that reflected the “opportunity
” (Cutrona & Russell, 1987) and one that said “to relax and have fun.”
As Figure 5.1 indicates, several of these response options had highly uneven
distributions (> 90% selected one value); therefore, only the variables that were more
han a 90:10 split were used to predict health outcomes. These
included seeking informational support (80% said yes), emotional support (80% said
yes), esteem support (61% yes), the opportunity to nurture (75% said yes), and to relax
yes). In addition, all of the nine response options were added
together (including an “other” category) to create an index variable that reflected all the
of support from the
perceived perceived usefulness of the site for respondents was
also measured. This was a scale variable created by averaging the responses to three
100
80%
80%
95%
75%
75%
72
individual survey questions. Using a seven-point Likert scale that ranged from strongly
disagree to strongly agree, respondents rated how useful the site is to them; how much
they trust what they read on the site; and whether they’ve followed recommendations
they’ve read on the site. Principal components analysis with varimax rotation was used to
determine whether the data could be reduced to a single scale measure. The three items
loaded onto one factor with Eigenvalue = 2.06, and factor loadings of .77, .86 and .85
respectively. The scale was reliable with a Cronbach’s alpha (α) of .77. See Table 5.3
below for descriptive data.
Table 5.3: Descriptive statistics for frequency and intensity variables
Dimension Mean Median Mode Standard
Deviation
Skewness
Membership
length
4.72 5.00 6 1.250 -.367
Session length 2.95 3.00 2 1.362 .225
Reasons 4.78 5.00 6 1.486 -.504
Site usefulness 2.95 3.00 2 1.362 .225
Participation
Several of the variables that measured participation had highly skewed
distributions making them inappropriate for regression analysis. For example,
89% of respondents had asked for support and 97% had provided support at least once.
Moreover, 89% of the sample had both asked for and provided advice on these sites.
Therefore, the predictor variables for participation were limited to the frequency of
support that was asked for and provided. These variables had a distribution that was
approximately normal. However, they could not be combined into a scale because it did
not quite meet the requirement for reliability, instead it had a reliability score of α = .67.
73
Descriptive statistics for these two participation variables are indicated in Table 5.4
below.
Table 5.4: Descriptive statistics for participation variables
Variable Mean Median Mode Standard
deviation
Skewness
Amount of support
asked for
2.61 2.00 2 1.311 .941
Amount of support
provided
4.60 5.00 6 1.546 -.624
Health Outcomes
Health outcomes were measured through a variety of survey questions designed to
assess the impact of participation on respondents’ health-related knowledge, attitudes,
and behaviors. These measures are described in more detail below.
Knowledge. Knowledge measures were created for the survey from pregnancy
health information posted on the Centers for Disease Control and Prevention’s web site:
“Having a Healthy Pregnancy…Pregnancy Tips (A-Z)” (retrieved from
www.cdc.gov/ncbddd/bd/abc.htm). These items asked respondents whether they agreed
or disagreed (a dichotomous response option) with some basic pregnancy health
information, such as “it’s fine for a pregnant women to have x-rays taken,” and “folic
acid prevents the risk of birth defects of the brain and spine.” There were 10 such items
included on the survey and these items were added together to create an index variable of
prenatal-care knowledge. However, these items were not included in the analysis due to
low variance in the responses. The mean score on the cumulative knowledge index was
was 9.73., the median and mode were 10, the standard deviation was .54, and skewness
was -2.32. Further, each individual item also had a similar lack of variance with most of
74
the sample consistently getting the answers to the knowledge items correct (e.g. > 95%).
However, even though the variance is small, one additional knowledge measure will be
retained, which asked respondents if they had learned anything new from participating on
the site (75% of pregnant women indicated they had).
Attitudes. The attitude items measured respondents’ attitudes towards prenatal
health and being pregnant. These measures focused on how the women felt about making
lifestyle changes for the health of their baby and how they felt about the physical changes
that accompany being pregnant. These questions asked respondents to rate their
agreement with each item on a 7-point Likert scale that ranged from strongly disagree to
strongly agree, with a higher score indicating stronger agreement with an attitude. Tests
of normality (e.g. Kolmogorov-Smirnov) indicate that the attitude outcome variables
were not normally distributed. However, Leech, Barrett, and Morgan (2008) suggest that
if the skewness statistic is less than plus or minus one (< +/- 1) the variable is
approximately normal. Therefore this criterion of normalcy is used throughout the
analyses. Using this standard, the skewness statistic suggests that six of these variables
are approximately normal and therefore they are suitable for Ordinary Least Squares
regression. See Table 5.5 for descriptive statistics for the attitude outcomes.
Principle components factor analysis, with varimax rotation, was conducted on
these six attitude items for data reduction purposes. A minimum Eigenvalue criterion of 1
was established, and the factors were rotated with varimax rotation which yielded the
final loadings. While the factor analysis produced two factors that met the minimum
criterion, only one scale was retained in the final analysis because it met the minimum
75
reliability criterion of Cronbach’s alpha > .70. This scale included two items that reflect a
positive attitude toward the physical changes of pregnancy (α = .78). The first item (Q1)
said, “I enjoy the physical changes that being pregnant brings,” and the second item (Q8,
reverse-coded) said, “I feel so uncomfortable in my pregnant body.” Item one had a
factor loading of .82 and item eight loaded at .80. The new scale variable had a mean of
4.68, median of 5, mode of 6, standard deviation of 1.65, and a skewness statistic of -
.432. In addition to this scale variable, two other attitude variables were regressed on the
predictor variables. These include item seven, “I look forward to getting pregnant again,”
and item nine, “I am always looking for ways to be healthier while I’m pregnant.”
Table 5.5: Descriptive statistics of attitude outcomes
Mean Median Mode Standard
deviation
Skewness
1. “I enjoy the physical
changes that pregnancy
brings.”
4.81 5.00 5 1.69 -.665
2. “I don’t mind changing
my lifestyle for the well-
being of my baby.”
6.36 7.00 7 .942 -1.811
3. I am worried that I will
do something wrong while
I’m pregnant that will hurt
my baby.”
4.18 5.00 5 1.85 -.289
4. “Most pregnancies result
in the births of healthy
babies.”
5.87 6.00 6 1.18 -1.452
5. “Being pregnant is an
inconvenience I can’t wait
to be finished with.”
Reverse-coded.
6.06 7.00 7 1.37 -1.805
6. “It seems that there are so
many things that may be
dangerous to my
pregnancy.”
4.93 5.00 5 1.55 -.653
7. “I look forward to getting
pregnant again.”
4.06 4.00 1 2.20 -.126
76
Table 5.5, Continued
Mean Median Mode Standard
deviation
Skewness
8. “I feel so uncomfortable
in my pregnant body.”
Reverse-coded.
4.56 5.00 7 1.95 -.300
9. “I am always looking for
ways to be healthier while
I’m pregnant.”
5.61 6.00 6 1.01 -.374
10. “I don’t mind giving up
certain food or drinks while
I’m pregnant if it will
reduce my risks for
pregnancy complications.”
6.44 7.00 7 .824 -2.218
Respondents’ overall attitude toward their pregnancy was also assessed as an
outcome variable. Using a seven-point Likert scale that was based on Mistry et al.’s
(2007) study, women were asked how often they felt six different emotions when
thinking about their pregnancy. The six emotions were being nervous, calm and peaceful,
downhearted and blue, so down in the dumps that nothing can cheer you up, happy, and
worried. Response options ranged from never (1) to all the time (7). The central tendency
and standard deviations of these outcomes are depicted in Table 5.6. One variable had an
especially skewed distribution (“so down in the dumps that nothing can cheer you up”)
with most women never experiencing this feeling. Therefore, this variable was excluded
from the analysis. However, another similar item also reflecting depression (feeling
“downhearted and blue”) while somewhat skewed (skewness = 1.00) was retained in the
analysis. Furthermore, these two depression variables were correlated (r = .62, p < .01)
suggesting construct validity.
77
Table 5.6: Descriptive statistics for overall attitude toward pregnancy
“How often do you feel
the following ways when
thinking about your
pregnancy?” (1-7 scale)
Mean Median Mode Standard
deviation
Skewness
Very nervous 3.94 4.00 4 1.723 -.020
Calm and peaceful 4.64 5.00 4 1.364 -.194
Downhearted and blue 2.14 2.00 2 1.135 1.003
So down in the dumps that
nothing can cheer you up
1.53 1.00 1 .989 2.436
Happy 5.79 6.00 6 1.088 -.999
Worried 4.67 5.00 4 1.533 -.221
Principal components factor analysis was conducted on the five remaining
attitudes. A minimum Eigenvalue criterion of 1 was established, and the factors were
rotated with varimax rotation which yielded the final loadings. The factor analysis
produced one factor that met both the minimum Eigenvalue criterion, as well as the
minimum reliability criterion of Cronbach’s alpha > .70. This scale was an anxiety scale
that reflected being nervous (item one) and worried (item six) when thinking about one’s
pregnancy (α = .81, factor loadings of .88 and .89 respectively). The resulting scale was
also fairly normally distributed with a mean of 4.30, median of 4.00, mode of 4.00,
standard deviation of 1.50 and a skewness statistic of -.093. In addition to this scale, two
other overall attitudes were used as outcomes. The first was feeling depressed and the
second was feeling happy when thinking about the pregnancy.
Behavior. Behavior was assessed through 11 survey questions that measured how
participation on the site has impacted a variety of pregnancy-related behaviors. Using a 7-
point Likert scale that ranged from never to daily, respondents were asked how often they
had done any of 11 behaviors that either are or are not recommended by health
78
professionals during pregnancy, as a result of something they read on the web site. The
recommended behaviors included taking prenatal vitamins, exercising, eating fruits
and/or vegetables, seeing your doctor or obstetrician, and asking your doctor for further
information. The behaviors that were not recommended included drinking alcohol, using
illegal/street drugs, smoking cigarettes, eating sushi, drinking coffee, and drinking soda.
While all of these behaviors are not equivalently adverse for a pregnancy, they are used
as barometers of health orientation. Additional questions also assessed whether
respondents had spoken with their doctor or sought additional information on the Internet
about anything they had read on the site. Several of these items were strongly negatively
skewed, suggesting that most women engaged in healthy behavior throughout their
pregnancy. However, using a criterion of +/-1 for skewness, this statistic suggests that
five of these variables do not deviate too far from the normal curve, and therefore are still
suitable for OLS regression (Table 5.7).
Principal components factor analysis was conducted on these five behavior items.
A minimum Eigenvalue criterion of 1 was established, and the factors were rotated with
varimax rotation which yielded the final loadings. While the factor analysis produced two
factors that met the minimum criterion, only one scale was retained in the final analysis
because it also met the minimum reliability criterion of Cronbach’s alpha > .70. This
scale included two items (10 and 11 from Table 5.7) each with a factor loading of .93,
that reflect the behavioral dimension of doctor visits (α = .84). The mean for this scale
variable was 3.04, the median and mode were 3, standard deviation was .931 and
skewness was .826.
79
Table 5.7: Descriptive statistics for behavior outcomes
“As a result of something
you read on the web site,
how often have you done
the following since you’ve
been pregnant?” (1-7
scale)
Mean Median Mode Standard
deviation
Skewness
1. Drank alcohol. Reverse-
coded.
6.84 7.00 7 .475 -3.630*
2. Used illegal/street drugs.
Reverse-coded.
7.00 7.00 7 .000 *
3. Smoked cigarettes.
Reverse-coded.
6.60 7.00 7 1.375 -3.524*
4. Exercised. 4.42 5.00 6 1.894 -.465
5. Drank coffee. Reverse-
coded.
5.27 6.00 7 2.156 -.897
6. Drank soda. Reverse-
coded.
3.33 2.50 2 2.114 .615
7. Taken prenatal vitamins. 6.54 7.00 7 1.475 -3.167*
8. Ate fruits and/or
vegetables.
6.52 7.00 7 .990 -3.610*
9. Ate sushi. Reverse-coded. 6.80 7.00 7 .608 -3.179*
10. Seen your doctor or
obstetrician.
3.17 3.00 3 .898 .292
11. Asked your doctor for
further information.
2.93 3.00 3 1.103 .780
*Not tested
The other three variables that had an approximately normal distribution (exercise,
soda and coffee drinking) were individually regressed on the predictor and control
variables (i.e. they were not part of a scale). Six of these behavior outcome variables
(drinking alcohol, using illegal drugs, smoking, taking prenatal vitamins, eating fruits and
or vegetables, and eating sushi) were so strongly skewed that they were excluded from
analysis entirely. These items had at least 88% of the sample agreeing with an anchor
point, reflecting a lack of variance, which would have prohibited the detection of any
significant associations.
80
An additional survey item assessed behavioral change as a result of participation
in the SNS. This was a more general question in which respondents were asked, “As a
result of your participation in the online community, since you’ve been pregnant have
you…” and then they were provided with a list of 12 behavioral items that included
statements such as “changed what you eat; changed what you drink; changed how you
interact with your doctor”; etc. A principal components factor analysis was conducted on
the 12 behavior change items; however, the analyses did not produce any factors that
were both reliable and valid. Therefore, to ensure maximum reliability of the results each
of these behavior change items were run individually as dependent variables in
hierarchical logistic regression analysis. In addition, an index was created with all 12
items that measured overall change and, using hierarchical multiple linear regression, this
variable was also regressed on the control and predictor variables. This index was fairly
normally distributed, with a mean of 6.38, a median of 6.00, a standard deviation of
2.652, and a skewness statistic of -.061. Cronbach’s alpha (α) for the index is .71.
Table 5.8: Behavior change distributions
“As a result of your participation in this community, since you’ve
been pregnant have you…”
Yes
“Changed what you eat?” 59%
“Changed what you drink?” 60%
“Changed your exercise routine?” 34%
“Changed your mind about any medical procedures…?” 24%
“Worried more about your pregnancy?” 51%
“Changed how you interact with your doctor?” 28%
“Spoken with your doctor about something you read on this web
site?”
51%
“Sought additional information on the Internet about something you
read on this web site?”
84%
“Spoken with your friends or family about something you read on
this web site?”
83%
81
Table 5.8, Continued
“Sought a second opinion from another doctor/health care provider?” 17%
“Felt less stressed?” 69%
“Felt more confident in how you take care of yourself?” 80%
Analysis
Statistical Package for the Social Sciences (SPSS) version 17.0 statistical software
was used to test the hypotheses related to the content analysis and survey data.
Descriptive statistics summarized the content of messages on the research sites, and chi-
square and t-tests were used to test for statistically significant differences in the content
analysis data. The survey analyses relied on both hierarchical multiple linear regression
and hierarchical logistic regression to examine how well the independent variables, such
as social support and social capital, predicted change in the outcome variables, such as
attitudes towards the pregnancy or health-related behaviors, while controlling for other
socio-demographic variables. The variables that were controlled for included health
status, education level, income, first pregnancy, minority status, marital status, working
full-time, age, and networking site from which the member was recruited. Correlations
were examined among all the variables to determine if there was a risk of
multicollinearity, but none of the correlations were near or exceeded a threshold of
Pearson’s r = .80, a threshold level that has been identified for multicollinearity (Licht,
1995). In the hierarchical multiple regression models, the predictor variable was always
forced into the model as the first block, with the second block of variables, the socio-
demographic variables, entered into the model using stepwise selection. The variables
were entered into the model stepwise with an entry value of p < .05 and an exit value of
82
p < .10. Similarly, in the hierarchical logistic regression analyses, the forward selection,
likelihood-ratio method was used with the same entry and exit values. In all of the
regression analyses, the model with the highest adjusted R-squared will be presented,
assuming the predictor variable was significantly associated with the outcome.
Conducting the impact analyses with the social network metrics as predictor
variables was a two-step process. First, the network data was converted into an
appropriate format to export into the social network software. Network analyses were run
in that software (UCINET), then the resulting metrics were imported back into the SPSS
file and regression analyses were conducted using these data as predictor variable
83
CHAPTER 6
SURVEY RESULTS
Sample Characteristics
The survey sample was restricted to adult women who were registered members
of one of eight pregnancy social networking sites. The majority of the sample was
Caucasian (75%) with a mean age of 31 years. Seventy-five percent of respondents also
had at least some college education: 32% had some college or a two-year college degree,
26% had a four-year college degree, and 17% had a Master’s or Doctoral/professional
degree. Thirty-eight percent of the sample had a full-time job outside the home, while
28% were not currently working. Approximately one-half of the sample had a household
income of $50,000 or greater per annum. The location of respondents was categorized
according to the four US Census Bureau Regions, plus an International category. The
breakdown was as follows: 1) Northeast, 12%; 2) Midwest, 18%; 3) South, 29%; 4)
West, 19%, and 5) International, 8%. Survey respondents were largely in good health,
with 60% reporting their health status to be very good or excellent and 40% of the sample
was pregnant at the time of the survey. Table 6.1 reports the complete demographic
description of the sample.
The analyses that examine the health impact of participation on the sites use a
sub-sample of pregnant women only; whereas, the network analyses examine the entire
sample. Of the respondents who were pregnant (n = 114), more than half of them (54%)
reported that this was their first pregnancy. Forty-six percent of pregnant respondents
were in the second trimester of pregnancy, while 25% were in the first trimester, and 27%
84
in the third. Further, the majority of these pregnant women were under the regular care of
a physician (96%).
Table 6.1: Characteristics of the survey sample
> = 288
Mean age 31 years
Age range 19- 59
Ethnicity
White/Caucasian 75%
African American/Black 2%
Hispanic/Latino 4%
Asian American/Pacific Islander 1%
Native American 1%
Other 2%
Marital Status
Married 68%
Unmarried partner 12%
Single 3%
Separated/divorced 2%
Other <1%
Education
Less than high school 1%
High school/GED 10%
Some college 24%
2-year college degree 8%
4-year college degree 26%
Master’s degree 13%
Doctoral or professional degree 4%
Employment status
Full-time job outside home 38%
Part-time job outside home 9%
Work from home 7%
Full-time student 4%
Not currently working 28%
Annual household income
Under $20,000 5%
$20,000 - $49,999 26%
$50,000 - $79,999 25%
$80,000 or more 28%
85
Table 6.1, Continued
Location (based on US Census Bureau Regions)
US - Northeast 12%
US – Midwest 18%
US – South 29%
US – West 19%
International 8%
Health Status
Excellent 10%
Very good 50%
Fair 24%
Poor or very poor 1%
Current number of children
0 34%
1 30%
2 or more 23%
Currently pregnant 40%
Analyses were conducted to examine whether there were significant differences
between the socio-demographic characteristics of the respondents who were pregnant at
the time of the survey, compared to those who were not. Two significant differences were
found. The women who were pregnant were significantly younger (29.18 vs. 32.48, t
(240.50) = 4.16, p <. 001) and had fewer children (.64 vs. 1.59, t (248) = 5.378, p <. 001)
than those who were not.
The sample consisted of fairly avid users of the Internet and social networking
sites, in particular. Sixty-four percent of the sample had been on the Internet every day
during the previous week, and 69% of the sample had visited an online social networking
site every day during the previous week. No other media were used as frequently as the
Internet: 44% of the sample watched entertainment TV every day and 30% watched TV
news every day, however, only 14% listened to talk radio every day and 8% read the
newspaper every day.
86
RQ5: How is frequency and intensity of use of a health-related social networking site
associated with health outcomes? Specifically, are frequency and intensity of use
significantly associated with:
(i) knowledge of prenatal health;
(ii) attitudes towards pregnancy and prenatal health; and
(iii) health behavior?
Hierarchical stepwise multiple regression and hierarchical logistic regression were
used to test the associations between the predictor and outcome variables. Health status,
education level, income, whether this was a first pregnancy, minority status, marital
status, full-time employment, age, and the different pregnancy social networking site of
which respondents were members were also controlled for in the models. Several
predictor variables were used to assess how frequency and intensity of use were
associated with outcomes; therefore, the findings will be presented in the following order.
First, the association between frequency of use and health outcomes are presented,
followed by the findings examining how type of support sought was associated with
outcomes (a measure of intensity). Finally, the self-reported perceived usefulness of the
site and its association with health outcomes are examined (a second intensity measure).
Frequency of Use and Health Outcomes
Knowledge. Frequency of use was measured with two questions that assessed
length of membership and time spent using the site in one sitting. Hierarchical logistic
regression analysis tested whether these predictor variables were associated with
respondents’ indicating that they had learned something new from participating on the
87
web site. The results show that neither of these measures was associated with respondents
learning anything new.
Attitudes. Three attitudes were regressed on the predictor variables. These were: a
positive attitude toward the physical changes of pregnancy; a positive attitude toward
getting pregnant again; and a positive attitude toward being healthy. As Table 6.2
indicates, both of the frequency of use measures were associated with a positive attitude
toward getting pregnant again. However, while length of site membership was positively
associated with getting pregnant again, time spent online was negatively associated with
this item. In addition, length of site membership was positively associated with attitude
toward being healthy, while time spent on the site at one time did not have a significant
association with this item. This significant standardized betas for the best fitting model
predicting attitude toward being healthy were minority status, β = .20, p < .05; member
of Site G, β = -.29, p < .01; and membership length, β = .20, p < .05. The overall model
was significant at p < .01, with F (3, 95) = 5.10 and adjusted R
2
= .11.
Table 6.2: Standardized betas indicating significant associations between frequency of
site use and a positive attitude toward getting pregnant again
Membership
length on S
S
Time spent
online at once
Health status -- .21*
Education -- --
Income -- --
First pregnancy .26** .22*
Minority status -- --
Married respondents -.22* -.22*
Fulltime worker -- --
Age -.34*** -.39***
Site B -.27**
Site F .20*
88
Table 6.2, Continued
Membership
length on S
S
Time spent
online at once
Frequency variable .21* -.17*
Overall F (df) 8.08*** (5, 93) 8.78*** (6, 93)
Adjusted R
2
.27 .32
+p < .10 * p < .05 ** p < .01 ***p < .001
Note: Only the significant web sites are listed in the tables, however all
of them were tested for significance.
In addition to the three specific attitudes described above, the relationship
between the predictor variables and respondents’ overall attitude toward their pregnancy
was also tested. Specifically, the predictor variables were tested to determine whether
they were associated with respondents’ feeling anxious, depressed, or happy about their
pregnancy. Length of membership was not associated with any of these attitudes, while
minutes on the site was significantly associated with feeling less depressed about the
pregnancy (adjusted R
2
= .08, F (2, 98) = 5.14, p < .01). The significant standardized
betas included income, β = - .27, p < .01, and minutes on the site at once, β = - .22,
p < .05. These results indicated that respondents with higher incomes and those who
spend more time on the site were less depressed about their pregnancy.
Behavior. Of the absolute measures of behavior that were tested (doctor visits,
exercise, coffee consumption, and soda consumption), length of site membership was
significantly associated with doctor visits (adjusted R
2
= .15, F (3, 95) = 7.03, p < .001).
The significant standardized betas in the model included health status, β = -.19, p < .05,
working full-time, β = -.25, p < .01, and membership length, β = .29, p < .01. These
results indicate that respondents who had better health and worked full-time went to the
89
doctor less often, while respondents who were members of the SNS for a longer period of
time visited the doctor more often.
Spending more time online at once was the only variable (β = .22, p < .05) in a
model that was positively associated with drinking soda (adjusted R
2
= .04, F (1, 98) =
4.84, p < .05). In addition, spending more time online at once also trended toward being
significantly negatively associated with exercise (adjusted R
2
= .06, F (2, 97) = 4.00, p <
.05. The variables in the model include working full-time, β = .24, p < .05, and spending
more time online at once, β = - .17, p < .10.
Spending time on the web site and membership length were also associated with
two behavior change measures. Hierarchical logistic regressions showed that length of
web site membership was significantly associated with speaking to a doctor about
something the respondent had read on the web site (overall χ
2
= 19.25 (2), p < .001). The
significant predictors in the model were working full-time, AOR = .33, p < .05 and
membership length, AOR = 1.82, p < .001. These results suggest that women who
worked full-time were nearly 70% less likely to ask their doctors about something they
read on the web site, while with each increase in length of membership, respondents were
more than 80% more likely to ask their doctor a question about something they read on
the web site. Membership length was also marginally associated with seeking a second
opinion, AOR = 1.59, p < .10, overall χ
2
= 4.01 (1), p < .05.
More time spent online was marginally associated with two behavior change
measures. Respondents who spent more time online were less likely to change their
exercise as a result of something they read on the web site (AOR = .76, p < .10, overall χ
2
90
= 3.26 (1), p < .10) and felt more confident as a result of something they read on the web
site (AOR = 1.42, p < .10, overall χ
2
= 3.42 (1), p < .10).
Reasons for Support Seeking and Health Outcomes
Health outcomes were first regressed individually on five facets of social support-
seeking that had the most moderate distributions, including informational, emotional,
esteem, opportunity for nurturance, and relaxation, that respondents may have sought on
the web site. Subsequently, an index variable was created with all eight support options
that respondents may have sought from the site and this index was also used to predict
outcomes.
Knowledge. Of the five individual facets tested, only the pursuit of esteem support
was associated with respondents indicating that they had learned something new from the
web site (overall χ
2
(2) = 8.86, p < .01). A hierarchical stepwise logistic regression found
income, AOR = .82, p < .05, and the pursuit of esteem support, AOR = 2.35, p < .10
significantly predicted learning something new. The results suggest that with each
increase in income, respondents were nearly 20% less likely to learn something new,
while respondents who sought esteem support were more than twice as likely to have
learned something new on the site. The overall index of support sought was not
associated with knowledge.
Attitudes. While the overall index of support sought was not associated with any
attitudes related to the pregnancy, several of the specific support facets had significant
associations with attitudes. Of the three specific attitudes that were examined (positive
physical attitude, positive attitude toward getting pregnant again, positive attitude toward
91
being healthy), the seeking of esteem support was negatively associated with having a
positive attitude toward the physical changes of pregnancy (adjusted R
2
= .08, F (2, 97) =
5.36, p < .01) and it was also negatively associated with looking forward to getting
pregnant again (adjusted R
2
= .33, F (7, 92) = 7.87, p < .001). Table 6.3 lists all of the
significant co-variants in the models. The seeking of informational support was also the
only other variable in this series that was marginally negatively associated with a positive
attitude toward getting pregnant again, indicating that looking for informational support
was associated with a less positive attitude toward a subsequent pregnancy.
Table 6.3: Standardized betas indicating significant associations between seeking esteem
support and pregnancy-related attitudes
Positive
attitude
toward getting
pregnant
again
Positive
attitude
toward being
healthy
Health status -- .20*
Education -- --
Income -- --
First pregnancy -- .20*
Minority status -- --
Married respondents .21*
Fulltime worker -- -.17*
Age -- -.41***
Site B -.26** -.26**
Seeking esteem
support
-.20* -.16+
Overall F 5.36 ** (2. 97) 7.87*** (7, 92)
Adjusted R
2
.08 .33
+p < .10 * p < .05 ** p < .01 ***p < .001
There were several significant associations between the seeking of various kinds
of support and respondents’ overall attitudes toward their pregnancy. As table 6.4
indicates, hierarchical stepwise regression analysis showed that seeking emotional
92
support was significantly associated with feeling depressed, anxious, and less happy
about the pregnancy. The seeking of esteem support was also associated with feeling
depressed about the pregnancy (adjusted R
2
= .08, F (2, 98) = 5.29, p < .01). The
significant standardized betas in the model included health status, β = - .23, p < .01, and
seeking esteem support, β = .21, p < .01. Only the seeking of informational support was
marginally associated with a positive attitude toward the pregnancy. Respondents who
sought informational support tended to be more likely to be happy about the pregnancy
(adjusted R
2
= .08, F (2, 98) = 5.34, p < .01). The significant standardized betas in the
model included health status, β = .29, p < .01, and seeking informational support, β = .16,
p < .10.
Table 6.4: Standardized betas indicating significant associations between seeking
emotional support and attitudes toward the pregnancy
Depressed Anxious Happy
Health status -.20* -.28** .24*
Education -- --
Income -- --
First pregnancy -- .19*
Minority status -- --
Married respondents --
Fulltime worker -- -.38***
Age -- --
Site G -.19*
Seeking emotional
support
.25** .22** -.21*
Overall F 6.20 ** (2, 98) 9.03*** (5, 95) 6.33** (2, 98)
Adjusted R
2
.09 .29 .10
+p < .10 * p < .05 ** p < .01 ***p < .001
Behavior. Interestingly, seeking four of the five facets of support were positively
associated with drinking more soda, as was the overall index of support sought. Seeking
informational (adjusted R
2
= .07, F (1, 98) = 7.94, p < .01, β = .27, p < .01) and emotional
93
support (adjusted R
2
= .05, F (1, 98) = 6.45, p < .01, β = .25, p < .01), an opportunity to
nurture (adjusted R
2
= .05, F (1, 98) = 6.45, p < .01, β = .25, p < .01), and an opportunity
for relaxation (adjusted R
2
= .04, F (1, 98) = 5.31, p < .05, β = .23, p < .05) were all
significantly associated with drinking soda. For each of these variables, the seeking of
support was the only variable in the model (in addition to the constant), and all the
models were significant. Similarly, the model for the overall index of support sought had
an adjusted R
2
= .12, F (1, 98) = 14.58, p < .001, and the index was also the only variable
in the model, β = .36, p < .001. In addition, seeking esteem support was also negatively
associated with exercise (adjusted R
2
= .10, F (2, 97) = 6.25, p < .01). The significant
standardized betas in the model included income, β = -.23, p < .05, and seeking esteem
support, β = -.28, p < .01.
Of the 12 behavior change measures, opportunity for nurturance was associated
with the most change. As Table 6.5 shows, hierarchical stepwise logistic regression
analyses found that seeking an opportunity for nurturance was significantly associated
with a respondent changing her drinking habits, speaking with the doctor about
something she read on the web site, seeking additional information online, speaking with
friends and family about the web site, and feeling less stress. In fact, seeking every kind
of support except emotional support was associated with at least a few behavior changes.
Only the seeking of emotional support was not associated with any behavior change.
Furthermore, the overall index of support sought significantly predicted behavior change
on three outcome variables: changing what respondents drank, speaking with their family
and friends about the web site, and feeling more confident.
94
Table 6.5: Adjusted odds ratios indicating significant associations between seeking an
opportunity for nurturance and behavior change
Changed
what they
drank
Spoke
with
doctor
about web
site
Sought
additional
info.
online
Spoke
with
family/
friends
about site
Felt less
stress
Health status -- -- -- --
Education -- -- -- --
Income -- -- -- --
First pregnancy -- -- -- 4.15*
Minority status -- -- -- --
Married
respondents
Fulltime worker .28** .32** -- --
Age -- -- --
Site D .11+
Site E .24*
Seeking
opportunity for
nurturance
2.51+ 2.89* 5.49** 2.79+ 2.59*
Overall χ
2
(df) 14.15**
(3)
16.70***
(3)
7.97** (1) 7.56* (2) 3.92* (1)
+p < .10 * p < .05 ** p < .01 ***p < .001
However, when examining how the seeking of each kind of support predicted an
overall index of behavior change, only the opportunity for nurturance and seeking an
opportunity for relaxation and fun were significantly associated with overall change.
Hierarchical stepwise regression analysis found the best fitting model for opportunity for
nurturance and overall behavior change had an adjusted R
2
= .12, F (3, 97) = 5.69, p <
.001. The significant standardized betas in this model were working full-time, β = -.29, p
< .01, being a member of Site F, β = - .21, p < .05, and seeking an opportunity to nurture,
β = .22, p < .05. Similarly, the model for seeking an opportunity to relax and have fun
had an adjusted R
2
= .11, F (3, 97) = 5.28, p < .01, with the significant standardized betas
of working full-time, β = - .32, p < .01, being a member of Site F, β = - .26, p < .01, and
95
opportunity to relax and have fun, β = .20, p < .05. The overall index of support sought
was also significantly associated with overall behavior change (adjusted R
2
= .13, F (3,
97) = 6.06, p < .001). The significant standardized betas in the model included working
full-time, β = -.30, p < .01, being a member of Site F, β = -.22, p < .05, and the overall
index of support sought, β = .24, p < .05.
Perceived Usefulness of the Site and Health Outcomes
The last series of analyses that examined how intensity of site use was associated
with health outcomes regressed health outcomes on a variable that reflected respondents’
perceived usefulness and trustworthiness of the site.
Knowledge. Respondents’ perception that the site was useful was associated with
their learning something new from the site (overall χ
2
(4)
= 24.04, p < .001). Hierarchical
logistic regression analysis found the best fitting model included health status, AOR =
3.06, p < .0; age, AOR = .84, p < .01; Site B, AOR = .19, p < .05; and perceived
usefulness, AOR = 3.05, p < .01. The model suggests that younger respondents in better
health who found the site useful and were not members of Site B were more likely to
report learning something new.
Attitudes. Perceived usefulness of the site was generally associated with positive
pre-natal attitudes. Hierarchical stepwise regression found perceived usefulness was
associated with a positive attitude toward trying to be healthy (adjusted R
2
= .13, F (3,
96) = 5.87, p < .001). Significant standardized betas in the model included being a
minority, β = .21, p < .05, being a member of Site G, β = - .31, p < .001, and perceiving
the site as useful, β = .21, p < .05. Perceived usefulness of the site also predicted being
96
happy about the pregnancy, adjusted R
2
= .09, F (2, 98) = 6.20, p < .01, with only health
status, β = .29, p < .01 and usefulness, β = .20, p < .05 as predictors in the model.
Similarly, usefulness was also associated with feeling less depressed about the pregnancy
(adjusted R
2
= .07, F (2, 98) = 4.50, p < .05). The model included two variables,
education and site usefulness, which were both negatively associated with depression, β =
- .26, p < .01 and β = -.20, p < .05, respectively.
Behavior. As indicated in Table 6.6, hierarchical stepwise logistic regression
found perceived usefulness of the site significantly predicted seven out of 12 (58%)
behavior changes. The more useful respondents found the site, the more likely they were
to change what they ate and drank, to have spoken with their doctor about the web site, as
well as their family friends, to have sought a second opinion from another health care
professional and to have less stress and more confidence. Furthermore, the association
between site usefulness and overall behavior change is also reflected in the significant
relationship between these two variables (adjusted R
2
= .22, F (2, 98) = 14.72, p < .001).
The best-fitting model included two significant co-variants, working full-time, β = -.27,
p< .01, and perceived site usefulness, β = .42, p < .001. This model suggests that
respondents who worked full-time were less likely to change while those who found the
site useful were more likely to change.
97
Changed
what
they ate
Changed
what
they
drank
Spoke
with
doctor
about
web site
Spoke
with
family
and
friends
about web
site
Sought
second
medical
opinion
Felt less
stress
Felt
more
confident
Health status -- -- -- -- -- -- --
Education -- -- -- -- -- -- --
Income -- -- 1.23* -- -- -- --
First pregnancy -- -- -- 3.59* -- --
Minority status -- -- -- -- -- -- --
Marital status -- -- -- -- -- -- --
Fulltime worker -- .21** .34* -- .34* -- --
Age -- -- -- -- -- -- --
Site D -- .06* -- -- -- -- --
Perceived site
usefulness
2.18** 2.87*** 3.21*** 1.91* 2.78** 1.72* 2.07**
Overall χ
2
(df) 9.69**
(1)
25.41***
(3)
26.50***
(3)
9.43**
(1)
10.97**
(2)
4.77* (1)
6.78**
(1)
+p < .10 * p < .05 ** p < .01 ***p < .001
Table 6.6.: Adjusted odds ratios indicating significant associations between perceived site usefulness and
behavior change
98
RQ6: How is participation on a health-related social networking site associated with
health outcomes? Specifically, is participation significantly associated with:
(i) knowledge of prenatal health;
(ii) attitudes towards pregnancy and prenatal health; and
(iii) health behavior?
Participation was assessed by the frequency with which members asked for
support or provided support on the site during the previous month. Hierarchical stepwise
regression and hierarchical logistic regression were used to test the associations between
the predictor and outcome variables, while controlling for health status, education,
income, whether this was a first pregnancy, minority status, marital status, full-time
employment, age, and the web site from which the member was recruited.
Knowledge. Participation on the site was not associated with learning something
new.
Attitudes. Frequency of support sought was not associated with any pre-natal
attitudes; however, frequency of support provided was associated with two positive
attitudes. First, hierarchical stepwise regression analysis found that respondents who
provided support more often had a more positive attitude toward getting pregnant again
(adjusted R
2
= .25, F (5, 95) = 7.26 p < .001). The significant standardized beta
coefficients in the model included income, β = .23, p < .05; first pregnancy, β = .29,
p < .01; being married, β = -.24, p < .01; age, β = -.41, p < .001; and frequency of support
provided, β = .16, p < .10. These findings suggest that respondents looked forward to
getting pregnant again if they had a higher income, this was their first pregnancy, and
99
they provided support more often, while being married and older were negatively
associated with a positive attitude toward another pregnancy. The second attitude that
frequency of providing support was associated with was feeling less depressed about the
pregnancy (adjusted R
2
= .10, F (2, 94) = 6.42, p < .01). Higher education, β = - .30, p <
.01, and frequency of providing support, β = - .23, p < .05 were the significant
standardized beta coefficients in the model, indicating that a higher education level and
being more active in providing support on the site were both negatively associated with
being depressed about the pregnancy.
Behavior. Seeking support was marginally associated with one behavior change
outcome. The more support that respondents sought, the more respondents reported
changing what they drank as a result of something they read on the web site (overall χ
2
(3) = 13.98, p < .01). The significant variables in the model included being married, AOR
= 3.92, p < .10, working full-time, AOR = .31, p < .05, and frequency of seeking support,
AOR = 1.46, p < .10. Providing more support, on the other hand, was associated with six
behavioral change outcomes. Respondents who provided more support drank more soda,
were more likely to have changed what they drank generally, spoke to a doctor about
something on the web site, sought additional information online, spoke to family and
friends about the web site, and felt more confident. In fact, the relationship between
providing support and health outcomes was further reflected in the significant association
between providing support and the overall behavior change index (adjusted R
2
= .12, F
( 3, 93) = 5.39, p < .01). The model included three significant standardized beta
coefficients, working full-time, β = - .34, p < .001, being a member of Site F, β = -.23,
100
p < .05, and frequency of providing support, β = .19, p < .05, indicating that not working
full-time, not being a member of Site F, and providing more support significantly
predicted more overall change.
RQ7a: How is the perception of social support on a health-related social networking site
associated with health outcomes? Specifically, is the perception of global social support
significantly associated with:
(i) knowledge of prenatal health;
(ii) attitudes towards pregnancy and prenatal health; and
(iii) health behavior?
Hierarchical stepwise multiple regression and hierarchical logistic regression
were used to examine the associations between the global measure of social support (the
average of all six dimensions from the Social Provisions Scale) and health-related
outcomes. The models also controlled for significant associations with health status,
education, income, whether this was a first pregnancy, minority status, marital status,
full-time employment, age, and the respondents membership in the different social
networking site from which members were recruited.
Knowledge. The knowledge outcome measured how well the perception of social
support predicted the acquisition of new knowledge on the social networking sites
(SNSs). Results show that social support was not significantly associated with this
outcome.
Attitudes. The analyses examined how well the perception of social support
predicted three attitudes: 1) a positive attitude toward the physical changes of pregnancy;
101
2) a positive attitude toward getting pregnant again; and 3) a positive attitude toward
being healthy. As Table 6.7 indicates, social support was significantly associated with a
positive attitude toward getting pregnant again and a positive attitude toward being
healthy (β = .18, p < .05 and β = .20, p < .05 respectively).
Table 6.7: Standardized betas indicating significant associations between social support
and prenatal attitudes
Looking forward
to getting
pregnant again
Always trying to be
healthier
Health status .21* --
Education -- --
Income -- --
First pregnancy .26** --
Minority status -- .21*
Marital status -.22* --
Fulltime worker -- --
Age -.33*** --
Site B -.23*
Site G -.28**
Global social support .18* .20*
Overall F (df) 7.74*** (7, 92) 5.80*** (3, 96)
Adjusted R
2
.32 .13
+p < .10 * p < .05 ** p < .01 ***p < .001
Additional analyses examined how the independent variables were associated
with respondents’ overall attitude toward their pregnancy. Hierarchical stepwise
regression tested whether social support predicted women feeling anxious, happy, or
depressed about their pregnancy. The results show that health status, whether this was a
first pregnancy, working full-time, being a member of Site G, and social support were
significantly associated with respondents feeling anxious about their pregnancy (adjusted
R
2
= .29, F (5, 94) = 9.10, p < .001). Standardized beta coefficients for the four predictor
variables included in the model were: health status, β = - .30, p < .001, first pregnancy,
102
β = .20, p < .05, working full-time, β = - .40, p < .001, being a member of Site G, β = -
.19, p < .05, and social support, β = .19, p < .05. These coefficients suggest that women
feel anxious about their pregnancy if they are in poorer health, are pregnant for the first
time, do not work full time, are not a member of Site G, and perceive social support on
the social networking site.
There was also a trend toward greater social support predicting greater happiness
about the pregnancy (adjusted R
2
= .10, F (3, 96) = 4.46, p < .01), although the
standardized beta coefficient for social support didn’t quite reach statistical significance:
social support, β = .17, p < .10, first pregnancy, β = .21, p < .05, and health status, β =
.26, p < .01. This model suggests that having social support, being pregnant for the first
time, and being healthier all predict women’s happiness about their pregnancy. Social
support did not significantly predict depression.
Behavior. Several behaviors were measured, both absolutely (e.g. “how often
have you done the following as a result of something you read on the SNS…”) and also
in terms of relative change (e.g., “since you’ve been a member of this community, have
you changed [behavior X]…”). Four direct measures of behavior were tested: 1) doctor
visits; 2) exercise; 3) coffee consumption; and 4) soda consumption. Of these measures,
social support was marginally significantly associated with seeing one’s doctor, while
working full time was negatively associated with this outcome (adjusted R
2
= .07, F (2,
97) = 4.72, p < .01). The standardized beta coefficients were: social support, β = .17, p <
.10, and working full time, β = -.26, p< .01.
103
Hierarchical logistic regression analysis examined how well social support
predicted respondents’ behavioral change on 12 outcomes since they began participating
in the online community. Global social support predicted change on 42% or 5 of the 12
outcomes. As reported in Table 6.8, social support predicted change related to: 1) eating;
2) drinking; 3) speaking to one’s doctor about something that was posted on the SNS; 4)
feeling less stress; and 5) feeling more confident about taking care of oneself.
Table 6.8: Adjusted odds ratios indicating associations between social support and
behavioral change as a result of participating on the SNS
Changed
what you
eat
Changed
what you
drink
Spoke
with
doctor
about web
site
Felt less
stress
Felt more
confident
Health status -- -- -- -- --
Education -- -- -- -- --
Income -- -- 1.35*** -- --
First pregnancy -- -- -- -- --
Minority status -- -- -- -- --
Marital status -- -- -- -- --
Fulltime
worker
-- .26** -- -- --
Age -- -- -- -- --
Site D -- .07* -- -- --
Social support 2.20* 3.58** 6.41*** 2.86* 3.27**
Overall χ
2
(df) 4.23* (1) 19.21***
(3)
23.32***
(2)
6.41**
(1)
6.12** (1)
+p < .10 * p < .05 ** p < .01 ***p < .001
An overall index of behavior change was also regressed on social support. The
results show that social support predicted behavioral change, while working full-time was
negatively associated with change (adjusted R
2
= .13, F (2, 97) = 8.66, p < .001).
Standardized beta coefficients for the two predictors were: working full-time, β = - .23, p
< .05, and social support, β = .33, p < .001.
104
RQ7b: Are there differences in how the specific dimensions of informational and
emotional support are associated with:
(i) knowledge of prenatal health;
(ii) attitudes towards pregnancy and prenatal health; and
(iii) health behavior?
The associations between these two dimensions of social support and outcomes
are quite similar to each other and to the global measure of social support. As Table 6.9
indicates, the perception of global social support and the perception of informational
support were both significantly associated with 11 outcomes, with one difference
between them, while emotional support was significantly associated with nine outcomes,
and it differed from the global measure on three outcomes. The perception of global
support predicted happiness about the pregnancy, while informational and emotional
support did not. Emotional support also did not predict visits to the doctor or changes to
respondents’ eating habits, as did both the global measure and informational support
individually. However, the perception of emotional support was significantly associated
with seeking additional information online, which the other two measures were not.
105
Table 6.9: Summary of significant associations between perception of social support and
outcomes
Perception of
informational
support
Perception of
emotional
support
Overall
perception of
global
support
Attitudes
Looking forward to getting
pregnant again
Always trying to be healthier
+
Anxiety about pregnancy
Happiness about pregnancy
+
Behaviors
Doctor visits scale
+
+
Changed eating
Changed drinking
Spoke to doctor about web site
Sought additional information
online
Sought second medical opinion
+
Felt less stress about pregnancy
Felt more confident in taking care
of oneself
Overall behavior change index
Total Significant Associations 11 9 11
+
Association at the level of p < .10
RQ8a. How is social capital on a health-related social networking site associated with
health outcomes? Specifically is the perception of global social capital associated with:
(i) knowledge of prenatal health;
(ii) attitudes towards pregnancy and prenatal health; and
(iii) health behavior?
Hierarchical multiple linear regression and hierarchical logistic regression were
used to examine the associations between the overall measure of social capital (the
106
average of the bonding and bridging sub-scales from the participant survey) and health-
related outcomes. The models also controlled for significant associations with health
status, education, income, whether this was a first pregnancy, minority status, marital
status, full-time employment, age, and the specific web site of which the respondent was
a member.
Knowledge. The knowledge outcome measured how well social capital predicted
the acquisition of new knowledge on the social networking sites. The results of a
hierarchical logistic regression show that health status, age, and social capital were
significantly associated with respondents indicating they had learned something new
from participating on the site (overall χ
2
(3) = 13.07, p < .01). With each increase in
health status, respondents were more than twice as likely to have learned something new
from the site (AOR = 2.39, p < .05) and with each increase in social capital, respondents
were more than one-and-a-half time more likely to have learned something new (AOR =
1.87, p < .10). As for age, with each unit increase in age, respondents were more than
10% less likely to have learned something new from the site (AOR = .88, p < .05).
Attitudes. The analyses examined how well social capital predicted three specific
attitudes: 1) a positive attitude toward the physical changes of pregnancy; 2) a positive
attitude toward getting pregnant again; and 3) a positive attitude toward being healthy.
Further, analyses examined if social capital was associated with women’s overall
attitudes of anxiety, depression, and happiness about their pregnancy. The results show
only one significant association. Social capital predicted a positive attitude toward trying
to be healthy (adjusted R
2
= .13, F (3, 96) = 5.79, p < .001). The best fitting model
107
included three significant standardized beta coefficients, minority status, β = .21, p < .05,
being a member of Site G, β = - .28, p < .01, and global social capital, β = .20, p < .05.
The results indicate that being a minority and perceiving social capital on the site were
associated with trying to be healthier while being a member of Site G was not.
Behavior. Stepwise logistic regression analyzed the association between social
capital and behavior. As Table 6.10 reports, social capital was significantly associated
with changing seven behaviors: 1) eating; 2) drinking; 3) speaking with doctor about the
SNS; 4) seeking additional information online; 5) speaking with friends and family about
the SNS; 6) feeling less stress; and 7) feeling more confident in how the respondents took
care of themselves. Furthermore, a hierarchical stepwise regression also showed a
significant association between social capital and overall behavior change (adjusted R
2
=
.22, F (3, 97) = 10.60, p < .001). The standardized beta coefficients were: first pregnancy,
β = .20, p < .05, working full time, β = - .27, p < .01, and social capital, β = .45, p < .001,
suggesting that women who were pregnant for the first time, those without full-time jobs,
and those with more social capital changed their overall behavior more as a result of their
participation on the SNS.
108
Table 6.10: Adjusted odds ratios indicating associations between social capital and behavioral change as a result
of participating on the SNS
Changed
what you
eat
Changed
what you
drink
Spoke with
doctor
about S
S
Sought
additional
info.
online
Spoke
with
family or
friends
about the
S
S
Felt less
stress
Felt more
confident
Health status -- -- -- -- -- -- --
Education -- -- -- -- -- -- --
Income -- -- 1.28* -- -- -- --
First pregnancy -- -- -- -- 5.46** -- --
Minority status -- -- -- -- -- -- --
Marital status -- -- -- -- -- -- --
Fulltime worker .41* .20** .27* -- -- -- --
Age -- -- 1.05 -- -- -- --
Site D .06 -- -- -- --
Social capital 2.14** 2.83** 4.85*** 2.16* 2.39* 2.21** 3.07***
Overall χ
2
(df) 9.78** (2) 22.18***
(3)
31.12***
(3)
4.31* (1) 9.80** (2) 7.66** (1) 11.52***
(1)
+p < .10 * p < .05 ** p < .01 ***p < .001
109
RQ8b. Are there differences in how the specific dimensions of bonding and bridging
social capital associated with health outcomes? Specifically, what are the differences in
their associations with:
(i) knowledge of prenatal health;
(ii) attitudes towards pregnancy and prenatal health; and
(iii) health behavior?
Similar to the previous research questions, hierarchical stepwise regression and
logistic regression were used to examine the associations between the bonding and
bridging sub-scales and health-related outcomes. The models also controlled for
significant associations with health status, education, income, whether this was a first
pregnancy, minority status, marital status, full-time employment, age, and the web site of
from which respondents were recruited.
Knowledge. Of the two sub-scales, only bonding social capital was associated
with learning something new from the social networking site. The results of a hierarchical
logistic regression show that health status, age, and social capital were associated with
respondents indicating they had learned something new from participating on the site
(overall χ
2
(3) = 12.99, p < .01). The three significant co-variants in the model had the
following adjusted odds ratios: health status, 2.39, p < .05, age, AOR = .87, p < .01, and
social capital, AOR = 1.65, p < .10, suggesting that younger, healthier women with
greater bonding social capital learned more new things from participating on the web site.
Attitudes. Both bonding and bridging social capital were associated with a
positive attitude toward being healthy. As Table 6.11 shows, each model also contained
110
the same three significant co-variants: minority status, being a member of Site G, and
social capital, suggesting that respondents of a minority ethnicity, with greater social
capital, who are not members of Site G were more likely to have a positive attitude about
being healthy.
Table 6.11: Standardized beta coefficients indicating significant associations between
bonding and bridging social capital and a positive attitude toward being healthy
Positive attitude
toward being
healthy
Positive attitude
toward being
healthy
Health status -- --
Education -- --
Income -- --
First pregnancy -- --
Minority status .22* .20*
Marital status -- --
Fulltime worker -- --
Age -- --
Site G -.28* -.29*
Bridging social capital .19* --
Bonding social capital .19*
Overall (df) 5.57*** (3, 96) 5.59***
Adjusted R
2
.12 .12 (3, 96)
+p < .10 * p < .05 ** p < .01 ***p < .001
Behavior. Hierarchical logistic regression analyses show that bonding and
bridging social capital were associated with the same number of behavioral change items
as the global social capital measures (seven of 12); however, not all of the measures were
redundant across the two dimensions. Both bonding and bridging social capital were
associated with changes in eating and drinking habits, speaking with a doctor, and
speaking with family and friends about the web site, feeling less stress and more
confidence in how the respondents took care of themselves. Bridging social capital also
predicted seeking additional information online (overall χ
2
(2) = 6.85, p < .01). The
111
significant variables in the equation were first pregnancy, AOR = 5.43, p < .001 and
bridging social capital, AOR = 2.84, p < .01, suggesting that respondents experiencing
their first pregnancy were nearly five-and-a-half times more likely to search for
additional information online, and that with each increase in bridging social capital,
respondents were more than two-and-a-half times more likely to look for additional
information online. Bonding social capital was not associated with searching for
additional information online; however, it was associated with respondents’ changing
their mind about a medical procedure they were once going to undergo (overall χ
2
(2) =
7.63, p < .05). There were two significant co-variants in the best-fitting model, working
full-time, AOR = .34, p < .05 and bonding social capital, AOR = 1.67, p < .10, indicating
that respondents who worked full-time were 66% more likely not to change their minds
about a procedure while with each increase in perceived bonding social capital on the
site, respondents were 67% more likely to do the opposite.
Further support for the association between bridging and bonding social capital
and behavior change was also evidenced in the hierarchical stepwise regression between
these subscales and the overall index of behavior change. While both subscales were
significantly associated with predicting change on this index, as Table 6.12 shows, the
standardized beta coefficient was larger for bridging social capital and the overall chi-
square for the model was also larger, and accounted for more variance in the dependent
variable.
112
Table 6.12: Standardized betas indicating significant associations between bonding and
bridging social capital and overall behavior change
Bonding social
capital
Bridging social
capital
Health status -- --
Education -- --
Income -- --
First pregnancy -- .18*
Minority status -- --
Fulltime worker -.26** -.27**
Age -- --
Behavior change index .37*** .48***
Overall F 10.87*** 12.63*** (3, 97)
Adjusted R
2
.17 .26
+p < .10 * p < .05 ** p < .01 ***p < .001
113
CHAPTER 7
SOCIAL
ETWORK A
ALYSIS METHODOLOGY
Social network analysis was used to examine how the ties between members were
associated with participants’ knowledge, attitudes, and behaviors related to pregnancy
and prenatal care. In addition social network analysis was used to measure the
sociological perspective of social support (social integration), as well as social capital.
Data for the social network analysis were collected in two ways.
First, the data were collected from the content analysis described in a Chapter 3.
Social network data collected in this way included the screenname and socio-
demographic characteristics of everyone active in the network during September 2008, as
well as the type of support sought and received on two sites for one month. This dataset
indicates that Site A had 199 nodes (network members) with 933 ties (connections
between two members), while Site B had 465 nodes with 659 ties. The data from these
sites were used to analyze the research questions related to the structure of the sites, and
the methodological and theoretical questions about social capital and social support.
In addition, the survey collected social network data through two survey questions
that asked respondents to provide the screennames for the 10 people that she provides
support to most often and the 10 people that provide her with support most often. From
these data, two networks were constructed for each social networking site from which
members were recruited. This resulted in 16 networks in total (eight sites X two support
networks). The first support network reflected the flow of outgoing support, with
respondents indicating the other members to whom they provide support. The second
114
support network reflected incoming support, with respondents indicating the other
members that provide support to them.
Unfortunately, few survey respondents provided social network data. Across all
eight social networking sites from which respondents participated, less than 10% of the
pregnant respondents at any one site completed the social network questions on the
survey. This resulted in a sub-sample of 35 pregnant women who provided social
network data on the survey, which represents 28% of all pregnant survey respondents.
Despite the small sub-sample size, the data from these measures were used to examine
associations between social network measures of social support and social capital and
health outcomes.
Measures
Survey Data
In addition to measuring social support and social capital with self-reported
survey scales, theorists have suggested several different social network measures that can
be used to assess these constructs (e.g. Berkman, 1998; Borgatti et al., 1998; Hall &
Wellman, 1985; Wellman, 1982). Among these measures are both egocentric network
measures and sociometric measures of overall network cohesion. Egocentric network
analysis examines a local network around a particular individual (ego), specifically
looking at the people (alters) to whom this ego is connected (Borgatti et al., 1998; Scott,
2007; Valente et al. 2004). Measures of the overall network (sociometric measures) look
at the overall connectivity of network members and the resulting network structure.
Sociometric measures require obtaining social network data from everyone (or almost
115
everyone) in a bounded network and then these local measures are aggregated to get
global network measures (Valente at al., 2004).
In the current analyses, only the content analysis provided complete social
network data, at both the egocentric and sociometric levels, for one month of activity on
two social networking web sites. The survey data is limited by incomplete network data
because not everyone active on the sites participated in the survey and not everyone who
participated in the survey completed the social network measures. Therefore, on the
survey data an ego may have named her alters, but there may not be any corresponding
data from those alters to indicate reciprocity. Therefore, the network data collected from
the survey does not show all linkages, which means that overall network cohesion
measures could not be accurately calculated from these data.
The survey data, therefore, relied on egocentric network techniques to examine
social support and social capital. Specifically, the current analyses used measures of
degree to predict health outcomes as a strategy to better understand the relationship
between social support, social capital and health. Degree was selected as the social
network measure because it is one that is more resilient to missing network data than
others (Borgatti et al., 2006; Costenbader & Valente, 2003). This measure is described in
detail below.
Degree
Degree is one of the most popular measures of what network analysts call
“centrality.” At the egocentric level, measures of centrality can be calculated without
referencing the entire network. Instead, analyses examine the direct connections each
116
individual (node) has to those around her. A node with a higher degree score indicates
that they are the individuals who are the most well connected, as reflected in the number
of ties they have, in the network.
Specifically, degree refers to the number of alters (other network members) to
whom an ego is directly connected. In an asymmetric network, such as the ones in this
analysis, there are two measures of degree. An asymmetric network indicates that not all
ties are reciprocal. Person A might provide Person B with support, but Person B might
not provide Person A with support. In other words, support might only flow one way
between two individuals who are connected (a dyad). This results in an asymmetrical
network. In asymmetric networks, the two measures of degree are indegree and
outdegree. Indegree is a measure of how many other network members are providing
support to a particular ego, whereas outdegree is the support that an ego is providing to
other network members (her alters). Degree has a positive relation to social capital and
social support because the more people an ego is connected to, the more resources to
which she has access and the more opportunities for the exchange of social support
(Borgatti et al. 1998). However, somewhat contradictory, greater degree does not
necessarily correlate with more positive health behaviors. For example, popular middle
school students, those with higher indegree, are more likely to smoke than their less
popular peers (Alexander et al., 2001; Valente et al., 2004). Therefore, the literature
regarding the association between degree – social capital – health is paradoxical.
Degree is also often used to measure the opinion leaders in a network. Recall
from Chapter 2 that opinion leaders are members of a network that have considerable
117
influence over other network members. An Opinion Leader is usually a network member
with the highest indegree, a measure reflecting their popularity; however in the current
study Opinion Leaders are identified somewhat differently.
As mentioned above, in this project there were two support networks constructed
from the survey data. One network reflected outgoing support and the other reflected
incoming support. As Table 7.1 indicates, from these data four dimensions of degree
were measured: indegree and outdegree on the two networks. Of these four measures,
Opinion Leaders, contrary to other studies, are the network members with the highest
outdegree measure in the Incoming Support Network. A higher indegree measure in the
Incoming Support Network might be an indicator of popularity because it indicates the
member who is receiving the most support (as self-reported); however, it also reflects
being on the receptive end of the support exchange dyad which is not the most influential
position. Instead, it is the individual with the higher outdegree measure in the Incoming
Support Network that would be more influential. This measure indicates the individual
that network members recognize as providing them with support; therefore, it is the
support recipients indicating those from whom they receive the most advice.
Table 7.1: Social network measures used to predict health outcomes
Incoming support network Outgoing support network
Indegree How many members provide a
node with support.
How many nodes are indicated as
receiving support from other
members (as reported by other
network members, not self-
reported).
Outdegree How many nodes are indicated
as providing support to other
members (as reported by other
members, not self-reported).
How many nodes to whom a
member provides support.
118
In the Outgoing Support Network, the network members with a higher outdegree
identify those members who provide the most support to other members, as self-reported.
These individuals might also be looked at as Opinion Leaders of sorts, although the data
is susceptible to social-desirability bias as these individuals are essentially identifying
themselves as popular. The higher indegree scores in this network reflect the members
who are receiving the most support from other members, as reported by the support
providers. Generally, the Outgoing Support Network reflects the flow of support as
constructed by the support-providers while the Incoming Support Network reflects the
flow of support as constructed by the support-recipients.
The measures of degree were normalized so that scores could be compared across
networks. Once the network metrics were calculated, the data were imported into SPSS
and used in regression analyses.
Content Analysis Data
Social network data collected from the content analysis was used to evaluate the
flow of support and the structure of the network. While measures of degree were also
calculated from these data, other metrics, including those that assess overall network
cohesion were calculated because these data were much more comprehensive and
represented nearly the entire network. Furthermore, as with degree discussed above,
many of the measures that were assessed can be used as indicators of both social support
and social capital. Therefore, these measures will be discussed as they pertain to both
constructs. The measures will be briefly introduced and then elaborated below.
119
Social support and social capital were primarily assessed through measures of
centralization and tie strength. First, overall network centralization measures such as
degree centrality, closeness, and betweenness were used to examine the accessibility of
ties to one another, or more specifically, the accessibility of support. These measures,
along with density, examine the sites’ overall integration and how well-connected
everyone is in the network. Generally, the more connected everyone is, the greater the
potential for social support and social capital, with one exception. Density may have
inverse relationship with social capital, which will be discussed below.
Second, tie strength was assessed by examining how frequently support dyads
exchanged support, the reciprocity of the exchange, as well as the multiplexity of the ties,
meaning how many different kinds of support was provided across the dyads. The
heterogeneity of the support dyads was also examined by comparing the socio-
demographic characteristics of the support-provider and support-recipients and assessing
their diversity or similarity on dimensions such as age, race, employment status, and
geographic location. More diversity suggests weaker ties with access to more
informational support and bridging social capital, while more similarity across the ties
suggests stronger ties, with access to more emotional support and bonding social capital.
Centrality and Centralization
Centrality and centralization are measures that examine whether there are a few
key individuals at the center of the network or whether all of the connections are diffused
more among more members. Centrality and centralization are similar measures with one
important difference. Centrality measures how central a specific node (individual) is at
120
the local level, in terms of their ties and network position in their local “neighborhood.”
Centralization, on the other hand, looks at how much the overall network is structured
around a few key people (Scott, 2007). Centrality, therefore, is an individual-level
measure while centralization is a network-level measure. Three of the most common
measures of centrality and centralization are degree, closeness, and betweenness. While
degree is considered a local centrality measure, which means it is less susceptible to
missing data (Valente, in press), closeness and betweenness rely on data from the entire
network for their calculation. Since degree has already been discussed, the other two
measures are explained below.
Closeness. Closeness measures how close a node is to everyone else in the
network. This measure can be calculated at the individual level, as a measure of
centrality, with each node having a closeness score, and it can also be assessed at the
network level, which examines how much the network is centralized—or structured—
around this metric. At the individual level, closeness indicates how well-connected an
individual is to other network members; therefore, an individual who is closer to other
people will have more social capital and greater access to social support. Similarly, at the
network level, greater closeness centralization suggests greater social capital and
availability of social support. The particular measure of closeness that this study uses is
Valente & Foreman’s (1998) measure of integration and radiality. Conceptually, these
measures are similar to closeness described above; however, computationally they are
calculated slightly differently resulting in a different measure. Integration is a measure of
how efficiently an individual can be reached by others in a network and radiality refers to
121
how efficiently an individual can reach others in the network. Integration and radiality are
essentially incoming and outgoing measures of closeness. This particular measure was
chosen because it can calculate a network closeness centralization score on a network that
is not completely connected and it works well with asymmetric networks.
Betweenness. Betweenness is a measure that assesses how many times a member
lies on the shortest path connecting all other network members. A high level of
betweenness suggests the cohesiveness of the overall network and, as with degree and
closeness, the greater the integration of the network, the more social capital and
opportunities for the exchange of social support.
Density
Density refers to the proportion of pairs of alters that are connected compared to
all possible connections. Unlike the other measures, it has a negative relation to social
capital because too much redundancy in an individual’s network connections does not
foster social capital (Borgatti et al., 1998). However, density can have positive
associations with social support. For example, studies have shown that close-knit
networks exchange more affective and instrumental support (Barnes, 1954 as cited in
Heaney & Israel, 2002).
Tie Strength
Tie strength was measured by the frequency of contact between two nodes, the
reciprocity of the tie, and the multiplexity of support provisions. All of these measures
are positively associated with tie strength. Frequency of contact refers to how many
support exchanges occurred between two nodes. Hybrid reciprocity, the specific
122
reciprocity measure, is a ratio of all reciprocal ties to all actual ties, and multiplexity
refers to how many different kinds of support were provided between two nodes.
To determine whether weak ties were more likely to exchange informational
support, the support messages were examined between ties that had only exchanged
support once to determine whether informational or emotional support was more likely to
be provided. Further, the analyses also examined whether members that provided only
one kind of support to their alter (e.g. support-recipient) at a time—regardless of how
many interactions they had—were more likely to provide informational than emotional
support.
Heterogeneity
Heterogeneity of the support dyads was examined to better understand tie strength
and bonding and bridging social capital. Heterogeneity was measured by comparing the
socio-demographic characteristics of the support-provider and support-recipients.
Characteristics such as age, race, employment status, geographic location, and pregnancy
experience were compared.
Analysis
Both social network datasets were first imported into a SPSS data file, then
converted and analyzed in UCINET 6.00 Version 1.00 software (Borgatti et al., 2002).
Once the network metrics were calculated, these data were then exported back into one
SPSS data file. The health outcomes were then regressed on the network measures for all
of the survey respondents, regardless of the network in which they participated.
Additional social network analyses were conducted solely in UCINET.
123
CHAPTER 8
SOCIAL
ETWORK A
ALYSIS RESULTS
This chapter begins with one last question about the impact of participation on a
pregnancy-related social networking web site. Specifically, this question ties together the
dissertation’s focus on health impacts and social network analysis to examine how
network measures are associated with health.
RQ9. How are social network measures of social support and social capital, on a health-
related social networking site, associated with health outcomes? Specifically, are these
measures associated with:
(a) knowledge of prenatal health;
(b) attitudes towards pregnancy and prenatal health; and
(c) health behavior?
As mentioned in the previous chapter, there were two support networks examined
from the survey data, incoming support and outgoing support. These analyses will use the
social network measures of indegree and outdegree to predict health outcomes on each
network.
Incoming Support >etwork
Indegree. Results of hierarchical stepwise regression analyses indicate that
indegree was not associated with any knowledge or attitude outcomes; however, there
were several associations between indegree and behavior. Greater indegree was
associated with more visits to the doctor (adjusted R
2
=.46, F (3, 30) = 10.17, p < .001, β
= .55, p < .001) and, as Table 8.1 indicates, it was also negatively associated with
124
exercising (adjusted R
2
=.33, F (3, 30) = 6.34, p < .01, β = -.58 p < .001). Results of
hierarchical logistic regression analyses also indicate that indegree was associated with
one behavior change. With each unit change in indegree, members were more than 30%
more likely to change what they drank (overall χ
2
(1) = 5.14, p < .05). The only
significant variable in the model was indegree, with an AOR = 1.36, p < .10.
Table 8.1: Standardized betas indicating significant associations between indegree and
health behaviors
Doctor visits Exercise
Health status -- --
Education -- --
Income .51*** -.74***
First pregnancy -- --
Minority status -- --
Marital status --
Fulltime worker -- -.40*
Age -- --
Site C .28*** --
Indegree .55*** -.58***
Overall F (df) 10.17*** (3, 30) 6.34** (3, 30)
Adjusted R
2
.46 .33
+p < .10 * p < .05 ** p < .01 ***p < .001
Outdegree. In the Incoming Support Network, there were no associations between
outdegree and knowledge or attitude outcomes, and one marginally significant
association with a behavioral change outcome. A hierarchical stepwise logistic regression
analysis shows outdegree was marginally associated with respondents being less likely to
change anything they drank as a result of something they read on the web site (overall χ
2
(1) = 4.53, p < .05, AOR = .03, p < .10). Outdegree was the only variable in the model,
and these results suggest that with each increase in outdegree in the Incoming Support
Network, respondents were nearly 100% not likely to change what they drank.
125
Outgoing Support >etwork.
Indegree. Members with high indegree in this network were members who were
receiving a lot of support as indicated by the support providers. This Outgoing Support
Network reflects the flow of support from the support-provider’s perspective, whereas the
Incoming Support Network, reflects the flow of support from the support-receivers
perspective. In the current network, there were no associations between indegree and
knowledge and one significant association between indegree and an attitude. Indegree
was the only variable in a hierarchical stepwise regression model associated with
members having a positive attitude toward getting pregnant again (adjusted R
2
= .06, F
(1, 32) = 3.17, p < .10, β = .30, p < .10).
There were some significant associations between indegree and behavior. This
measure was marginally associated with members being less likely to say that they
changed how they interact with their doctor as a result of their participation on the web
site. Hierarchical stepwise logistic regression analysis produced a model that was
significant for changing doctor interaction (overall χ2 (2) = 10.01, p < .01). The best
fitting model included two significant co-variants, working full-time, AOR = .09, p < .01,
and indegree, AOR = .85, p < .10. This model suggests that respondents who work full-
time and those who are receiving more support were less likely to change how they
interacted with their doctor. Similarly, indegree was also negatively associated with
respondents feeling confident in how they take care of themselves (overall χ2 (2) = 5.73,
p < .05). Indegree had an adjusted odds ratio of .86, p < .10 and income had an adjusted
odds ratio of 1.74, p < .10, indicating that respondents with higher income were nearly
126
75% more likely to feel confident in how they take care of themselves as a result of
something they read on the web site, while each increase in indegree was associated with
a 14% decrease in the likelihood that respondents felt confident in how they take care of
themselves.
Moreover, indegree in this Outgoing Support Network was also negatively
associated with cumulative behavior change. A hierarchical stepwise regression (adjusted
R
2
= .23, F (3, 30) = 4.28, p < .01) showed that income, β = .51, p < .01, and first
pregnancy, β = .33, p < .05, were positively associated with the behavior change index
while indegree, β = -.45, p < .01, was not. Women who were pregnant for the first time,
with higher incomes, were more likely to change, while women who received more
support were less likely to change.
Outdegree. In the Outgoing Support Network, unlike indegree that was positively
associated with a positive attitude toward getting pregnant again, greater outdegree was
negatively associated with this outcome. A hierarchical stepwise regression (adjusted R
2
= .20, F (2, 31) = 5.18, p < .01) showed that first pregnancy and outdegree were
significantly associated with this outcome. The standardized beta coefficients were: first
pregnancy, β = .35, p < .05, and outdegree, β = -.32, p < .05, suggesting that women who
were pregnant for the first time were more positive towards an additional pregnancy
while those providing more social support to other network members were less likely to
have a positive attitude. Greater outdegree in this network was also negatively associated
with feeling anxious about the pregnancy (adjusted R
2
= .62, F (6, 27) = 9.95, p < .001).
The best-fitting model contained six significant co-variants and their standardized beta
127
coefficients were as follows: health status, β = - .36, p < .01; working full-time, β = -.29,
p < .05; being a member of Site C, β = - .37, p < .01, being a member of Site E, β = - .48,
p < .001, being a member of Site G, β = - .43, p < .01, and outdegree, β = - .21, p < .10.
These results suggest that members in better health, who work work full-time, who
provide more support are less likely to be anxious about their pregnancy. In addition,
these data also suggest that certain web sites may foster more feelings of anxiousness
than others.
Outdegree in the Outgoing Support Network was associated with some of the
same behavioral outcomes as indegree in the Incoming Support Network. This similarity
suggests that providing and receiving support may have similar health impacts.
Outdegree in the Outgoing Support Network was marginally associated with respondents
going to the doctor (adjusted R
2
= .16, F (2, 31) = 4.17, p < .05). The significant
standardized beta coefficients in the model were income, β = .43, p < .05, and outdegree,
β = .31, p < .10, indicating that respondents with higher income who provided more
support were more likely to go to the doctor. Outdegree was also negatively associated
with exercise (adjusted R
2
= 31, F (2, 31) = 8.35, p < .001). The best fitting model
consisted of two variables, income, β = -.59, p < .001, and outdegree, β = - .32, p < .05,
indicating that respondents with higher incomes and those who provided more support as
reflected in the outdegree measure, exercise less often. Hierarchical stepwise logistic
regression found one association between outdegree and behavioral change. Outdegree in
the Outgoing Support Network predicted respondents changing what they drank (overall
χ
2
(2) = 10.94, p < .01). There were two significant variables in the model, working full-
128
time, AOR = .10, p < .10, and outdegree, AOR = 1.90, p < .10, indicating a trend toward
statistical significance and suggesting that respondents who worked full-time were less
likely to change their drinking habits while with each increase in support provided,
respondents were nearly twice as likely to change what they drank. Finally, hierarchical
multiple regression shows that outdegree in the Outgoing Support Network was
significantly associated with overall behavior change as measured on the behavior
changed index (adjusted R
2
= .15, F (2, 32) = 3.87, p < .05). The significant standardized
beta coefficients in the model included income, β = .40, p < .05, and outdegree, β = .33, p
< .10, suggesting that respondents with higher incomes changed on more change
measures overall, as did members who provided the most support to others.
RQ10: Are social network measures of social capital and social support more strongly
associated with survey scales of social capital or social support?
This research question examined how similar the associations between survey
measurements of social support and social capital with health outcomes were to the social
network measures of degree (also indicative of these constructs) with health outcomes.
Since degree can reflect both social support and social capital, these analyses were
designed to see whether degree could essentially be claimed as more indicative of one or
the other construct. As with the previous research question, social network measures
included indegree and outdegree in both incoming and outgoing support networks. These
networks were constructed from survey data. Table 8.2 indicates the significant
associations between the predictor and the outcome variables.
129
Table 8.2: Significant associations between social support, social capital, social network
measures and health outcomes
Incoming & outgoing
support
Outcome Global
social
support
Global
social
capital
Out-
degree
Indegree
Knowledge
New knowledge
+
Attitudes
Looking forward to getting
pregnant again
+ - +
Always trying to be
healthier
+ +
Anxiety
+ -
Happiness
+
Behaviors
Doctor visits scale
+ + +
Exercise
- -
Changed eating
+ +
Changed drinking
+ + +/- +
Changed doctor interaction
-
Spoke to doctor about SNS
+ +
Spoke to family or
friends about SNS
+
Sought additional
information online
+
Sought second
medical opinion
Felt less stress
+ +
Felt more confident
in self-care
+ + -
Overall behavior
change index
+ + + -
Total Healthy Associations 11 10 8 5
- Negative association + Positive association
130
Of the significant associations presented in Table 8.2 the perceived social support
measure was associated with the most outcomes, 11 in all. Of these 11 associations, four
of them were similar to the indegree and outdegree social network measures. The social
capital measure, on the other hand, was significantly associated with ten outcomes and
two of them were redundant to the social network measures. Therefore, on sheer quantity
of redundant associations, the measures of indegree and outdegree parallel the significant
associations of social support.
Pearson correlations between the social network measures and the survey scales
of social support and social capital were also calculated. These correlations are listed in
Table 8.3 and support the previous finding that social support is more strongly associated
with the social network measures. Of the four correlations that compared social support
and social capital with the network measures, two of them were significant at the p < .05
level. The two correlations show a modest, but significant, relationship between social
support and indegree (r = .21, p < .05) in the Incoming Support Network, and outdegree
(r = .25, p < .05) in the Outgoing Support Network.
Social capital, by contrast, was only marginally and less strongly correlated with
three social network measures. In the Incoming Support Network, indegree and social
capital were correlated at r = .17, p < .10. In the Outgoing Support Network, social
capital and outdegree were similarly correlated, r = .17, p < .10, and social capital and
indegree were negatively correlated, r = -.l7, p < .10. It should also be noted that the two
survey scales of global social support and social capital were strongly correlated, with a
Pearson’s correlation of .81, p < .001 (n = 246). The survey scales of informational social
131
support and bridging social capital were modestly correlated at r = .66, p < .001 (n= 246)
and emotional social support and bonding social capital were correlated at r = .79, p <
.001 (n = 246).
Table 8.3: Pearson’s correlations between social support, social capital, and social
network measures
Incoming support network
N = 116
Outgoing support network
N = 97
Indegree Outdegree Indegree Outdegree
Global social support
N = 258
.21* .09 -.10 .25*
Global social capital
N = 246
.17+ .05 -.17+ .17+
+ p < .10 * p < .05 ** p < .01 *** p < .001
RQ11: How does the structure of the network change based on the type of social support
provided?
Five support networks were constructed for two social networking sites based on
the type of social support that was provided from one member to another. One network
measured overall support, while four others measured each dimension of support
(informational, emotional, esteem, and relationship), excluding tangible support of which
none was provided on either site. Centrality measures of degree, closeness, and
betweenness were measured along with overall network centralization, density, and
reciprocity across each of the five networks to examine how the network structure
changed based on the type of support provided.
On Site A the active network for the one month period of September 2008
consisted of 199 nodes with 927 ties between them, whereas Site B had 465 active
members with 659 ties between them. Table 8.4 reports the measures for mean in- and
out-degree, closeness, and betweenness centrality, network centralization, density, and
132
reciprocity. Despite some similar trends across both social networking sites, Table 8.4
shows that there were also some strong differences between the two communities.
As Table 8.4 indicates, the most prominent similarity across both networks was that the
network became less centralized and less dense as the dimension of support provided
varied, from informational to emotional to esteem to relationship. This is also visually
depicted in Figures 8.1 through 8.10. On the first dimension of support, informational
support, there were some key members holding the network together by either sending or
receiving a large amount of support; however, these central figures dropped out of the
network as the less common dimensions of support were provided, resulting in a more
decentralized network. Furthermore, not only did central network members drop out of
the support network as less common forms of support were exchanged, many members
overall also dropped out of the network. Across both social networking sites, about only
15% of the network members that were actively involved in exchanging informational
support were also active in the relationship support network. Consequently, both social
networking sites suggest that there is a hierarchy of social support. On both sites, the
informational support network had the most active members, followed by the emotional
support network. With regards to the last two kinds of support, esteem and relationship,
the social networking sites differed in terms of the size of these respective networks. On
site A there were more members involved in the exchange of esteem support, then
relationship support, whereas on Site B it was the reverse. Furthermore, there were no
members who participated in the exchange of tangible support. This hierarchy suggests
an inverse relationship between the demands of the support provision and the number of
133
Table 8.4: Comparison of five support networks on Social Networking Sites A and B
Overall
network
Informational
support
Emotional
support
Esteem
support
Relationship
support
Site A Site B Site A Site B Site A Site B Site A Site B Site A Site B
Number of nodes 199 465 181 417 170 364 41 35 29 75
Statistics
Out- and indegree mean 6.14 1.59 4.65 1.25 3.45 .98 .41 .07 .12 .13
Normalized out and
indegree mean
.35 .07 .26 .07 .35 .07 .05 .01 .03 .01
Outdegree – s.d. 10.99 2.67 8.76 2.13 6.31 1.66 1.17 .23 .40 .47
Indegree – s.d. 13.95 3.02 10.75 2.43 7.73 2.05 1.76 .47 .62 .50
Normalized integration
and radiality mean
27.799 2.078 26.996 1.201 25.159 .808 3.506 3.305 2.709 .955
Normalized
betweenness mean
.35 .02 .40 .01 .45 .00 .05 .04 .01 .00
etwork centralization (percent)
Outdegree 4.84 1.14 3.80 1.17 4.52 1.23 1.22 .42 .48 .35
Indegree 5.97 .80 4.70 .96 4.83 1.01 1.72 1.28 1.75 .35
Integration
(incloseness)
43.12 15.53 39.09 13.88 35.34 8.74 31.64 26.87 15.69 5.88
Radiality (outcloseness) 15.21 15.75 18.34 11.88 20.00 8.35 14.98 4.67 4.59 4.51
Betweenness 7.05 1.14 8.92 .51 7.30 .35 .54 1.15 .26 .04
Average density/ s.d. .03/.24 .003/.07 .02/.20 .003/.06 .02/.07 .002/.05 .00/.07 .00/.01 .00/.03 .00/.02
Hybrid reciprocity .07 .006 .06 .01 .05 .01 .00 .00 .00 .00
134
people willing to provide it. Informational support may require less time and resource
investment than the provision of emotional support, esteem support, and relationship
support.
An examination of the structure of both networks also reveals that the networks
were not dense, and they became quite sparse as the dimension of support changed. On
each network, there were many more potential ties that could be added. However, of the
two networks, Site A was a denser network with more active members participating in
the exchange of support. The overall support network on Site A had 199 nodes/ members
with 927 ties between them, while Site B had 465 nodes/ members with 659 ties between
them. While both sites had fairly low density measures, on Site A the density measure
was 3% whereas on Site B it was less than 1%. In other words, on Site A 3% of all
possible ties had been made, whereas on Site B less than 1% of all possible ties had been
made. This suggests that not everyone on the site is providing support to everyone else,
and that members pick and choose whom to support and how.
Consistent with the density differences between the social networking sites, the
sites also had large differences related to out and indegree measures. On Site A, the mean
degree sore was 6.14 for the overall support network, indicating that each member, on
average, made connections to six other members during the one-month period. By
contrast, on Site B the mean degree measure of 1.59 means that each member was
connected to less than two other members during the one-month period.
135
Figures 8.1 – 8.10: Sociometric diagrams depicting the provision of overall social support
as well as four specific dimensions on Sites A and B.
Figure 8.1: All support on Site A
Figure 8.2: All support on Site B
Figure 8.3: Informational support on Site A Figure 8.4: Informational support on Site B
Figure 8.5: Emotional support on Site A Figure 8.6: Emotional support on Site B
136
Figure 8.7: Esteem support on Site A Figure 8.8: Esteem support on Site B
Figure 8.9: Relationship Support on Site A Figure 8.10: Relationship Support on Site B
Site A was also more centralized overall on indegree, with measures ranging from
5.97% to 1.75%. When a site is more centralized on indegree it suggests that there are a
few people at the center of the network receiving more support than others. In other
words, the receipt of support is not evenly spread throughout the network. On Site B,
however, the support networks were usually more centralized on outdegree, although not
very strongly, with the exception of esteem support and relationship support. On Site B,
the higher measures of outdegree centralization suggest that the provision of support is
not evenly distributed throughout the network: there are some more central members that
are more active support-providers than others. However, on both Site A and B esteem
support is more centralized on indegree suggesting that there are a few members in
particular receiving more esteem support than others. This is, in fact, visible in Figures
137
8.7 and 8.8. Finally, in the Relationship Network, Site B has the same in- and outdegree
centralization measure of less than 1%, suggesting that the network is not centralized on
either of these measures.
The differences in the closeness measures (as indicated by integration and
radiality) and the betweenness measures from Site A to Site B continue to underscore
how generally dissimilar the two networks were in terms of connectivity. Table 8.4
shows how Site A was more centralized than Site B on betweenness (7.05 versus 1.14)
and integration (43.12 versus 15.53) while both Site A and B had nearly identical
radiality / outcloseness measures (15.21 versus 15.75). The difference in betweenness
measures suggests that Site A had more members who occupied strategic bridging
positions connecting the overall network together, whereas Site B was generally less
structured in this way. Similarly the integration measure on Site A was higher than on
Site B indicating that the network was more tightly connected through incoming support
than was Site B. However, the parity in radiality across the two networks suggests that
the sites were similar in terms of how structured they were in the flow of outgoing
support. Radiality suggests that both networks were similarly integrated on outgoing
support, meaning that there were a similar number of steps connecting the entire
networks together in the flow of outgoing support. However, for Site B, the nearly
identical centralization measures for integration and radiality suggest that the network
was not more tightly held together by members who were either providing or receiving
support whereas on Site A the network was clearly more tightly integrated on integration/
incoming support.
138
Similarly, while both social networking sites also had fairly low reciprocity, on
Site A the reciprocity in the overall network was 7%, while on Site B it was less than 1%.
Essentially support only flowed one way on Site B, whereas on Site A there was some
minimal reciprocal support provision between members—primarily informational
support with 6% reciprocity.
RQ12: Is bonding or bridging social capital more strongly cultivated on health-related
social networking sites?
Using the entire sample (both pregnant and non-pregnant women), > = 288, a
paired samples t-test compared the means of the bonding and bridging social capital
scales that were administered to respondents on the survey. The 5-pt Likert scales asked
respondents to indicate their agreement with a series of statements that reflected the
dimensions of social capital and ranged from strongly disagree (1) to strongly agree (5).
The results of the test show that there was significantly more bridging social capital than
bonding social capital on the sites (4.13 versus 3.42, t (245) = -18.01 p < .001). Further,
when each site was examined individually, across eight sites with sample sizes ranging
from three to 100, each site consistently had significantly more bridging social capital
than bonding social capital.
Survey respondents also completed scales assessing their perception of the
availability of social support on the sites. This Social Provisions Scale (Russell &
Cutrona, 1987) measured five dimensions of social support, including informational and
emotional support. Further corroborating the bonding and bridging measurements, the
mean survey score for perceived informational support was significantly greater than the
139
mean score on the perceived emotional support scale. On a Likert-type scale that ranged
from one to four, where one is strongly disagree and four is strongly agree, the mean
score for informational support was 3.09 versus 2.70 for emotional support (t (258) =
15.96, p < .001).
Finally, bonding and bridging social capital were also analyzed with social
network analysis, yielding similar results. Using data that were collected from the content
analysis, two complete networks were constructed for Site A and B of active members for
one month. The network data included socio-demographic characteristics of the members
and the frequency and type of support exchanged. Overall, the social network analyses
corroborate the findings from the content analysis: these are social networking sites
comprised of ties with infrequent repetitive contact that largely provide informational
support. Below, the two networks of which social network data was collected by content
analysis are discussed in more detail.
Site A
As illustrated in Figure 8.11, 80% of the support dyads (n = 745) on Site A
included a one-time support provision from one member to another; 13% (n = 121)
provided support to the same person twice; 4% (n =36) did it three times; and less than
1% (n = 3) provided support to the same person four times or more in a one-month
period. Furthermore, very few of the ties were reciprocal. This network had a hybrid
reciprocity measure of .07, indicating that seven percent of all ties in the network were
reciprocated. Instead, support was primarily flowing one way.
140
Figure 8.11: Frequency of support provisions to same member on Site A
Across all 1223 messages from one month, nearly half the messages, 49% (n =
604), contained only one dimension of support, while 39% (475) contained two
dimensions. Only 3% (39) of the messages contained three or more dimensions of
support, and 8% (n=103) contained none of the five dimensions of support. Further, the
strength of the ties is also illustrated by how many different dimensions of support were
provided from one member to another, also known as the tie multiplexity. Results from
this analysis further support the assertion that this was a weak tie network. Forty-two
percent (n = 393) of the support-providers across the 927 dyads only provided one
dimension of support to the same recipient, regardless of how many times she provided
support to this particular member, 46% (n = 427) provided two dimensions of support,
and 5% (n = 43) provided three or more. Seven percent (n = 64) of the support dyads
provided none of the five support dimensions.
Site B
As Figure 8.12 illustrates, 91% of support-providers (n = 598) provided support
only one time to a specific recipient; 7% (n = 46) provided support to the same person
80%
13%
4%
1%
One time
Two times
Three times
Four or more
141
twice; 2% (n = 12) did it three times; and less than 1% (n = 3) provided support to the
same person four times or more in a one-month period. Furthermore, very few of the ties
were reciprocal. This network had a hybrid reciprocity measure of .006, indicating that
less than 1% of all ties in the network were reciprocated. Instead, support largely flowed
one way.
Figure 8.12: Frequency of support provisions to same member on Site B
Across all 739 support-provision messages that were sent in one month, a similar
number contained one and two dimensions of support. Forty-four percent of these
messages (n = 325) contained one dimension of support and 45% (332) contained two
dimensions of support. Only 6% (42) contained three dimensions of support, and 5% (n =
40) contained none of the five dimensions of support. Furthermore, differences were
compared across support dyads and revealed a similar trend. Forty percent (n = 262) of
the support-providers across the 659 dyads only provided one dimensions of support to
the same recipient, regardless of how many times she provided support to this particular
member. Forty-eight percent (n = 319) provided two dimensions of support and 7% (n =
49) provided three. Therefore, the evidence demonstrates that the ties in this network
91%
7%
2% 1%
One time
Two times
Three times
Four or more
142
were weak ties in which significantly more informational support flowed between dyads
that had few repeat interactions with each other.
RQ13: How is the strength of a tie associated with the exchange of social support,
specifically:
(a) Are weak ties more likely to exchange informational support?
(b) Are strong ties more likely to exchange emotional support?
On both social networking sites, in messages between weak ties — those wherein
support providers only provided support one time to the same member — there were
significantly more messages that had informational and emotional support than those that
did not. On Site A, 76% (555) of these messages contained informational support which
was significantly more than messages without this kind of support (555 versus 177, χ
2
(1)
= 195.20, p < .001), while 60% (436) of the messages contained emotional support (436
versus 296, χ
2
(1) = 26.78, p < .001).
Similarly, the results for Site B show that 82% (495) of the messages between
weak ties contained informational support which was significantly more than the
messages that did not provide it (495 versus 103, χ
2
(1) = 256.96, p < .001). Further, 63%
(373) of messages provided emotional support, which was significantly more than those
messages without it (373 versus 225, χ
2
(1) = 36.63, p < .001).
Despite the prevalence of both informational and emotional support, when ties
exchanged only one kind of support, regardless of how many interactions they had,
significantly more of the ties provided informational support than not, while significantly
less of the ties provided emotional support. On Site A, 69% (270) of the ties provided
143
informational support, which was significantly more than the 31% (123) of the ties that
did not ( 270 versus 123, χ
2
(1) = 54.99, p < .001). Interestingly, the proportions were the
exact opposite for those who provided emotional support: 31% (120) provided emotional
support and 69% (273) did not (χ
2
(1) = 59.57, p < .001). Similarly, on Site B, 76% of all
ties, wherein only one kind of support was ever provided, provided significantly more
informational support than not (198 versus 64, χ
2
(1) = 68.53 p < .001), whereas
significantly less of the ties provided emotional support than not (58 versus 204, χ
2
(1) =
81.36, p < .001).
To answer the second part of this research question, whether strong ties were
more likely to provide emotional support, chi square analyses tested whether ties that had
three or more contacts provided significantly more emotional support. The analyses of
this sub-sample of ties show that there was not significantly more emotional support
provided in these messages. On Site A, of the 238 messages that were exchanged
between ties in which the support-provider sent supportive messages three or more times
to the same recipient, there was no significant difference in the amount of messages that
contained emotional support versus those that did not. Forty-seven percent (112) of these
messages contained emotional support and 53% (126) did not (χ
2
(1) = .82, NS). There
was, however, significantly more informational support, than not, exchanged in these
messages (184 versus 54, χ
2
(1) = 71.01, p < .001). Similarly, on Site B, of the 48
messages that involved multiple exchanges between the same dyads, there was no
significant difference in the amount of emotional support provided: 46% (22) of the
messages provided emotional support while 54% (26) did not (χ
2
(1) = .33, NS). The
144
exchange of informational support was similar, as well: 48% (23) provided informational
support and 52% (25) did not (χ
2
(1) = .08, NS).
RQ14: Across which, if any, socio-demographic variables do the ties exhibit diversity?
Analyses were conducted in SPSS examining differences in the socio-
demographic characteristics of the support dyads that comprised the overall networks.
The network data were obtained through the content analysis of the support messages on
two sites for one month. Analyses examined how these dyads differed in age, race,
employment status, geographic location, and pregnancy experience. Specifically, the
analysis sought to exam whether similar women were primarily speaking with other
similar women, or whether there was bridging across any of these characteristics.
Age. There was a significant difference in the age of advice-providers and advice-
receivers in the support dyads; however, the results were not consistent across both sites.
On Site A, a paired samples t-test shows that the mean age of advice-providers was
significantly younger than that of advice-receivers (26.52 versus 27.07 t (827) = -2.061, p
< .05). Meanwhile, on Site B, the mean age of advice providers was 26.11 and for
receivers it was 23.33. A paired samples t-test shows that this difference was not
significant (t (17) = 1.37, NS). The low sample size, however, was problematic and was
due to so few women making their age available on this site. Descriptive statistics were
run on the entire network and they showed the mean age of advice-providers was 27.56,
median = 27, mode = 34, s.d = 6.14, n = 106 and advice-receivers had a mean = 25.20,
median = 23, mode = 22, s.d. = 5.42, n = 82. A one sample t-test was conducted to
enlarge the sample size by comparing the mean age of advice providers to the mean age
145
of advice receivers. This difference was significant, suggesting that the advice-providers
on Site B were significantly older than the advice-receivers (27.56 versus 25.20, t (105) =
3.95, p < .001).
Race. On Site A, there were 273 (29% of all dyads) dyads in which the race/
ethnicity information of both participants was coded as being either Asian-American /
Asian, African-American / Black, European American / White, or Hispanic / Latina.
Eighty-four percent (n = 228) of these dyads were of the same race, with all the dyads but
two consisting of Caucasian women. Thirteen percent (n = 35) of the dyads consisted of a
Hispanic and Caucasian woman, and the remainder consisted of a variety of mixed-race
pairings.
On Site B, there were 157 dyads (24% of all dyads) in which the race of both
participants was coded. The missing data is primarily a result of members not posting a
photo of themselves because this was the primary means of coding race, or because
members blocked access to their profile by making it accessible only to their friends.
Seventy-seven percent (121) of these dyads consisted of women of the same race, and
this was almost exclusively Caucasian women communicating with other Caucasian
women (99%, 120), with one African American dyad. Twenty-three percent of the dyads
in which race was recorded consisted of mixed-race communicators, with the large
majority being African American and Caucasian women (22%, n = 35), while one dyad
consisted of a Caucasian and Hispanic woman.
Employment status. There was some diversity in the dyads when it came to
employment status. On Site A, there were 135 dyads in which the employment status of
146
both members was discernible and therefore coded. Employment status was coded as
full- or part-time; working from home; student; and not working. Forty-two percent of the
dyads (n = 57) on Site A had no bridging across this variable. These dyads largely
consisted of two women with full-time jobs (34%, n = 46) and a much smaller proportion
of two women who weren’t currently working (4%, n = 6) as well as student to student
support (4%). Fifty-eight percent (n = 78) of the dyads, then, consisted of two women
with different employment statuses. The majority (31%, n = 31) of these dyads consisted
of a combination of unemployed and fully employed women. In fact, support flowed
almost equally in both directions across these dyads. Sixteen percent (n = 21) of the
dyads contained support flowing from a full-time worker to an unemployed woman and
15% (n = 20) of the dyads had support flowing the other way.
Similarly on Site B, there were 147 dyads in which employment status of both
members was coded. Fifty-four percent (n = 80) of these dyads featured similarly
employed women. Fifty percent (73) of these dyads consisted of two fully employed
women, and 4% (7) of these dyads consisted of two women who were both currently not
working outside the home. Twenty-one percent (31) of advice flowed from an
unemployed woman to one who was working full-time, while four percent of dyads (7)
had support flowing the other way. The remaining dyads consisted of a mix of women
who had full and part-time jobs, those who worked at home, students, and those currently
not working outside the home.
Location. The sites show the most diversity across geographic location. Of the
844 dyads on Site A in which geographic location was coded, 19% (n = 162) consisted of
147
women living in the same region. The locations were coded using the four US Census
Bureau Regions: Northeast, Midwest, South, West; plus an International category. The
greatest amount of homogeneity in location was with the International category, where
7% (n = 58) of the dyads lived. This was followed by the Southern US, (6%, n = 47); the
Midwest (4%, n = 30); the West (2%, n = 20); and finally the US Northeast, where only
seven (>1%) of the dyads lived. Members that lived internationally hailed from five
countries: Australia, Canada, England, Hong Kong, and Lebanon. Of the 58 dyads in
which both members lived internationally, no more than 2% (n = 17) lived in the same
country (Canada).
On Site B, there were 256 dyads in which the location of both members was
coded. There was a fair amount of bridging across regions, with only 27% (n = 70) of the
support dyads living in the same area. As with Site A, the US region with the largest
amount of dyads living in it was the US South, with 17% of all dyads (n = 43) living
there. Four percent of all dyads (n = 11) lived in the Northeast and in the West (n=9), 2%
(n=6) lived in the Midwest, and less than 1% (n = 1) had both members living abroad.
Pregnancy experience. Pregnancy experience was coded as members having had
either zero, one, or more than one pregnancy. For this analysis, if the coder was unable to
tell whether the member was or had been pregnant before, they were coded as not having
pregnancy experience. Approximately half of the members on each of the sites were
receiving support from other members who had had a different number of pregnancies
than the support-receiver. Interestingly, it was not always the members with more
experience who were providing support to those with less experience.
148
On Site A, 47% (n= 439) of the support dyads consisted of women who had a
similar number of pregnancies. Interestingly, the most advice was provided from
members who had been pregnant more than once to other members who had also been
pregnant more than once (27%, n = 250). Twenty percent (n = 189) of the similarly-
matched members consisted of women with only one pregnancy providing support to
each other. As for the dyads of women with different pregnancy experience, on Site A
there was nearly an equivalent number of women who had had more than one pregnancy
providing advice to women pregnant for the first time, as there were “first-timers”
providing advice to women who had had more than one pregnancy. Twenty-four percent
(n = 220) of the dyads consisted of women who had been pregnant multiple times
providing advice to those who had been pregnant only once and 25% (n = 228) of the
dyads consisted of women who had been pregnant only once providing advice to those
who had been pregnant multiple times.
On Site B, there was less diversity in the pregnancy experience of the support
dyads. Fifty-nine percent of the dyads consisted of women with similar pregnancy
experience, with the majority of these dyads, 53% (n = 349) consisting of both advice
providers and receivers who had been pregnant only once. On Site B, the second most
common dyad was more diverse and consisted of pregnancy experts providing advice to
members pregnant for the first time. Twenty-one percent (n = 137) of the dyads reflected
this kind of diversity. Ten percent (n = 65) of the dyads included members with only one
pregnancy providing advice to pregnancy experts. The remainder of the dyads consisted
of women with other diverse pregnancy experiences.
149
CHAPTER 9
DISCUSSIO
A
D IMPLICATIO
S
The Internet is now being positioned at the center of a decentralized health care
system that relies on new technology to support a patient-centered model of health-care
(Hughes, Joshi, & Wareham, 2008). This new model of health care is one that promotes
collaboration between patients, caregivers, medical professionals and other stakeholders
(Sarasohn-Kahn, 2008). Social networking sites represent an online platform that
supports collaboration between patients and other health-care stakeholders. On social
networking sites, there is opportunity for patients to share knowledge and support to
facilitate better health outcomes for group members. Few studies, however, have
examined the health impact of participation in health-related social networking sites.
Therefore, this dissertation attempted to fill the gap in the literature related to
participation in health-related social networking sites and health outcomes, specifically
examining outcomes related to pregnancy.
Social networking sites that focused on the discussion of pregnancy were selected
in part because of their nearly exclusive interest to women and the unique nature and
prevalence of pregnancy as a health issue. There are more than 4,000,000 live births in
the U.S. each year, resulting in a large population of women who may be in need of
social support and health information. Research suggests that women are more likely than
men to use the Internet for health information (Fox & Follows, 2003; Krommer & Rainie,
2002) and women are more likely to participate in social networking sites (Hargittai,
2007). Studies have also shown that peer-to-peer influence is stronger among same sex
150
dyads (Christakis & Fowler, 2007; 2008). Therefore, pregnancy-related social networking
sites were selected for their unique appeal to women and the important health outcomes
that may be associated with women’s participation on them.
The goals of this dissertation were three-fold. First, on a practical level, this
dissertation sought to examine pregnancy-related social networking sites as a source of
health information and social support. Using a content analysis, the project identified the
health topics that were of concern to this sample of pregnant women along with the
dimensions of social support that were exchanged. Further, using a survey this
dissertation examined what kind of health-related impact participation on social
networking sites had on members. The survey of users examined not only how frequency
and intensity of participation were associated with health-related outcomes, but also
focused specifically on the role of social support and social capital as facilitating
outcomes.
Second, on a theoretical level, this dissertation attempted to untangle the related
constructs of social support and social capital. This dissertation examined these
constructs employing several different approaches. Using data from the survey analysis,
the study examined how social support and social capital were associated with health-
related outcomes for group members. The study also used a social network analysis to
examine how the support network on the web sites changed according to the type of
support that was provided. The specific dimensions of informational and emotional
support and bridging and bonding social capital were also examined in detail. The study
investigated whether bonding or bridging social capital was more prevalent on the sites
151
and, similarly, the study examined whether more informational or emotional support was
provided. In addition, this dissertation looked at how tie strength was associated with the
exchange of informational and emotional support. Finally, the study also examined the
diversity of network ties, as another indicator of tie strength.
On a methodological level, this dissertation compared survey measures of both
social support and social capital to better understand how these constructs are related and
the differences captured in their measurements. In addition, this study compared social
network measures of social support and social capital to the survey measurements of the
same constructs to examine whether social network measures are correlated more with
social support or social capital.
Major Findings and Implications
The first set of research questions focused on the content of pregnancy-related
social networking sites. The purpose of these questions was to better understand the
socio-demographic characteristics of this population of Internet users, the health
information that was being circulated, and the dimensions of social support that were
most frequently exchanged.
Content Analysis
Not surprisingly, this population of Internet users is young (mean age of 27),
American (55%), Caucasian (46%), married (45%), and pregnant (80%). Further, these
figures are likely higher due to missing data (members did not always post this
information). These numbers are consistent with other American studies that have found
152
Caucasians are online more than other racial groups (e.g. Fox & Jones, 2009) and the
younger sample is to be expected given the health topic at the center of this study.
In September 2008, across two pregnancy-related social networking sites, 572
solicitations for support were posted. The sample of support responses consisted of 1,965
responses, with a mean of 3.44 responses per support request. While, there was variety of
health issues that were addressed in these forums, the two most common health concerns
were establishing a pregnancy diagnosis (11%) and labor and delivery (9%). It’s
interesting that the two most common health issues for which support was sought are also
book-ends to pregnancy. This suggests that women are most in need of support when
they first suspect a pregnancy. This seems understandable, especially because most of
these women are pregnant for the first time (72%) and they may be unsure of their health
status. Since pregnancy is a health issue that is typically self-diagnosed with a pregnancy
test or self-assessment of other physical symptoms, the demand for support at the
beginning of a pregnancy might also reflect the support-seeker’s uncertainty of a correct
diagnosis. Finally, the beginning of a pregnancy is also typically marked by an onslaught
of symptoms, such as fatigue or pregnancy cravings, symptoms that may also provoke
support requests.
The second most common health topic for which women sought support was
labor and delivery, another major transitional point during pregnancy. Labor and delivery
represents the “grand finale” of this health condition and it also a critical time of intense
activity for everyone involved. This finding that the beginning and end of pregnancy
instigate the most support requests suggests that it is the change in health status or
153
condition that is critical for support-seekers. It is the transition from a non-pregnant
woman to a pregnant woman and then back again where women require the most support.
Of the different kinds of health information sought (including symptoms,
complications, treatment, diagnosis, risk factors, and prevention) the most common was
information related to symptoms (37%) followed by complications (30%). In addition,
17% of health information messages sought treatment information and 14% sought
information related to a diagnosis. The health information that members sought the least
was information related to risk factors (11%) and prevention or screening information
(5%). These data suggest that many support-seekers need help coping with their
pregnancy symptoms. This was also reflected in the data presented previously regarding
the specific health topics for which support was sought. Six of the top 10 health topics
were related to pregnancy symptoms, including baby movements; morning sickness;
other miscellaneous symptoms; weight gain, general aches and pains; and
stomach/digestion complaints. Since support-seekers are foremost interested in
information related to symptoms and complications, the focus of these sites seems to be
on managing physical symptoms and potential crisis. Symptoms and complications are
typically health problems that are happening in the moment, whereas information related
to risk factors may avert complications, yet few support-seekers are interested in this kind
of information. However, this makes sense, because participation in a social support
community likely does not happen prior to the onset of a problem, but more so when
individuals are in the midst of one, as is reflected here.
154
Finally, the content analysis showed that the most common kind of social support
that group members sought was informational support, followed by emotional support.
This reflects other findings in which informational and emotional support are the two
most popular kinds of support exchanged on social support sites (Dunham et al., 1998;
Finn, 1999; Han & Belcher, 2001). There were far fewer requests for esteem and
relationship support, and none for tangible support, such as money or maternity products.
The lack of requests for tangible support is not surprising given that this is an online
environment where participants might never meet face-to-face to facilitate the exchange
of tangible support.
The findings from the content analysis also indicate that the support sought was
not always provided. Specifically, the data show that 87% of messages sought
informational support, but only 77% of responses provided information. In contrast, only
49% of messages sought emotional support, yet 58% of responses provided it. Esteem
and relationship support, however, each had a reciprocal number of messages that were
seeking and providing it. The discrepancy between the informational and emotional
support seeking and provision suggests that when a member asked for information that
could not be provided, emotional support was substituted instead. Support substitution
may be the hallmark of an attentive community as it might suggest that members try to
help one another even if they do not have the resource that support-seekers are
requesting.
155
Social >etworking Member Survey
The survey analyses focused on examining how participation on a pregnancy-
related social networking site was associated with health-related outcomes. The analyses
focused on three sets of predictor variables: 1) frequency and intensity of participation; 2)
social support; 3) and social capital. Each series of analyses used both hierarchical
multiple linear regression and hierarchical logistic regression to examine how the
independent variables predicted health-related knowledge, attitudes, and practices. The
findings for each series of variables are discussed below.
The survey sample was similar to that indicated in the content analysis results.
The majority of the sample was Caucasian (75%), with a mean age of 31 years. It was a
fairly educated sample with 75% having gone to college for some period of time. Forty
percent of the sample was also currently pregnant.
Frequency and Intensity of Participation and Health Outcomes. Frequency of
participation on the sites was measured with two variables — length of membership on
the site and time spent on the site in one sitting. Intensity of participation was measured
with three sub-sets of variables: 1) seeking five different kinds of support on the sites and
an overall index variable of all support-seeking dimensions; 2) a scale variable measuring
perceived usefulness of the site; and 3) the frequency of seeking and providing support.
Of all these predictor variables, the perceived usefulness of the site was
significantly associated with the most positive health outcomes, 12 in all. As Table 9.1
indicates, perceiving the site as useful was significantly associated with respondents
learning something new, having a positive attitude toward being healthy, feeling less
156
Table 9.1: Significant association between participation variables and outcomes
Perceived
usefulnesss
Membership
length
Time spent
on site at
once
Knowledge
Learning something new
+
Attitudes
Subsequent pregnancy
+ –
Being healthy
+ +
Depression about pregnancy
– –
Happiness about pregnancy
+
Behaviors
Doctor visits
+
Drinking soda
+
Exercise
–
Change what you eat
+
Change what you drink
+
Change your exercise
–
Spoke with doctor about SNS
+ +
Spoke with friends/family about SNS
+
Sought second medical opinion
+ +
Less stress
+
More confident in taking care of
oneself
+ +
Overall behavior change
+
TOTAL HEALTHY OUTCOMES 12 4 3
+ : positive association – : negative association
depressed and happier about their pregnancy, changing what they ate or drank, speaking
with their doctor and family and friends about the web site, seeking a second medical
opinion, and feeling less stress and more confidence in how they take care of themselves,
and, finally, perceived usefulness was associated with an overall behavior change index.
157
In contrast, the quantitative measurements of time spent on the site or
membership length were associated with far fewer positive outcomes. The longer
someone was a member of a social networking site, the more positive their attitude
toward a subsequent pregnancy and toward being healthy, and the more likely they spoke
with a doctor about something they read on the web site, and sought a second medical
opinion. Time spent on the web site in one sitting predicted respondents being less
depressed about their pregnancy, having more doctor visits, and feeling more confident in
how they take care of themselves. However, more time spent on the site in one sitting
was also associated with a negative attitude toward a subsequent pregnancy, drinking
more soda, and exercising less.
One could draw several implications from these results. First, there may not be a
direct correlation between time spent on a health-related social networking site and health
status. The current findings suggest that a qualitative perception of the usefulness of the
site is a better indicator of how much of an impact a site will have on a member. In
addition, the negative health impacts that were associated with the predictor variable of
time spent on the site in one sitting also suggest that there are some effects from this
medium that might be distinct from its content. Specifically, more time spent on the site
in one sitting was associated with drinking soda and exercising less. These are two
outcomes that might easily result from the time spent sitting at the computer, as opposed
to anything a respondent might read in the support messages.
Next, a series of regressions tested how five individual facets of support seeking
(informational, emotional, esteem, opportunity for nurturance and relaxation) were
158
associated with health outcomes, as well as an index variable of all support-seeking
dimensions, with a higher number indicating that a member went to the site in search of
more dimensions of support. Interestingly, as Tables 9.2 shows, of all the five facets and
the support-seeking index, the opportunity for nurturance was associated with the most
positive impacts, six in all. Respondents who participated on a pregnancy-related social
networking site because they wanted to help others were more likely to change what they
drank, speak with a doctor and family and friends about the web site, seek additional
information online, feel less stress, and change more overall as measured by the behavior
change index. This finding is somewhat paradoxical because these are the members who
participated on the site to have an impact on other members — by helping them, or
“nurturing’ them — yet these findings show that members who sought to help others
were also helped themselves.
All of the support-seeking variables, except for the seeking of esteem support,
were associated with the negative outcome of drinking more soda. Respondents who
sought informational support, emotional support, opportunity for nurturance and
relaxation and the overall index of support were all more likely to drink more soda as a
result of their participation on the web site. Furthermore, seeking each facet of support
was also consistently associated with at least one negative outcome, such as drinking
more soda, but some of the support-seeking variables predicted two and even four
negative outcomes. Opportunity for nurturance and relaxation and the index of support
sought were each only associated with one negative outcome, drinking soda. Seeking
informational support was associated with drinking soda and having a negative attitude
159
Info.
support
Emo.
support
Esteem
support
urture Relax SS Index
Knowledge
Learn something new
+
Attitudes
Positive attitude toward physical changes
–
Positive attitude toward subsequent
pregnancy
– –
Depression about pregnancy
+ +
Anxiety about pregnancy
+
Happy about pregnancy
+ –
Behaviors
Drinking soda
+ + + + +
Exercise
–
Changed what you drink
+ +
Spoke with doctor about SNS
+
Sought additional information online
+
Spoke with friends/family about SNS
+ +
Felt less stress
+
More confident in taking care of oneself
+
Overall behavior change index
+ + +
TOTAL HEALTHY OUTCOMES 1 0 1 6 1 4
+ : positive association – : negative association
Table 9.2: Significant associations between dimension of support sought and outcomes
160
toward a subsequent pregnancy. Seeking esteem and emotional support, however, were
each associated with four negative outcomes. Seeking esteem support predicted a
negative attitude toward the physical changes of pregnancy and toward a subsequent
pregnancy, more depression about the pregnancy, and less exercise. Members who sought
emotional support were more likely to be depressed and anxious, and less happy about
their pregnancy; in addition to drinking more soda.
These support-seeking findings suggest several things. First, it is not surprising
that seeking various kinds of support is associated with negative outcomes. The seeking
of support is reflective of a member who needs help and is potentially struggling with her
pregnancy. Most of the negative outcomes that support-seeking was associated with were
attitudinal, not behavioral (with the exception of drinking soda), underscoring that
members were seeking support perhaps to help them cope psychologically with their
pregnancy. Furthermore, additional longitudinal data is necessary to assess the
effectiveness of members’ support-seeking. With the current cross-sectional data, we
cannot assess the impact that received support may have had on members. Future
research should measure received support, as opposed to perceptions of support, to better
assess that relationship.
Also, from these data it is evident that support-seeking is not additive. Seeking
more dimensions of support from a social networking site is not associated with more
health-related impacts. However, seeking more dimensions of support on the sites was
associated with more behavioral impact than seeking any one kind of support
161
individually. There was one exception though. The seeking of an opportunity for
nurturance was associated with more behavioral impacts than the index.
Participating in the site to relax and have fun was associated with the fewest
number of outcomes, only drinking soda and overall behavior change. This suggests that
the more members participated in the site for fun, the more they drank soda, and the more
they changed their overall behaviors. While this may sound paradoxical, this suggests
that a small group of members who participated on the sites to relax and have fun
consistently changed on several different behavioral outcomes. While there was not a
significant amount of change on each of the individual outcomes, cumulatively when all
the variables were looked at together, there was significant change.
Participation on the site was also assessed by the frequency that members sought
or provided support. As Table 9.3 indicates, while seeking support was significantly
associated with change on one outcome — support-seekers were more likely to change
something that they drank — providing support was associated with eight positive
outcomes. Similar to the finding above about opportunity for nurturance, these data show
that members who provided support the most often were more likely to change than
members who provided support less often. Providing support was significantly associated
with eight positive outcomes (and one negative one, drinking soda), including a positive
attitude toward a subsequent pregnancy, less depression about their current pregnancy,
changing what they drank, speaking with their doctor and family and friends about the
web site, seeking additional information online, being more confident in how they take
care of themselves and also their overall behavior change.
162
Table 9.3: Significant associations between support seeking and providing and outcomes
Seeking
Support
Providing
Support
Attitudes
Subsequent pregnancy
+
Depression about pregnancy
–
Behaviors
Drinking soda
+
Changed what you drink
+ +
Spoke with doctor about SNS
+
Sought additional information online
+
Spoke with friends/family about SNS
+
More confident in taking care of oneself
+
Overall behavior change
+
TOTAL HEALTHY OUTCOMES 1 8
+ : positive association – : negative association
These findings suggest that the impact of participation on support web sites is
stronger for support-providers than for support-seekers. This signifies a major
paradigmatic shift: instead of focusing on how support messages influence their audience,
these data suggest an additional focus on how support messages influence their authors.
Although scholars have previously suggested the importance of the written mode of
online communication (Baym, 2002; Braithwaite et al., 1999; Weinberg et al., 1995;
Wright & Bell, 2003), the focus has typically been on how the written mode of
communication facilitates the support-seeker, not the support-provider. However, these
data suggest that the time that support-providers put into thinking about and writing a
response may strengthen the author’s resolve to follow their own advice. For example, if
support-providers advocate certain attitudes or behaviors in their responses, they may
experience cognitive dissonance if they do not follow their own advice. As a result,
163
support-providers may ultimately be more influenced by their own advice than their
intended audience.
Social Support and Social Capital and Health Outcomes. As Table 9.4 indicates,
both social support and social capital were associated with a number of positive outcomes
and no negative outcomes. The global measures of social support and social capital were
both associated with a positive attitude toward being healthy, changing what a respondent
ate and drank, speaking with a doctor about the web site, feeling less stress and more
confidence in how a respondent takes care of herself, and the overall behavior change
index. In addition to these shared significant associations, there were several differences
between the two constructs. For instance, global social support was also associated with a
positive attitude toward a subsequent pregnancy, less anxiety, more happiness about the
current pregnancy, and more doctor visits. Global social capital was also uniquely
associated with seeking additional information online and speaking with friends and
family about the web site.
The differences in the health-related impact of these two constructs are primarily
due to their impact on attitudes. While the behavioral impact is similar across these two
constructs, the perception of social support as measured by the Social Provisions Scale
(Russell & Cutrona, 1984) was associated with more positive attitudes than the
perception of social capital, as measured by the Internet Social Capital Scales (Williams,
2006). This difference in attitudinal effects may be a reflection of the more externally
directed focus of social capital, in which it is often conceptualized as a means to achieve
164
+ : positive association – : negative association
Table 9.4: Significant associations between the perception of social support, social capital, and outcomes
Global
social
support
Info.
support
Emo.
support
Global
social
capital
Bridging
social
capital
Bonding
social
capital
Knowledge - Learn something new
+ +
Attitudes
Positive attitude toward subsequent preg.
+ + + + +
Positive attitude toward being healthy
+ + + +
Anxiety about pregnancy
– + +
Happy about pregnancy
+
Behaviors
Doctor visits
+ +
Changed what you eat
+ + + + +
Changed what you drink
+ + + + + +
Changed mind about medical procedure
+
Spoke with doctor about SNS
+ + + + + +
Sought additional information online
+ + +
Spoke with friends/family about SNS
+ + +
Sought second medical opinion
+
Felt less stress
+ + + + + +
More confident in taking care of oneself
+ + + + + +
Overall behavior change index
+ + + + + +
TOTAL HEALTHY OUTCOMES 11 10 8 10 9 10
165
another end such as a new job, a new social contact, or neighborhood safety. Social
support may be more internally directed because it functions as a coping mechanism,
which may explain why it was more significantly associated with attitudes. Irrespective
of the causal factors of these differences, these data are provocative for illustrating that
there are some experiential differences between these two constructs at the individual
level.
Interestingly, the differences in outcomes between the specific dimensions of
informational and emotional support and bonding and bridging social capital also
reflected the difference in attitudinal effects. Behaviorally these dimensions of social
support and social capital had similar impact, but attitudinally both dimensions of social
support were associated with more attitudes than were the dimensions of social capital.
However, the perception of informational and emotional support was associated with
increased anxiety about the pregnancy, an association that did not exist for the bonding
and bridging dimensions of social capital. These associations between informational and
emotional support and anxiety are somewhat counterintuitive because one would think
that the perception of support within one’s social network would decrease anxiety, not
increase it. Further research is needed to examine this association. It makes more intuitive
sense that the perception of bonding or bridging social capital was not associated with
increased anxiety because social capital is a resource, which suggests that its availability
would not significantly increase anxiety. On the contrary, one would expect it to be
associated with decreased anxiety.
166
In sum, these data suggest that, despite the fair amount of overlap in the impact of
social support and social capital on health outcomes, there were some experiential
differences between these two constructs. The differences were reflected in the
associations between social support and attitudinal outcomes, which were not found for
social capital. Moreover, much of the literature around social capital suggests that it is
social support framed in an economic model of profit. This notion of social capital
facilitating financial and behavioral advantages (i.e. of “getting ahead”) may be reflected
in its impact on behaviors and less so on attitudes. Social capital suggests that people are
resources that facilitate outcomes; therefore, social capital may tend to be more
outwardly directed than social support.
Social >etwork Measures of Degree and Health Outcomes. Analyses were also
conducted to test how social network measures of degree were associated with outcomes.
Degree has been theorized to reflect both social capital and social support because it
reflects the direct connections that a network member has to other network members.
These direct connections suggest a network member’s potential access to social support
and social capital. Network measures of indegree (how many network members were
providing support to a particular member) and outdegree (how many other network
members an individual provided with support) were calculated for both an Incoming
Support Network, which reflected support that respondents received, and an Outgoing
Support Network, which reflected support that respondents provided. In essence, the
Incoming Support Network reflects the flow of support from the support-receiver’s
167
perspective while the Outgoing Support Network reflects the flow of support from the
support-provider’s perspective.
As Table 9.5 shows, in the Incoming Support Network, greater indegree was
associated with three behavioral outcomes. Members who received more support, as
measured by indegree, were more likely to go to the doctor and change what they drank,
but they also exercised less often. Outdegree, in the Incoming Support Network, reflected
individuals who provided more support as indicated by the support-recipients. Therefore,
greater outdegree may be an indicator of opinion leadership. This measure was associated
with one negative outcome: greater outdegree was negatively associated with being less
likely to change what a member drank. These findings suggest that although receiving
support was minimally associated with positive effects, individuals identified as opinion
leaders showed no positive outcomes as a result of participation on the site. However, the
literature on opinion leaders focuses more on opinion leaders as influencing change in
other members while not necessarily being at the forefront of change themselves (Valente
& Pumpuang, 2007).
The findings from the Outgoing Support Network, specifically the outdegree
measure, support the previous survey findings that the provision of support is influential
in changing the support-provider, perhaps more so than the support-recipient. While
providing support was associated with some negative outcomes, including not looking
forward to another pregnancy and exercising less, the positive associations indicated that
providing support was associated with less anxiety about the pregnancy, more doctor
visits, changing drinking behaviors, and overall behavior change.
168
Table 9.5: Significant associations between social network measures of degree and
outcomes
Incoming support
network
Outgoing support
network
Indegree Outdegree Indegree Outdegree
Attitudes
Positive attitude toward
subsequent pregnancy
+ –
Anxiety about pregnancy
–
Behaviors
Doctor visits
+ +
Exercise
– –
Changed what you eat
Changed what you drink
+ – +
Changed doctor interaction
–
More confident in taking
care of oneself
–
Overall behavior change
index
– +
TOTAL HEALTHY
OUTCOMES
2 1 3 4
+ : positive association –: negative association
While these data are limited because of the small sub-sample of survey
respondents who provided social network data (n = 35), the significant findings are
provocative, especially in light of previous findings indicating the association of the
provision of support and outcomes. Moreover, their significance lies in their exploratory
value and suggests the need for future research with a larger sample to size to more fully
examine these, and other, associations.
169
Social
etwork Analyses
The last series of analyses primarily used social network analysis to examine the
structure of the support networks and its implications for the exchange of social support
and the creation of social capital.
Social >etwork Indicators: Valid Measures of Social Support or Social Capital?
Both the social support and social capital literatures have suggested that social
network measures can be used to measure these constructs. The problem is that scholars
have suggested using the same or similar network measures as indicators of both social
support and social capital. One of the objectives of the current study was to compare
survey measurements of participants’ self-perceptions of social support and social capital
and identify whether the social network measures are more valid measurements of one or
the other construct.
As indicated in the discussion of survey results above, when network measures of
degree were used to predict health outcomes, the significant outcomes were more similar
to the outcomes predicted by the survey measures of social support. Moreover, when
measures of social support and social capital were correlated with measures of indegree
and outdegree, the correlations were stronger with social support. Therefore, these data
suggest that social network measures of degree are better indicators of social support and
less so of social capital. While this may be due to the topic under consideration it is
nevertheless intriguing.
Moreover, further consideration must be given to the implications of relying on
social network measures of only degree as indicators of social support and/or social
170
capital. As discussed in an earlier chapter, degree is a measure of local centrality. This
dissertation was able to use degree as a fairly reliable measure because it would not be as
impacted as some other measures by missing data (Costenbader & Valente, 2003).
Measures of degree are obtained by examining the links of each node, without reference
to the overall structure of the network (Valente, in press). As a consequence, degree does
not capture the overall connectivity of a network. Therefore, as a sole indicator, degree
may be minimizing social capital as a consequence of not considering the connectivity of
the entire network.
A more accurate social network measure of social capital would take into account
the overall connectivity of the entire network. Again, the current study precluded the use
of other measures due to missing data. While social support might be considered a
“product” that is more easily identified as it flows through a network, social capital
reflects the overall connectedness of a network that then enables products or commodities
(like social support) to flow. This suggests that you can have social capital without
having social support, but you cannot have social support without having social capital.
In other words, social capital may be the necessary structure that facilitates the flow of
social support.
Suggesting that social capital might be better captured by overall network data
does not necessarily imply that social capital is more accurately described as a property
of groups, not individuals. Nevertheless, using social network measures to evaluate a
complete network would provide a better assessment of social capital because it would
take into account all of the ties within a network. The current findings are noteworthy,
171
though, because they suggest that social support and social capital at the individual level
are not redundant, as measured by their unique patterns of associations with health
outcomes. Both survey scales of social support and social capital captured individual-
level data and despite the strong correlation of their scales (r = .81, p < .001) they were
not significantly associated with all of the same impacts. The distinct pattern in their
differences showed that social capital was not as strongly associated with attitudes as was
social support.
>etwork Structure and Social Support
The structure of the network changed discernibly with the dimension of social
support being provided. This analysis examined five dimensions of social support,
informational, emotional, esteem, relationship, and tangible. The social network analysis
suggests that as the demands of providing each type of support increased, the network
had fewer members and became less centralized and less dense. Essentially, the network
started to deteriorate. Most of the network measures indicating the connectivity of the
network steadily decreased as the support provided moved from informational, to
emotional, to esteem, to relationship, to tangible (of which there was not a network
because no one provided that dimension of support, a dimension of support that requires
the most effort on behalf of support-providers). These findings suggest that there may be
a hierarchy of social support. This hierarchy was more explicit on Site A, the denser
network.
One additional point of interest, while the social network analysis illustrates that
Site B is a less dense social network, with fewer provisions of social support and
172
therefore lower social capital than Site A, there are no significant differences in
respondents’ self-reported perceptions of social support and social capital from the
survey scales. Since there are no significant differences in survey measures of global
social capital, bonding social capital, bridging social capital, or overall social support it is
unlikely that all of these scales are inaccurate, but more likely that overall connectivity is
not important or not perceptible to members. Members may only be concerned with
whether their individual support-seeking needs are being met.
Bridging and Bonding Social Capital, Weak and Strong Ties
The current findings indicate that pregnancy-related social networking sites
cultivate more bridging social capital than bonding social capital. The analyses also
showed that these were networks of weak ties with infrequent repetitive contact that
largely exchanged informational support. Across both sites, an average of 85% of the
support dyads was one-time connections. Few dyads communicated with each other more
than once and less than 7% of the ties were reciprocal (with each tie providing support to
the other) on each site, further suggesting weak tie networks. Moreover, the ties had little
multiplexity, with an average of 88% of the dyads across both sites not providing more
than two dimensions of support to the same member. Of this 88%, a nearly equal
percentage provided one (41%) or two support dimensions (47%) to the same member.
While this dissertation supported previous research that weak ties are more likely
to provide informational support (e.g., Wellman & Gulia, 1997) the study could not
confirm that strong ties were more likely to provide emotional support. Of the ties that
had three or more contacts, there was no significant difference in the amount of
173
emotional support that they provided. Moreover, analyses of the strong ties on Site A
actually showed that they were more likely to provide informational support than not. It
should be noted that there are some methodological issues that make it more difficult to
assess the communication content of messages between strong ties. In order to become a
strong tie, members have to first be a weak tie; therefore, initial communication messages
may have to be excluded from analysis to focus only on later messages. Additional
measurements may be needed to more precisely map the ties’ relational development to
better understand whether they are weak or strong.
The benefit of weak tie networks is that they connect individuals from diverse
backgrounds who can provide each other with a range of information that is not available
through an individual’s strong ties. Since strong ties link individuals who are generally
more homogenous, it is weak ties that provide individuals access to diverse information
(Granovetter, 1973). The strength of weak ties is ultimately the paradoxical notion that
“…interpersonal ties with dissimilar others, despite being ‘weak’ in terms of frequency of
contact and personal trustworthiness, are ‘strong’ in terms of providing expertise”
(Chaffee & Mutz, 1988, p. 31).
In terms of diversity, analyses indicated that the ties were fairly homogeneous
with respect to age and race. On average, in 81% of the ties the race of the women was
the same, primarily Caucasian. In addition, while analyses indicated a statistical
difference in the age of support-providers and receivers, practically there was little
difference. Both support-providers and receivers across both sites were in their mid- to-
late twenties. This lack of diversity, of course, is not surprising given the health topic.
174
Across both sites, there was the most diversity in terms of geographic location and
employment status. On average, 77% of the ties consisted of members living in different
regions. There was also a fair amount of heterogeneity with regards to employment
status. On average, 52% of the dyads consisted of women with different employment
status, with the most common link between fully employed and unemployed members.
While weak ties with diverse others are more valuable for informational support,
similarity between two group members may predict more effective emotional support.
Bippus (2001) found that one of the criteria that individuals used to judge emotional
comforting was the skillfulness of their support-giver in communicating that he or she
related to them and “showed that they had faced similar problems and made the
distressed person feel as though their problems were not unique” (p.307). Other studies
have shown that similarity predicts interpersonal attraction, relational development and
judgments of source credibility (Adamic, Buyukkokten, & Adar, 2003; Craig, Igiel,
Wright, Cunningham, & Ploeger, 2007; Preece, 1999; Wright, 2000).
While these social networking sites are a valuable resource of information for
pregnant women, they may not be providing the emotional support that members are also
seeking. Eighty percent of members indicated that they visited a social networking site
for both informational and emotional support. Interestingly, data from the survey
analyses indicated that seeking emotional support on the sites was associated with more
negative outcomes than was seeking informational support. In fact, seeking emotional
support on the social networking sites was not associated with any positive outcomes, but
it was associated with four negative outcomes, being more depressed, anxious, and less
175
happy about the pregnancy, and drinking more soda. The negative health-related
outcomes of seeking emotional support on the sites might be due to the sites being
primarily weak tie networks that might not be meeting members’ emotional needs.
While intriguing, this premise obviously requires further research. The lack of
multiplexity in online relationships is not too dissimilar from offline relationships where
people also get specialized support from a variety of ties (Wellman & Guilia, 1997;
Wellman & Frank, 2001). Wellman & Gulia (1997) suggest that the idea of offline
communities that are broadly supportive across multiple dimensions might be purely
nostalgic. Furthermore, they suggest that the Internet “has continued the trend of
fostering specialized relationships” (p. 4). Therefore, the data collected in the study are an
example of specialized support relationships that flourish in both online and offline
environments.
Methodological Limitations and Implications
This dissertation triangulated three methodologies to examine how social support
and social capital are constructed on social networking sites, and how participation on
these sites is associated with health-outcomes. Triangulating methodologies proved to be
a worthwhile strategy for verifying the validity of results. For example, the social
network analysis supported findings from the survey data, and the content analysis
supported findings from the social network analysis, etc. There were methodological
limitations for each methodology. The significance of these limitations for the current
study and future research are discussed below.
176
Content Analysis
An overall limitation of the content analysis was that only two pregnancy-related
social networking sites were coded for one month. Future research should examine more
sites with different health issues, including mental health issues. The analysis also
focused on sites primarily used by women only, additional studies should examine sites
that are more popular with men only as well as mixed-gender sites.
In addition, The were challenges in coding the socio-demographic characteristics
of network members on social networking sites because many members chose not to
display their profile photograph or to indicate even basic information, such as the regions
where they lived, in their personal profile. This resulted in a fair bit of missing data for
this part of the content analysis. In future research, more complete data may be collected
if researchers either collaborate with the social network gatekeepers and / or work with an
automated web crawl program that might be able to better extract the data. However, both
of these strategies will doubtlessly increase the cost of conducting research.
Survey
An important methodological implication that arose out of this study involved the
strategy of recruiting survey participants. While it appears that social networking sites
with a more active network are the more desirable population to recruit because of their
interest in the topic, a challenge in posting a survey recruitment message to an online
board is that on the more active boards the recruitment message gets buried that much
faster in all the other new message postings. Therefore, a smaller and perhaps less active
network might be a better target sample for researchers; however, this might then
177
introduce the issue of generalizability. Clearly, there are tradeoffs involved between
recruiting members from a large, very active site or a smaller, more intimate community.
Furthermore, it was sometimes difficult to gain the approval of social network
moderators to post the survey recruitment message in the online community. Several sites
refused to allow the recruitment invitation to be posted with the defense of “protecting”
their members. Other sites allowed the message to be posted but only in a forum
specifically singled out for solicitations. Most sites preferred to have their participation in
the study kept anonymous, while two sites in particular, americanpregnany.org and
babyfit.com, asked that the study acknowledge their participation. The extra layer of
difficulty in coordinating with these online forums is that the researcher is very much
dependent on electronic communication. There were typically no phone numbers
available to contact the gatekeepers, thus making the process to secure approval slow.
Examining the health impacts of participation on a social networking site also
posed challenges in designing a survey instrument that could adequately identify health
outcomes. Since the flow of conversation on a network is dynamic and exposure to any
specific message thread is not guaranteed, studies of online social networks must include
a range of measures to assess health-related impacts. This study attempted to measure
both absolute behaviors, e.g. “Since you’ve been a member of this site have you [insert
behavior here]…” and behavior change measures, “Since you’ve been a member of this
site have you changed [insert behavior here]…” Both measures had their weaknesses.
Assessing health behaviors was also challenging for this specific population because
pregnant women who join an online network for this health issue are already more highly
178
ego-involved in the topic and are generally aware of the good and bad behaviors that they
should be engaging in or avoiding. Hence, many of the absolute measures were highly
skewed. Therefore, the behavior change measures as a group elicited more significant
findings; however, many of these measures are non-directional and therefore do not
explicitly indicate a positive outcome. In fact, some of them could indicative a negative
change. For example, the significance of members changing what they ate or drank is not
an overtly informative measure, yet it does indicate a behavioral impact. In addition,
there are other outcomes, like doctor visits, which are being recognized as positive
outcomes; however, in truth there could be such a thing as too many unnecessary doctor
visits. Another limitation of these measures was the reliance on self-reported data for the
behavioral outcomes. The collection of medical data, data that reflects actual pregnancy
outcomes would also be useful to resolve some of these measurement issues. .
A final major limitation was the incomplete social network data that was collected
on the member survey. From the study’s inception, the response rate for this data was
uncertain because of the information the survey questions asked of respondents, which
was to name the other members with whom they exchange support. In anticipation of a
low response rate for this section of the survey, social network data was also collected
from the content analysis, despite the fact that the content analysis social network data
lacks the behavioral outcomes of the survey data.
Survey respondents did express concern through the online recruitment message
posting that they felt “uncomfortable” providing the screennames of their fellow network
members, and feared violating their privacy, despite being assured that the data would be
179
anonymous. This is a difficult obstacle to overcome, unless researchers can partner with
the social network gatekeepers and perhaps find another way to collect this data.
However, in order to accurately test the health impacts of participating in an online social
network about a health issue, complete network data should be collected in future studies.
Future Research
While the current study has achieved its research objectives, it also suggests
several areas of inquiry for future research. Below, some of the most provocative areas
for future research are discussed.
Social Support Substitution
Findings from the content analysis suggest that when network members did not
have the resources to provide the particular kind of support requested, they often
substituted another kind of support. In particular, this study found that emotional support
may have been substituted for informational support. Future research should examine the
impact of such support substitutions on the support-recipients. Does receiving any
dimension of support help the support-recipient cope, or are there some dimensions of
support that are more acceptable substitutions than others? Furthermore, is support
substitution more appropriate for ties of a particular intensity? For example, is it more
acceptable (i.e., has less negative outcomes) for support group members to substitute one
dimension of support for another, than for a spouse? In addition, what is the role of
communication channel? Is support substitution more acceptable in online situations, but
less so in face-to-face situations?
180
Hierarchy of Social Support
The social support networks in the current study suggested a hierarchy of social
support. This hierarchy seemed to reflect a negative relationship between the demands of
the support provision and the strength of the network, such that the greater the demands
of the support provision, the weaker the network. The social network analysis showed
that there was a denser and more centralized network providing informational support,
and as the dimension of support changed from emotional to esteem support to
relationship support the network became less dense and less centralized. This proposal
rests on some untested assumptions. First, the study used Cutrona and Suhr’s (1992)
typology of social support that consists of five dimensions including, informational,
emotional, esteem, relationship, and tangible support. Future research should test whether
participants perceive that the demands of each type of support increase along a
continuum that begins with informational support and ends with tangible support.
Furthermore, there are a myriad of social support typologies that have been proposed.
Additional research should test the hierarchy hypothesis with another typology to assess
whether is reliable across different typologies. The focus on pregnancy-related sites, as
mentioned previously, is also a limitation for the social network analysis. Future research
should study more sites with different health focuses. Finally, future studies should test
the hierarchy hypothesis with different health issues in different settings to better
understand the factors that are associated with it.
181
Health Impact of Social >etworking Sites
A major finding that resulted from the survey analyses was that there was a
stronger impact on health outcomes for the support-providers than for the support-
receivers. Providing support more frequently and a higher outdegree measure predicted
more health outcomes than those seeking support and those with higher indegree,
reflecting received support. Future research should explore the specific causal
mechanisms that might facilitate these outcomes. For example, research should examine
whether the association with more outcomes is a by-product of the channel—the act of
writing out a response as some researchers have suggested (Baym, 2002; Weinberg et al.,
1995; Wright & Bell, 2003) — or whether there are other mechanisms at work, such as
cognitive dissonance, involvement, etc. In addition, research should explore the
characteristics of support-providers that might make them more or less vulnerable to
impacts of the medium.
Future research on online health content should also untangle the effects of the
health content versus the effects of the medium. This is an issue that Eveland (2003)
addressed in proposing a mixed attributes approach to the study of media. One of
Eveland’s critiques was that too much media effects research focused on the content of
the medium rather than also considering other influential factors of a particular medium.
In the current study, one of the negative outcomes that was associated with some of the
predictor variables was drinking more soda and exercising less. However, what remains
unknown is whether these negative outcomes are a result of something that was posted on
the web site (the media content) or if these negative outcomes are a result of too much
182
time sitting in front of the computer. Future research, perhaps through the use of a
controlled experiment, should address this issue.
Social >etwork Analysis, Social Support, and Social Capital
Future research needs to use data from a complete social network to better
measure the social support and social capital on the sites, especially assessing network-
level metrics, and how they may be associated with health outcomes. In addition, future
research should test how other network indicators of social support and social capital line
up with survey scales of the same constructs. Comparisons could also be made with other
social support scales beyond the Social Provisions Scale (Rusell & Cutrona, 1984) and
other social capital scales beyond the Internet Social Capital Scales (Williams, 2006) to
better understand the relationship between the two constructs, and if the findings hold
constant across different measurement tools. In addition, future research should examine
how other scale measurements of social support and social capital compare.
More work can also be done in examining the characteristics of strong ties in
online networks. They present a more difficult sub-sample of ties to assess because of the
stages of relational development that ties have to go through to become strong. All ties
start off as weak ones; therefore, studies need to carefully consider how they can ensure
that they are examining a sample of strong ties.
Finally, the current study has significant implications for health care providers
who may be planning their own social network site as a patient outreach strategy. Future
research would benefit from a collaboration of researchers and practitioners in the
creation of a social network site so that data could be gathered from a closed network and
183
health practitioners could begin to integrate social media as a strategy for a more patient-
centered model of health care.
Conclusion
This dissertation attempted to fill the gap in the literature related to participation
in health-related social networking sites and health outcomes, specifically related to
pregnancy. This dissertation focused on extending the literature in three ways: 1)
practically, by assessing the health impacts of participation on these sites; 2)
theoretically, by extending the literature related to social capital and social support in
online environments; and 3) methodologically by testing several measurements of the
same and similar constructs against each other to facilitate a better understanding of their
validity. The findings from this study suggest that social support and social capital, while
similar and sometimes overlapping constructs, are experienced differently at the
individual level. For example, greater social support predicted more significant attitudinal
outcomes than did social capital. Furthermore, social network analysis provides a rich set
of theories and methodology can provide valuable insight into the structure of networks
and the flow of resources among network members. Finally, this dissertation supports the
contention that social media represent a new and important way for patients to exchange
health information. Emblematic of Medicine 2.0, social media may be the face of a new
kind of health care interaction, one in which patients are able to pool their knowledge and
experiences together so that they may navigate the health care system with more
confidence and self-efficacy.
184
REFERENCES
Adler, P.S., & Kwon, S.W. (2002). Social capital: Prospects for a new concept. Academy
of Management Review, 27(1), 17-40.
Alexander, C., Piazza, M., Mekos, D., & Valente, T.W. (2001). Peer networks and
adolescent cigarette smoking: an analysis of the national longitudinal study of
adolescent health. Journal of Adolescent Health, 29, 22-30.
Aloise-Young, P.A., Graham, J.W., & Hansen, W.B. (1994). Peer influence on smoking
initiation during early adolescence: a comparison of group members and group
outsiders. Journal of Applied Psychology, 79(2), 218-287.
Anderson, J.G., Rainey, M.R., & Eysenbach, G. (2003). The impact of cyberhealthcare
on the physician-patient relationship. Journal of Medical Systems, 27(1), 67- 84.
Aneshensel, C. S., & Stone, J. D. (1982). Stress and depression: A test of the buffering
model of social support. Archives of General Psychiatry, 39, 1392-1396.
Atkinson, N. L., Saperstein, S. L. & Pleis, J. (2009). Using the internet for health-related
activities: Findings form a national probability sample. Journal of Medical
Internet Research, 11(1). doi:10.2196/jmir.1035
Baker, L., Wagner, T. H., Singer, S. & Bundorf, M. K. (2003). Use of the internet and e-
mail for health care information: Results from a national survey. American
Medical Association, 289(18), 2400-2406
Ball-Rokeach, S. J. (1998). A theory of media power and a theory of media use: Different
stories, questions, and ways of thinking. Mass Communication and Society,
1(1/2), 5-40.
Ball-Rokeach, S.J. & Jung, J.Y. (2003). Media system dependency to communication
infrastructure: A review of the evolution of MSD theory and a proposal for a new
Concept. Paper for the International Conference on Mass Media in the Era of
Globalization, Marketing and Hi-Technology.
Barnes, J.A. (1954). Class and committees in a Norwegian island parish. Human
Relations, 7, 39-58.
Barrera. M, Jr. (1986). Distinctions between social support concepts, measures, and
models. American Journal of Community Psychology, 14(4), 413-445.
185
Bass, S.B. (2003). How will Internet use affect the patient? A review of computer
network and closed Internet-based system studies and the implications in
understanding how the use of the Internet affects patient populations. Journal of
Health Psychology, 8(1), 25-38.
Bass, S.B., Ruzek, S.B., Gordon, T.F., Fleisher, L., McKeown-Conn, N., & Moore, D.
(2006). Relationship of Internet health information use with patient behavior and
self-efficacy: Experiences of newly diagnosed cancer patients who contact the
National Cancer Institute’s Cancer Information Service. Journal of Health
Communication, 11, 219-236.
Baum, F. E. & Ziersch, A. M. (2003). Social capital. J Epidemiol Community Health, 57,
320-323.
Baym, N. K. (2002). Interpersonal life online. In L. Lievrouw & S. Livingstone, (Eds.)
The Handbook of >ew Media (pp. 62-76). Thousand Oaks, CA: Sage.
Beck, C.T. (2001). Predictors of postpartum depression: An update. >ursing Research,
50, 275-285.
Berkman, L. F., Glass, T., Brissette, I., & Seeman, T. E. (2000). From social integration
to health: Durkheim in the new millennium. Social Science & Medicine, 51, 843-
857.
Berkman, L. F. (1995). The role of social relations in health promotion. Psychosomatic
Medicine, 57, 245-254.
Berkman, L.F. (1984). Assessing the physical health effects of social networks and social
support. Annual Review of Public Health, 5, 413-432.
Berkman, C. F., & Syme, S. L. (1979). Special networks, host resistance and mortality: A
nice year follow-up of Alameda Country residents. American Journal of
Epidemiology, 109, 186-204.
Bettinger, J.S., Adler, N.E., Curriero, F.C. & Ellen, J.M. (2004). Risk perceptions,
condom use, and sexually transmitted diseases among adolescent females
according to social network position. Sex. Transm. Dis., 31, 575-579.
Bippus, A.M. (2001). Recipients’ criteria for evaluating the skillfulness of comforting
communication and the outcomes of comforting interactions. Communication
Monographs, 683(3), 301-313.
186
Boase, J. & Wellman, B. (2006). Personal relationships: On and off the Internet. In A.L.
Vangelisti & D. Perlman (Eds.) The Cambridge Handbook of Personal
Relationships. Cambridge: Cambridge University Press. (709-723).
Boneva, B., Kraut, R., & Frohlich, D. (2001). Using e-mail for personal relationships:
The difference gender makes. American Behavioral Scientist, 45(3), 530-548.
Borgatti, S.P., Everett, M.G. and Freeman, L.C. 2002. Ucinet for Windows: Software for
Social Network Analysis. Harvard, MA: Analytic Technologies.
Borgatti, S. P. & Jones, C. (1998). Network measures of social capital. Connections,
21(2),-36.
Bourdieu, P. (1986). The forms of capital. In J. Richardson (Ed.), Handbook of theory
and research for the sociology of education. New York: Macmillan.
boyd, d. m., & Ellison, N. B. (2007). Social network sites: Definition, history, and
scholarship. Journal of Computer-Mediated Communication, 13(1), article 11.
Braithwaite, D.O., Waldron, V.R., & Finn, J. (1999). Communication of social support in
computer-mediated groups for people with disabilities. Health Communication,
11(2), 123-151.
Broom, A. (2005). Virtually he@lthy: The impact of Internet use on disease experience
and the doctor-patient relationship. Qualitative Health Research, 15(3), 325-345.
Burleson, B. R., & MacGeorge, E. L. (2002). Supportive communication. In M. L. Knap
& J. A. Daly (Eds.) Handbook of interpersonal communication (pp. 374-424).
Thousand Oaks, CA: Sage.
Bylund, C., Gueguen, J., Sabee, C., Imes, R., Li, Y., & Sanford, A. (2007). Provider-
patient dialogue about internet health information: An exploration of strategies to
improve the provider-patient relationship. Patient Education and Counseling,
66(3), 346-352.
Caplan, G. (1974). Support systems and community mental health. New York: Behavioral
Publications.
Cassell, M.M., Jackson, C., & Cheavront, B. (1998). Health communication on the
Internet: An effective channel for health behavior change? Journal of Health
Communication, 3, 71-79.
187
Centers for Disease Control and Prevention. (2009). Maternal and infant health research:
pregnancy complications. Retrieved October 6, 2009 from
http://www.cdc.gov/reproductivehealth/maternalinfanthealth/
PregComplications.htm
Chafee, S.H. & Mutz, D.C. (1988). Comparing mediated and interpersonal
communication data. In R.P. Hawkins, J.M. Wiemann, & S. Pingree (Eds).
Advancing communication science: Merging mass and interpersonal processes
(pp.19-43). Newbury Park, CA: Sage.
Chomitz, V.R., Cheung, L.W.Y., & Lieberman, E. (1995). The role of lifestyle in
preventing low birth weight. The Future of Children, 5(1): 121-138.
Christakis, N. A. & Fowler, J. H. (2007). The spread of obesity in a large social network
over 32 years. The >ew England Journal of Medicine, 357, 370-379.
Christakis, N. A. & Fowler, J. H. (2008). The collective dynamics of smoking in a large
social network. The >ew England Journal of Medicine, 358, 2249-2258.
Cline, R. J. W. & Haynes, K. M. (2001). Consumer health information seeking on the
internet: The state of the art. Oxford University Press, 16(2): 671-692.
Cohen, J. (2001). Social relationships and susceptibility to the common cold. In C. D.
Ryff & B. H. Singer (Eds.), Emotion, social relationships, and health (pp. 221-
233). New York: Oxford University Press.
Cohen, J., Gottlieb, B. H., & Underwood, L. G. (2000). Social relationships and health.
In. S. Cohen, L. G. Underwood, & B. H. Gottlieb (Eds.), Social support
measurement and intervention: A guide for health and social scientists (pp. 3-25).
New York: Oxford University Press.
Cohen, J. (1988). Psychosocial of the role of social support in the etiology of physical
disease. Health Psychology, 7, 269-297.
Cohen, S. & Wills, T. A. (1985). Stress, social support, and the buffering hypothesis.
Psychological Bulletin, 98, 310-357.
Cohen, S, & Hoberman, H. (1983). Positive events and social supports as buffers of life
change stress. Journal of Applied Social Psychology, 13, 99-125.
Coleman, J. S. (1990). Foundations of social theory. Cambridge, MA: Belknap Press of
Harvard University Press.
188
Conway, K. S., & Kennedy, L.F. (2004). Maternal depression and the production of
infant health. Southern Economic Journal, 71(2), 260-286.
Costenbader, E., & Valente, T.W. (2003). The stability of centrality measures when
networks are sampled. Social >etworks, 25, 283-307.
Craig. E., Igiel, M., Wright, K. B., Cunningham, C.B., & Ploeger, N. (2007). Will you be
my friend?: Computer-mediated relational Development on Facebook.com. Paper
presented at the annual meeting of the International Association of
Communication Association, San Francisco, May 25-28.
Cronenwett, L.R. (1985). Network structure, social support, and psychological outcomes
of pregnancy. >ursing Research, 34(2), 93-99.
Cutrona, C.E., & Russell, D. (1987). The provisions of social relationships and adaptation
to stress. In W.H. Jones & D. Perlman (Eds.), Advances in personal relationships,
(Vol. 1, pp. 37-67). Greenwich Conn: JAI Press.
Cutrona, C.E., & Suhr, J.A. (1992). Controllability of stressful events and satisfaction
with spouse support behaviors. Communication Research, 19, 154-174.
Cutrona, C.E., Suhr, J.A., & MacFarlane, R. (1990). Interpersonal transactions and the
psychological sense of support. In Steve Duck and Roxane Silver (Eds.), Personal
relationships and social support, (pp. 30-45). London: Sage Publications.
Dejin-Karlsson, E., Hanson, B.S., Ostergren, P., Lindgren, A., Sjoberg, N., & Marsel, K.
(2000). Association of a lack of psychosocial resources and the risk of giving
birth to small for gestational age infants: a stress hypothesis. BJOG: An
International Journal of Obstetrics and Gynecology, 107(1): 89-100.
Diaz, J.A., Griffith, R.A., Ng, J.J., Reinert, S.E., Friedmann, P.D., & Moulton, A.W.
(2002). Patients’ use of the Internet for medical information. Journal of General
Internal Medicine, 17, 180-185.
Dunham, P. J., Hurshman, A., Litwin, E., Gusella, J., Ellsworth, C., & Dodd, P. W. D.
(1998). Computer mediated social support: Single young mothers as a model
system. American Journal of Community Psychology, 26(2), 281-306.
Dunkel-Schetter, C., Sagrestano, L. M., Feldman, P., & Killingsworth, C. (1996). Social
support and pregnancy: A comprehensive review focusing on ethnicity and
culture. In G. R. Pierce, B. R. Sarason, & I.G. Sarason (Eds.), Handbook of social
support and the family (pp. 375-412). New York, NY: Plenum Press.
189
Edlin, M. (2003). Nurture the needs of women at risk for prenatal complications:
Educational pregnancy programs targeting minority members and those with risk
factors help deliver optimal outcomes. Managed Healthcare Executive, 13(7), 32.
Ellison, N. B., Steinfield, C. & Lampe, C. (2007). The benefits of facebook “friends:”
Social capital and college students’ use of online social network sites. Journal of
Computer-Mediated Communication, 12, 1143-1168.
Ennett, S.T., & Bauman, K.E. (1993). Peer group structure and adolescent cigarette
smoking: A social network analysis. Journal of Health Soc. Behavior, 34, 226-
236.
Eveland, W.P. (2003). A “mix of attributes” approach to the study of media effects and
new communication technologies. Journal of Communication, 5(3), 395-410.
Eysenbach, G. (2008). Medicine 2.0: Social networking, collaboration, participation,
apomediation, and openness. Journal of Medical Internet Research, 10(3). doi:
10.2196/jmir.1030.
Feldman, P.J., Dunkel-Schetter, C., Sandman, C.A., Wadhwa, P.D. (2000). Maternal
social support predicts birth weight and fetal growth in human pregnancy.
Psychosomatic Medicine, 62(5):715-725.
Feinstein, A.R., & Cicchetti, D.V. (1990). High agreement but low kappa: I. The
problems of two paradoxes. Journal of Clinical Epidemiology, 43, 543-549.
Fishbein, M., & Cappella, J.N. (2006). The role of theory in developing effective health
communications. Journal of Communication, 56: S1-S17.
Flanagin, A. J., & Metzger, M. J. (2001). Internet use in the contemporary media
environment. Human Communication Research, 27(1), 153-181.
Ford, K., Hoyer, P., Weglicki, L., Kershaw, T., Schram, C., & Jacobson, M. (2001).
Effects of a prenatal care intervention on the self-concept and self-efficacy of
adolescent mothers. The Journal of Perinatal Education, 10(2): 15-22.
Fox, S., & Jones, S. (2009, June 11). The social life of health information: Americans’
pursuit of health takes place within a widening network of both online and offline
sources. A Pew Internet and American Life Project. Retrieved October 19, 2009
from http://www.pewinternet.org/Reports/2009/8-The-Social-Life-of-Health-
Information.aspx?r=1.
190
Fox, S. (2006, October 29). Online health search 2006. A Pew Internet and American
Life Project. Retrieved October 19, 2009 from
http://www.pewinternet.org/Reports/2006/Online-Health-Search-2006.aspx?r=1.
Fox, S. (2005, May 17). Health information online. A Pew Internet and American Life
Project. Retrieved October 19, 2009 from
http://www.pewinternet.org/Reports/2005/Health-Information-Online.aspx?r=1.
Fox, S., & Fallows, D. (2003, July 16). Internet health resources. A Pew Internet and
American Life Project. Retrieved October 19, 2009 from
http://www.pewinternet.org/Reports/2003/Internet-Health-Resources.aspx?r=1.
Fox, S., & Rainie, L. (2000, November 26). The online health care revolution: How the
Web helps Americans take better care of themselves. A Pew Internet and
American Life Project. Retrieved October 19, 2009 from
http://www.pewinternet.org/~/media//Files/Reports/2000/PIP_Health_Report.pdf.
pdf.
Garton, L., Haythornthwaite, C. & Wellman, B. (1997). Studying online social networks.
Journal of Computer-Mediated Communication, 3(1). Retrieved from
http://jcmc.indiana.edu/vol3/issue1/garton.html
Gerber, B.S., & Eiser, A.R. (2001). The patient-physician relationship in the Internet age:
Future prospects and the research agenda. Journal of Medical Internet Research,
3(2). doi: 10.2196/jmir.3.2.e15
Gittell, R., & Vidal, A. (1998). Community organizing: Building social capital as a
development strategy. Thousand Oaks: Sage.
Granovetter, M. (1973). The strength of weak ties. The American Journal of Sociology,
78(6), 1360.
Granovetter, M. (1974). Getting a job: A study of contacts and careers. Chicago:
University of Chicago Press.
Hall, A., & Wellman, B. (1985). Social networks and social support. In S. Cohen, & S.
L. Syme (Eds.), Social support and health, (pp.23-41). Orlando, FL: Academic
Press.
Hargittai, E. (2007). Whose space? Differences among users and non-users of social
network sites. Journal of Computer-Mediated Communication, 13(1), article 14.
Hawe, P. & Shiell, A. (2000). Social capital and health promotion: A review. Social
Science & Medicine, 51, 871-885.
191
Heaney, C. A., & Idrael, B. A. (2002). Social networks and social support. In K. Glanz,
B. K. Rimer, & F. M. Lewis (Eds.), Health behavior and health education (p.
185). San Francisco, CA: Jossey-Bass.
House, J. S., Umberson, D., & Landis, K. R. (1988). Structures and processes of social
support. Annual Review of Sociology, 14, 293-318.
House, J. S., Landis, K. R., & Umberson, D. (1988). Social relationships and health.
Science, 241, 540-545.
House, J. S. (1987). Social support and social structure. Soc. Forum, 2(1), 135-146.
House, J. S., Kahn, R. L., McLeod, J. D., & Williams, D. (1985). Measures and concepts
of social support. In S. Cohen, & S. L. Syme (Eds.), Social support and health
(pp. 83-108). Orlando, FL: Academic Press, Inc.
House, J. S. (1981). Work stress and social support. Reading, Mass: Addison-Wesley.
Houston, T. K. & Allison, J. J. (2002). Users of Internet health information: Differences
by health status. Journal of Medical Internet Research, 4(2). doi:
10.2196/jmir.4.2.e7.
Hughes, B., Joshi, I. & Wareham, J. (2008). Health 2.0 and medicine 2.0: Tensions and
controversies in the field. Journal of Medical Internet Research, 10(3). doi:
10.2196/jmir.1056
Katz, E. & Lazarsfeld, P. F. (2006). Personal influence: The part played by people in the
flow of mass communications. New Brunswick (USA): Transaction Publishers.
Kavanaugh, A.L., Reese, D.D., Carroll, J.M., & Rosson, M.B. (2005). Weak ties in
networked communities. The Information Society, 21(2), 119-131.
Kawachi, I., Subramanian, S.V., & Kim, D. (2008). Social capital and health: A decade
of progress and beyond. In I. Kawachi, S.V. Subramanian, & D. Kim (Eds.),
Social capital and health (pp. 1-26). New York, NY: Springer.
Kawachi, I., Kennedy, B. P, & Glass, R. (1999). Social capital and self-rated health: A
contextual analysis. American Journal of Public Health, 89(8), 1187-1193.
Kawachi, I., Kennedy, B. P., Lochner, S. M., & Prothrow-Stith, D. (1997). Social capital,
income inequality, and mortality. American Journal of Public Health, 87(9),
1491-1498.
192
Kim, D., Subramanian, S.V., & Kawachi, I. (2008). Social capital and physical health. In
I. Kawachi, S.V. Subramanian, & D. Kim (Eds.), Social capital and health (pp.
139-190). New York, NY: Springer.
Krause, N. (1990). Stress, support, and well-being in later life: Focusing on salient social
roles. In M. A. Stephens, J. H. Crowther, S. E. Hobfoll, & D. L. Tennenbaum
(Eds.), Stress and coping in later-life families (pp. 71-97). New York:
Hemisphere Publishing Company.
Lakon, C.M., Godette, D.C., & Hipp, J.R. (2008). Network-based approaches for
measuring social capital. In I. Kawachi, S.V. Subramanian, & D. Kim (Eds.),
Social capital and health (pp. 63-81). New York, NY: Springer.
Leech, N.L., Barrett, K.C., & Morgan, G.A. (2008). SPSS for intermediate statistics: Use
and interpretation. New York: Taylor & Francis.
Licht, M.H. (1995). Multiple regression and correlation. In L.G. Grimm & P.R. Yarnold
(Eds.) Reading and understanding multivariate statistics, (pp. 19-64).
Washington, DC: American Psychological Association.
Liese, L.H., Snowden, L.R., & Ford, L.K. (1989). Partner status, social support, and
psychological adjustment during pregnancy. Family Relations, 38(3): 311-316.
Lin, N. (1999). Social networks and status attainment. Annual Review of Sociology, 25,
467-487.
Loury G. C. (1977). A dynamic theory of racial income differences. In P. A. Wallace, &
A. M. La Mond (Eds.), Women, minorities, and employment discrimination,
(pp.153-186). Lexington, MA: Heath.
Luke, D.A, & Harris, J.K. (2007). Network analysis in public health: History, methods,
and applications. Annual Review of Public Health, 28, 69-93.
March of Dimes (2009). Premature birth. Retrieved October 6, 2009 from
http://www.marchofdimes.com/prematurity/21198_10734.asp
McPherson, M., Smith-Lovin, L., & Cook, J.M. (2001). Birds of a feather: homophily in
social networks. Annual Review of Sociology, 27, 415-444.
Mistry, R., Steven, G.D., Sareen, H., De Vogli, R., & Halfon, N. (2007). Parenting-
related stressors and self-reported mental health of mothers with young children.
American Journal of Public Health, 97, 1261-1268.
193
Miyata, K. (2002). Social Support for Japanese Mothers Online and Offline. In Barry
Wellman and Caroline Haythornthwaite (Eds.), The Internet in everyday life. (pp.
520-548). Malden, MA: Blackwell Publishers.
Neaigus, A., Friedman, S.R., Curtis, R., Des Jarlais, D.C., Furst, R.T., Jose, B., Mota, P.,
Stepherson, B., Sufian, M., Ward, T., et al. (1994). The relevance of drug
injectors’ social and risk networks for understanding and preventing HIV
infection. Social Science & Medicine, 38 (1), 67-78.
Neuendorf, K. A. (2002). The content analysis guidebook. Thousand Oaks: Sage
Publications.
Nigthoujam, P. (2008). The top 10 social networking sites for women. Retrieved [online]
September 10, 2009 from www. http://mashable.com/2008/05/11/top-10=social-
networking-sites-for-women/
Norris, P. (2004). The bridging and bonding role of online communities. In P. N.
Howard, & S. Jones (Eds.), Society online: The internet in context (pp. 31-41).
Thousand Oaks: Sage.
Pearce, N. & Davey Smith, G. (2003). Is social capital the key to inequalities in health?
American Journal of Public Health, 93(1): 122-129.
Perrault, W.D., & Leigh, L.E. (1989). Reliability of nominal data based on qualitative
judgments. Journal of Marketing Research, 26, 135-148.
Piotrow, P. T. Kincaid, D. L., Rimon, J. O. & Rinehart, W. (1997). Health
communication: Lessons from family planning and reproductive health. Westport,
CT: Praeger.
Portes, A. (1998). Social capital: Its origins and applications in modern sociology. Annual
Review of Sociology, 24, 1-24.
Potter, W.J., & Levine-Donnerstein, D. (1999). Rethinking validity and reliability in
content analysis. Journal of Applied Communication Research, 27, 258-284.
Preece, J. J. (1999). Empathetic communities: Reaching out across the Web. Interaction,
2, 32-43.
Putnam, R.D. (2000). Bowling alone: The collapse and revival of American community.
New York: Simon & Schuster.
Putnam, R. D. (1995). Bowling Alone: America’s declining social capital. J. Democr. 6:
65-78.
194
Putnam, R. D. (1993). The prosperous community: Social capital and public life. Am.
Prospect, 13, 35-42.
Rogers, E. M. (2003). Diffusion of innovations (5
th
ed.). New York: Free Press.
Rodgers, S., & Chen, Q. (2005). Internet community group participation: Psychosocial
benefits for women with breast cancer. Journal of Computer-Mediated
Communication, 10(4), article 5.
Rourke, L., Anderson, T., Garrison, D.R., & Archer, W. (2001). Methodological issues in
the content analysis of computer conference transcripts. International Journal of
Artificial Intelligence in Education, 12, 8-22.
Russell, D., & Cutrona, C.E. (1984). The provisions of social relationships and adaptation
to stress. Paper presented at the annual meetings of the American Psychological
Association, Toronto, Canada, August.
Sable, M.R., & Wilkinson, D.S. (2000). Impact of perceived stress, major life events and
pregnancy attitudes on low birth weight. Family Planning Perspectives, 32(6),
288-294.
Sarason, B. R., Sarason, I.G., & Gurung, R.A.R (2001). Close personal relationships and
health outcomes: A key to the role of social support. In Steve Duck (Ed.)
Handbook of personal relationships (2
nd
ed.), pp. 547-573.
Sarasohn-Kahn, J. (2008, April). The wisdom of patients: Health care meets online
social media. California HealthCare Foundation. Retrieved October 19, 2009
from http://www.chcf.org/topics/chronicdisease/index.cfm?itemID=133631.
Sayil, M., Gure, A., & Ucanok, Z. (2006). First time mothers’ anxiety and depressive
symptoms across the transition to motherhood: Associations with maternal and
environmental characteristics. Women & Health, 44(3), 61-77.
Schaffer, M. A., & Lia-Hoagberg, B. (1997). Effects of social support on prenatal care
and health behaviors of low-income women. JOG>> Clinical Studies, 26(4), 433-
440.
Scott, J. (2007). Social network analysis: A handbook. Los Angeles: Sage Publications.
Seeman, T. E. (2001). How do others get under our skin? Social relationships and health.
In C. D. Ryff & B. H. Singer (Eds.), Emotion, social relationships, and health
(pp. 189-210). New York: Oxford University Press.
195
Shade, L.R. (2004). Bending gender into the net: Feminizing content, corporate interests,
and research strategy. In Philip N. Howard & Steve Jones (Eds.) Society online:
The Internet in context (pp.57- 70). Thousand Oaks, CA: Sage.
Shirky, C. (2008). Here comes health. [video podcast]. Retrieved from
http://www.health2con.com on October 6, 2009.
Spiegel, D., & Kimerling, R. (2001). Group psychotherapy for women with breast cancer:
Relationships among social support, emotional expression, and survival. In C. D.
Ryff & B. H. Singer (Eds.), Emotion, social relationships, and health (pp. 97-
123). New York: Oxford University Press.
Stansfeld, S.A. (2006). Social support and social cohesion. In Michael Marmot and
Richard Wilkinson (Eds.) Social determinants of health, pp. 148-171.
Surowiecki, J. (2004). The wisdom of crowds. New York: Anchor Books.
Takahashi, Y., Uchida, C., Miyaki, K., Sakai, M., Shimbo, T. & Nakayama, T. (2009).
Potential benefits and harms of a peer support social network service on the
internet for people with depressive tendencies: Qualitative content analysis and
social network analysis. Journal of Medical Internet Research, 11(3).
doi:10.2196/jmir.1142
Tejada-Vera, B., & Sutton, P.D. (2009). Births, marriages, divorces, and deaths:
Provisional data for 2008. >ational Vital Statistics Reports, 57(19). Hyattsville,
MD: National Center for Health Statistics.
Turner, R. J., Frankl, G., & Levin, D. (1983). Social support: Conceptualization,
measurement, and implications for mental health. In J. Greenley (Ed), Research in
community and mental health (Vol. 3, pp. 67-111). Greenwich, CT: JAI Press.
Valente, T. W. (1995). >etwork models of the diffusion of innovations. Cresskill, New
Jersey: Hampton Press, Inc.
Valente, T.W. (2002). Evaluating health promotion programs. New York: Oxford
University Press.
Valente, T.W. (in press). Modeling influence: Social networks, behavior, and health.
New York: Oxford University Press.
Valente, T. W. & Davis, R. L. (1999). Accelerating the diffusion of innovations using
opinion leaders. A>>ALS, AAPSS, 566, 55-67.
196
Valente, T.W., & Foreman, R.K. (1998). Integration and radiality: measuring the extent
of an individual’s connectedness and reachability in a network. Social >etworks,
20, 89-105.
Valente, T. W., Gallaher, P. & Mouttapa, M. (2004). Using social networks to understand
and prevent substance use: A transdisciplinary perspective. Substance Use &
Misuse, 39(10), 1685-1712.
Valente, T.W. & Pumpuang, P. (2007). Identifying opinion leaders to promote behavior
change. Health Education & Behavior, 34(6), 881-896.
Valente, T.W., & Saba, W.P. (2001). Campaign exposure and interpersonal
communication as factors in contraceptive use in Bolivia. Journal of Health
Communication, 6(4), 303-322.
Valente, T.W., & Unger, J., & Johnson, A.C. (2005). Do popular students smoke? The
association between popularity and smoking among middle school students.
Journal of Adolescent Health, 37, 323-329.
Van der Gaag, M., & Webber, M. (2008). Measurement of individual social capital. In I.
Kawachi, S.V. Subramanian, & D. Kim (Eds.), Social capital and health (pp. 29-
49). New York, NY: Springer.
Ventura, S.J., Mosher, W.D., Curtin, S.C., Abma, J.C., & Henshaw, S. (2000). Trends in
pregnancies and pregnancy rates by outcome: Estimates for the United States,
1976-96. National Center for Health Statistics. Vital Health Statistics, 21(56).
Viera, A.J., & Garrett, J.M. (2005). Understanding interobserver agreement: The kappa
statistic. Family Medicine, 37, 360-363.
Viswanath, K. (2008). Social capital and health communications. In I. Kawachi, S. V.
Subramanian, & D. Kim (Eds.), Social capital and health, (pp. 259-272). New
York, NY: Springer.
Walther, J.B., & Boyd, S. (2002). Attraction to computer-mediated support. In Carolyn
A. Lin & David J. Atkin, (Eds.) Communication and society: Audience adoption
and uses (pp. 153-188). Cresskill, NJ: Hampton Press.
Weinberg, N., Schmale, J.D., Uken, J., & Wessel, K. (1995). Computer-mediated support
groups. Social Work with Groups, 17, 43-55.
Wellman, B. (1981). Applying network analysis to the study of support. In Benjamin
Gottlieb (Ed.), Social networks and social support (pp. 171-200).
197
Wellman, B., & Frank, K. (2001). Network capital in a multilevel world: Getting support
from personal communities. In N. Lin, K. Cook, & R.S. Burt (Eds.), Social
capital: Theory and research (pp. 233-273).
Wellman, B., & Gulia, M. (1997). Net surfers don’t ride alone: Virtual communities as
communities. In P. Kollock, & M. Smith (Eds.) Communities and cyberspace.
Wellman, B., & Wortley, S. (1990). Different strokes from different folks: Community
ties and social support. The American Journal of Sociology, (96)3, 558-588.
White, M., & Dorman, S.M. (2001). Receiving social support online: implications for
health education. Health Education Research, 16(6), 693-707
Williams, D. (2006). On and off the ‘Net: Scales for social capital in an online era.
Journal of Computer Mediated Communication, 11(2), article 11.
Wright, K. (2000). Perceptions of on-line support providers: An examination of perceived
homophily, source credibility, communication and social support within on-line
support groups. Communication Quarterly, 48(1), 44-59.
Wright, K. B. and Bell, S. B. (2003). Health-related support groups on the internet:
linking empirical findings to social support and computer-mediated
communication theory. Journal of Health Psychology, 8(1), 39-54.
Ybarra, M. & Suman, M. (2006). Help seeking behavior and the internet: A national
survey. International Journal of Medical Informatics, 75, 29-41.
Ybarra, M. & Suman, M. (2008). Reasons, assessments and actions taken: Sex and age
differences in uses of Internet health information. Health Education Research,
23(3), 512-521.
Yuan, Y.C., & Gay, G. (2006). Homophily of network ties and bonding and bridging
social capital in computer-mediated distributed team. Journal of Computer-
Mediated Communication, 11(4), article 9.
Ziersch, A.M. (2005). Health implications of access to social capital: Findings from an
Australian study. Social Science & Medicine, 61(10), 2119-2131.
198
APPE
DIX A
SUPPORT-SEEKER CODE SHEET
1.
ame of web site (Circle one):
a. DS b. WTE c. Other:
2. Date of message seeking advice :_ __(mo) / __ __ (day)/ _ __ (year)
3. Subject line of initiating message:
4.
umber of replies to message:
5. Did this message contain an explicit request for support?
a. No [If no, then skip to Question 24]
b. Yes
Informational Support:
6. Was informational support (includes advice, factual input, and feedback on
actions) sought in this message?
a. No [If no, then skip to Question 12]
b. Yes
7. If informational support was sought, what kind of assistance was needed?
(Circle all that apply.)
a. Suggestions/Advice (Someone to offer ideas and suggest actions)
b. Referral (Referral to some other source of help or information)
c. Situation Appraisal (Someone to reassess or redefine the situation)
d. Teaching (Someone to explain the information, facts, or news about a
situation or about skills needed to deal with the situation)
e. Other:
8. Was health information sought?
a. No [If no, then skip to Question 11]
b. Yes
9. If health informational support was sought, what kind of health
information was the focus of this message? (Circle all that apply.)
a. Prevention/screening (e.g. how disease/injury can be prevented)
b. Risk Factors (e.g. pre-disease factors, e.g. age, lifestyle, etc.)
c. Symptoms (e.g., pregnancy related symptoms)
d. Diagnosis (e.g., whether one is pregnant or not)
199
e. Treatment (e.g., medical course of action)
f. Complications (e.g., unanticipated complications of pregnancy)
g. Other:
10. For health informational support, what was the health topic of this
message? (Circle all that apply.)
a. Baby movements
b. Baby – Birth defects (e.g. Down’s Syndrome/ cleft palette, etc.)
c. Baby – “Daily Health” (e.g. breastfeeding; food; diapers; etc.)
d. Baby - Circumcision
e. Braxton Hicks Contradictions
f. Determining/confirming pregnancy
g. Emotional issues (e.g. mood swings; depressions)
h. Exercise
i. Food/drink safety (questions related to food or drinks that can or cannot
be consumed while pregnant)
j. Health insurance
k. Labor and Delivery
l. Maternal drug use (illegal), including drug testing
m. Morning sickness
n. Nutrition (questions related to eating healthy)
o. Pregnancy complications (any medical complication requiring special
treatment) (Circle all that apply below.)
a. gestational diabetes
b. pre-eclampsia
c. breech baby
d. Bleeding/spotting
e. Ectopic pregnancy
f. Abdominal pain/cramping
g. Other:
h. Other:
p. Pregnancy termination (Circle all that apply below.)
a. Abortion
b. Miscarriage – before 20weeks
c. Miscarriage (death of baby) – after 20 weeks
d. Miscarriage – data unknown
q. Prenatal testing (ultrasound; quad screen; amniocentesis; blood tests;
etc.)
r. Sexual and intimacy issues (concerns about being sexual while pregnant)
s. Weight gain (also includes things like coping with stretch marks, etc.)
t. Other:
u. Other:
200
11. If informational support was sought, other than health-related, what was the
topic of this message? (Circle all that apply.)
a. Baby names
b. Baby products
c. Child care
d. Employment-related (e.g. maternity leave)
e. Financial issues
f. Gift registry
g. Hospital procedures
h. Physician interaction
i. WIC program
j. Not applicable (e.g. health information was sought)
k. Other:
l. Other:
Emotional Support:
12. Was emotional support (emotional support and understanding) sought in this
message?
a. No [If no, then skip to Question 15]
b. Yes
13. If emotional support was sought, what kind of assistance was asked for
(either explicitly or implicitly)? (Circle all that apply.)
a. Affection (“Virtual” physical contact, including hugs, etc.)
b. Sympathy (Someone to feel sorrow or regret for the recipient’s situation)
c. Understanding/empathy (Someone to understand the situation perhaps
through personal experience)
d. Encouragement (Someone to provide the recipient with hope and confidence)
e. Prayers (Someone to pray with the recipient)
f. Other:
14. If emotional support was sought, what was the topic of this message? (Circle
all that apply.)
Support for dealing with…
a. Husband/significant other
b. Other family members
c. Workplace/professional issues
d. Concerns about money/financial issues
e. Health care professionals
f. Anxieties about health of pregnancy
201
g. Other concerns about being pregnant/having a baby (i.e. not health-
related)
h. Fear of parenting
Esteem Support:
15. Was esteem support (expression of regard for one’s skills, abilities, and
intrinsic value) sought in this message?
a. No [If no, then skip to Question 18]
b. Yes
16. If esteem support was sought, what kind of assistance was needed? (Circle all
that apply.)
a. Validation
(Someone to express agreement with the recipient’s perspective on the
situation, e.g.: “Yes, you were right to redo the test.”)
b. Relief of blame
(Someone to alleviate the recipient’s feelings of guilt about the situation,
e.g.: “There’s nothing you could have done to prevent that complication
from happening.”)
c. Other:
17. If esteem support was sought, what was the topic of this message? (Circle all
that apply.)
Support for dealing with…
a. Husband/significant other
b. Other family members
c. Workplace/professional issues
d. Concerns about money/financial issues
e. Health care professionals
f. Anxieties about health of pregnancy
g. Other concerns about being pregnant/having a baby (i.e. not health-
related)
h. Fear of parenting
i. Other:
j. Other:
Social Network Support:
18. Was Social
etwork Support** (a sense of belonging among people with
similar interests and concerns) sought in this message?
a. No [If no, then skip to Question 21]
b. Yes
202
** We could generally say that participation in any of these online communities is an
effort to seek social network support. In these questions, we are specifically looking at
the four kinds of social network support specified below.
19. If social network support was sought, what kind of assistance was needed?
(Circle all that apply.)
a. Access (Someone to provide recipient with access to new companions,
including another support group, e.g.: “I found out my baby has a genetic
disorder, is there another support group for this?”)
b. Friendship (An explicit request to make friends and participate in the
group; Would also include any exchange of email addresses; or
request/offer to talk at a later date or in a private conversation)
c. Physical presence (Someone looking to meet other moms in same
geographic area)
d. Companions (Someone who needs to be reminded of the availability of
companions that they can talk to about a problem)
e. Other:
20. If social network support was sought, what was the topic of this message?
(Circle all that apply.)
Support for dealing with…
a. Husband/significant other
b. Other family members
c. Workplace/professional issues
d. Concerns about money/financial issues
e. Health care professionals
f. Anxieties about health of pregnancy
g. Other concerns about being pregnant/having a baby (i.e. not health-
related)
h. Fear of parenting
i. Other:
j. Other:
Tangible Support:
21. Was tangible support (actual, physically-needed goods and services) sought in
this message?
a. No [If no, then skip to Question 24]
b. Yes
22. If tangible support was sought, what kind of assistance was needed? (Circle
all that apply.)
a. Loan (A request to lend the recipient something, including money)
203
b. Gift (A request to give the recipient something, including money)
c. Direct task (A request to perform a task directly for the recipient, e.g.
babysitting)
23. If tangible support was sought, what was the topic of this message? (Circle all
that apply.)
a. Baby products (Things for the baby, e.g. toys, clothes, car seats, bottles,
etc.)
b. Maternity products (Things for the mother, e.g. maternity clothes, breast
pump, etc.)
c. Money
d. Babysitting
e. Other:
f. Other:
24. Briefly describe the advice being sought or the contents of the message.
Socio-demographic Characteristics of Advice-Seeker
25. SCREE
AME:
26. Public Profile?
a. Yes
b. No
32. Location
a. US
b. International
c. Unable to tell
27. Headshot?
a. Yes
b. No
33. State/Country
a. US State or:
b. International Country:
c. Unable to tell
28. Gender
a. Female
b. Male
c. Unable to tell
34. Age
(Write number or Unable to Tell)
Or b. Unable to tell
29. Race (select one)
a. Asian American
b. African American
c. Caucasian
d. Hispanic
35. Is advice-seeker pregnant?
a. Yes
b. No
c. Unable to tell
d. N/A (male) [If n/a, skip to Q 37]
204
e. Other
f. Unable to tell
30. Marital status (select one)
a. Married
b. Unmarried partner
c. Single
d. Divorced or Separated
e. Other
f. Unable to tell
36. Has advice-seeker been pregnant before?
a. Yes
b. No [If no, skip to Q38]
c. Unable to tell [If UTT, skip to Q38]
31. Employment Status
a. F/T outside home
b. P/T outside home
c. Works in home
d. Student
e. Not working
f. Unable to tell
37. How many other biological kids does advice-
seeker have?
(Write number or Unable to Tell)
Or b. Unable to tell
38. Please record the first 10 friends that the recipient lists on their home page.
(Write down their screen name only.)
i.
ii.
iii.
iv.
v.
vi.
vii.
viii.
ix.
x.
39. Did the Advice-seeker/Recipient post a follow-up message to her original post?
a. No [If no, then skip to Q51- Respondent demographics.]
b. Yes
205
40. Within the first 10 responses received from other members, how many follow-
ups did the Advice-Seeker post? (Write number here):
For each follow-up message posted, first provide a response #, that is number the
message in order that it was added to the discussion. These numbers should be
continuous with the response #s provided to the other responses (i.e. responses from other
people). The purpose is so that we can place this message in chronological order with the
rest of the responses.
Then, please indicate the purpose/general contents of each message:
Follow-up #1:
41. Response # (in order that this message was received):
42. What was the general purpose of the message? (Circle all that apply.)
a. To thank respondents for their advice
b. To clarify something she wrote in her first post
c. To ask for additional support, circle all that apply:
i. Emotional
ii. Informational
iii. Esteem
iv. Network
v. Tangible
c. To defend herself from negative comments
d. Other:
e. Other:
Follow-up #2:
43. Response # (in order that this message was received):
44. What was the general purpose of the message? (Circle all that apply.)
a. To thank respondents for their advice
b. To clarify something she wrote in her first post
c. To ask for additional support, circle all that apply:
i. Emotional
ii. Informational
iii. Esteem
iv. Network
v. Tangible
c. To defend herself from negative comments
d. Other:
e. Other:
Follow-up #3:
45. Response # (in order that this message was received):
206
46. What was the general purpose of the message? (Circle all that apply.)
a. To thank respondents for their advice
b. To clarify something she wrote in her first post
c. To ask for additional support, circle all that apply:
i. Emotional
ii. Informational
iii. Esteem
iv. Network
v. Tangible
c. To defend herself from negative comments
d. Other:
e. Other:
Follow-up #4:
47. Response # (in order that this message was received):
48. What was the general purpose of the message? (Circle all that apply.)
a. To thank respondents for their advice
b. To clarify something she wrote in her first post
c. To ask for additional support, circle all that apply:
i. Emotional
ii. Informational
iii. Esteem
iv. Network
v. Tangible
c. To defend herself from negative comments
d. Other:
e. Other:
Follow-up #5:
49. Response # (in order that this message was received):
50. What was the general purpose of the message? (Circle all that apply.)
a. To thank respondents for their advice
b. To clarify something she wrote in her first post
c. To ask for additional support, circle all that apply:
i. Emotional
ii. Informational
iii. Esteem
iv. Network
v. Tangible
c. To defend herself from negative comments
d. Other:
e. Other:
207
APPE
DIX B
SUPPORT-PROVIDER CODE SHEET
51. Screen
ame: 73. Screen
ame:
52. Date of Response:
53. Response#:
63. Pregnant right now?:
1. Yes
2. No
3. Unable to tell
4. N/A (male)[skip to Q65]
74Date of Response:
75.Response#:
85.Pregnant right now?:
1. Yes
2. No
3. Unable to tell
4. N/A (male)[skip to Q86]
54. Public Profile:
1. Yes
2. No
64. Been pregnant before?
1. Yes
2. No [skip to Q66]
3. Unable to tell -skip to 66
76. Public Profile:
1. Yes
2. No
86.Been pregnant before?
1. Yes
2. No [skip to Q88]
3. Unable to tell -skip to 88
56. Gender:
1. Female
2. Male
3. Unable to tell
65. # of Bio Kids (incl. 0): 78.Gender:
1. Female
2. Male
3. Unable to tell
87. # of Bio Kids (incl. 0):
57. Age:
Or 2. Unable to tell
67. Info Support:
1. Suggestion/advice
2. Referral
3. Situation appraisal
4. Teaching
5. None
79. Age:
Or 2. Unable to tell
89.Info Support:
1.Suggestion/advice
2.Referral
3.Situation appraisal
4.Teaching
5. None
58. Race:
1. Asian American
2. African American
3. Caucasian
4. Hispanic
5. Other
6. Unable to tell
68. Emotional Support:
1. Affection
2. Sympathy
3. Underst /empathy
4. Encouragement
5. Prayer
6. None
80. Race:
1. Asian American
2. African American
3. Caucasian
4. Hispanic
5. Other
6. Unable to tell
90Emotional Support:
1. Affection
2. Sympathy
3. Underst /empathy
4. Encouragement
5. Prayer
6. None
59. Marital Status:
1. Married
2. Unmarried partner
3. Single
4. Divorced/Separated
5. Other
6. Unable to tell
69. Esteem Support:
1.Compliment
2.Validation
3.Relief of blame
4.None
81.Marital Status:
1. Married
2. Unmarried partner
3. Single
4. Divorced/Separate
5. Other
6. Unable to tell
91.Esteem Support:
1.Compliment
2.Validation
3.Relief of blame
4.None
60. Location:
1. US
2. International
3. Unable to tell
70.
etwork Support
1. Access 2. Friendship
3. Physical presence
4. Companions 5. None
82. Location:
1. US
2. International
3. Unable to tell
92.
etwork Support
1. Access 2. Friendship
3. Physical presence
4. Companions 5. None
61. State/Country: 3.utt
1. US State OR:
2. Int. Country:
71. Tangible Support
1. Loan
2. Gift
3.Task
4. None
83. State/Ctry: 3.UTT
1. US State OR:
2. Int. Country:
93. Tangible Support
1. Loan
2. Gift
3.Task
4. None
62. Employment status:
1. F/T outside home
2. P/T outside home
3. Works in home
4. Student
5. Not working
6. Unable to tell
72. Were there negative
comments (e.g. criticism,
sarcasm, disagreement,
etc.)?
1. Yes
2. No
84. Employment stats:
1.F/T outside home
2. P/T outside home
3.Works in home
4.Student
5.Not working
6. Unable to tell
94. Were there negative
comments (e.g. criticism,
sarcasm, disagreement,
etc.)?
1. Yes
2. No
(Circle all that apply.) (Circle all that apply.) (Circle all that apply.)
208
APPE
DIX C
MEMBER SURVEY
Communication Ecology:
1. In the past week, on how many days did you…
[response options range from 0 to 7]
Read a newspaper?
Watch television news or documentary programs?
Watch entertainment television (i.e. not news or documentary programs)?
Listen to news or talk radio?
Use the Internet, other than email and social networking sites?
Visit online social networking sites?
2. On a scale from 1 to 10, where 1 is not at all, and 10 is completely, when
searching for pregnancy-related health information, how much do you trust
each of the following information sources?
Television – News programs and documentaries
Television – Other types of programming
Newspapers
Books about pregnancy/motherhood
Magazines about pregnancy/motherhood
Internet—online “chat rooms” or social networking sites about
motherhood/pregnancy
Internet—informational web pages about pregnancy (but not chat rooms)
Internet—Other
Radio
Friends
Family
Your doctor/obstetrician (your primary doctor who’s taking care of your
pregnancy)
3. On a scale from 1 to 10, where 1 is “not at all important” and 10 is “very
important” how much do you rely on each of the following for pregnancy-
related health information?
Television – News programs and documentaries
Television – Other types of programming
209
Newspapers
Books about pregnancy/motherhood
Magazines about pregnancy/motherhood
Internet—online “chat rooms” or social networking sites about
motherhood/pregnancy
Internet—informational web pages about pregnancy (but not chat rooms)
Internet—Other
Radio
Friends
Family
Your doctor/obstetrician (your primary doctor who’s taking care of your
pregnancy)
Frequency and Intensity of Participation:
4. How many days per week do you typically log into [INSERT WEB SITE’S
NAME HERE]? [0-7 response options]
5. On average, roughly how long do you spend on the site at one time?
15 minutes or less
Between 16 and 30 minutes
Between 31 and 45 minutes
Between 46 minutes and one hour
More than 1 hour
6. Approximately, how long have you been a member of [INSERT WEB SITE’S
NAME]?
3 weeks or less
Between 1 and 3 months
Between 4 and 6 months
Between 6 and 9 months
More than 9 months
I’m not a registered member
7. What are your main reasons for going to this web site (check all that apply)?
To get health information about my pregnancy
To get emotional support
To look for people who live in my area for the possible exchange of
babysitting, baby-related products (e.g. old clothes, toys, etc.), maternity
products or other products or services
To get referrals to someone else (e.g. a health care professional; another web
site; etc.)
To share pregnancy stories with others who are in the same situation
Other:
210
8. On a scale of 1 to 10, where 1 = strongly disagree, and 10 = strongly agree,
please indicate your agreement with the following statements.
This web site has been very useful to me.
I trust the information I read on this web site.
I have followed recommendations I’ve read on this web site.
9. Since you’ve been a member of this web site, have you ever posted a
question/asked for advice?
No (If no, skip to Q13)
Yes
10. In the past month, approximately how many times have you posted a
question / asked for advice on this site? [response options: 1 to 10; More than
10]
11. After you asked for advice / posted a question on this web site, how useful
were the responses you received?
[Responses on 1-10 likert scale from 1= not useful at all to 10 = very useful]
12. After you asked for advice / posted a question on this web site, did you follow
the advice that you were offered?
No
Yes
Some of it, but not all
13. Since you’ve been a member of this web site, have you ever responded to
someone else’s question?
No (If no, skip to Q15)
Yes
14. In the past month, approximately how many times have you responded to
someone else’s comment or question (e.g. by providing advice, etc.)?
[response options: 1 to 10; More than 10]
15. Do you participate regularly (e.g. at least once a week) in any other online
chat room/social network web site focused on pregnancy?
No (If no Skip to Q17)
Yes
211
16. If yes, what are the names of the other web sites [ADD ADDITIONAL
NAMES]?
Cafemom.com Dailystrength.com
Mayasmom.com Momjunction.com
Storknet.com Whattoexpect.com
Other:
Maternal Mental Health:
17. On a scale from 1 to 10, where 1 = “never” and 10 = “all the time,” how often
do you feel the following ways when thinking about your pregnancy?
Very nervous
Calm and peaceful
Downhearted and blue
So down in the dumps that nothing can cheer you up
Happy
Stressed
Worried
Bonding and bridging social capital:
Bonding subscale:
18. Please indicate your agreement with the following statements: (adapted from
Williams, 2006) [Insert scale]
a. There are several people in this online community that I trust to help solve
my problems.
b. There is someone in this online community that I can turn to for advice
about making very important decisions.
c. There is no one in this online community that I feel comfortable talking to
about intimate personal problems (reversed).
d. When I feel lonely there are several people in this online community I can
talk to.
e. I do not know people in this online community well enough to get them to
do anything important. (reversed)
f. The people I interact with in this online community would help me fight
an injustice.
g. If I needed an emergency loan of $100 there is someone in this online
community I can turn to.
212
Bridging subscale:
a. Interacting with people in this online community makes me want to try
new things.
b. Interacting with people in this online community makes me interested in
what people unlike me are thinking.
c. Interacting with people in this online community makes me feel like part
of a larger community.
d. Interacting with people in this online community reminds me that
everyone in the world is connected.
e. I am willing to spend time to support general activities in this online
community.
f. Interacting with people in this online community gives me new people to
talk to.
g. In this online community, I chat with new people all the time.
h. I feel like I am part of this online community.
i. This online community is a good place to be.
j. I am interested in what goes on in this online community.
Social Support/Social Provisions Scale:
19. In answering the following questions, think about your relationships with
the other members of this online pregnancy community. Please indicate to
what extent each statement describes your current relationships with the
members of this community. (1 to 4 scale from strongly disagree to strongly
agree).
There are people I can depend on in this online community to help me if I really
need it.
I feel that I do not have close personal relationships with other people in this
online community. (reverse coded)
There is no one I can turn to in this online community for guidance in times of
stress. (reverse coded)
There are other people in this online community who depend on me for help.
There are other people in this online community who enjoy the same social
activities that I do.
Other people in this online community do not view me as competent. (reverse
coded)
I feel personally responsible for the well-being of another person in this online
community.
In this online community, I feel part of a group of people who share my beliefs
and attitudes.
In this online community, I feel part of a group of people who respect my skills
and abilities.
213
If something went wrong, no one in this online community would come to my
assistance.
I have close relationships with people in this online community that provide me
with a sense of emotional security and well-being.
There is someone in this online community I could talk to about important
decisions in my life.
I have relationships in this online community where my competence and skills are
recognized.
There is no one in the online community who share my interests and concerns.
(reverse coded)
There is no one in the online community who really relies on me for their well-
being. (reverse coded)
There is a trustworthy person I could turn to in this online community for advice
if I were having problems.
I feel a strong emotional bond with at least one other person in this online
community.
There is no one I can depend on in this online community for aid if I really need
it. (reverse coded)
There is no one I feel comfortable talking about problems with in this online
community.
There are people in this online community who admire my talents and abilities.
I lack a feeling of intimacy with another person in this online community.
(reverse coded)
There is no one in this online community who like to do the things I do. (reverse
coded)
There are people in this community I can count on in an emergency.
In this online community no one needs me to care for them. (reverse coded)
Health Outcomes (Knowledge, attitudes, and behaviors):
Knowledge:
20. Have you learned any new health information from this online community?
No
Yes
Not sure/can’t remember
21. Please indicate your agreement with the following statements. (response
options: dichotomous agree/disagree)
a. It’s okay for a pregnant woman to have one alcoholic drink per day.
(decoy)
b. Caffeine consumption should be limited during pregnancy.
c. Infant formula is the healthiest food you can feed a baby. (decoy)
214
d. Cigarette smoking during pregnancy increases the chances of premature
birth, certain birth defects, and infant death.
e. Women should drink extra fluids throughout pregnancy.
f. Folic acid prevents the risk of birth defects of the brain and spine.
g. For a healthy pregnancy, vigorous exercise is recommended (decoy).
h. It’s fine for a pregnant woman to have x-rays taken. (decoy)
i. Folic acid should be taken both before and during pregnancy.
j. Over-the-counter cough and cold remedies should be avoided during
pregnancy.
k. Doctors recommend sleeping on your back when you’re pregnant. (decoy)
Attitudes towards pre-natal health:
22. On a scale from 1 to 10 where 1 = strongly disagree and 10 = strongly agree,
please indicate your agreement with the following statements.
a. I enjoy the physical changes that being pregnant brings.
b. I don’t mind changing my lifestyle for the well being of my baby.
c. I am worried that I will do something wrong while I’m pregnant that will hurt
my baby.
d. Most pregnancies result in the births of healthy babies.
e. Being pregnant is an inconvenience that I can’t wait to be finished with.
(reverse)
f. It seems that there are so many things that I have to be aware of that may
potentially be dangerous to my pregnancy.
g. I look forward to getting pregnant again.
h. I feel so uncomfortable in my pregnant body. (reverse)
i. I am always looking for ways to be healthier during my pregnancy.
j. I don’t mind giving up certain foods or drinks during my pregnancy if it will
reduce my risks for pregnancy complications.
Behaviors:
23. On a scale from 1 to 10, where 1 = “never” and 10 = “every day”, as a result
of something you read on this web site, how often have you done the
following:
a. drank alcohol
b. consumed illegal drugs
c. smoked cigarettes
d. started exercising
e. drank coffee
f. drank soda
g. taken prenatal vitamins
h. eaten fruits and vegetables
215
i. ate sushi
j. seen your doctor/obstetrician regularly
24. As a result of your participation in this online community have you: (yes/no
responses)
a. Changed what you eat?
b. Changed what your drink?
c. Changed your exercise routine?
d. Changed your mind about any medical procedures that you were, at one
time, going to undergo?
e. Felt more worried about your pregnancy?
f. Changed how you interact with your own doctor?
g. Spoken with your doctor about information you read on this web site?
h. Sought additional information online about something you read on this
web site?
i. Spoken to your friends or family about something you read on this web
site?
j. Sought a second opinion from another doctor / health care practitioner?
k. Felt less stress?
l. Felt more confident in how you take care of yourself?
25. Think about the last recommended health behavior that you read about in
this online community. Did you try this health behavior yourself?
Yes [If yes, skip to Q33]
No
26. If no, what prevented you from trying it? (Check all that apply).
I didn’t have the money
I didn’t have the time
It didn’t seem like something I really needed to do
I didn’t feel like I had the skills or abilities to do it
I needed more information
The area I live in doesn’t have the resources I would need to do this particular
behavior
Other:
Demographics and Background Information:
27. Are you currently pregnant?
No [Skip to Q 40]
Yes. If yes,
a. How many weeks pregnant are you?
b. When is your due date? [MONTH/DATE/YEAR]
216
28. Are you under the regular care of an obstetrician/doctor?
Yes
No
29. Do you have any other children?
a. No [Skip to Q 42]
b. Yes
30. How many children do you currently have?
31. At what age did you, or will you, have your first child? [response option will
be open-ended plus one N/A option]
32. Do you have family that lives nearby that helps you (or will help you) with
your child/children and family responsibilities on a regular basis (e.g. at least
once per week)? (Yes/No Response options)
33. In general, would you say that your health is…
Excellent
Very good
Fair
Poor
Very Poor
Unsure
34. How old are you?
35. How would you describe yourself?
African American/Black
Asian American
Caucasian/White
Hispanic/Latina
Other:
36. What is your marital status?
Single
Married
Divorced
Living with partner/significant other
Other:
37. What is your highest level of education achieved?
Some high school or less
Completed high school
217
Some college/trade school
Completed college
Graduate school
38. Do you have a job (e.g. paid employment)?
Yes, I have a full-time job outside the home
Yes, I have a part-time job outside the home
Yes, I have a job that allows me to work from home
I am a full-time student
No, currently I am not working
39. What is your annual household income?
Less than $25,000
$25,000 to $49,999
$50,000 to $74,999
$75,000 to $99,999
$100,000 or more
40. What state do you live in?
[LIST ALL STATES]
Social Network Analysis:
[SHOULD BE ABLE TO SAVE ALL DATA UP TO THIS POINT IN CASE PEOPLE
DROP OUT OF SURVEY DURING THE FOLLOWING SET OF QUESTIONS]
ow, we’d like to find out a little more about how many members are
actively engaged in conversation. Please note that all personally identifying
information will be removed to ensure your anonymity.
41. What is your screen name that you use on [INSERT WEB SITE NAME
HERE]?
42.
ame up to 10 people in this community who provide advice to you most
often (or who you chat with most often). (Please provide their screen names—
note all information will be kept confidential.)
Screen
ame Kind of advice provided (check all that apply)
a. emotional support b) tangible support (define)
c. referral to another person or organization d)
informational support
43.
ame up to 10 people in this community you provide advice to most often.
218
(Please provide their screen names—note all information will be kept confidential.)
Screen
ame Kind of advice provided (check all that apply)
a) emotional support b) tangible support (define)
c. referral to another person or organization d)
informational support
44.
ame up to 10 people on your “friends” list whom you regularly interact
with online? (This could be through messages; “hugs;” “reading their journal;
etc.)
Thank you for your contribution to this research. We appreciate your time and
interest, and wish you luck on your pregnancy!
Abstract (if available)
Abstract
A large number of Americans search the Internet for health information. Moreover, Internet users are no longer solely passive consumers of online health content, but they are also active producers as well. Social media, such as social networking sites, are online communities where patients can pool their knowledge and experience to manage their health care more confidently. However, little is known about the impact of participation on these kinds of sites. This dissertation used three methodologies to explore the health-related impact of these sites. First, a content analysis was conducted with a sample of 572 support-seeking messages and 1,965 support-providing messages from two sites during one month. Members were largely American, Caucasian, and married. They sought information on a range of health topics, but especially issues related to the beginning and end of pregnancy. Further, informational support and emotional support were most in demand. Second, using survey data collected from 114 pregnant members of these sites, this dissertation tested whether participation was associated with health outcomes. The survey also compared the impact that perceptions of social support and social capital had on outcomes such as prenatal knowledge, attitudes, and health behaviors. Results show that members who provided more support tended to be the most influenced by their participation on the site. Further, social support was associated with more attitudinal outcomes than was social capital, suggesting some experiential differences in these constructs. Third, a social network analysis compared social network metrics with survey scales of social support and social capital. The results showed that social network metrics were correlated more with social support, suggesting further differences between these constructs.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
The formation and influence of online health social networks on social support, self-tracking behavior and weight loss outcomes
PDF
Relationship formation and information sharing to promote risky health behavior on social media
PDF
The effect of Twitter social support on health outcomes and its mediators: a randomized controlled trial of a social support intervention on Twitter for patients affected by cancer
PDF
Benefits of social networking in online social support groups
PDF
Contagious: social norms about health in work group networks
PDF
The power of social media narratives in raising mental health awareness for anti-stigma campaigns
PDF
The influence of social networking sites on high school students' social and academic development
PDF
Building personal wellness communities: meaningful play in the everyday life of a network society
PDF
Communication and social support of parents of children treated at Childrens Hospital Los Angeles
PDF
The social groups approach to quitting smoking: An examination of smoking cessation online and offline through the influence of social norms, social identification, social capital and social support
PDF
Mobile self-tracking for health: validating predictors, effects, mediator, moderator, and social influence
PDF
Motivation for participating in virtual religious communities: developing social capital
PDF
School-based Health Centers 2.0: Using social media to increase utilization of UMMA Wellness Center Services
PDF
Social network influences on depressive symptoms among Chinese adolescents
PDF
Social capital and community philanthropy: the impact of social trust and social networks on individual charitable behavior and community foundation development
PDF
Network influences in health initiatives: multimedia games for youth in Peru
PDF
Doing family on Facebook: connecting communicative affordances of mobile media, parental practices, and social capital
PDF
The social meaning of sharing and geocoding: features and social processes in online communities
PDF
A network analysis of online and offline social influence processes in relation to adolescent smoking and alcohol use
PDF
The egocentric social networks of Black adolescents and sources of racial socialization messages
Asset Metadata
Creator
Hether, Heather Jane
(author)
Core Title
Social media and health: social support and social capital on pregnancy-related social networking sites
School
Annenberg School for Communication
Degree
Doctor of Philosophy
Degree Program
Communication
Publication Date
11/05/2009
Defense Date
09/24/2009
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
health,Internet,OAI-PMH Harvest,social capital,social media,social networking,social support
Place Name
USA
(countries)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Murphy, Sheila T. (
committee chair
), McLaughlin, Margaret (
committee member
), Valente, Thomas W. (
committee member
)
Creator Email
heatherb52@hotmail.com,hether@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-m2715
Unique identifier
UC1422409
Identifier
etd-Hether-3339 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-275595 (legacy record id),usctheses-m2715 (legacy record id)
Legacy Identifier
etd-Hether-3339.pdf
Dmrecord
275595
Document Type
Dissertation
Rights
Hether, Heather Jane
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Repository Name
Libraries, University of Southern California
Repository Location
Los Angeles, California
Repository Email
cisadmin@lib.usc.edu
Tags
health
Internet
social capital
social media
social networking
social support