Using a human factors engineering perspective to design and evaluate communication and information technology tools to support depression care and physical activity behavior change among low-inco...

PDF

Download

Open document

Flip pages

Summarize

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content 1

USING A HUMAN FACTORS ENGINEERING PERSPECTIVE TO DESIGN AND
EVALUATE COMMUNICATION AND INFORMATION TECHNOLOGY TOOLS TO
SUPPORT DEPRESSION CARE AND PHYSICAL ACTIVITY BEHAVIOR CHANGE
AMONG LOW-INCOME LATINO PATIENTS WITH DIABETES

A DISSERTATION
SUBMITTED TO THE GRADUATE SCHOOL
AT THE UNIVERSITY OF SOUTHERN CALIFORNIA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
(INDUSTRIAL AND SYSTEMS ENGINEERING)

MAGALY RAMIREZ
DECEMBER 2016
2
Acknowledgements

I would like to express sincere gratitude to the people and institutions that supported me
on my journey to earning a Ph.D. First, I would like to thank my advisor, Dr. Shinyi Wu,
for her expert knowledge and guidance that helped me to identify a topic and
successfully complete my dissertation research. Second, I would like to thank Dr.
Mansour Rahimi, Dr. Najmedin Meshkati, Dr. Donna Spruijt-Metz, and Dr. Todd
Schroeder for serving on my qualifying exam and/or dissertation committee.

My gratitude also goes to the institutions that provided me with financial support: the
National Science Foundation for my Graduate Research Fellowship, the National
Institute of Neurological Disorders and Stroke for the Diversity Supplement via the Los
Angeles Stroke Prevention/Intervention Research Program in Health Disparities, and the
Daniel J. Epstein Institute for a pilot study grant.

Finally, I would like to thank Dave Prince for planting the seed of pursuing a Ph.D. when
I was an undergraduate student at the University of Washington, my parents for their
hard work and sacrifices that enabled me to pursue an education, my husband for his
willingness to move to California and his unrelenting patience and encouragement, and
my baby for motivating me to work with diligence to finish my dissertation before his
arrival.

3
Table of Contents

Chapter 1 Introduction ................................................................................................... 5
References ................................................................................................................... 8
Chapter 2 Automated Remote Monitoring of Depression: Acceptance Among Low-
Income Patients in Diabetes Disease Management .................................................. 10
Abstract ...................................................................................................................... 10
Introduction ............................................................................................................... 12
Methods ...................................................................................................................... 16
Study Design and Participants ................................................................................. 16
Survey-Based Measures of ATA Call Acceptance ................................................... 16
ATA Call Completion Rates ..................................................................................... 18
Statistical Analysis ................................................................................................... 18
Results ....................................................................................................................... 20
Sample Characteristics ............................................................................................ 20
First Analysis: Patient Acceptance of ATA Calls Over Time .................................... 22
Second Analysis: Factors Predicting Patient Acceptance of ATA Calls .................. 23
Discussion ................................................................................................................. 25
Principal Findings .................................................................................................... 25
Limitations ................................................................................................................ 26
Comparison With Prior Work ................................................................................... 28
Conclusions ............................................................................................................. 29
Acknowledgements .................................................................................................. 30
Abbreviations ............................................................................................................ 30
References ................................................................................................................. 31
Appendix .................................................................................................................... 37
Chapter 3 Patient Preferences for Text Messaging Intervention Features to
Support Physical Activity Behavior Change: A Discrete Choice Experiment with
Low-Income Latino Patients with Diabetes ................................................................ 43
Abstract ...................................................................................................................... 43
Introduction ............................................................................................................... 44
Methods ...................................................................................................................... 47
Step 1: Identify Attributes and Levels to Describe TMIs .......................................... 48
Step 2: Construct Choice Sets and Design Survey .................................................. 49
Step 3: Conduct Survey to Measure Preferences .................................................... 51
Step 4: Analyze the Data ......................................................................................... 52
Results ....................................................................................................................... 54
Respondent Characteristics ..................................................................................... 54
Importance of Attributes ........................................................................................... 54
Preferred Levels ...................................................................................................... 55
Influence of Respondent Characteristics ................................................................. 55
Discussion ................................................................................................................. 56
Principal Results and Comparison with Prior Work ................................................. 56
Limitations ................................................................................................................ 58
Conclusions ............................................................................................................. 59
Acknowledgements .................................................................................................. 59
4
References ................................................................................................................. 59
Appendix .................................................................................................................... 64
Chapter 4 Phone Messaging for Physical Activity and Social Support Prompting
Among Low-Income Latino Patients with Diabetes: A Randomized Pilot Study ... 68
Abstract ...................................................................................................................... 68
Introduction ............................................................................................................... 70
Methods ...................................................................................................................... 73
Study Design ............................................................................................................ 73
Recruitment .............................................................................................................. 73
Interventions ............................................................................................................ 75
Study Measures ....................................................................................................... 80
Analysis ................................................................................................................... 83
Results ....................................................................................................................... 84
Process Feasibility ................................................................................................... 84
Participant Characteristics ....................................................................................... 87
Technical Feasibility (PM and PM+FF Only) ........................................................... 88
Perceived Usefulness .............................................................................................. 90
Potential Effectiveness ............................................................................................ 94
Discussion ................................................................................................................. 97
Process Feasibility ................................................................................................... 97
Technical Feasibility ................................................................................................ 98
Perceived Usefulness .............................................................................................. 99
Potential Effectiveness .......................................................................................... 101
Limitations .............................................................................................................. 103
Conclusions ........................................................................................................... 104
Acknowledgements ................................................................................................ 105
Abbreviations .......................................................................................................... 105
References ............................................................................................................... 106
Appendix .................................................................................................................. 111
Chapter 5 Conclusion ................................................................................................. 114
References ............................................................................................................... 118

5
Chapter 1 Introduction

Communication and information technology (CIT) tools are increasingly being developed
for use by patients to help them manage their health and health care. The Agency for
Healthcare Research and Quality defines health CIT tools as any electronic tool,
technology, or system that: (1) is primarily designed to interact with health information
users or consumers; (2) both uses and provides health information; and (3) is used for
the purpose of helping consumers manage their health or health care [1]. While it has
been demonstrated that health CIT tools can improve health outcomes [2], a Cochrane
review found that those designed to support adults with diabetes have resulted in little to
no effect on biological, cognitive, behavioral, or emotional outcomes [3]. In addition, it is
common to observe discontinued use of these tools by adults with diabetes [4].

The current health literature approach to design and evaluation of health CIT tools for
individuals with diabetes is a major problem that may be contributing to the observed
lack of effectiveness and sustained user engagement. Published reports of studies
involving health CIT interventions seldom discuss design approaches; it appears that
most of the tools are designed in an ad hoc way [2], [3]. Of the limited reports that do
discuss design approaches, they are driven almost exclusively by a causal modeling
approach wherein a theoretical model that causally links behavioral determinants,
behavior, physiological and biochemical variables, and health outcomes is used to
specify techniques that will change the behavioral determinants [5]. The techniques are
then translated into system specifications. In non-technology interventions, this approach
tends to yield effective interventions [6]–[8]. When applied to technology-based
interventions, however, the approach may not sufficient for ensuring user acceptance
6
and sustained use of the tools [9]. Moreover, evaluations of health CIT tools tend to
focus on effects on health outcomes [3], largely ignoring important aspects of the
systems that may affect performance. Because the need, interest, and demand for
health CIT tools is expected to continue to grow in the future [10], rigorous design and
evaluation approaches must be utilized in order to make health CIT tools that will have a
significant impact on patients’ health and health care.

This dissertation used a human factors perspective to design and evaluate health CIT
tools that support adults with diabetes. Human factors is an important aspect of research
at the intersection of Industrial and Systems Engineering and health care because of the
central role of humans in the health care system. It is humans – clinicians, patients, and
families – who are called on to perform health care work in order to improve health and
well-being. Human factors is a science concerned with the understanding of interactions
among humans and other elements of a system, and the discipline that takes this
knowledge and uses it to specify, design, and test systems that optimize effectiveness,
productivity, safety, and satisfaction [11], [12]. Whereas current health CIT design
approaches, at best, center the design process on behavioral theories, the key principle
in the field of human factors is to center the design process on the user. In addition,
human factors evaluations not only measure changes in performance (for example,
health outcomes), but also investigate aspects of the systems that may affect
performance such as acceptability, usability, performance of the user or human-machine
system, and safety [13]. A human factors approach to design and evaluation thus
involves studying users in order to understand their needs and preferences,
consolidating this understanding to develop system specifications, and requesting user
response to design solutions. Such an approach to design and evaluation of health CIT
7
may increase the likelihood that patients’ interaction with the tools is one that enhances
performance, increases user satisfaction, and increases safety.

The target population in this dissertation is urban, low-income, predominantly Latino
adults with diabetes who are part of the safety-net system of health care. Despite the
promise of health CIT to effectively support health and health care, little is known about
the impact of these tools on highly vulnerable populations. In the United States, Latinos
have a 66 percent higher risk of developing diabetes compared to non-Latino whites
[14]. They are also 1.5 times more likely to die from the disease. The burden of illness is
even greater among urban, low-income, Latino patients who receive care in safety-net
health systems [15]. Even though this population stands to benefit the most from
additional support, they are generally understudied in the health CIT literature.

The first part of this dissertation involved evaluating the design of an automated
telephonic assessment (ATA) system that periodically called patients to assess
depression symptoms, monitor treatment adherence, prompt self-care behaviors, and
inquire about patients’ needs for provider contact [16]. The ATA system was tested in the
Diabetes-Depression Care-management Adoption Trial (DCAT) [17]. The manuscript in
Chapter 2 presents secondary DCAT data analyses performed to answer the following
research questions: What is patients’ acceptance of ATA calls over time? What factors
predict patients’ long-term acceptance of ATA calls? Patient ATA call acceptance was
measured in terms of ATA call completion rates, patients’ willingness to use ATA calls,
preferred mode of reach, perceived ease of use, usefulness, non-intrusiveness, and
privacy/security.

8
While the ATA system called patients periodically, the second part of this dissertation
involved designing and evaluating a user-centered health CIT tool that provided more
frequent ongoing support for a key aspect of diabetes self-care: engagement in physical
activity (PA). The communication modality used was short text/voice messaging
(ST/VM). First, patient needs for PA behavior change support were investigated,
potential system configurations derived, and patient preferences elicited using a discrete
choice experiment. The manuscript in Chapter 3 presents results that answer the
research question: What configuration of ST/VM system features for PA behavior
change support do patients most prefer? Patient preferences were then translated into
system specifications and a pilot test of the resulting ST/VM system conducted to gauge
patients’ response. The manuscript in Chapter 4 presents results that answer the
research question: What is the feasibility, acceptance, and potential effectiveness of a
ST/VM system to support PA behavior change? Feasibility was assessed in terms of
process and technology feasibility; acceptance in terms of patients’ perceived usefulness
of the ST/VM system; and potential effectiveness in terms of daily steps, social support,
body mass index, and self-efficacy.

The major findings and contributions of the three manuscripts that comprise this
dissertation are summarized in Chapter 5.

References

[1] K. Pal, S. Eastwood, S. Michie, A. Farmer, M. Barnard, R. Peacock, B. Wood, J.
Inniss, and E. Murray, “Computer-based diabetes self-management interventions
for adults with type 2 diabetes mellitus (review),” Cochrane Libr., no. 3, 2013.

9
[2] J. Piette, “Interactive behavior change technology to support diabetes self-
management: where do we stand?,” Diabetes Care, vol. 30, no. 10, pp. 2425–32,
Oct. 2007.
[3] H. B. Jimison, M. Pavel, A. Parker, and K. Mainello, “The Role of Human
Computer Interaction in Consumer Health Applications: Current State, Challenges
and the Future,” pp. 163–187, 2015.
[4] F. D. Davis, “User acceptance of information technology: system characteristics,
user perceptions and behavioral impacts,” Int. J. Man. Mach. Stud., vol. 38, no. 3,
pp. 475–487, 1993.

10
Chapter 2 Automated Remote Monitoring of Depression:
Acceptance Among Low-Income Patients in Diabetes Disease
Management

Abstract

Background: Remote patient monitoring is increasingly integrated into health care
delivery to expand access and increase effectiveness. Automation can add efficiency to
remote monitoring, but patient acceptance of automated tools is critical for success.
From 2010 to 2013, the Diabetes-Depression Care-management Adoption Trial (DCAT)–
a quasi-experimental comparative effectiveness research trial aimed at accelerating the
adoption of collaborative depression care in a safety-net health care system–tested a
fully automated telephonic assessment (ATA) depression monitoring system serving
low-income patients with diabetes.
Objective: The aim of this study was to determine patient acceptance of ATA calls over
time, and to identify factors predicting long-term patient acceptance of ATA calls.
Methods: We conducted two analyses using data from the DCAT technology-facilitated
care arm, in which for 12 months the ATA system periodically assessed depression
symptoms, monitored treatment adherence, prompted self-care behaviors, and inquired
about patients’ needs for provider contact. Patients received assessments at 6, 12, and
18 months using Likert-scale measures of willingness to use ATA calls, preferred mode
of reach, perceived ease of use, usefulness, non-intrusiveness, privacy/security, and
long-term usefulness. For the first analysis (patient acceptance over time), we computed
descriptive statistics of these measures. In the second analysis (predictive factors), we
collapsed patients into two groups: those reporting “high” versus “low” willingness to use
11
ATA calls. To compare them, we used independent t tests for continuous variables and
Pearson chi-square tests for categorical variables. Next, we jointly entered independent
factors found to be significantly associated with 18-month willingness to use ATA calls at
the univariate level into a logistic regression model with backward selection to identify
predictive factors. We performed a final logistic regression model with the identified
significant predictive factors and reported the odds ratio estimates and 95% confidence
intervals.
Results: At 6 and 12 months, respectively, 89.6% (69/77) and 63.7% (49/77) of patients
“agreed” or “strongly agreed” that they would be willing to use ATA calls in the future. At
18 months, 51.0% (64/125) of patients perceived ATA calls as useful and 59.7% (46/77)
were willing to use the technology. Moreover, in the first 6 months, most patients
reported that ATA calls felt private/secure (75.9%, 82/108) and were easy to use (86.2%,
94/109), useful (65.1%, 71/109), and nonintrusive (87.2%, 95/109). Perceived
usefulness, however, decreased to 54.1% (59/109) in the second 6 months of the trial.
Factors predicting willingness to use ATA calls at the 18-month follow-up were perceived
privacy/security and long-term perceived usefulness of ATA calls. No patient
characteristics were significant predictors of long-term acceptance.
Conclusions: In the short term, patients are generally accepting of ATA calls for
depression monitoring, with ATA call design and the care management intervention
being primary factors influencing patient acceptance. Acceptance over the long term
requires that the system be perceived as private/secure, and that it be constantly useful
for patients’ needs of awareness of feelings, self-care reminders, and connectivity with
health care providers.
Trial Registration: ClinicalTrials.gov NCT01781013
12
Introduction

In late 2014, the Centers for Medicare and Medicaid Services (CMS) issued new rules
that expanded provider reimbursements beginning in 2015 for remote monitoring of
Medicare beneficiaries [1]. Telemedicine and telehealth–or, what Kvedar and colleagues
refer to collectively as “connected health”–capitalize on advances in health information
technology (HIT) to remotely provide health care services, information, health education,
and self-management support [2]. A number of studies have demonstrated the potential
of these technologies to increase access and quality of care while decreasing health
care costs [3-8].

Given the mounting evidence for the clinical and cost effectiveness of connected health,
the CMS ruling is likely to boost interest in its adoption. Attempts to improve patient care
with connected health, however, will be futile unless patients accept these technologies.
Prior studies suggest that individuals who do not accept technologies simply will not use
them [9,10]. Indeed, so critical is user acceptance that it has been regarded as “the
pivotal factor in determining the success or failure of an information system” [11]. In an
editorial review of connected health technologies to support behavior changes for self-
management, Piette [12] remarks that patients’ discontinued use, which results from a
lack of acceptance, has largely hindered large-scale implementation. Therefore, it is
clear that patient acceptance has important implications for the broader domain of
connected health, since patients who do not accept (and thus do not use) these
technologies will not realize the full benefits of them, resulting in a loss for both patients
and payers.

13
This study investigates patient acceptance of an automated telecommunications system
designed to facilitate depression care management of low-income patients with diabetes
in a safety-net care system [13,14]. There is evidence of significant disparities in receipt
of depression treatment in low-income, uninsured, and minority populations. These
groups are less likely to receive depression care [15-19], show greater treatment
discontinuation [20], and experience higher rates of clinically significant depression.
Patient barriers to depression care influence detection and treatment processes. For
example, minority patients are less likely to voluntarily report depressive symptoms, may
view depression as a moral weakness or character flaw instead of an illness, may be
more likely to ascribe symptoms of depression to a physical illness [21,22], and may
refuse or discontinue treatment due to stigma [23]. Non-adherence to depression
treatment in minority groups with diabetes is common, due in part to side effects of
diabetes medications [24-26]. Further exacerbating the challenges are cost and complex
patient-provider interactions inherent in caring for patients with comorbid chronic
illnesses. For instance, prioritizing among competing demands may negatively affect
initiation and long-term follow-up of depression management in primary care [27-34].

To address these issues in order to accelerate the adoption of evidence-based
depression care [24,35], we designed an advanced automated telephonic assessment
(ATA) system. It had the capability to inquire –via periodic telephone calls to patients–
about important aspects of depression care using a combination of the following six
modules: monitoring for depression, assessing pain, assessing adherence to
antidepressant medications, assessing psychotherapy practice, prompting depression
self-care activities, and allowing patients to request contact from a clinician [14]. The
14
ATA system was fully integrated into an existing disease management registry (DMR),
which allowed it to automatically select these modules and the frequency of calls
depending on individual patient clinical data: results from previous ATA calls or clinical
assessments (depressed patients were called monthly, nondepressed patients
quarterly), whether patients had an active antidepressant medication prescription, and
whether patients were receiving psychotherapy. It also allowed patients to indicate their
preferences for language (English or Spanish), call days and times, and receiving human
calls instead of machine calls. If a call was not answered, the ATA system attempted
again three times per day for seven days (morning, afternoon, and evening). As a whole,
the design allowed the ATA system to individually customize calls to focus on patients’
specific needs and preferences rather than having patients adapt to standard
comprehensive assessments–in essence, illustrating the philosophy of patient-centered
care.

Moreover, the ATA system facilitated timely, proactive follow-up by clinicians and staff.
Data captured on the ATA calls were automatically assessed and the results sent to the
DMR for clinician and staff review [14]. Notifications, tasks, and alerts were triggered in
response to specific issues identified from the ATA calls: patient requests for contact,
high depression scores, non-adherence to antidepressant medications, or suicidal
ideation.

The ATA system was tested in the Diabetes-Depression Care-management Adoption
Trial (DCAT) [13]. DCAT was a 12-month, quasi-experimental comparative effectiveness
research trial conducted in collaboration with the Los Angeles County Department of
15
Health Services (LACDHS) with the aim of comparing different approaches for
accelerating the adoption of collaborative team depression care in routine safety-net
primary care practice. The study was conducted in its ambulatory care clinics serving
low-income, racially/ethnically diverse (but primarily Latino) patients. It tested three
depression care delivery models: usual care (UC), supported care (SC), and technology-
facilitated care (TC). UC represented the status quo, whereby primary care providers
(PCPs) and their staff initiate the translation and adoption of depression care evidence.
Both SC and TC included care teams of the LACDHS disease management program
(DMP) for the first six months of the trial to support diabetes care as well as depression
care using evidence-based protocols [36]. After six months, patients returned to their
PCPs for care. The difference between SC and TC was that the latter utilized the ATA
system for 12 months to facilitate automated depression screening and monitoring, and
timely follow-up by clinicians and staff. The provider notifications, tasks, and alerts
generated by the ATA system were sent to DMP teams during the first six months of
DCAT and to PCPs and their staff during the second six months.

If such automated remote screening and monitoring of depression–and more broadly,
connected health–is to be integrated into mainstream health care delivery to help reach
the important policy goal of expanding access to high-quality, effective, and efficient
care, an understanding of patient technology acceptance is urgently needed. Studies on
remote assessment and monitoring via connected health, however, continue to overlook
this important research area [35,37]. Those that do touch upon elements of patient
acceptance tend to be cross-sectional and operationalize the construct using measures
16
of patient satisfaction with care, which in itself reveals little about technology acceptance
or how to design the system to improve patient acceptance.

The present study echoes the information technology literature [38-40] by measuring
technology acceptance as patients’ willingness to use ATA calls as part of their
depression care. Moreover, this study is longitudinal, which allows for an understanding
of how patient acceptance may change over time. Finally, to inform future design
choices for automated remote depression monitoring technology, the evaluation includes
several system characteristics that may explain why patients accept or reject the
technology. Thus, in sum, the study was undertaken to (1) determine patient acceptance
of ATA calls for remote depression screening and monitoring over time, and (2) identify
what factors predict long-term patient acceptance of ATA calls.

Methods
Study Design and Participants

To answer the research questions, we analyzed data collected from patients in the TC
arm of DCAT. English-Spanish bilingual interviewers administered assessments of
technology acceptance at 6, 12, and 18 months. Thus, patients received two
assessments during the study and one assessment 6 months after the study had ended.
Survey-Based Measures of ATA Call Acceptance

DCAT defined technology acceptance as patients’ reported willingness to use ATA calls
in the future as part of their depression care. The measurement was administered at 6,
12, and 18 months. DCAT also assessed additional measures of ATA call design
17
characteristics: perceived ease of use (7 items), perceived usefulness (6 items),
perceived nonintrusiveness (3 items), and perceived privacy/security (1 item). DCAT
administered these assessments at 6 and 12 months. Moreover, patients’ preference for
mode of reach (1 item) was assessed at 6, 12, and 18 months. Finally, at 18 months,
patients were asked about their long-term perceived usefulness of ATA calls (3 items).
All measures were assessed on a 5-point Likert scale. Table 1 provides the exact
wording used in the DCAT assessments.

Table 1 Measures of patient ATA call acceptance
Domain of
measurement
Items Administration
Willingness to
use ATA calls

a

To what extent do you agree or disagree with the following
statement?
6, 12, and 18
months
You would not mind receiving automated calls as part of your
depression care in the future.
Perceived
ease of use
b,
c

“How often would you say…” 6 and 12
months the language used by Amy
c
in the calls was easy for you to
understand?
Amy’s voice on the call was loud enough to hear without
straining?
Amy was speaking too fast on the automated call?
you were clear on how to respond to Amy’s questions?
you had difficulty answering the questions when asked to press
buttons on your phone?
giving answers to a real person would have been easier than
giving answers to the automated operator Amy?
Amy had difficulty understanding you when you responded
verbally?
Perceived
usefulness
b, c

“How often would you say…” 6 and 12
months the call made you feel confident that your nurse or social worker
knew how you were doing?
the calls made you feel like your nurse of social worker was
more accessible?
the calls by Amy were just as effectiveness in reporting your
feelings as an in-person visit with your care provider?
the antidepressant medication questions asked by Amy
reminded you to take your medications?
the problem-solving skills questions asked by Amy reminded you
to use these skills?
the calls reminded you to do things like a physical activity or a
fun activity?
Perceived
nonintrusiven
“How often would you say…” 6 and 12
months you enjoyed receiving the calls?
18
ess
b
you felt the calls were a bother?
the length of the calls seemed about right?
Perceived
privacy/
security
a

To what extent do you agree or disagree with the following
statement?
6 and 12
months
You feel automated calls are private and/or secure.
Preferred
mode of
reach
To what extent do you agree or disagree with the following
statement?
6, 12, and 18
months
Instead of receiving automated calls, you would prefer to call the
automated service at your convenience.
Long-term
perceived
usefulness
a

To what extent do you agree or disagree with the following
statements?
18 months
The automated calls helped you be more aware of how you are
feeling.
The automated calls reminded you to take care of your health,
such as doing exercise.
The automated calls helped you stay better connected with your
doctors, nurses or social worker.
a
Patients responded using a 5-point Likert scale of agreement (1=strongly disagree, 2=disagree,
3=neutral, 4=agree, and 5=strongly agree)
b
Patients responded using a 5-point Likert scale of frequency (1=never, 2=rarely, 3=about half the
time, 4=usually, and 5=always)
c
“Amy” was the persona of the ATA calls

ATA Call Completion Rates
We assessed the rate of completed ATA calls for three time periods: 0 to 6 months, 7 to
12 months, and 0 to 12 months. An ATA call was defined as complete if it reached the
patient and recorded answers to the depression assessment questions: PHQ-2 or PHQ-
9, whichever was asked.
Statistical Analysis

We conducted two analyses: one to determine patient acceptance of ATA calls for
remote depression screening and monitoring over time, and the other to identify what
factors predict long-term patient acceptance of ATA calls. Sample characteristics and
sample sizes for each analysis are shown in the Results section (Table 2).

For the first analysis (patient acceptance over time), we included the DCAT TC arm
patients who provided responses for a given survey-based measure at each of the
19
measurement periods. By excluding patients who did not meet this criterion, we were
able to estimate changes more accurately for each measure over time. We computed
descriptive statistics of all measures. For those measures consisting of multiple items,
we computed the average points across items and rounded the average to the nearest
integer. Furthermore, we conducted a paired t test to determine if there was a significant
difference between the ATA call completion rates from 0 to 6 months and from 6 to 12
months. We also used Spearman rank correlation to test the association between the
ATA call completion rate of months 0 to 12 and the survey-based measures of ATA call
acceptance.

In the second analysis (predictive factors), we used a different criterion to select patients
from among the pool of TC arm patients: patients who responded to the questions of
willingness to use ATA calls at 18 months and at least once at 6 or 12 months or both. If
patients answered the question at both 6 and 12 months, we computed the average for
use in the analysis. The 125 patients in this sample were collapsed into two groups:
those reporting “high” willingness to use ATA calls at 18 months (Likert scale response
of 4 or higher) and those reporting “low” willingness to use ATA calls at 18 months (all
other response categories). We compared the descriptive statistics for the two groups:
patient sociodemographic characteristics, health conditions, health care utilization, and
ATA call completion rate. We also compared their responses for perceived ease-of-use,
perceived usefulness, perceived non-intrusiveness, perceived privacy/security,
preference of ATA call mode, and long-term perceived usefulness. If patients completed
assessments of these measures at both 6 and 12 months, we computed the average of
the two for use in the analysis. To compare the two groups of patients, we used
20
independent t tests for continuous variables and Pearson chi-square tests for categorical
variables. Next, we jointly entered independent factors found to be significantly
associated with 18-month willingness to use ATA calls at the univariate level into a
logistic regression model with backward selection to identify predictive factors. Then, we
performed a final logistic regression model with the identified significant predictive
factors and reported the odds ratio estimates and 95% confidence intervals. All analyses
were conducted at 0.05 significance level (2-tailed) using IBM SPSS software, version
22.0.

Results
Sample Characteristics

Table 2 provides the characteristics of patients in the two samples used in the two
analyses. The majority of patients were female, Latino, and preferred Spanish as their
primary language. The characteristics of the two samples were not significantly different
from one another. A comparison of these samples with the rest of the patients in DCAT
TC excluded from the analyses did reveal significant differences in characteristics (see
Tables A-1 and A-2 in Appendix). Compared to the rest of DCAT TC, the two samples
had a greater proportion of Latinos, reported a higher willingness to use ATA calls at 6
and 12 months, and had a higher ATA call completion rate. The sample for the second
analysis also had lower blood sugar values, better diabetes self-care, and reported
higher perceived ease-of-use and perceived non-intrusiveness at 6 and 12 months
compared to the rest of patients in the TC arm of DCAT.

21
Table 2 Patient characteristics for samples in the two analyses (no statistically significant
difference between the two samples)
Characteristic
Sample for first
analysis
Sample for second
analysis
N Statistics
a
N Statistics
a

Female 109 72 (66.1%) 125 80 (64.0%)
Age 109 51.94 (9.01) 125 51.31 (8.81)
Latino 109 105 (96.3%) 125 116 (92.8%)
Spanish as preferred language 109 93 (85.0%) 125 104 (83.2%)
Married 109 49 (45.0%) 125 55 (44.0%)
PHQ-9 (range 0–27, higher=more severe
depression)
b, c

109 5.73 (4.93) 125 5.65 (4.60)
Total number of socioeconomic stressors
c
109 2.28 (1.56) 125 2.37 (1.46)
SCL-20, mean score
c, d
109 0.54 (0.53) 125 0.51 (0.48)
SF-12 mental (general population=50,
higher=better)
c, e

109 50.54 (9.15) 125 51.08 (9.03)
Time with diabetes in years 107 10.15 (7.42) 124 9.98 (7.05)
On insulin treatment
c
109 82 (75.2%) 125 89 (71.2%)
BMI
c, f
109 32.93 (6.55) 125 32.75 (6.16)
A1C value
c, g
108 8.87 (1.39) 124 8.72 (1.39)
Low-density lipoprotein cholesterol
c
108 167.08
(36.20)
124 168.44
(36.60)
Whitty-9 diabetes symptoms (range 1–5,
1=none to 5=every day)
c

109 1.64 (0.54) 125 1.62 (0.49)
Number of diabetes complications
c
109 1.26 (0.89) 125 1.22 (0.79)
Toolbert diabetes self-care in the past 7
days (range 0–7)
c

109 4.63 (0.98) 125 4.65 (1.01)
Diabetes emotional burden (range 1–5,
1=not a problem to 5=very burdensome)
c

109 2.53 (1.35) 125 2.48 (1.37)
Diabetes regime distress (range 1–5, 1=not
a problem to 5=very burdensome)
c

109 2.19 (1.14) 125 2.13 (1.17)
Self-rated health (range 1–5, 1=poor to
5=excellent)
c

109 2.29 (0.60) 125 2.34 (0.60)
Chronic pain
c
109 17 (15.6%) 125 24 (19.2%)
SF-12 physical (general population=50,
higher=better health)
c, e

109 43.18 (9.62) 125 43.17 (9.49)
Sheehan disability scale (range 0–10,
0=none to 10=extremely)
c

109 2.21 (2.34) 125 2.14 (2.26)
Number of ICD-9 diagnosis
c,h
108 8.60 (4.50) 124 8.46 (4.46)
Number of clinic visits
c
107 10.44 (5.61) 124 10.56 (5.64)
Number of emergency room visits
c
41 1.33 (0.61) 44 1.33 (0.60)
Number of hospitalizations
c
15 1.47 (0.83) 18 1.39 (0.78)
Willingness to use
c
109 4.02 (0.93) 125 4.00 (1.08)
Perceived ease-of-use
c
109 4.05 (0.56) 125 4.12 (0.50)
Perceived usefulness
c
109 3.63 (0.89) 125 3.69 (0.90)
Perceived nonintrusiveness
c
109 4.20 (0.87) 125 4.29 (0.84)
Perceived privacy/security
c
109 4.10 (1.11) 125 4.17 (1.08)
22
Preference of ATA call mode
c
109 3.82 (1.06) 125 3.58 (1.32)
Long-term perceived usefulness 76 3.71 (0.92) 125 3.74 (0.99)
ATA call completion rate
c
108 0.70 (0.26) 123 0.74 (0.24)
a
Values are numbers (column percentages) for categorical variables and mean (SD) for
continuous variables;
b
Patient Health Questionnaire, 9 items;
c
Assessment at 6 or 12
months. If both were available, then the average was taken;
d
Symptoms CheckList, 20
items;
e
Short-Form Health Survey, 12 items;
f
Body mass index;
g
Glycated hemoglobin
test;
h
International Classification of Diseases, 9th Revision

First Analysis: Patient Acceptance of ATA Calls Over Time

Figure 1 illustrates patient acceptance of ATA calls over time. In the first 6 months of the
trial, 90% (69/77) of patients reported a high willingness to use ATA calls. At 12 and 18
months, however, the proportion of patients reporting a high willingness to use ATA calls
decreased to 64% (49/77) and 60% (46/77), respectively. After 6 months in the trial, 83%
(62/75) of patients agreed or strongly agreed that they would prefer to receive automated
calls rather than calling the ATA system at their convenience. The proportion of patients
reporting this decreased to 40% (30/75) and 29% (22/75) at 12 and 18 months,
respectively. Throughout the trial, most patients agreed or strongly agreed that ATA calls
felt private/secure (82/108 at 6 months, 89/108 at 12 months). At 6 months, 86.2%
(94/109) of patients reported that ATA calls were usually or always easy to use. This
number decreased to 78.0% (85/109) at 12 months. The proportion of patients reporting
that ATA calls were usually or always useful decreased from 65.1% (71/109) at 6 months
to 54.1% (59/109) at 12 months. At the 18-month follow-up, 51.0% (64/125) of patients
agreed or strongly agreed that the ATA calls were useful. At 6 months, most patients
87% (95/109) perceived that ATA calls were usually or always nonintrusive. More
patients perceived the calls to be intrusive after 12 months in the trial as is evident from
a decrease in the proportion of patients who reported otherwise 76% (83/109).

23

Figure 1 Patient acceptance of ATA calls over time

The ATA call completion rate was 72.6% and 67.8% at 6 and 12 months, respectively–
the difference between the two was not statistically significant (P=.10). In investigating
the associations between the ATA call completion rate of months 0 to 12 (70.2%) and
the various survey-based acceptance measures, only two measures were statistically
significant: (1) perceived ease-of-use (Spearman correlation coefficient=0.25, P=.008)
and (2) perceived non-intrusiveness (Spearman correlation coefficient=0.27, P=.004).
Second Analysis: Factors Predicting Patient Acceptance of ATA Calls

We compared patients who reported, at 18 months, a high willingness to use ATA with
patients reporting low willingness to use ATA calls to determine how the two groups
differed in terms of the various sociodemographic characteristics, health conditions,
24
health care utilization, and ATA-related measures listed in Table 2. Table 3 provides
results for characteristics where there was a statistically significant difference between
the two groups. See Table A-3 in Appendix for full results.

Table 3 Characteristics of patients reporting high versus low willingness to use ATA calls
at 18 months
Characteristic
High willingness to
use ATA calls at 18
months
Low willingness to
use ATA calls at 18
months
P
b

N Statistics
a
N Statistics
a

Toolbert diabetes self-care in
the past 7 days (range 0–7)
c

74 4.81 (0.95) 51 4.43 (1.05) .03
Willingness to use
c
74 4.17 (1.00) 51 3.75 (1.16) .04
Perceived usefulness
c
74 3.84 (0.82) 51 3.49 (0.97) .03
Perceived nonintrusiveness
c
74 4.42 (0.65) 51 4.09 (1.03) .05
Perceived privacy/security
c
74 4.42 (0.91) 51 3.81 (1.22) .003
Long-term perceived
usefulness
74 4.07 (0.91) 51 3.25 (0.91) <.001
a
Values are numbers (column percentages) for categorical variables and mean (SD) for
continuous variables;
b
Two-sample t test;
c
Patients’ response at 6 or 12 months. If
patients provided responses at 6 and 12 months, then the average of these was used.

When we compared patients who reported a high versus a low willingness to use ATA
calls at 18 months, we found six factors to be significantly associated with patients’
reported willingness to use ATA calls. Patients with a high willingness to use ATA calls
at 18 months (1) had better diabetes self-care (P=.03) and (2) reported a higher
willingness to use ATA calls while in the study (P=.04); they also reported (3) higher
perceived usefulness (P=.03), (4) non-intrusiveness (P=.05), and (5) privacy/security
(P=.003) while in the study. Moreover, patients who reported a high willingness to use
ATA calls at 18 months also reported (6) higher long-term perceived usefulness
(P<.001). We jointly entered the six factors into a logistic regression model with
backward selection to identify predictive factors. The results revealed that two factors
jointly predicted willingness to use ATA calls at 18 months: perceived privacy/security
25
(odds ratio OR=1.59, P=.017, 95% CI [1.09, 2.33]) and long-term perceived usefulness
(OR=2.77, P<.001, 95% CI [1.65, 4.63]).

Discussion
Principal Findings

The promises of connected health to efficiently improve access and quality of care [2],
rest upon the assumption that patients will readily accept the technologies. Our study on
safety-net patient acceptance of automated depression screening and monitoring using
ATA calls has important findings suggesting that assumption may be questionable. In
the first 6 months of the trial, most patients were accepting of ATA calls and perceived
the calls to be private/secure, easy to use, useful, and nonintrusive. Over time, however,
patients’ acceptance and their positive perception of ATA call characteristics decreased–
although call completion rates remained high and steady. One explanation may be that
since ATA call results and prompts for follow-up were sent to DMP care teams during the
first 6 months of the trial and to PCPs thereafter, timely follow-up by the latter might have
been challenging due to their busy practice loads. Thus, although patients continued to
complete ATA calls in the second half of the trial, their PCPs may not have responded to
their needs in a timely manner thereby leading them to doubt the value of ATA calls.
Furthermore, patients’ acceptance and their perception of ATA call characteristics may
also reflect an improvement in their depressive symptoms over time. That is, patients
with improved depressive symptoms–due, possibly, to the intervention itself–may no
longer perceive the benefits of the ATA calls. We investigated this hypothesis and found
that there was no statistically significant difference in the percentage of patients reporting
high usefulness and high willingness to use ATA calls among those with improved
26
symptoms, no change in symptoms, or worse symptoms. It may be, however, that our
sample size was not large enough to detect these differences.

Another important finding in our study was the identification of two factors that
significantly predicted patients’ long-term acceptance of ATA calls: the perception that
ATA calls are private/secure and the long-term perceived usefulness of ATA calls. These
two factors could be potentially modified to improve patients’ willingness to use ATA calls
as part of their depression care.
Limitations

This study has limitations worth noting. First, we used two different samples for the
analyses. For the sample used to determine patient acceptance of ATA calls over time,
we included only those patients in the DCAT TC arm who responded to ATA-related
measures at each of the corresponding measurement periods. For the sample used to
identify factors that predict long-term patient acceptance of ATA calls, we included only
those patients in the TC arm who answered the question on willingness to use ATA calls
at 18 months and at least once at 6 or 12 months. We chose to accommodate two
sample sizes for our study in order to maximize the sample sizes for both analyses,
although this may have introduced additional bias.

Second, although the two different samples for the analyses were not significantly
different from each other, they were both somewhat different from the rest of TC patients
who were excluded from the analyses because they did not answer any of the ATA-
related questions. Samples used in the analyses reported a slightly higher willingness to
use ATA calls at 6 and 12 months than TC patients excluded from the analyses.
27
However, it is not likely that this limitation affected our findings because only a small
percentage (about 10%) of patients excluded from the analyses refused to engage with
ATA calls. Nearly 90% of them reported that they could not answer the ATA-related
questions because they did not receive or did not remember receiving ATA calls, or they
received calls but did not answer because they were unavailable.

Furthermore, the small sample size of 125 patients reporting on willingness to use ATA
calls limits the robustness of our findings of factors predicting long-term patient
acceptance of ATA calls. Future studies should validate the generalizability of our
findings.

A final limitation is that in the analysis of factors predicting long-term acceptance, we
defined acceptance as patients’ self-reported willingness to use ATA calls at 18 months
instead of using a more objective measure such as ATA call completion rate. This may
seem to be a better indicator of patient acceptance, but since we were interested in
learning about patients’ long-term acceptance, we did not have the ATA call completion
rate at 18 months(the intervention was only for 12 months). Moreover, in our qualitative
study of DCAT TC patients with incomplete ATA calls, we discovered that patients were
actually willing to take the ATA calls, but were unable to do so mainly because of
nonintervention related reasons, including not being available, the ATA system having
the wrong phone number, or experiencing connection issues [41]. For this reason, we
assumed that if patients did not complete ATA calls, it was not due to a lack of
acceptance. Therefore, given the DCAT study design and the practical reasons for
patients not answering ATA calls, we chose to follow the Patient Technology Acceptance
28
Model (PTAM) [39] and define acceptance as self-reported willingness (ie, intention) to
use the technology.
Comparison With Prior Work

The finding that patients are generally accepting of ATA calls, albeit in the short term, is
a promising start to our understanding of patients’ perception of such technologies.
Because automated depression screening and monitoring technology is emerging, little
is known about patients’ acceptance of it. Related studies of connected health
technologies [42], including those focused on depression care [37,43-54], uncritically
regard acceptance as patient satisfaction with care, which tells us little about why
patients accept or reject the technology or how system design features affect patient
acceptance. This study is significant in the connected health literature for depression
care in that it utilizes measures from the literature of user acceptance of new
technologies [11,55,56]. These user acceptance measures allow us to derive new
knowledge that helps not only to explain why the ATA system is acceptable or not to
patients, but also to understand how we may improve patient acceptance through the
design of the system.

Numerous studies on connected health applications have reported a drop in technology
usage over time [57-67]. Unlike these studies, we found that patients’ completion of ATA
calls was high and constant throughout the trial. As mentioned above, the main reasons
patients reported for incomplete ATA calls were not related to the intervention [41]. In
fact, we found in an analysis not included in this paper that the survey-based measures
of acceptance were not statistically significant predictors of ATA call completion rates.
Nonetheless, as reported in the Results section, the ATA call completion rate was
29
positively correlated with perceived ease of use and perceived non-intrusiveness. The
significance of the former factor is in agreement with the PTAM. However, the finding
that patients continued to complete ATA calls over time despite a general decrease in
acceptance is surprising. Future research is needed to determine whether it was the
special characteristics of the study population (ie, predominantly urban, low-income
Latinos) or the technology design (ie, outbound calls to patients) that resulted in this
finding.

The PTAM sheds light on factors that increase the likelihood that patients will be willing
to use connected health technologies. Among a myriad of potential factors, the main
ones predicting patient acceptance are perceived usefulness, perceived ease of use,
subjective norm, and health care knowledge. Others have similarly reported that
perceived usefulness and perceived ease of use are the main driving forces of patient
technology acceptance [11,68,69]. Likewise, we found that long-term perceived
usefulness of ATA calls significantly predicted patient acceptance of automated
depression screening and monitoring. A new predictor of acceptance suggested in our
analysis was patients’ perception that calls were private/secure. Future patient
technology acceptance research should consider this factor in the technology design and
should validate the finding.
Conclusions

In the short term, safety-net ambulatory care patients with diabetes are generally
accepting of ATA calls for depression screening and monitoring. How patient acceptance
can be sustained over time is an important topic for future investigation. In order to
increase the odds that patients will accept ATA calls over the long term, especially for
30
sensitive mental health conditions, the system should gauge patient perception of its
privacy/security. Moreover, it is critically important that the technology not only be
aligned with patients’ needs, but also be perceived as useful for them over the long term.
Based on the items measuring long-term usefulness, future research should focus on
designing and testing technologies that help patients be more aware of how they are
feeling, remind them to take care of their health, and help them stay better connected
with their health care providers.
Acknowledgements

This project was supported with a grant from the US Department of Health and Human
Services, Office of the Assistant Secretary for Planning and Evaluation, grant number
1R18AE000054-01. Dr Shinyi Wu was the Principal Investigator. Ms Magaly Ramirez
received support through the National Science Foundation Graduate Research
Fellowship Program. We acknowledge the LACDHS clinics, providers, and patients who
participated in DCAT; the research team for their contribution to this study; and
4PatientCare for assisting with the development of the ATA calls.
Abbreviations

ATA: automated telephonic assessment
BMI: body mass index
CMS: Centers for Medicare and Medicaid Services
DCAT: Diabetes-Depression Care-management Adoption Trial
DMP: disease management program
DMR: disease management registry
HIT: health information technology
31
ICD-9: International Classification of Diseases, 9th Revision
LACDHS: Los Angeles County Department of Health Services
PCP: primary care provider
PHQ: Patient Health Questionnaire
SC: supported care
SCL-20: Symptom Checklist, 20 items
SF-12: Short-form Health Survey, 12 items
TC: technology-facilitated care
UC: usual care

References

1. Medicare program; revisions to payment policies under the physician fee schedule,
clinical laboratory fee schedule, access to identifiable data for the Center for Medicare
and Medicaid Innovation models & other revisions to Part B for CY 2015. §42 CFR Parts
403, 405, 410, 411, 412, 413, 414, 425, 489, 495, and 498. 2014.
2. Kvedar J, Coye MJ, Everett W. Connected health: a review of technologies and
strategies to improve patient care with telemedicine and telehealth. Health Affairs 2014
Feb;33:(2)194–199. PMID: 24493760
3. Dang S, Dimmick S, Kelkar G. Evaluating the evidence base for the use of home
telehealth remote monitoring in elderly with heart failure. Telemed E J Health
2009;15(8):783–796. PMID: 19831704
4. Antonicelli R, Testarmata P, Spazzafumo L, Gagliardi C, Bilo G, Valentini M. Impact of
telemonitoring at home on the management of elderly patients with congestive heart
failure. J Telemed Telecare 2008;14(6):300–305. PMID: 18776075
5. Polisena J, Tran K, Cimon K, Hutton B, McGill S, Palmer K. Home telemonitoring for
congestive heart failure: a systematic review and meta-analysis. J Telemed Telecare
2010;16(2):68–76. PMID: 20008054
6. Clark R, Inglis S, McAlister F, Cleland J, Stewart S. Telemonitoring or structured
telephone support programmes for patients with chronic heart failure: systematic review
and meta-analysis. BMJ 2007;334:942. PMID: 17426062
32
7. Kulshreshtha A, Kvedar J, Goyal A, Halpern E, Watson A. Use of remote monitoring
to improve outcomes in patients with heart failure: a pilot trial. Int J Telemed Appl
2010;2010(3). PMID: 23737229
8. Darkins A, Ryan P, Kobb R, Foster L, Edmonson E, Wakefield B. Care
coordination/home telehealth: the systematic implementation of health informatics, home
telehealth, and disease management to support the care of veteran patients with chronic
conditions. Telemed J E Health 2008;14(10):1118–1126. PMID: 19119835
9. Venkatesh V, Morris MG, Davis GB, Davis FD. User acceptance of information
technology: toward a unified view. MIS Quart 2003;27(3):425–478. DOI:
10.2307/30036540
10. Taylor S, Todd PA. Understanding information technology usage: a test of competing
models. Inf Syst Res 1995;6(2):144–176. DOI: 10.1287/isre.6.2.144
11. Davis FD. User acceptance of information technology: system characteristics, user
perceptions and behavioral impacts. Int J Man Mach Stud 1993;38(3):475–487. DOI:
10.1006/imms.1993.1022
12. Piette JD. Interactive behavior change technology to support diabetes self-
management: where do we stand? Diabetes Care 2007 Oct;30(10):2425–2432. PMID:
17586735
13. Wu S, Ell K, Gross-Schulman SG, Sklaroff LM, Katon WJ, Nezu AM, Lee PJ,
Vidyanti I, Chou CP, Guterman JJ. Technology-facilitated depression care management
among predominantly Latino diabetes patients within a public safety net care system:
comparative effectiveness trial design. Contemp Clin Trials 2014 Mar;37(2):342–354.
PMID: 24215775
14. Wu S, Vidyanti I, Liu P, Hawkins C, Ramirez M, Guterman J, Gross-Schulman S,
Myerchin L, Ell K. Patient-centered technological assessment and monitoring of
depression for low-income patients. J Ambul Care Manage 2014;37(2):138–147. PMID:
24525531
15. Alegria M, Chatterji P, Wells K, Cao Z, Chen CN, Takeuchi D, Jackson J, Meng XL.
Disparity in depression treatment among racial and ethnic minority populations in the
United States. Psychiatr Serv 2008;59(11):1264–72. PMID: 25577670
16. Sclar DA, Robison LM, Skaer TL. Ethnicity/race and the diagnosis of depression and
use of antidepressants by adults in the United States. Int Clin Psychopharmacol
2008;23(2):106–9. PMID: 18301125
17. Simpson SM, Krishnan LL, Kunik ME, Ruiz P. Racial disparities in diagnosis and
treatment of depression: a literature review. Psychiatr Q 2007;78(1):3–14. PMID:
17102936
33
18. Arean PA, Unutzer J. Inequalities in depression management in low-income,
minority, and old-old adults: a matter of access to preferred treatments? J Am Geriatr
Soc 2003;51(12):1808–9. PMID: 14687363
19. Stockdale SE, Lagomasino IT, Siddique J, McGuire T, Miranda J. Racial and ethnic
disparities in detection and treatment of depression and anxiety among psychiatric and
primary health care visits. 2008;46(7):668–77. PMID: 18580385
20. Sanchez-Lacay JA, Lewis-Fernandez R, Goetz D, Blanco C, Salman E, Davies S,
Liebowitz M. Open trial of nefazodone among Hispanics with major depression: efficacy,
tolerability, and adherence issues. Depress Anxiety 2011;13(3):118–24. PMID:
11387731
21. Heithoff K. Does the ECA underestimate the prevalence of late-life depression? J Am
Geriatr Soc 1995;43(1):2–6. PMID: 7806734
22. Lyness JM, Cox C, Curry J, Conwell Y, King DA, Caine ED. Older age and the
underreporting of depressive symptoms. J Am Geriatr Soc 1995;43(3):216–21. PMID:
7884106
23. Sirey JA, Bruce ML, Alexopoulos GS, Perlick DA, Raue P, Friedman SJ, Meyers BS.
Perceived stigma as a predictor of treatment discontinuation in young and older
outpatients with depression. Am J Psychiatry 2001;158(3):479–81. PMID: 11229992
24. Ell K, Katon W, Xie B, Lee PJ, Kapetanovic S, Guterman J, Chou CP. Collaborative
care management of major depression among low-income, predominantly Hispanics with
diabetes: a randomized controlled trial. Diabetes Care 2010;33(4):706-13. PMID:
20097780
25. Cabassa LJ, Hansen MC, Palinkas LA, Ell K. Azucar y nervios: explanatory models
and treatment experiences of Hispanics with diabetes and depression. Soc Sci Med
2008;66(12):2413–24. PMID: 18339466
26. Huang ES, Brown SE, Zhang JX, Kirchhoff AC, Schaefer CT, Casalino LP, Chin MH.
The cost consequences of improving diabetes care: the community health center
experience. Jt Comm J Qual Patient Saf 2008;34(3):138–46. PMID: 18419043
27. Dwight-Johnson M, Unutzer J, Sherbourne C, Tang L, Wells KB. Can quality
improvement programs for depression in primary care address patient preferences for
treatment? Med Care 2001;39(9):943–44. PMID: 11502951
28. Miranda J, Duan N, Sherbourne C, Schoenbaum M, Lagomasino I, Jackson-Triche
M, Wells KB. Improving care for minorities: can quality improvement interventions
improve care and outcomes for depressed minorities? results of randomized, controlled
trial. Heal Serv Res 2003;38(2):613–30.
34
29. Oxman TE, Dietrich AJ, Williams JW Jr, Kroenke K. A three-component model for
reengineering systems for the treatment of depression in primary care. Psychosomatics
2002;43(6):441–50. PMID: 12444226
30. Oxman TE, Dietrich AJ, Schulberg HC. The depression care manager and mental
health specialist as collaborators within primary care. Am J Geriatr Psychiatry
2003;11(5):509–16. PMID: 14506084
31. Badamgarav E, Weingarten SR, Henning JM, Knight K, Hasselblad V, Gano A Jr,
Ofman JJ. Effectiveness of disease management programs in depression: a systematic
review. Am J Psychiatry 2003;160(12):2080–90. PMID: 14638573
32. Henke RM, McGuire TG, Zaslavsky AM, Ford DE, Meredith LS, Arbelaez JJ.
Clinician- and organization-level factors in the adoption of evidence-based care for
depression in primary care. Heal Care Manag Rev 2008;33(4):289–99. PMID: 18815494
33. Post EP, Kilbourne AM, Bremer RW, Solano FX Jr, Pincus HA, Reynolds CF 3rd.
Organizational factors and depression management in community-based primary care
settings. Implement Sci 2009;4(84). PMID: 20043838
34. Nutting PA, Gallagher K, Riley K, White S, Dickinson WP, Korsen N, Dietrich A. Care
management for depression in primary care practice: findings from the RESPECT-
Depression trial. Ann Fam Med 2008;6(1):30–7. PMID: 18195312
35. U.S. Preventive Services Task Force. Screening for depression in adults: U.S.
Preventive Services Task Force recommendation statement. Ann Intern Med
2009;151(11):784–792. PMID: 19949144
36. Bodenheimer T, Wagner EH, Grumbach K. Improving primary care for patients with
chronic illness: the chronic care model, Part 2. J Am Med Assoc 2002;288:1909–1914.
PMID: 12377092
37. Kroenke K, Theobald D, Norton K. Effect of telecare management on pain and
depression in patients with cancer: a randomized trial. JAMA 2014;304(2):163–171.
PMID: 20628129
38. Hu PJ, Chau PYK, Sheng ORL, Tam KY. Examining the technology acceptance
model using physician acceptance of telemedicine technology. J Manage Inform Syst
1999 Fall;16(2):91–112.
39. Or CKL, Karsh BT, Severtson DJ, Burke LJ, Brown RL, Brennan PF. Factors
affecting home care patients’ acceptance of a web-based interactive self-management
technology. J Am Med Informatics Assoc 2011;18(1):51–59. PMID: 21131605
40. Legris P, Ingham J, Collerette P. Why do people use information technology? A
critical review of the technology acceptance model. Inf Manag 2003;40(3):191–204. DOI:
10.1016/S0378-7206(01)00143-4
35
41. Vidyanti I, Wu B, Wu S. Low-income minority patient engagement with automated
telephonic depression assessment and impact on health outcomes. Quality of Live
Research 2015;24(5):1119-1129. DOI: 10.1007/s11136-014-0900-8
42. Williams TL, May CR, Esmail A. Limitations of patient satisfaction studies in
telehealthcare: a systematic review of the literature. Telemed J E Health 2001;7(4):293–
316. PMID: 11886667
43. Clarke G, Eubanks D, Reid E, Kelleher C, O'Connor E, DeBar LL, Lynch F, Nunley S,
Gullion C. Overcoming depression on the Internet (ODIN): a randomized controlled trial
of an Internet depression skills intervention program. J Med Internet Res 2002
Dec;4(3):E14. PMID: 12554545
44. Christensen H, Griffiths KM, Jorm AF. Delivering interventions for depression by
using the Internet: randomised controlled trial. BMJ 2004 Jan;328(7434):265. PMID:
14742346
45. Andersson G, Bergstrom J, Hollandare F, Carlbring P, Kaldo V, Ekselius L. Internet-
based self-help for depression: randomised controlled trial. Br J Psychiatry
2005;187(5):456–461. PMID: 16260822
46. Lynch DJ, Nagel R, Tamburrino M. Telephone counseling for patients with minor
depression: preliminary findings in a family practice setting. J Fam Pract 1997;44(3):293.
PMID: 9071250
47. Hunkeler EM, Meresman JF, Hargreaves WA, Fireman B, Berman WH, Kirsch AJ,
Groebe J, Hurt SW, Braden P, Getzell M, Feigenbaum PA, Peng T, Salzer M. Efficacy of
nurse telehealth care and peer support in augmenting treatment of depression in primary
care. Arch Fam Med 2009;9:700–708. PMID: 10927707
48. Mohr DC, Likosky W, Bertagnolli A, Goodkin DE, Van Der Wende J, Dwyer P, Dick
LP. Telephone-administered cognitive-behavioral therapy for the treatment of depressive
symptoms in multiple sclerosis. J Consult Clin Psychol 2000;68(2):356–361. PMID:
10780138
49. Miller L, Weissman M. Interpersonal psychotherapy delivered over the telephone to
recurrent depressives. A pilot study. Depress Anxiety 2002 Jan;16(3):114–117. PMID:
12415535
50. Lynch D, Tamburrino M, Nagel R, Smith MK. Telephone-based treatment for family
practice patients with mild depression. Psychol Rep 2004;94(3):785–792. PMID:
15217028
51. Lieberman MA, Winzelberg A, Golant M, Wakahiro M, DiMinno M, Aminoff M,
Christine C. Online support groups for Parkinson’s patients: a pilot study of
effectiveness. Soc Work Health Care 2006;42(2):23–38. PMID: 16390834
36
52. Spek V, Nyklícek I, Smits N, Cuijpers P, Riper H, Keyzer J, Pop V. Internet-based
cognitive behavioural therapy for subthreshold depression in people over 50 years old: a
randomized controlled clinical trial. Psychol Med 2007 Dec;37(12):1797–1806. PMID:
17466110
53. Warmerdam L, van Straten A, Twisk J, Riper H, Cuijpers P. Internet-based treatment
for adults with depressive symptoms: randomized controlled trial. J Med Internet Res
2008 Jan;10(4):e44. PMID: 19033149
54. Meyer B, Berger T, Caspar F, Beevers CG, Andersson G, Weiss M. Effectiveness of
a novel integrative online treatment for depression (Deprexis): randomized controlled
trial. J Med Internet Res 2009 Jan;11(2):e15. PMID: 19632969
55. Davis FD. Perceived usefulness, perceived ease of use, and user acceptance of
information technology. MIS Quart 1989;13(3):319. DOI: 10.2307/249008
56. Dillon A, Morris M. User acceptance of new information technology: theories and
models. Annu Rev Inform Sci Technol 1996;31:3–32.
57. Glasgow R, Boles S, McKay H, Feil E, Barrera M. The D-net diabetes self-
management program: long-term implementation, outcomes and generalization results.
Prev Med 2003;36(4):410–419. PMID: 12649049
58. Lorig K, Ritter P, Laurent D, Plant K. Internet-based chronic disease self-
management: a randomized trial. Med Care 2006;44:964–971. PMID: 17063127
59. Heisler M, Piette J. “I help you, and you help me”: facilitated telephone peer support
among patients with diabetes. Diabetes Educ 2005;31(6):869–879. PMID: 16288094
60. Fisher L, Chesla CA. The family and disease management in Hispanic and
European-American patients with type 2 diabetes. Diabetes Care 2000;23(3):267–272.
PMID: 10868849
61. Balas EA, Krishna S, Kretschmer RA, Cheek TR, Lobach DF, Boren SA.
Computerized knowledge management in diabetes care. Med Care 2004 Jun;42(6):610–
621. PMID: 15167329
62. Murray E, Burns J, See T, Zai R, Nazareth I. Interactive health communication
applications for people with chronic disease. Cochrane Database Syst Rev
2005;4:CD004274. PMID: 16235356
63. Young AS, Chaney E, Shoai R, Bonner L, Cohen AN, Doebbeling B, Dorr D,
Goldstein MK, Kerr E, Nichol P, Perrin R. Information technology to support improved
care for chronic illness. J Gen Intern Med 2007;22(3):425–430. PMID: 18026812
37
64. Dorr D, Bonner LM, Cohen AN, Shoai RS, Perrin R, Chaney E, Young AS.
Informatics systems to promote improved care for chronic illness: a literature review. J
Am Med Inform Assoc 2007;14(2):156–163. PMID: 17213491
65. Tate D, Jackvony E, Wing R. Effects of Internet behavioral counseling on weight loss
in adults at risk for type 2 diabetes. JAMA 2003;289:1833–1836. PMID: 12684363
66. Tate D, Wing R, Winett R. Using Internet technology to deliver a behavioral weight
loss program. JAMA 2001;285(9):1172–1177. PMID: 11231746
67. Glasgow RE, La Chance PA, Toobert DJ, Brown J, Hampson SE, Riddle MC. Long-
term effects and costs of brief behavioural dietary intervention for patients with diabetes
delivered from the medical office. Patient Educ Couns 1997 Nov;32(3):175–184. PMID:
9423499
68. Jimison H, Gorman P, Woods S, Nygren P, Walker M, Norris S, Hersh W. Barriers
and drivers of health information technology use for the elderly, chronically ill, and
underserved. Evid Rep Technol Assess (Full Rep) 2008 Nov. PMID: 19408968
69. Davis F, Bagozzi R, Warshaw P. User acceptance of computer technology: a
comparison of two theoretical models. Manage Sci 1989;35(8):982–1003. DOI:
10.1287/mnsc.35.8.982

Appendix

Table A-1 Comparison of samples in the first analysis to the rest of patients in the TC
arm of DCAT
Characteristic
Sample for First
Analysis
Rest of Patients in
the TC Arm of DCAT P
b
N Statistics
a
N Statistics
a

Female 109 72 (66.1%) 333 201 (60.4%) 0.29
Age 109 51.94 (9.01) 333 52.80 (8.87) 0.38
Latino 109 105 (96.3%) 332 295 (88.9%) 0.02
Spanish as preferred language 109 93 (85.0%) 333 268 (80.5%) 0.26
Married 109 49 (45.0%) 333 174 (52.3%) 0.19
PHQ-9 (range 0–27,
higher=more severe
depression)
b, c

109 5.73 (4.93) 333 5.48 (5.05) 0.66
Total number of socioeconomic
stressors
c

109 2.28 (1.56) 333 2.35 (1.59) 0.72
SCL-20, mean score
c, d
109 0.54 (0.53) 333 0.51 (0.52) 0.62
SF-12 mental (general 109 50.54 (9.15) 333 50.96 0.70
38
population=50, higher=better)
c, e
(10.09)
Time with diabetes in years 107 10.15 (7.42) 330 10.61 (7.14) 0.57
On insulin treatment
c
109 82 (75.2%) 333 229 (68.8%) 0.20
BMI
c, f
109 32.93 (6.55) 333 32.93 (6.81) 1.00
A1C value
c, g
108 8.87 (1.39) 332 9.12 (1.60) 0.14
Low-density lipoprotein
cholesterol
c,

108 167.08
(36.20)
332 171.82
(39.51)
0.27
Whitty-9 diabetes symptoms
(range 1–5, 1=none to 5=every
day)
c

109 1.64 (0.54) 333 1.60 (0.49) 0.45
Number of diabetes
complications
c

109 1.26 (0.89) 333 1.26 (0.97) 0.96
Toolbert diabetes self-care in
the past 7 days (range 0–7)
c

109 4.63 (0.98) 333 4.42 (1.08) 0.08
Diabetes emotional burden
(range 1–5, 1=not a problem to
5=very burdensome)
c

109 2.53 (1.35) 333 2.49 (1.50) 0.79
Diabetes regime distress (range
1–5, 1=not a problem to 5=very
burdensome)
c

109 2.19 (1.14) 333 2.19 (1.37) 0.98
Self-rated health (range 1–5,
1=poor to 5=excellent)
c

109 2.29 (0.60) 333 2.37 (0.71) 0.25
Chronic pain
c
109 17 (15.6%) 333 58 (17.4%) 0.66
SF-12 physical (general
population=50, higher=better
health)
c, e

109 43.18 (9.62) 333 43.85 (9.50) 0.52
Sheehan disability scale (range
0–10, 0=none to 10=extremely)
c

109 2.21 (2.34) 333 2.14 (2.43) 0.81
Number of ICD-9 diagnosis
c, h
108 8.60 (4.50) 332 7.96 (4.14) 0.18
Number of clinic visits
c
107 10.44 (5.61) 317 9.54 (5.56) 0.15
Number of emergency room
visits
c

41 1.33 (0.61) 90 1.23 (0.42) 0.35
Number of hospitalizations
c
15 1.47 (0.83) 33 1.09 (0.26) 0.10
Willingness to use
c
109 4.02 (0.93) 153 3.56 (1.40) 0.002
Perceived ease-of-use
c
109 4.05 (0.56) 146 3.98 (0.66) 0.36
Perceived usefulness
c
109 3.63 (0.89) 145 3.60 (0.94) 0.81
Perceived non-intrusiveness
c
109 4.20 (0.87) 146 4.02 (1.05) 0.13
Perceived privacy/security
c
109 4.10 (1.11) 148 4.14 (1.12) 0.81
Preference of ATA call mode
c
109 3.82 (1.06) 152 3.42 (1.53) 0.01
Long-term perceived usefulness 76 3.71 (0.92) 71 3.68 (1.13) 0.84
ATA call completion rate
c
108 0.70 (0.26) 303 0.50 (0.33) <.001
a
Values are numbers (column percentages) for categorical variables and mean (SD) for
continuous variables
b
Chi-square test for categorical variables and two-sample t-test for continuous variables
c
Assessment at 6 or 12 months. If both were available, then the average was taken.
d
Symptoms CheckList, 20 items
e
Short-Form Health Survey, 12 items
39
f
Body mass index
g
Glycated hemoglobin test
h
International Classification of Diseases, 9th Revision

Table A-2 Comparison of samples in the second analysis to the rest of patients in the TC
arm of DCAT
Characteristic
Sample for Second
Analysis
Rest of Patients in
the TC Arm of DCAT P
b
N Statistics
a
N Statistics
a

Female 125 80 (64.0%) 333 201 (60.4%) 0.29
Age 125 51.31 (8.81) 317 53.09 (8.91) 0.06
Latino 125 116 (92.8%) 332 295 (88.9%) 0.02
Spanish as preferred language 125 104 (83.2%) 333 268 (80.5%) 0.26
Married 125 55 (44.0%) 333 174 (52.3%) 0.19
PHQ-9 (range 0–27,
higher=more severe
depression)
b, c

125 5.65 (4.60) 317 5.50 (5.18) 0.78
Total number of socioeconomic
stressors
c

125 2.37 (1.46) 317 2.31 (1.63) 0.72
SCL-20, mean score
c, d
125 0.51 (0.48) 317 0.52 (0.54) 0.89
SF-12 mental (general
population=50, higher=better)
c, e

125 51.08 (9.03) 317 50.76
(10.18)
0.76
Time with diabetes in years 124 9.98 (7.05) 313 10.70 (7.27) 0.35
On insulin treatment
c
125 89 (71.2%) 317 222 (70.0%) 0.81
BMI
c, f
125 32.75 (6.16) 317 33.00 (6.97) 0.73
A1C value
c, g
124 8.72 (1.39) 316 9.20 (1.59) 0.003
Low-density lipoprotein
cholesterol
c

124 168.44
(36.60)
316 171.53
(39.57)
0.45
Whitty-9 diabetes symptoms
(range 1–5, 1=none to 5=every
day)
c

125 1.62 (0.49) 317 1.60 (0.50) 0.65
Number of diabetes
complications
c

125 1.22 (0.79) 317 1.27 (1.01) 0.57
Toolbert diabetes self-care in
the past 7 days (range 0–7)
c

125 4.65 (1.01) 317 4.40 (1.07) 0.02
Diabetes emotional burden
(range 1–5, 1=not a problem to
5=very burdensome)
c

125 2.48 (1.37) 317 2.50 (1.50) 0.90
Diabetes regime distress (range
1–5, 1=not a problem to 5=very
burdensome)
c

125 2.13 (1.17) 317 2.21 (1.37) 0.58
Self-rated health (range 1–5,
1=poor to 5=excellent)
c

125 2.34 (0.60) 317 2.36 (0.71) 0.81
Chronic pain
c
125 24 (19.2%) 333 51 (16.1%) 0.43
SF-12 physical (general 125 43.17 (9.49) 317 43.88 (9.54) 0.48
40
population=50, higher=better
health)
c, e

Sheehan disability scale (range
0–10, 0=none to 10=extremely)
c

125 2.14 (2.26) 317 2.17 (2.46) 0.91
Number of ICD-9 diagnosis
c, h
124 8.46 (4.46) 316 7.98 (4.14) 0.29
Number of clinic visits
c
124 10.56 (5.64) 300 9.44 (5.53) 0.06
Number of emergency room
visits
c

44 1.33 (0.60) 87 1.09 (0.27) 0.31
Number of hospitalizations
c
18 1.39 (0.78) 30 1.09 (0.27) 0.14
Willingness to use
c
125 4.00 (1.08) 137 3.53 (1.34) 0.002
Perceived ease-of-use
c
125 4.12 (0.50) 130 3.91 (0.70) 0.005
Perceived usefulness
c
125 3.69 (0.90) 129 3.54 (0.93) 0.17
Perceived non-intrusiveness
c
125 4.29 (0.84) 130 3.92 (0.70) 0.003
Perceived privacy/security
c
125 4.17 (1.08) 132 4.07 (1.14) 0.47
Preference of ATA call mode
c
125 3.58 (1.32) 136 3.59 (1.41) 0.98
Long-term perceived usefulness 125 3.74 (0.99) 22 3.45 (1.18) 0.24
ATA call completion rate
c
123 0.74 (0.24) 288 0.48 (0.32) <.001
a
Values are numbers (column percentages) for categorical variables and mean (SD) for
continuous variables
b
Chi-square test for categorical variables and two-sample t-test for continuous variables
c
Assessment at 6 or 12 months. If both were available, then the average was taken.
d
Symptoms CheckList, 20 items
e
Short-Form Health Survey, 12 items
f
Body mass index
g
Glycated hemoglobin test
h
International Classification of Diseases, 9th Revision

Table A-3 Characteristics of patients reporting high versus low willingness to use ATA
calls at 18 months
Characteristic
High willingness to
use ATA calls at 18
months
Low willingness to
use ATA calls at 18
months
P
b
N Statistics
a
N Statistics
a

Female 74 52 (70.3%) 51 28 (54.9%) 0.08
Age 74 51.97 (8.56) 51 50.35 (9.16) 0.41
Latino 74 71 (95.9%) 51 45 (88.2%) 0.10
Spanish as preferred language 74 65 (87.8%) 51 39 (76.5%) 0.10
Married 74 32 (43.2%) 51 23 (45.1%) 0.84
PHQ-9 (range 0–27, higher=more
severe depression)
b, c

74 5.62 (4.58) 51 5.69 (4.69) 0.94
Total number of socioeconomic
stressors
c

74 2.46 (1.45) 51 2.25 (1.48) 0.38
SCL-20, mean score
c, d
74 0.52 (0.47) 51 0.50 (0.51) 0.84
SF-12 mental (general
population=50, higher=better)
c, e

74 50.40 (9.18) 51 52.07 (8.81) 0.31
Time with diabetes in years 73 9.88 (7.12) 51 10.14 (7.01) 0.84
41
On insulin treatment
c
74 53 (71.6%) 51 36 (70.6%) 0.90
BMI
c, f
74 32.54 (6.20) 51 33.06 (6.16) 0.65
A1C value
c, g
74 8.71 (1.30) 50 8.73 (1.53) 0.91
Low-density lipoprotein
cholesterol
c

74 172.03
(37.36)
50 166.10
(35.68)
0.56
Whitty-9 diabetes symptoms
(range 1–5, 1=none to 5=every
day)
c

74 1.59 (0.46) 51 1.67 (0.53) 0.34
Number of diabetes
complications
c

74 1.27 (0.83) 51 1.15 (0.73) 0.42
Toolbert diabetes self-care in the
past 7 days (range 0–7)
c

74 4.81 (0.95) 51 4.43 (1.05) 0.03
Diabetes emotional burden
(range 1–5, 1=not a problem to
5=very burdensome)
c

74 2.50 (1.37) 51 2.46 (1.37) 0.87
Diabetes regime distress (range
1–5, 1=not a problem to 5=very
burdensome)
c

74 2.14 (1.14) 51 2.13 (1.23) 0.96
Self-rated health (range 1–5,
1=poor to 5=excellent)
c

74 2.37 (0.59) 51 2.30 (0.62) 0.52
Chronic pain
c
74 13 (17.6%) 51 11 (21.6%) 0.58
SF-12 physical (general
population=50, higher=better
health)
c, e

74 43.48 (9.53) 51 42.72 (9.51) 0.66
Sheehan disability scale (range
0–10, 0=none to 10=extremely)
c

74 2.13 (2.08) 51 2.15 (2.53) 0.97
Number of ICD-9 diagnosis
c, h
74 8.30 (4.02) 50 8.69 (5.08) 0.63
Number of clinic visits
c
74 10.97 (5.75) 50 9.96 (5.47) 0.33
Number of emergency room
visits
c

29 1.44 (0.69) 15 1.10 (0.28) 0.07
Number of hospitalizations
c
11 1.45 (0.82) 7 1.29 (0.76) 0.67
Willingness to use
c
74 4.17 (1.00) 51 3.75 (1.16) 0.04
Perceived ease-of-use
c
74 4.17 (0.50) 51 4.05 (0.51) 0.17
Perceived usefulness
c
74 3.84 (0.82) 51 3.49 (0.97) 0.03
Perceived non-intrusiveness
c
74 4.42 (0.65) 51 4.09 (1.03) 0.05
Perceived privacy/security
c
74 4.42 (0.91) 51 3.81 (1.22) 0.00
3
Preference of ATA call mode
c
74 3.66 (1.35) 51 3.47 (1.29) 0.43
Long-term perceived usefulness 74 4.07 (0.91) 51 3.25 (0.91) <.00
1
ATA call completion rate
c
72 0.75 (0.23) 51 0.73 (0.25) 0.56
a
Values are numbers (column percentages) for categorical variables and mean (SD) for
continuous variables
b
Chi-square test for categorical variables and two-sample t-test for continuous variables
c
Assessment at 6 or 12 months. If both were available, then the average was taken.
d
Symptoms CheckList, 20 items
e
Short-Form Health Survey, 12 items
42
f
Body mass index
g
Glycated hemoglobin test
h
International Classification of Diseases, 9th Revision
43
Chapter 3 Patient Preferences for Text Messaging Intervention
Features to Support Physical Activity Behavior Change: A
Discrete Choice Experiment with Low-Income Latino Patients
with Diabetes

Abstract

Background: Interest in using text message interventions (TMIs) to support patient self-
care behavior change is growing. To ensure that such services provide optimal support,
they must be designed with an underlying theoretical framework of behavior change and
with consideration of patient preferences.
Objective: To estimate, from the patient perspective, the importance of TMI features to
support physical activity (PA) behavior change among urban, low-income Latino patients
with diabetes.
Methods: A discrete choice experiment (DCE) was used to measure patient preferences
for five TMI attributes. In a survey, respondents were provided with descriptions of two
hypothetical TMIs and were asked to indicate which TMI they preferred. Respondents
were asked to make a total of seven such comparisons. Respondents were recruited in
person from a diabetes management program (DMP) of a safety-net ambulatory care
clinic; clinicians referred participants to the research assistant after routine clinic visits.
The data were analyzed using conditional logistic regression models to identify whether
each of the five attributes was important to patients at the 0.05 level of significance.
Results: From December 2014 to August 2015, 125 respondents completed the survey.
Frequency of text messaging and PA behavior change education were considered
significantly important attributes (the former being more important than the latter). PA
44
goal setting, feedback on PA performance, and social support were not significantly
important.
Conclusions: The findings suggest that developers of TMIs to support PA behavior
change among urban, low-income, Latino adults with diabetes should consider patient
preferences for frequency of text messaging. In addition, educational content is an
important feature of TMIs for this population. Future research is needed to validate and
expand our findings.

Introduction

Diabetes self-care is critical for improving glycemic control and reducing the risk of
complications that contribute to morbidity and mortality (1). Self-care entails healthy
eating, engaging in physical activity (PA), monitoring blood glucose, and taking diabetes
medications. Diabetes self-management education (DSME) prepares individuals with the
knowledge, skills, and abilities necessary to accomplish these behaviors (2). DSME
improves clinical outcomes, health status, and quality of life (3-9). Nonetheless, Latinos
are less likely to receive DSME compared to non-Latino whites (10). In addition,
knowledge deficiencies are observed among urban, low-income, Latino safety-net
patients with diabetes (11). It is not surprising then that Latinos report worse diabetes
self-care behaviors compared to non-Latino whites (10).

Many barriers preclude Latinos from receiving DSME, including health system factors
such as lack of access to care and provision of health education; provider factors such
as language and culture differences, and ineffective interpersonal communication; and
patient-level factors such as low health literacy and numeracy, and cultural differences in
45
health perceptions (12-18). Navigator programs addressing these challenges are
successful, but their cost and labor intensity makes them difficult to implement widely
(19-22). Readily available communication technologies, such as automated text
messaging, are alternative modalities for delivering DSME with the potential to reach,
engage, and activate self-care behaviors among Latinos. Technologies can target most
aspects of DSME, use the patient’s preferred language, use content designed with an
understanding of patients’ cognitive abilities, address patients’ needs while also
considering preferences, and be delivered at a relatively low cost. A systematic review
found that text message interventions (TMIs) delivering health education made
significant improvements in glycemic control among individuals with type 2 diabetes
(T2D) (23). However, most studies were conducted in Asia, limiting the generalizability of
the findings.

There is a need for research to design, implement, and evaluate the effectiveness of
using technologies such as text messaging to provide health education to vulnerable
racial/ethnic minorities and to prompt their self-care behaviors (10, 24). Regarding
design, the current evidence base provides insufficient guidance. Although partial
guidance is sometimes drawn from behavior change theories, designing TMIs based on
behavior change theories alone is not sufficient for addressing a major challenge in TMI
research: lack of sustained patient engagement (25). Research suggests that individuals
are more likely to use technologies that address their needs and preferences (26).
Therefore, a more sensible approach may be to design TMIs with an underlying behavior
change theory to ensure the intervention contains components predicted to facilitate the
behavior change process, and also with consideration for patient preferences to increase
46
the likelihood they will use the technology. A design approach that incorporates both
theory of change and input from the target audience may yield TMIs that are both
effective and engaging.

While a vast amount of literature already exists to inform the theoretical frameworks for
TMIs (27), less is known about patient preferences for TMIs. Sarkar et al. (28) conducted
a survey to investigate preferences for diabetes self-management support, but only
assessed preferences for different delivery modes. Belmon et al. (29) also conducted a
survey to explore the opinions of young adults for attributes of a mobile health (mHealth)
app to promote PA. However, participants were Dutch and mostly female, highly
educated, and physically active, which makes it difficult to generalize the results. Patrick
et al. (30) evaluated a TMI for weight loss. During TMI development, they conducted a
focus group to assess, among other things, preferences for type and frequency of
messaging. The report focused on the evaluation, and so details on preferences for
frequency of messaging were not provided.

We conducted a study to address the gap in the literature on patient preferences so that
we could design a TMI that incorporates both behavior change theories and input from
the target audience. The focus is on urban, low-income, Latino adults with diabetes in
order to explore the potential of using TMIs to reduce disparities in the receipt of DSME
and engagement in self-care behaviors. Additionally, we focus on TMIs for PA behavior
change support given that people with T2D who exercise regularly have improved blood
glucose control and insulin sensitivity, yet only 28 percent of Latinos with diabetes are
sufficiently active; Latinos perceive PA as one of the most difficult aspects of diabetes
47
self-care; and the majority of DSME delivered via text messaging has targeted other self-
care behaviors (23, 31-39). The primary objective was to estimate, from the patient
perspective, the importance of TMI features in supporting PA behavior change. The
secondary objective was to investigate how feature preferences varied by patient
characteristics (such as age, gender, and education).

Methods

We used a discrete choice experiment (DCE) to investigate patient preferences. A DCE
is a quantitative approach for eliciting individual preferences for products/services (40).
DCEs originated in marketing research, but have been increasingly used for eliciting
patient preferences for health service delivery (41-42). We chose to conduct a DCE
because it would allow us to systematically uncover how patients value TMI features.

In a DCE, a product/service is described by attributes. Each attribute in turn has different
levels. For example, the attribute of a healthcare delivery system, appointment waiting
time, could have the levels 3–6, 7–10, or 11–14 days (41). Using a survey, individuals
are asked to state their preferences for hypothetical alternatives of the product/service.
Alternatives are described by attributes and differ by attribute levels. Thus, each
alternative is a different combination of attribute levels. Responses are used to
determine whether the attributes significantly influence preferences, the relative
importance of the attributes, and which attribute levels are preferred. The DCE in this
study comprised the four steps described below.

48
Step 1: Identify Attributes and Levels to Describe TMIs

We derived attributes from a set of 26 behavior change strategies used in similar
interventions (43) and then formed a subset of 12 attributes by excluding those that did
not meet at least two of three criteria: 1) attribute is linked to a theoretical framework, 2)
relatively strong evidence that attribute can improve PA behavior, and 3) attribute can
address PA barriers among Latinos.

Because DCE guidelines suggest using six or fewer attributes (44), the remaining 12
attributes were combined and/or re-expressed as four attributes: 1) PA goal setting, 2)
feedback on PA performance, 3) PA behavior change education, and 4) social support
(see Table A-1 in Appendix). We added a fifth attribute, frequency of messaging,
because we hypothesized that message frequency would be important to patients asked
to evaluate TMIs.

To minimize the cognitive burden on survey respondents, we assigned two levels per
attribute. DCE guidelines suggest using between two and five levels (44). To comply with
guidelines that levels cover the full range of product/service possibilities, we reviewed
published studies and consulted experts to understand how the five attributes are
typically operationalized and selected the two most salient ways as the levels.

Based on a pilot with six individuals who work with our target population, attribute and
level descriptions were adjusted for clarity and concision as shown in Table 1.

49
Table 1. TMI attributes and levels included in the DCE survey
Attribute Level 1
Level 2
PA goal setting
Patient’s doctor recommends PA
goals
Patient selects their own,
personalized PA goals
Feedback on
PA
performance
Patient receives feedback on his
or her individual performance
Patient’s performance is
compared to that of other patients
PA behavior
change
education
Patient’s doctor recommends the
educational content
Patient specifies what type of
educational content they want to
receive
Social support
Family members learn how to
offer support
Patient meets other patients so
they could support one another
Frequency of
messaging
Patient’s doctor recommends
how often patient should receive
messages
Patient specifies how often they
want to receive messages

Step 2: Construct Choice Sets and Design Survey

To construct the choice sets (generally two or more hypothetical product/service
alternatives), we first generated an experimental design to specify the attribute-level
combinations (i.e., which alternatives) respondents would evaluate in the survey. A full-
factorial design would require respondents to evaluate 32 alternatives (5 attributes at 2
levels each: 25 = 32). To make the survey more manageable, DCE macros available on
SAS software were used to construct a D-efficient experimental design that would
require respondents to evaluate fewer combinations while minimizing variances of the
parameter estimates (45). The resulting experimental design consisted of 12
combinations (i.e., alternatives).

Next, the DCE macros were used to place the 12 alternatives into pairs in a way that
would allow us to estimate all parameters (45). Each pair represented a choice set for
respondents to evaluate. For the pilot test, we developed several surveys, each with a
different number of choice sets generated by SAS. We asked the six pilot test
50
respondents selected in Step 1 to indicate at what point they felt too burdened to
continue evaluating choice sets and determined that respondents could reasonably be
expected to evaluate up to seven choice sets.

Based on the pilot finding, we designed a survey with seven questions that
corresponded to the seven choice sets (see Table A-2 in Appendix). For each of the
seven survey questions, we designed two cards, one for each alternative (called
“Program A” and “Program B”). Each card depicted a picture of a male or female
(depending on respondent’s gender) and was available in English and Spanish. Each
card had five sections that used pictures and text to describe the attributes. Every card
contained all five attributes, but the attribute levels could differ from card to card. The
attribute-level combinations depicted in each card corresponded directly to the 12
alternatives. Figure 1 provides the cards for the first question.

51
Figure 1. Example of survey question cards. The top left sections of the cards describe
the attribute “PA goal setting.” The level for Program A is 1 (patient’s doctor
recommends PA goals). For Program B, the level is 2 (patient selects his or her own,
personalized PA goals). The top right sections describe the attribute “PA behavior
change education,” with level 1 (patient’s doctor recommends the educational content)
assigned to both Programs A and B. Three additional sections similarly depict the other
three attributes.

Step 3: Conduct Survey to Measure Preferences

The Health Sciences Institutional Review Board at the University of Southern California
approved the study procedures. Using Orme’s calculation (based on seven questions,
two alternatives per question, and two levels per attribute), we determined that the
minimum sample size to estimate a main effects model was 71 (46); we therefore set our
goal to 125 respondents.

We recruited respondents from an ambulatory care clinic of the Los Angeles County
Department of Health Services (LAC-DHS), a public safety-net health system. Any
Question #1 Program A
doctor
recommends what
educational messages
you should receive
Examples: exercise benefits,
reminders to be active, tips,
words of encouragement, etc.
doctor
recommends
physical activity goal
Example: 30 minutes per
day, 5 days per week
family
members
learn what to do or say
to encourage you to
lead a physically active
lifestyle
receive text messages only as
many times as you prefer
Physical
activity can
help you take
care of your
diabetes and
prevent
diabetes
problems.
Remember to
be active today.
receive a text
message each week
on how your activity
compares to that of
others in the
program
You did 90
minutes of
physical
activity. That
is better than
18 out of the
30 patients in
our program.
Try doing 10
more minutes
this week!
Question #1 Program B
receive text messages as many
times as your doctor
recommends
set your
own goal
for how much
physical activity to do
Example: 5 or 10 minutes a
day, and work up to more
time each week
meet others
in the program so
that you can encourage
one another
Physical
activity can
help you take
care of your
diabetes and
prevent
diabetes
problems.
Remember to
be active today.
receive a text
message each week
on how your physical
activity compares to
your goal
doctor
recommends what
educational messages
you should receive
Examples: exercise benefits,
reminders to be active, tips,
words of encouragement, etc.
You did 150
minutes of
physical
activity.
That's
excellent!
Try to do it
again next
week.
52
patient in the Diabetes Management Program (DMP), which serves approximately 1200
patients per year, was eligible to participate. After routine clinic visits, clinicians informed
patients that they were eligible to participate in a survey. If patients were interested, they
were referred to the research assistant (RA). Clinicians did not keep track of how many
patients were not interested. All patients who contacted the RA completed the survey.
Patients signed a consent form and received $10.

The RA administered the survey in person at the clinic. For each of the seven choice
questions, the RA placed two printed cards in front of the respondent and described
each alternative. The RA then asked respondents, “If you were going to join one of these
two programs to help you improve your physical activity, which one would you prefer?”
The RA recorded the seven responses by each respondent using a paper log.
Step 4: Analyze the Data

Methods to analyze DCEs put forth by Ryan et al. (40) and Kuhfeld (45) guided our
analysis. For each choice question, we assumed that the respondent would choose the
alternative that led to the higher utility. Thus, in a choice set consisting of two program
alternatives, 𝑖 and 𝑗, a respondent would choose program 𝑗 if 𝑈
!
𝑧
!
,𝑐 >𝑈
!
(𝑧
!
,𝑐), where
𝑈 represents the respondent’s latent utility, 𝑧 represents the attribute levels describing
the alternative, and 𝑐 represents the respondent’s characteristics. The latent utility is
𝑈=𝑉+ 𝜀, where 𝑉= 𝑓(𝑧,𝑐) is the deterministic component of utility and 𝜀 is the random
component.

The choice model is the difference in utilities between program alternatives 𝑖 and 𝑗.
Because we observed choice rather than differences in utilities, we used a binary
53
variable, 𝑦
!
, to reflect the 𝑛
!!
respondent’s choice of program. The form of the choice
model is thus 𝑦
!
= 𝛼+𝛽𝑧
!
+𝜕𝑐
!
+ 𝜀
!"
− 𝛼+𝛽𝑧
!
+𝜕𝑐
!
+ 𝜀
!"
, where 𝛼 is the constant
term, 𝛽 represents the part-worth utility for each attribute level, and 𝜕 represents the
influence of respondent characteristics on choice of program. This choice model
simplifies to 𝑦
!
=𝛽 𝑧
!
−𝑧
!
+ (𝜀
!"
− 𝜀
!"
). That is, a respondent’s choice of program is a
function only of the programs’ characteristics. We estimated the model using SAS
software conditional logistic regression. The coefficient estimates indicated (by statistical
significance) whether the corresponding attributes were important to patients when they
made decisions about their preferred TMI. The coefficient estimates also indicated (by
relative size) how important each attribute was in relation to others. A positive coefficient
for a given attribute indicated that a respondent preferred level 1 to level 2.

We can also assume that 𝛽 (i.e., the part-worth utility for each attribute level) depends
on 𝑐
!
. That is, 𝛽= 𝜋+𝜆𝑐
!
, where 𝜋 is a constant term and 𝜆 represents the influence of
respondent characteristics on part-worth utility. The choice model thus becomes
𝑦
!
= 𝛼+ 𝜋+𝜆𝑐
!
𝑧
!
+𝜕𝑐
!
+ 𝜀
!"
− 𝛼+ 𝜋+𝜆𝑐
!
𝑧
!
+𝜕𝑐
!
+ 𝜀
!"
, which simplifies to
𝑦
!
= 𝜋 𝑧
!
−𝑧
!
+𝜆𝑐
!
𝑧
!
−𝑧
!
+ 𝜀
!"
− 𝜀
!"
. We estimated this model using SAS software
conditional logistic regression to examine how preferences varied according to a
respondent’s age, gender, and educational attainment. A statistically significant
coefficient indicated that attribute preferences varied by respondent characteristic.

To assess the models’ goodness of fit, we used chi-square test statistics, which tests the
null hypothesis that the independent variables do not influence choice. All p-values
<0.05 were considered statistically significant.
54
Results
Respondent Characteristics

As shown in Table 2, 125 patients completed the survey. The average age was 52.6
years. Most respondents were Latino (99.2%), female (71.2%), preferred to speak
Spanish (85.6%), had less than a high school education (71.5%), and had been
diagnosed with diabetes 10.8 years prior.

Table 2 Respondent characteristics
Characteristic
N
a

Value
Latino, n (%)
125
124 (99.2)
Age (years), mean (SD)
124
52.6 (10.0)
Female, n (%)
125
89 (71.2)
Spanish as preferred language
125
107 (85.6)
Educational attainment
123

Less than high school, n (%)
88 (71.5)
High school graduate, n (%)
22 (17.9)
More than high school, n (%)
13 (10.6)
Annual household income
90

Less than $20,000, n (%)

70 (56.0)
$20,000 to $29,999, n (%)

15 (12.0)
$30,000 to $39,999, n (%)

5 (4.0)
Self-reported years since diabetes diagnosis (years), mean
(SD)
125
10.8 (9.0)
Level of comfort using text messaging
124

Very uncomfortable, uncomfortable, or neutral, n (%)

40 (32.3)
Comfortable or very comfortable, n (%)

84 (67.7)
a
Numbers are not all 125 because some respondents opted not to disclose the
information

Importance of Attributes

Table 3 presents the results of the main effects only model, with attributes listed in order
of decreasing part-worth utilities. The attributes “frequency of messaging” and “PA
behavior change education” were statistically significant. Respondents derived a greater
55
utility from the attribute “frequency of messaging” than “PA behavior change education”
(coefficient size 0.371 versus 0.280). Although the other attributes were not statistically
significant, “social support” had the next highest part-worth utility (coefficient size 0.103),
followed by “feedback on PA performance” (coefficient size 0.078) and “goal setting”
(coefficient size 0.056).

Table 3 Results from the main effects model
Variable Coefficient
a
Standard Error p
Frequency of
messaging
0.371 0.078 <0.001
PA behavior
change education
0.280 0.078 <0.001
Social support 0.103 0.081 0.21
Feedback on PA
performance
0.078 0.078 0.32
PA goal setting 0.056 0.078 0.47
a
A positive coefficient indicates that respondents preferred level 1 to level 2

Preferred Levels

Respondents preferred to have clinicians recommend both frequency of messaging and
the content of the PA behavior change education. While the other coefficients were not
statistically significant, their signs indicate that respondents preferred programs in which
their families learn how to provide support, feedback is based on individual PA
performance, and clinicians recommend PA goals to patients.
Influence of Respondent Characteristics

Only one statistically significant interaction was identified; namely, respondents with less
than a high school degree derived a greater utility for clinician-recommended PA goals
than patients with a high school degree or higher. The latter group derived a greater
utility from selecting their own, individualized PA goals.
56
Discussion

We conducted this study to estimate, from the patient perspective, the importance of TMI
features to support PA behavior change. Based on a previous large-scale research study
with DMP patients at LAC-DHS (47), we determined that our 125 study participants were
representative of the target population of urban, low-income, predominantly Latino adults
with diabetes.
Principal Results and Comparison with Prior Work

We found the attributes “frequency of messaging” and “PA behavior change education”
to be statistically significant features of a TMI to support patient PA behavior change
efforts. The former attribute is more important to patients than the latter. Future research
should examine frequency of messaging and its impact on participant engagement and
health outcomes (25). The finding that patients consider “PA behavior change education”
to be an important TMI attribute complements research findings that show that
educational content can facilitate behavior change among low-income, racial/ethnic
minorities (48-50) and confirms that the attribute needs to be a design requirement for
TMIs for our target population.

Patients did not consider the other three attributes – “social support,” “feedback on PA
performance,” and “PA goal setting” – to be highly important features of a TMI. A study
of young Dutch adults had a similar result relative to social support, which was a less
appreciated feature of an mHealth app for PA behavior change support (51). On the
other hand, goal setting and feedback on performance were positively rated mHealth
app features. Previous research demonstrates that social support, goal setting, and
feedback are positively associated with PA behavior (52-60). There is a need to increase
57
awareness among our target population that the three attributes are effective strategies
for improving PA behavior. TMIs that incorporate these attributes may then appear more
valuable to patients. Furthermore, we suggest that future TMIs include all five attributes
(although only two were statistically significant) in order to create not only TMIs that suit
patient preferences, but also that improve PA outcomes. In other words, we must identify
features that are “must haves” for patients, and then design TMIs that will satisfy both
patient preferences and the clinical need for tools that will have a high likelihood of being
effective.

Our findings suggest that patients with education at the high school level or above prefer
setting their own PA goals whereas those with less than a high school degree prefer that
goals be recommended. The finding is consistent with previous research findings that
individuals with lower levels of educational attainment prefer a passive role in decisions
related to their care (61-63). The finding may also reflect the diabetes knowledge gap
observed among Latino safety-net patients with less than a high school degree (11).
TMIs need both PA goal-setting capabilities to effectively serve our target population.

Furthermore, it is not surprising that our study respondents preferred levels that allowed
them to defer attribute settings to someone else because the study participants were
recruited from a DMP, which is reserved for patients with uncontrolled diabetes.
Research has found that Latinos and adults with severe diabetes are less likely to prefer
an active role in medical decision-making (61,63). Our research findings thus reveal the
broader need to empower patients to be actively engaged in decisions related to their
care, which, according to Barry and Edgman-Levitan, is the pinnacle of patient-centered
58
care (64). The implications of these findings for involving our target population in the
design, implementation, and evaluation of healthcare communication and information
technologies merits further investigation.
Limitations

Our study has limitations that require results to be interpreted with caution. First, we did
not formally evaluate response quality. Some researchers include a repeated survey
question, but we chose not to do so in order to define a number of choice sets that would
not be cognitively burdensome. Other researchers include an alternative whose attribute
levels are all better than those of the second alternative in the choice set. However,
there are no obvious choices in this DCE study. Instead, we compared our results with a
priori expectations of the coefficient signs. We knew from previous literature that Latino
adults generally take a more passive role in their healthcare, and so we expected the
signs of the coefficients for the attributes “frequency of messaging,” “PA goal setting,”
and “PA behavior change education” to be positive – indicating that patients prefer that
their doctors make recommendations. We also knew that, given the important role of
family in the Latino culture, we should expect the attribute “social support” to have a
positive coefficient which would indicate that the patients preferred family support to peer
support. Indeed, our study results confirmed our expectations.

Another limitation is that we used conditional logit models to analyze the survey
responses. Such models assume that responses are independent; in our study,
however, we obtained multiple responses from each patient. The responses from any
one patient have the potential to be correlated; however, conditional logit models do not

59
take this into account. Nevertheless, this approach for model estimation is typical and
generally produces unbiased results (45).
Conclusions

It is feasible to use a DCE to systematically quantify patient preferences for TMI features
to support PA behavior change. The most important features for patients are “frequency
of messaging” and “PA behavior change education.” Patients prefer clinicians to
recommend both message frequency and education content. Patients also prefer
clinician-recommended PA goals, feedback based on individual PA performance, and
opportunities for their families to learn how to offer them support. Given the novelty of
investigating patient preferences for TMI features, this study was mainly exploratory;
future research should validate and expand our findings. Nonetheless, we recommend
that designers be aware of these patient preferences when developing TMIs to support
PA behavior change among urban, low-income, Latino adults with diabetes.

Acknowledgements

This project was supported with a Research Supplement to Promote Diversity in Health-
Related Research from the National Institute of Neurological Disorders and Stroke,
National Institutes of Health, U54NS081764. We acknowledge the patients, staff, and
providers at the Roybal Comprehensive Health Center who participated in the study.
References
1. American Diabetes Association. Standards of medical care in diabetes--2014.
Diabetes Care. 2014;37:S14.

60
2. Powers M, Bardsley J, Cypress M, Duker P, Funnell M, Hess Fischl A, et al.
Diabetes self-management education and support in type 2 diabetes: a joint position
statement of the American Diabetes Association, the American Association of
Diabetes Educators, and the Academy of Nutrition and Dietetics. Diabetes Care.
2015;38:1–11.
3. Brunisholz K, Briot P, Hamilton S. Diabetes self-management education improves
quality of care and clinical outcomes determined by a diabetes bundle measure. J
Multidiscip Heal. 2014;7:533–42.
4. Weaver R, Hemmelgarn B, Rabi D. Association between participation in a brief
diabetes education programme and glycaemic control in adults with newly
diagnosed diabetes. Diabet Med. 2014;31:1610–4.
5. Steinsbekk A, Rygg L, Lisulo M, Rise M, Fretheim A. A group based diabetes self-
management education compared to routine treatment for people with type 2
diabetes mellitus. A systematic review with meta-analysis. BMC Heal Serv Res.
2012;12:213.
6. Duncan I, Birkmeyeter C, Coughlin S, Li Q, Sherr D, Boren S. Assessing the value
of diabetes education. Can J Diabetes. 2009;33:18–26.
7. Fan L, Sidani S. Effectiveness of diabetes self-management education intervention
elements: a meta-analysis. Can J Diabetes. 2009;33:18–26.
8. Ellis S, Speroff T, Dittus R, Brown A, Pichert J, Elasy T. Diabetes patient education:
a meta-analysis and meta-regression. Patient Educ Couns. 2004;52:97–105.
9. Norris SL, Lau J, Smith SJ, Schmid CH, Engelgau MM. Self-management education
for adults with type 2 diabetes: a meta-analysis of the effect on glycemic control.
Diabetes Care. 2002;25(7):1159–71.
10. Chen R, Cheadle A, Johnson D, Duran B. US trends in receipt of appropriate
diabetes clinical and self-care from 2001 to 2010 and racial/ethnic disparities in
care. Diabetes Educ. 2014;40(6):756–66.
11. Arora S, Marzec K, Gates C, Menchine M. Diabetes knowledge in predominantly
Latino patients and family caregivers in an urban emergency department. Ethn Dis.
2011;21(1):1–6.
12. Schillinger D, Bindman A, Wang F, Stewart A, Piette J. Functional health literacy
and the quality of physician-patient communication among diabetes patients. Patient
Educ Couns. 2004;52:315–23.
13. Escarce J, Kapur K. Access to and Quality of Health Care. In: Tienda M, Mitchell F
States, editors. Hispanics and the Future of America. Washington, D.C.: National
Academy Press; 2006. p. 410–46.
61
14. Lopez-Quintero C, Berry E, Neumark Y. Limited English proficiency is a barrier to
receipt of advice about physical activity and diet among Hispanics with chronic
diseases in the United States. J Am Diet Assoc. 2010;109(10): 1769-1774.
15. Documet P, Sharma R. Latinos’ health care access: financial and cultural barriers. J
Immigr Heal. 2004;6(1):5–13.
16. Juckett G. Caring for Latino Patients. Am Fam Physician. 2013;87(1):48–54.
17. Schillinger D, Grumbach K, Piette J, Wang F, Osmond D, Daher C, et al.
Association of health literacy with diabetes outcomes. JAMA. 2002;288(4):475–82.
18. Karter A, Ferrara A, Darbinian J, Ackerson L, Selby J. Self-monitoring of blood
glucose: language and financial barriers in a managed care population with
diabetes. Diabetes Care. 2000;23(4):477–83.
19. Deakin T, McShane C, Cade J, Al E. Group based training for self-management
strategies in people with type 2 diabetes mellitus. Cochrane Database Syst Rev.
2005;2:CD003417.
20. Jandorf L, Stossel L, Cooperman J, Al. E. Cost analysis of a patient navigation
system to increase screening colonoscopy adherence among urban minorities.
Cancer. 2013;119(3):612-20.
21. Sixta C, Ostwald S. Strategies for implementing a promotores-led diabetes self-
management into a clinic structure. Diabetes Educ. 2008;34(2):285–98.
22. Comellas M, Walker E, Movsas S, Al. E. Training community health promoters to
implement diabetes self-management support programs for urban minority adults.
Diabetes Educ. 2010;36:141–51.
23. Saffari M, Ghanizadeh G, Koenig HG. Health education via mobile text messaging
for glycemic control in adults with type 2 diabetes: a systematic review and meta-
analysis. Prim Care Diabetes. 2014;8(4):275–85.
24. López L, Grant RW. Closing the gap: eliminating health care disparities among
Latinos with diabetes using health information technology tools and patient
navigators. J Diabetes Sci Technol. 2012;6(1):169–76.
25. U.S. Department of Health and Human Services. Health Resources and Services
Administration. Using Health Text Messages to Improve Consumer Health
Knowledge, Behaviors, and Outcomes: An Environmental Scan. 2014. Available
from: http://www.hrsa.gov/healthit/txt4tots/environmentalscan.pdf
26. Or C, Karsh B-T, Severtson D, Burke L, Brown R, Brennan Flatley P. Factors
affecting home care patients’ acceptance of a web-based interactive self-
management technology. J Am Med Informatics Assoc. 2011;18(1):51–9.
27. Glanz K, Rimer B, Viswanath K, editors. Health Behavior and Health Education:
Theory, Research, and Practice. 4th ed. San Francisco, CA: Wiley; 2008.
62
28. Sarkar U, Piette J, Gonzales R, Lessler D, Chew LD, Reilly B, et al. Preferences for
self-management support: findings from a survey of diabetes patients in safety-net
health systems. Patient Educ Couns. 2008;70(1):102–10.
29. Belmon LS, Middelweerd A, Te Velde SJ, Brug J. Dutch young adults ratings of
behavior change techniques applied in mobile phone apps to promote physical
activity: a cross-sectional survey. JMIR mHealth and uHealth. 2015;3(4):e103.
30. Patrick K, Raab F, Adams MA, Dillon L, Zabinski M, Rock CL, et al. A text message-
based intervention for weight loss: randomized controlled trial. J Med Internet Res.
2009;11(1):e1.
31. Hu J, Amirehsani K, Wallace DC, Letvak S. Perceptions of barriers in managing
diabetes: perspectives of Hispanic immigrant patients and family members.
Diabetes Educ. 2013;39(4):494–503.
32. Nelson KM, Reiber G, Boyko EJ. Diet and exercise among adults with type 2
diabetes. Diabetes Care. 2002;25(10):1722-28.
33. Zanuso S, Jimenez a, Pugliese G, Corigliano G, Balducci S. Exercise for the
management of type 2 diabetes: a review of the evidence. Acta Diabetol.
2010;47(1):15–22.
34. Manders RJF, Van Dijk J-WM, van Loon LJC. Low-intensity exercise reduces the
prevalence of hyperglycemia in type 2 diabetes. Med Sci Sports Exerc.
2010;42(2):219–25.
35. Albright A, Franz M, Hornsby G, Kriska A, Marrero D, Ullrich I, et al. Exercise and
type 2 diabetes: American College of Sports Medicine and the American Diabetes
Association: joint position statement. Med Sci Sports Exerc. 2010;42(12):2282–303.
36. Hansen D, Dendale P, Jonkers R a M, Beelen M, Manders RJF, Corluy L, et al.
Continuous low- to moderate-intensity exercise training is as effective as moderate-
to high-intensity exercise training at lowering blood HbA(1c) in obese type 2
diabetes patients. Diabetologia. 2009;52(9):1789–97.
37. Umpierre D, Kramer CK, Leita CB, Gross JL, Ribeiro JP, Schaan BD. Physical
activity advice only or structured exercise training and association with HbA1c levels
in type 2 diabetes: a systematic review and meta-analysis. JAMA.
2011;305(17):1790-99.
38. O’Hagan C, De Vito G, Boreham CAG. Exercise prescription in the treatment of type
2 diabetes mellitus: current practices, existing guidelines and future directions.
Sports Med. 2013;43(1):39–49.
39. Balducci S, Zanuso S, Cardelli P, Salvi L, Bazuro A, Pugliese L, et al. Effect of high-
versus low-intensity supervised aerobic and resistance training on modifiable
cardiovascular risk factors in type 2 diabetes; the Italian Diabetes and Exercise
Study (IDES). PLoS One. 2012;7(11):e49297.
63
40. Ryan M, Gerard K, Amaya-Amaya M, editors. Using discrete choice experiments to
value health and health care. Springer; 2007.
41. Mühlbacher AC, Bethge S, Reed SD, Schulman KA. Patient preferences for
features of health care delivery systems: a discrete choice experiment. Health Serv
Res. 2016;51(2): 704-27.
42. Jan S, Mooney G, Ryan M, Bruggemann K, Alexander K. The use of conjoint
analysis to elicit community preferences in public health research: a case study of
hospital services in South Australia. Aust N Z J Public Health. 2000;24(1):64–70.
43. Abraham C, Michie S. A taxonomy of behavior change techniques used in
interventions. Health Psychol. 2008;27(3):379–87.
44. Orme B. Formulating Attributes and Levels in Conjoint Analysis. Sequim, WA:
Sawtooth Software, Inc; 2002.
45. Kuhfeld W. Marketing research methods in SAS. Cary, NC: SAS Institute Inc; 2005.
46. Orme B. Sample Size Issues for Conjoint Analysis. In: Getting Started with Conjoint
Analysis: Strategies for Product Design and Pricing Research. Madison, WI; 2010.
p. 57–66.
47. Wu S, Ell K, Gross-Schulman S, Sklaroff Myerchin L, Katon W, Nezu A, et al.
Technology-facilitated depression care management among predominantly Latino
diabetes patients within a public safety net care system: comparative effectiveness
trial design. Contemp Clin Trials. 2014;37(2):342–54.
48. Michie S, Jochelson K, Markham W, Bridle C. Low-income groups and behaviour
change interventions: a review of intervention content, effectiveness and theoretical
frameworks. J Epidemiol Community Health. 2009;63:610–22.
49. Marquez DX, McAuley E. Social cognitive correlates of leisure time physical activity
among Latinos. J Behav Med. 2006;29(3):281–9.
50. Coulter A, Ellins J. effectiveness of strategies for informing, educating, and involving
patients. BMJ. 2008;23:11–24.
51. Peek ME, Cargill A, Huang ES. Diabetes health disparities: a systematic review of
health care interventions. Med Care Res Rev. 2007;64(5):101S-56S.
52. Booth M, Owen N, Bauman A, Clavisi O, Leslis E. Social-cognitive and perceived
environment influences associated with physical activity in older Australians. Prev
Med. 2000;31:15-22.
53. Sternfeld B, Ainsworth B, Quesenberry C. Physical activity patterns in a diverse
population of women. Prev Med. 1999;28:313–23.
54. Treiber F, Baranowski T, Braden D, Strong W. Social support for exercise:
relationship to physical activity in young adults. Prev Med. 1991;20:737–50.
64
55. Sallis J, Hovell M, Hofstetter C. Predictors of adoption and maintenance of vigorous
physical activity in men and women. Prev Med. 1992;21:237–57.
56. Dishman R, Sallis J. Determinants and interventions for physical activity and
exercise. In: Bouchard C, Shepard R, Stephens T, Sutton J, McPherson B, editors.
Physical Activity, Fitness, and Health. Champaign, IL, England: Human Kinetics
Publishers; 1994. p. 214-238.
57. Felton G, Parsons M. Factors influencing physical activity in average-weight and
overweight young women. J Community Health Nurs. 1994;11:109–19.
58. Eyler a a, Brownson RC, Donatelle RJ, King a C, Brown D, Sallis JF. Physical
activity social support and middle- and older-aged minority women: results from a
US survey. Soc Sci Med. 1999;49(6):781–9.
59. Dombrowski S, Sniehotta F, Avenell A, Johnston M, MacLennan G, Araújo-Soares
V. Identifying active ingredients in complex behavioural interventions for obese
adults with obesity-related co-morbidities or additional risk factors for co-morbidities:
a systematic review. Health Psychol Rev. 2012;6(1):7–32.
60. Michie S, Abraham C, Whittington C, McAteer J, Gupta S. Effective techniques in
healthy eating and physical activity interventions: a meta-regression. Health
Psychol. 2009;28(6):690–701.
61. Levinson W, Kao A, Kuby A, Thisted RA. Not all patients want to participate in
decision making. A national study of public preferences. J Gen Intern Med.
2005;20(6):531–5.
62. Say R, Murtagh M, Thomson R. Patients’ preference for involvement in medical
decision making: A narrative review. Patient Educ Couns. 2006;60(2):102–14.
63. Arora NK, McHorney CA. Patient preferences for medical decision making: who
really wants to participate? Med Care. 2000;38(3):335–41.
64. Barry MJ, Edgman-Levitan S. Shared Decision Making — The Pinnacle of Patient-
Centered Care. N Engl J Med. 2012;366:780–1.

Appendix

Table A-1. Potential attributes evaluated against selection criteria
Potential Attributes (1)
Selection Criteria
Related
Attributes
Theoretical
Framework
(1)
Strength of
Evidence
(2,3)
Studies
Addressing
Barriers
Provide information
about behavior
a

IMB
b
(4)
PA behavior
change
65
education
Provide information on
consequences
TRA
c
, TPB
d
,
SCT
e
, IMB
b

Provide information
about others’ approval
TRA
c
, TPB
d
,
IMB
b

Prompt intention
formation
a

TRA
c
, TPB
d
,
SCT
e
, IMB
b
,
CT
f

Strongest (5)
PA goal
setting
Prompt barrier
identification
a

SCT
e
(6,7)
PA behavior
change
education
Provide general
encouragement
SCT
e

Set graded tasks SCT
e

Provide instruction
a
SCT
e
(5,8)
PA behavior
change
education
Model or demonstrate
the behavior
SCT
e

Prompt specific goal
setting
a

CT
f
Strongest (8)
PA goal
setting
Prompt review of
behavioral goals
a

CT
f
Strongest
PA goal
setting
Prompt self-monitoring
of behavior
a

CT
f
Strongest (8)
Feedback on
PA
performance
Provide feedback on
performance
a

CT
f
Strongest
Feedback on
PA
performance
Provide contingent
rewards
OC
g
(5)
Feedback on
PA
performance
Teach to use prompts
or cues
OC
g

Agree on behavioral
contract
a

OC
g
(8)
PA goal
setting
Prompt practice OC
g

Use follow-up prompts

Provide opportunities
for social comparison
a

SCT
e
(5)
Social
support
Plan social support or
social change
a

Social support
theories
(5,7-10)
Social
support
Prompt identification as
a role model

Prompt self-talk

Relapse prevention Relapse
66
prevention
therapy
Stress management
Stress
theories

Motivational
interviewing

Time management

a
Potential attributes meeting at least two selection criteria
b
Information-motivation-behavioral skills model
c
Theory of reasoned action
d
Theory of planned behavior
e
Social cognitive theory
f
Control theory
g
Operant conditioning

Table A-1 References
1. Abraham C, Michie S. A taxonomy of behavior change techniques used in
interventions. Health Psychol. 2008;23(7):379–87.

2. Michie S, Abraham C, Whittington C, McAteer J, Gupta S. Effective techniques in
healthy eating and physical activity interventions: a meta-regression. Health Psychol.
2009;28(6):690–701.

3. Dombrowski S, Sniehotta F, Avenell A, Johnston M, MacLennan G, Araújo-Soares V.
Identifying active ingredients in complex behavioural interventions for obese adults
with obesity-related co-morbidities or additional risk factors for co-morbidities: a
systematic review. Health Psychol Rev. 2012;6(1):7–32.

4. Marquez DX, McAuley E. Social cognitive correlates of leisure time physical activity
among Latinos. J Behav Med. 2006;29(3):281–9.

5. Wing RR, Jeffery RW. Benefits of recruiting participants with friends and increasing
social support for weight loss and maintenance. J Consult Clin Psychol.
1999;67(1):132–8.

6. Ickes MJ, Sharma M. A systematic review of physical activity interventions in
Hispanic adults. J Environ Public Health. 2012;2012:156435.

7. Pekmezi DW, Neighbors CJ, Lee CS, Gans KM, Bock BC, Morrow KM, et al. A
culturally adapted physical activity intervention for Latinas: a randomized controlled
trial. Am J Prev Med. 2009;37(6):495–500.

8. Wing R, Marcus M, Epstein L, Jawad A. A “family-based” approach to the treatment
of obese type II diabetic patients. J Consult Clin Psychol. 1991;59:156–62.

9. Gilliland SS, Azen SP, Perez GE, Carter JS. Strong in body and spirit: lifestyle
intervention for Native American adults with diabetes in New Mexico. Diabetes Care.
2002;25(1):78–83.
67

10. Brown S a., Hanis CL. Culturally competent diabetes education for Mexican
Americans: the Starr County study. Diabetes Educ. 1999;25(2):226–36.

Table A-2. Seven choice sets used in the survey
Choice
Set
Program
Alternative
Attribute Level
PA Goal
Setting

Feedback
on PA
Performance
PA
Behavior
Change
Education
Social
Support
Frequency
of
Messaging
1
A 1 2 1 1 2
B 2 1 1 2 1
2
A 1 1 2 1 2
B 2 2 1 1 1
3
A 2 2 2 1 1
B 1 2 1 2 2
4
A 1 2 1 2 2
B 1 1 2 1 1
5
A 1 1 1 2 1
B 2 2 2 2 2
6
A 2 1 2 2 2
B 1 2 1 1 2
7
A 1 2 2 2 1
B 2 1 1 1 2
68
Chapter 4 Phone Messaging for Physical Activity and Social
Support Prompting Among Low-Income Latino Patients with
Diabetes: A Randomized Pilot Study

Abstract

Background: Despite the promise of phone-based interventions to effectively support
diabetes self-management (DSM), little is known about their impact on the outcomes of
highly vulnerable populations such as urban, low-income, racial/ethnic minorities. And
while phone-based interventions have generally been successful at reaching and
engaging adults with diabetes, they have failed to do the same with family
members/friends (FF) whom are a promising source of ongoing support for DSM.
Objective: The objective was to investigate the feasibility, perceived usefulness, and
potential effectiveness of a short text or voice message (ST/VM) intervention to activate
1) physical activity (PA) behavior change among urban, low-income, Latino patients in
diabetes management and 2) supportive behaviors by their FF.
Methods: We conducted a 12-week pilot study in which participants were randomized
into one of three study arms: control, phone messaging (PM), and phone messaging
plus social support from FF (PM+FF). Participants were recruited in person from a
diabetes management program at a safety-net ambulatory care clinic. All participants
were given a pedometer and walking log for self-monitoring. Participants in the PM and
PM+FF arms received ST/VMs reminding them to review daily step goals and to self-
monitor; explaining the benefits of regular PA, importance of regular PA to their daily
lives, and ways to overcome commonly identified barriers to PA; asking them to report
69
on their PA performance; providing feedback based on their response. Participants in the
PM+FF identified a FF to receive ST/VMs with suggested behaviors that are perceived
as supportive by individuals making PA behavior changes. Participants received semi-
structured assessments in person at baseline, 6 weeks, and 12 weeks. They were asked
about the extent to which they perceived that the program enhanced their ability to make
PA behavior changes. The primary outcome measures were daily step counts and
perceived FF social support.
Results: Forty-two participants were randomized to a study arm. The retention rates
were 79% (11/14) for the control and PM arms, and 93% (13/14) for the PM+FF arm.
The majority of participants were female (67 percent) and preferred to speak Spanish
(76 percent). About half had less than a high school education (52 percent). Nearly 40
percent of participants in the PM and PM+FF study arms opted to receive voice
messages instead of text messages. Participants who opted to receive text messages
responded to 63 percent of text messages requiring a response. Those who opted to
receive voice messages responded 34 percent of the time. Participants perceived as
useful the guidance in self-regulation, namely: self-monitoring, goal setting, self-
instruction, feedback, and social support. All study arms had increased daily step counts
at 6 weeks; only PM and PM+FF arms had increased daily step counts at 12 weeks. All
study arms experienced an increase in perceived FF social support, but the increase
was highest within the PM+FF arm.
Conclusion: Text messaging may be a better mode of message delivery than voice
messaging for ensuring participant receipt of and engagement with messages.
Moreover, designing a ST/VM intervention based on self-regulation techniques is
feasible and perceived as useful by participants. Finally, such an intervention may
70
improve PA in terms of daily steps as well as perceived social support from FF that
participate in the intervention.

Introduction

Diabetes self-management (DSM) is necessary in order to improve glycemic
control and reduce the risk of developing diabetes-related complications that
contribute to morbidity and mortality [1]. Self-management behaviors include
eating healthy, engaging in physical activity (PA), monitoring blood glucose, and
taking diabetes medications. DSM education can be effective [2]–[8], but patients
require ongoing self-management support in order to maintain the initial gains
achieved through education [9]. In the health care system, there is limited funding
and staffing time for providing ongoing DSM support [10]. There are, however,
other resources to help fill this gap in care. Automated phone calls or text
messages are readily available communication modalities that can provide
ongoing support for all aspects of DSM and can reach large numbers of people at
a relatively low cost. A systematic review found that text messaging interventions
were effective at improving DSM [11]. Another promising source of ongoing
support are family members and close friends (FF). Self-management behaviors
involve changes in daily routines that happen at home and often in the presence
of FF. Given their ongoing and frequent contact with patients, FF can encourage
and take an active role in DSM. A systematic review found that social support
from family members can positively affect chronic disease self-management [12].
71

Despite the promise of phone-based interventions to effectively support DSM,
little is known about their impact on the outcomes of highly vulnerable
populations such as urban, low-income Latino adults. In the United States (US),
the risk of developing diabetes among Latinos is 66 percent higher compared to
non-Latino whites [13]. They are also 1.5 times more likely to die from the
disease. The burden of illness is even greater among urban, low-income, Latino
patients who receive care at safety-net health systems [14]. Even though this
population stands to benefit the most from additional DSM support, they are
generally understudied in the eHealth literature. To our knowledge, only one
study examined a text message-based DSM intervention with urban, low-income,
predominantly Latino adults, but it did not result in significant improvements in
blood glucose [15]. Another study found that it was feasible to use interactive
voice response (IVR) phone calls to support DSM among Spanish-speaking
patients in the US and Latin America [16]. However, the effectiveness of the
intervention was not reported.

Furthermore, interventions delivered via phones tend to simultaneously address
multiple aspects of DSM [11], [16]. Perhaps it may be more sensible to design
interventions that focus on the behaviors that are most challenging for the target
population. There is evidence that Latino adults perceive PA behavior change as
one of the most difficult aspects of DSM [17]. Indeed, only 28 percent of Latino
72
adults with diabetes report getting the recommended levels of PA [18]. This is
troubling given the strong evidence linking regular PA to improved blood glucose
control and insulin sensitivity among individuals with diabetes [19]–[25].

Lastly, while DSM interventions delivered via phone calls or text messages have
generally been successful at reaching and engaging individuals with diabetes,
they have failed to do the same with the FF of these individuals. Latino adults
with diabetes desire an increase in support from family members [17]. Family
members in turn are eager to provide support, but lack knowledge on how to do
so. Given the important role of FF in DSM, there is an opportunity for technology-
based interventions to address these unmet needs. The DSM intervention
delivered via IVR phone calls that was previously mentioned also involved calling
or sending an e-mail to FF to notify them of patients’ health status and/or to
instruct them how to provide support [16], [26]. However, the impact due to FF
participation was not reported.

Given the potential of phone-based interventions to reach and engage both
patients and FF, we investigated the feasibility, perceived usefulness, and
potential effectiveness of a short text or voice message (ST/VM) intervention to
activate 1) PA behavior change among urban, low-income, Latino patients in
diabetes management and 2) supportive behaviors by their FF.

73
Methods

The Health Sciences Review Board at the University of Southern California approved all
study procedures.
Study Design

We conducted a pilot study in which 42 participants were randomized with equal
probability into one of three study arms: control, phone messaging (PM), and phone
messaging plus social support from FF (PM+FF). The sample size was based on a
recommendation of 12 participants per group in pilot studies [27], and accounting for a
16 percent dropout rate experienced in one of our previous studies at the study site. An
online statistical computing web program (www.graphpad.com/quickcalcs) was used to
generate the randomization schedule. The study consisted of a 7-day baseline period
during which participants were asked not to change their walking habits, followed by a
12-week intervention period. Follow-up interviews were conducted at baseline, 6 weeks,
and 12 weeks. Participants received no monetary incentive for study enrollment. They
received a $10 gift card for each of follow-up interview they completed. Participating FF
received a $10 gift card at the end of the study.
Recruitment

Participants were recruited from the diabetes management program (DMP) at Edward R.
Roybal Comprehensive Health Center in East Los Angeles, which is part of the Los
Angeles County Department of Health Services Ambulatory Care Network – the second
largest safety-net care system in the US. The DMP provides team-based comprehensive
care to prevent diabetes-related complications among high-risk patients. After routine
clinic visits, DMP clinicians who were aware of the nature of the study and the eligibility
74
criteria referred patients to an English-Spanish bilingual research assistant (RA) located
in the clinic. The RA briefly explained the study to these patients. If a patient was
interested in learning more, the RA explained the study in detail. The RA then screened
for eligibility those patients who were interested in participating.
The inclusion criteria were: age greater than or equal to 18 years, diagnosis of type 2
diabetes (T2D) based on self-report, no medical conditions restricting patient from
beginning a walking program (judged by DMP clinicians prior to referral), preferred
language of English or Spanish, self-identifies as a Latino, ability to walk without the use
of assistive devices such as canes or walkers, available to attend three interviews at the
clinic, does not plan to move away from the region or be out of the country during the
next three months, and has a working mobile phone or landline phone where they can
receive regular ST/VMs for three months. Pregnant or breastfeeding women were
excluded from participating.

Eligible patients provided written informed consent. They were sent home with an
informational flyer to give to a FF who would participate in the study with them if need
be. The flyer explained the purpose of the study as well as eligibility, expectations, and
compensation for the FF. Participants were also given an Omrom HJ-321 Tri-Axis
Pedometer. This pedometer stores seven days of data and automatically resets to zero
at midnight. The RA explained to participants what a pedometer is, how to wear it, and
then instructed them to wear it throughout the day, except for sleeping and bathing. The
pedometer screen was set to display the time, and the RA instructed participants not to
change the screen display. They were instructed to record in a log each day the time
75
they put the pedometer on and the time they took it off. Finally, they were instructed not
to change their walking habits in the next seven days.

After the 7-day baseline period, participants returned to the clinic for the baseline (week
0) interview. The RA collected the log and pedometer. Participants were dismissed from
the study if there were less than three consecutive days of data stored in the pedometer
with at least 10 hours of self-reported pedometer wearing per day [28], the average
number of steps per day exceeded 8800 (indicating participant was sufficiently active)
[29], or if the participant was unable to secure a FF who was willing to participate in the
study with them.

The RA returned the pedometer to all participants continuing the study. They were
instructed to continue wearing it throughout the day for the next 12 weeks. The RA
provided verbal and written instructions for how to read daily step counts and review step
counts of the previous seven days. Participants also received a set of weekly walking
logs to record what time they put the pedometer on, what time they took it off, and the
number of steps they took each day.
Interventions

The intervention designs were informed by the results from our study of patient
preferences. We administered a discrete choice experiment (DCE) survey to 125
patients in the DMP. We selected five attributes to describe a text message-based
intervention: PA goal setting, feedback on PA performance, PA behavior change
education, social support, and frequency of text messaging. The first four attributes were
identified from the literature; each is linked to a behavior change theory and is supported
76
by evidence to be effective at changing PA behavior. Each attribute had two
options/levels that were identified from a combination of literature reviews, pilot testing,
and discussions among the authors. Based on the different combinations of the five
attributes and their two levels, there were 32 possible intervention design alternatives.

In the DCE survey, patients were presented with a hypothetical scenario comprised of
two intervention design alternatives and were asked to choose which one they preferred.
Each patient evaluated seven such pairs. After analyzing the responses using
conditional logistic regression, we were able to identify the intervention design that was
most preferred: patients are recommended PA goals (versus the alternative of self-
selecting personalized goals), they receive feedback based on their individual PA
performance (versus the alternative of performance being compared to patient peers),
clinicians recommend the content of the PA behavior change education (versus the
alternative of self-selecting desired content), family members learn how to offer support
(versus the alternative of engaging in patient peer support), and clinicians recommend
the frequency of messaging (versus the alternative of self-selecting desired frequency).

Thus, in this study, participants were recommended to use the pedometer and walking
log to self-monitor how many steps they walked each day. They were recommended to
gradually increase their daily steps over the course of 12 weeks until they reached
10,000 steps per day. They were also recommended to walk at a brisk pace for 3000 of
these steps, which roughly translates to 30 minutes per day. These recommendations
were contained in the walking logs that were handed to participants (including those in
the control arm).
77

Participants in the PM and PM+FF arms received ST/VMs throughout the 12-week
intervention period. Only participants in the PM+FF got to participate in the study with
the FF they identified at the baseline interview. FF also received ST/VMs throughout the
12-week study. Each person indicated whether they wanted to receive voice or text
messages and their preferred language (English or Spanish). The standard schedule of
message delivery days is presented in Table 1. All messages were sent at 9:00 a.m.
The standard schedule was adjusted if someone indicated a day or time conflict.

Table 1 Weekly schedule of ST/VMs
Recipient Sunday Tuesday Thursday Saturday
Participants in
PM and PM+FF
arms
Reminder to
review goals and
self-monitor
Educational
message
Educational
message
Request to
report on PA
performance

Feedback on PA
performance (if
participant
replied)
FF of
participants in
PM+FF arm

Educational
message
Educational
message

Once per week, participants received the same ST/VM that reminded them to review
their daily step goals for that week and to self-monitor their PA behavior using the
pedometer and walking log. Two times per week, participants received an ST/VM with
educational content. These educational messages, as well as those which were sent to
FF twice per week, were unique each time and were largely adapted from public material
available on the websites of the American Diabetes Association, National Institute of
Diabetes and Digestive and Kidney Diseases, and Healthy People 2020. The following
criteria were used to select the content for participants: it explained the benefits of
78
regular PA for individuals with T2D, the importance of regular PA to their daily lives, and
ways to overcome commonly identified barriers to PA. The criteria were based on
recommendations by Marquez and McAuley [30] for PA programs targeting Latinos. The
content for FF was derived based on supportive behaviors by FF that are perceived as
supportive by individuals making PA behavior changes [31]. Furthermore, once per
week, participants received the same ST/VM asking them to report on their PA
performance. If participants replied to this ST/VM, then they received another ST/VM
that provided feedback based on their response. These feedback ST/VMs were unique
each time. Examples of all ST/VMs are displayed in Table 2.

Table 2 Examples of ST/VMs
Recipient ST/VM Type Example
Participants
in PM and
PM+FF arms
Reminder to review
goals and self-monitor
Remember to review your daily step goals, wear your
pedometer, and fill out your walking log.
Educational message
Brisk walking can lower your blood sugar and improve
your A1C. Your doctor may instruct you to take fewer
diabetes pills or less insulin.

Brisk walking will leave you feeling better so you can do
activities you enjoy, such as spending quality time with
family/friends.

Walk first thing in the morning before your day gets too
busy. If you don’t have 30 minutes, look for three 10-
minute periods.
Request to report on
PA performance
How well did you do with your daily step goals in the
past 7 days? Reply with a number from 1 (not well at all)
to 5 (excellent).
Feedback on PA
performance
If response was 1, 2, or 3: Walking needs to be a
regular habit to produce benefits. Make an effort to
improve your walking in the next 7 days.

If response was 4 or 5: Great! Keep up your hard work,
and you will see that it will pay off. Increase your daily
goal by 1000 steps.
FF of
participants in
PM+FF arm
Educational message
Brisk walking can help lower the patient's blood sugar to
keep diabetes under control. Offer your support by
joining them on a brisk walk as often as you can.

79
Each participant could receive up to five ST/VMs per week. This was based on the
notion that the frequency of messaging should match the frequency of the expected
behavior (i.e., five times per week for PA) [32]. We chose to send FF two ST/VMs per
week with the intention that they would in turn perform supportive behaviors at least this
many times weekly.

All messages were written in English, and then translated to Spanish by a native
Spanish speaker. A second native Spanish speaker reviewed the translations for
accuracy. All voice messages were voice recorded in English and Spanish a native
Spanish-speaking woman who did not have contact with participants. We used Call-Em-
All to deliver voice messages and Google Voice to deliver text messages. Originally, we
used Call-Em-All to deliver both types of messages. Two weeks after recruitment began,
we switched all text message delivery to Google Voice because multiple participants
encountered technical problems when trying to opt in to receive the messages. A
representative from Call-Em-All informed us that some mobile phone carriers do not
support their platform. From participants’ perspective, the only difference after the switch
was that they received text messages from a different phone number. The voice
messages in Call-Em-All were scheduled in advance and were delivered automatically.
The text messages in Google Voice were manually delivered each day. A RA
continuously monitored both platforms and manually delivered feedback ST/VMs when
participants replied to ST/VMs requiring a response.

For participants who opted to receive voice messages, if Call-Em-All was unable to
reach a live answer or a voicemail, it continued to attempt three additional times. Eleven
80
weeks after recruitment began, we changed the number of retries to one. At this point in
the study, there were five participants who had been receiving voice messages. The
reason for the change was because some participants reported receiving an unusually
high amount of calls per week. Given that we were sending four to five voice messages
per week with three retries, we did not anticipate that if participants ignored calls or did
not have a voicemail set up on their phone, the number of calls could translate to up to
20 per week.
Study Measures

Interviews were not audio recorded. For unstructured questions, the RA wrote
participants’ responses. Spanish responses were translated to English. Interview
questions are available in the Appendix.
Process Feasibility

Process feasibility was assessed in terms of the following: sufficiency or restrictiveness
of eligibility criteria, recruitment and retention rates, non-compliance with wearing
pedometer as instructed, and length of time for interviews. If participants did not wear
pedometers for at least 10 hours per day on three consecutive days in the week leading
up to the 6- and 12-week follow-up interviews, then they were considered to be non-
compliant. Pedometers and walking logs were retrieved at the 6- and 12-week
interviews. The RA accessed the 7-day data storage in the pedometers to obtain the
daily step counts for the preceding seven days. Walking logs were used to determine
how many hours they wore the pedometers each day. If participants did not log this
information or if they forgot to bring their walking logs to the interview, they were asked
to recall the times they put the pedometer on and took it off in the preceding seven days.
81
Technical Feasibility

Technical feasibility was assessed in terms of the following: receipt of ST/VMs,
engagement with ST/VMs requiring a response, barriers to receipt of and engagement
with ST/VMs, and perceived usability of pedometers. Receipt of text messages was
based on self-report at 6- and 12-week follow-up interviews since Google Voice does not
provide these data. Receipt of voice messages was obtained from data provided by Call-
Em-All. We considered a voice message as received if the call reached a live answer or
a voicemail. We considered a participant to be engaged each time they responded with a
number between 1 and 5 to the messages asking them to report on their PA
performance. The aforementioned were examined in order to identify patterns that would
elucidate barriers to the receipt of and engagement with ST/VMs. In addition,
participants were asked unstructured questions during 6- and 12-week follow-up
interviews with the aim of uncovering such barriers. Participants were also asked
unstructured questions regarding the extent to which they perceived that using
pedometers was free of effort.
Perceived Usefulness

At the 6- and 12-week follow-up interviews, participants were asked a series of
unstructured questions regarding the extent to which they perceived that the program
enhanced their ability to make PA behavior changes. These questions inquired about
their thoughts on setting PA goals, self-monitoring, educational and feedback ST/VMs,
and the idea of using ST/VMs to communicate with patients about PA behavior change.
For participants in the PM+FF arms, the questions also inquired about supportive
behaviors exhibited by FF since the start of the program and their thoughts on the idea
of using ST/VMs to communicate with FF about patients’ PA behavior changes. All
82
participants were asked if they would be willing to participate in a similar program in the
future and if they would be willing to recommend the program to other patients. Both
were yes/no questions.
Potential Effectiveness

The following outcomes were used to assess potential effectiveness: daily step count,
body mass index (BMI), perceived FF social support, barriers self-efficacy, and exercise
self-efficacy. Participants received assessments at baseline, 6 weeks, and 12 weeks.
Daily step counts were obtained from the pedometer 7-day data storage. BMI was
calculated as weight (in kilograms) over height squared (in centimeters). A modified
version of the Social Support and Exercise Survey (SSES) [32] was used to measure
perceived FF social support. The original SSES asked individuals to evaluate how often
they perceived supportive behaviors separately from family and friends. In our study, we
asked participants to evaluate perceived social support from family and friends together.
Participants responded on a Likert scale from one (none) to five (very often), with higher
numbers indicating a higher perceived FF support. Likewise, we used modified versions
of the Barriers Self-Efficacy Scale (BSES) [33] and Exercise Self-Efficacy Scale (ESES)
[34]. The original scales asked individuals to indicate their level of confidence from zero
(not at all confident) to 100% (highly confident) in increments of 10. We changed the
response options to a four-point Likert scale ranging from one (not at all confident) to
four (very confident) with higher numbers indicating higher self-efficacy. In addition, we
changed the exercise frequency indicated in the BSES prompt from three times per week
to five times per week. We also changed the exercise duration indicated in the ESES
prompt from 40+ minutes per day to 30+ minutes per day.
83
Analysis
Qualitative Analysis

Participants’ responses to open-ended questions were analyzed separately at 6 and 12
weeks. They were also analyzed separately for the PM and PM+FF arms. The first
author categorized responses into predefined themes: barriers to receipt of ST/VMs,
barriers to engagement with ST/VMs, pedometer usability, ongoing behavior change
support, self-monitoring, goal-setting, self-reward, self-instruction, reporting, feedback,
unintended effects, and social support. All but the last theme was further divided into
subthemes that were derived from the responses of the participants. Social support was
divided into the subthemes that represent the broad types of supportive behaviors [35]:
emotional support, instrumental support, informational support, and appraisal support.
Exemplary responses from participants were reported. The second author checked the
results of the interview analysis for accuracy, and discrepancies were resolved via
discussion. Results from both follow-up periods and both intervention arms were
presented collectively. When we noted differences in results by follow-up period or by
study arm, these differences were reported.
Quantitative Analysis

ANOVA and Fisher’s Exact Test were used to identify differences in participant
characteristics among the three study arms. For each participant’s daily steps, we
computed average daily steps for consecutive days (at least three) wherein they wore
the pedometer for at least 10 hours per day. For each participant’s perceived FF support,
exercise self-efficacy, and barriers self-efficacy, we computed the average points across
items. Means and standard deviations for the three study arms were calculated for each
outcome measure. ANOVAs were used to identify differences in means among study
84
arms and paired t-tests were used to identify differences in means within study arms. In
addition, t-tests were used to assess the difference in the differences in means between
study arms. All analyses were conducted at the 0.05 significance level using SAS
University Edition.

Results
Process Feasibility

Participants were recruited from April to August 2015 and follow-up interviews were
conducted until November of 2015. Figure 1 depicts the flow of participants through the
study. Ninety-four percent (46/49) of potential participants who were screened were
eligible to participate. Two individuals did not meet the eligibility criterion of being
available to three interviews at the clinic because they did not have transportation (N=1)
or they would be working (N=1). Another individual was not eligible for the study because
they had plans to travel outside of the country in the next three months (N=1). Of the
potential participants who were eligible, 100 percent enrolled in the study.

After the 7-day baseline period, four participants were dismissed because their average
daily step counts were greater than 8800. Thus, 42 participants were randomized to a
study arm. The retention rates were 79% (11/14) for the control and PM arms, and 93%
(13/14) for the PM+FF arm. Three participants in the control arm were lost to follow-up:
one participant was no longer interested in participating and two did not respond to
contact attempts. In the PM arm, one participant was lost to follow-up due to a family
emergency, one was dismissed because of an injury unrelated to the study, and one was
dismissed for traveling outside of the country during the study. Only one participant in
85
the PM+FF arm was lost to follow-up after failed attempts to contact. Consequently, the
analysis of BMI, perceived FF social support, barriers self-efficacy, and exercise self-
efficacy included 11, 11, and 13 participants in the control, PM, and PM+FF arms,
respectively. Two participants in the PM+FF arm at the 6-week follow-up and one
participant in the control arm at the 12-week follow-up did not comply with wearing the
pedometer as instructed and so their daily step counts were not included in the analysis.
Thus, the analysis of daily steps included 10, 11, and 11 participants in the control, PM,
and PM+FF arms, respectively. The average duration of 6- and 12-week interviews were
16, 29, and 32 minutes for the control, PM, and PM+FF arms, respectively.
86

Figure 1 Participant flow diagram

While clinicians at Roybal DMP screened potential participants for inclusion criterion of
no medical conditions restricting patient from beginning a walking program, at the follow-
up interviews, three participants (one in PM and two in PM+FF) reported having medical
conditions precluding them from improving their PA behavior. These included a heart
condition, severe arthritis, and a hurt back from a previous car accident. In addition, two
Assessed for eligibility (n=49)
Enrolled (n=46)
Randomized (n=42)
Dismissed (n=4)
• Walked >8800 steps/day
during 7-day baseline period
Enrollment
Allocated to Control (n=14) Allocated to PM (n=14) Allocated to PM+SS (n=14)
Allocation
Lost to follow-up (n=3)
• No longer interested in
participating (n=1)
• Did not respond to contact
attempts (n=2)
Lost to follow-up (n=1)
• Family emergency (n=1)
Dismissed from study (n=2)
• Experienced injury
unrelated to study (n=1)
• Left the country (n=1)
Lost to follow-up (n=1)
• Did not respond to contact
attempts (n=1)
Follow-Up
Analyzed for BMI, social
support, and self-efficacy
(n=11)
Analyzed for daily steps
(n=10)
Analyzed for daily steps, BMI,
social support, and self-
efficacy (n=11)
Analyzed for BMI, social
support, and self-efficacy
(n=13)
Analyzed for daily steps
(n=11)
Analysis
Excluded (n=3)
• Not available for three
interviews at clinic (n=2)
• Will not be in L.A. area (n=1)
87
participants – one in each intervention arm – reported at the follow-up interviews having
experienced knee or ankle injuries as result of changes in their PA behavior.
Participant Characteristics

The characteristics of participants are presented in Table 3. There were no significant
differences among the study arms. The majority of participants were female (67 percent),
preferred to speak Spanish (76 percent), and owned a phone with the capacity to send
and receive text messages (86 percent). About half of participants had less than a high
school education (52 percent). On average, participants had been diagnosed with
diabetes for 12 years. Nearly 40 percent of participants in the PM and PM+FF study
arms opted to receive voice messages instead of text messages.

Table 3 Participant characteristics
Characteristics Total (n=42)
Control
(n=14)
PM (n=14)
PM+FF
(n=14)
Female, n (%) 28 (66.7) 9 (64.3) 7 (50.0) 12 (85.7)
Age, mean (SD) 52.3 53.2 (8.2) 53.0 (8.9) 50.4 (8.7)
Spanish as preferred
language, n (%)
32 (76.1) 10 (71.4) 12 (85.7) 10 (71.4)
Educational attainment
Less than high school n (%) 22 (52.4) 8 (57.1) 5 (35.7) 9 (64.3)
High school graduate n (%) 4 (9.5) 1 (7.1) 2 (14.3) 1 (7.1)
More than high school n (%) 16 (38.1) 5 (35.7) 7 (50.0) 4 (28.6)
Self-reported years since
diabetes diagnosis (years),
mean (SD)
11.6 (8.5) 11.6 (7.4) 11.4 (10.5) 11.9 (8.0)
Phone with text messaging
capability, n (%)
36 (85.7) 12 (85.7) 11 (78.6) 13 (92.9)
Opted to receive voice instead
of text messages, n (%)
11 (39.3) -- 6 (42.9) 5 (35.7)

Participants in the PM+FF arm invited immediate family, extended family, and friends to
participate in the study with them. The FF group included daughters (N=4), sons (N=2),
spouses (N=2), a parent (N=1), a sister (N=1), an aunt (N=1), a daughter-in-law (N=1), a
88
niece (N=1), and a friend (N=1). Only one FF (the only friend in the study) opted out of
the intervention. All FF opted to receive text messages instead of voice messages.
Technical Feasibility (PM and PM+FF Only)

At the 6- and 12-week interviews, all participants who opted to receive text messages
reported having received at least one text message from the study in the past six weeks.
On average, they reported receiving four to five text messages per week (reflected the
actual number of text messages sent). Figure 2 depicts weekly receipt of voice
messages for participants who opted to receive voice messages. On average,
participants were reached on 90 percent of the calls from weeks 0 to 6 and on 85
percent of the calls from weeks 7 to 12.

Figure 2 Participants’ receipt of voice messages

On average, throughout the 12-week study, participants who opted to receive text
messages responded to 63 percent of text messages requiring bi-directional interaction.
Those who opted to receive voice messages responded to 34 percent of voice
89
messages requiring a response. Individual, information, and system barriers made it
difficult for participants to reply to ST/VMs. Participants receiving text messages reported
that they did not understand the instructions (N=2) and that they did not know how to
reply to text messages (N=1). Participants receiving voice messages reported that they
were too busy to reply (N=2), they did not understand the instructions (N=2), and that
they were not given enough time to reply before the call ended (N=1). In addition, three
Spanish-speaking participants reported that they did not reply because they were
confused by an English message at the end of survey-type calls. These were
instructions that were automatically included by Call-Em-All. The instructions said, “To
give a numeric response, press the star key. To hear this message again, press 1. To
have your number placed on our Do Not Call list press three. To end this call press 9.” At
the time the study was being conducted, Call-Em-All was not able to play this message
in languages other than English.

Various device and individual factors made it difficult for participants to wear and use
pedometers to self-monitor their activity. Participants reported that pedometers
frequently fell off clips (N=3), and that they had difficulty navigating the various screen
display options (N=2). About 30 percent of participants lost their pedometer at least once
during the 12 weeks of the study. Participants also reported forgetting to wear the
pedometer (N=6), having difficulty reading the text on the screen because of their vision
(N=4), never learning to use the pedometer (N=3), and feeling physical discomfort while
wearing the pedometer on their waist (N=1). In addition, participants reported not
wanting to press the pedometer buttons for fear of “pushing the wrong button and
deleting something” (N=6).
90
Perceived Usefulness
Ongoing Behavior Change Support

Participants perceived ST/VMs to be a better alternative than going to the clinic for PA
behavior change support (due to work-related and transportation issues), and a good
way of being reminded to stay active in between clinic visits (N=5). One participant said,
“The doctor will tell me [during a clinic visit] to walk, but then we won’t discuss it again
until the next visit. I like that the messages constantly remind me.” Nonetheless, one
participant stated that it would have been helpful to periodically receive a call from a real
person.
Self-Monitoring

The pedometers and walking logs were useful tools for participants to self-monitor their
PA behavior. Most participants reported that wearing pedometers helped them to
objectively track their activity (N=21). They made statements such as, “The pedometer
helps me because it lets me know how much I am walking” or “Having a pedometer
keeps you from lying to yourself that you did walk enough.” Nine participants reported
regularly writing their steps in the walking logs. These individuals reported that doing so
helped them to keep a long-term record of their activity since pedometers only stored
seven days of data. This, in turn, allowed them to assess whether they were making
progress towards the long-term goal of 10,000 daily steps. For some participants, the
walking logs even helped them to identify patterns in their PA behavior. For example,
one participant said, “I like writing my steps because it lets me know what days I don’t
walk that much and what days I do walk a lot. I found out that on Saturdays, I walk the
most – that is because I go to parties and dance a lot.”
91
Goal-Setting

Overall, most participants reported that setting daily step goals motivated them to walk
everyday in order to achieve them (N=18); one person stated that she did not like having
goals because she felt “traumatized” if she could not meet them. In terms of the
recommended long-term goal of 10,000 steps per day, participants had mixed
perceptions. Some were satisfied with this goal (N=14), saying, “It’s a good idea
because it tells me what I need to work towards” or “Before, I didn’t know how many
steps to walk each day.” Other participants set an even greater goal (N=5), and still,
others felt that the goal was too ambitious (N=6). One person stated, “The 10,000 steps
goal is too much. I have to go walk, then rest, then walk, then rest. My back and my legs
hurt because of the arthritis.”
Self-Instruction

Information about their PA behavior prompted participants to self-instruct. After checking
pedometers for daily steps, participants reported that they would say to themselves that
they need to keep walking until reaching a certain step count, to try to walk more steps
the next day or in the upcoming week, and to think about how walking was going to
benefit them (N=9). One individual said, “When I don’t have enough steps, I tell myself
that I need to keep walking more. And when I don’t walk enough that day, I tell myself
that I need to walk more the next day.” Another stated that after looking at her steps in
the pedometer, she would say to herself, “Do I want to walk more or do I want to take
insulin?” This type of self-instruction motivated participants to engage in PA.

92
Reporting and Feedback

Participants stated that the ST/VMs asking them to report on their PA performance
prompted them to self-reflect and motivated them to improve their PA behavior in order
to provide a more favorable response the next time (N=3). And while one participant was
indifferent about the feedback messages received after reporting on their PA
performance, others reported that these messages motivated them to continue their
behavior change efforts and to increase their daily steps (N=9). Participants enjoyed
receiving positive feedback when they replied with a high number (N=6), saying, “It’s like
someone grading you, like you did good on your test.” In terms of the process for
reporting on PA performance, one participant mentioned that it was quick and easy to
reply with a single number instead of typing out a longer response. On the other hand,
another participant perceived this as a limitation. He was concerned that he would be
perceived as “lazy” after replying with a low number since he could not explain the
reason for his poor performance: an injury.
Instrumental Support

Most participants reported having received instrumental support. ST/VMs motivated
(N=17) and reminded participants to be active (N=22). One person described them as
“an alarm to go out and walk.” FF also provided instrumental support to participants in
the PM+FF arm by reminding or encouraging them to walk, asking if and/or how much
they have walked, or exercising with them (N=13). One person stated, “[My husband]
constantly asks me, ‘Did you walk already? If you haven’t, let’s eat dinner and then go.’
He walks with me, and then I forget that I am exercising because we begin talking.”
Other forms of instrumental support by FF were monitoring participants’ diet (N=2), and
creating opportunities for participants to walk by parking vehicles farther away from
93
destinations or by helping with household responsibilities (N=3). Speaking of her
husband, one participant said, “He helps me around the house so that I have time to
exercise... He tells me that he will watch our baby so that I could go walk with my
sister… I feel like our relationship has improved.”
Emotional Support

Only one participant in the PM+FF arm stated that her FF was a source of emotional
support – that is, she perceived her FF to care about her being physically active. On the
other hand, many participants in both the PM and PM+FF study arms perceived
emotional support from the receipt of ST/VMs from the program (N=13). One person
stated, “I like that someone is concerned and cares and takes the time to check on me. It
gives me more motivation. If someone is taking time to check on me, then I should take
time to walk.” Other participants similarly reported perceiving that the ST/VMs made
them feel like someone cared about them, in addition to perceiving that they were
interested in participants’ PA behavior change efforts and in helping them to have better
health.
Informational Support

Participants in both PM and PM+FF arms reported having received informational
support. Eight participants stated that the ST/VMs by the program taught them about the
benefits of PA for individuals with diabetes. One participant commented about PA, “I
thought it was just something you had to do, but I didn’t exactly know why you had to do
it... I didn’t know it was beneficial for my health.” Participants in PM+FF arm also
received informational support by their FF. They too talked to participants about why PA
94
was beneficial for their health, and gave them ideas of where and how to be active
(N=6).
Unintended Desirable Effects

Participants reported that by FF receiving phone messages, it made them reflect on their
own PA habits (N=5). FF had become motivated to be more active, and had made
changes in their PA behavior – mostly by joining participants on walks. The phone
messages also prompted FF to provide support not only to the study participant, but also
to other individuals in the family with diabetes (N=1). One participant stated, “His wife
also has diabetes, so he also motivated his wife to exercise more. He was helping both
of us.” Both of the aforementioned findings were unintended, but desirable effects of
including FF in the program.
Overall Satisfaction

Overall, participants were satisfied with the use of ST/VMs to activate PA behavior
change and supportive behaviors by their FF. At the 6- and 12-week follow-up
interviews, all participants reported that they would be willing to participate in similar
program in the future and that they would recommend the program to other patients.
Potential Effectiveness

Table 4 displays the baseline outcome measures, and the changes in these measures
for two time periods: weeks 0 to 6 and 6 to 12. There were no significant differences
among study arms at weeks 0, 6, and 12.

95
Table 4 Outcome Measures
Outcome
Measures
Control PM PM+FF
Wk 0,
Mean
(SD)
Mean Difference
(SD)
Wk 0,
Mean
(SD)
Mean Difference
(SD)
Wk 0,
Mean
(SD)
Mean Difference
(SD)
Wk 0
to 6
Wk 6
to 12
Wk 0
to 6
Wk 6
to 12
Wk 0
to 6
Wk 6
to 12
Daily Step
Count
3691
(892)
1915
(2308)*
-454
(1733)
3839
(1205)
2159
(1863)*
959
(2131)
4680
(2731)
1157
(1774)
192
(1988)
BMI
36.3
(5.7)
-0.1
(0.8)
0.2
(1.0)
34.7
(3.9)
0.2
(0.7)
-0.6
(1.5)
35.6
(10.3)
0.0
(1.2)
0.6
(0.6)**
Perceived
FF Social
Support
2.4
(0.9)
0.2
(0.6)
0.0
(0.6)
2.9
(0.9)
0.1
(0.8)
-0.1
(0.4)
2.9
(0.7)
0.4
(0.8)
0.1
(0.5)
Barriers
Self-Efficacy
2.7
(0.5)
0.2
(0.5)
0.1
(0.2)
2.9
(0.5)
-0.1
(0.4)
0.1
(0.3)
2.7
(0.5)
0.0
(0.6)
-0.1
(0.4)
Exercise
Self-Efficacy
3.2
(0.4)
-0.5
(0.7)
0.3
(0.6)
3.4
(0.6)
-0.2
(0.7)
0.0
(0.4)
3.1
(0.7)
-0.3
(0.8)
0.1
(0.7)
P<.05*
P<.01**
P<.001***

Daily Steps

From weeks 0 to 6, the control, PM, and PM+FF arms increased their daily walking by
1915 (SD=2308), 2159 (SD=1863), and 1157 (SD=1774) steps, respectively. The
increase was significant within the control arm (P=0.03) and the PM arm (P=0.02). The
PM arm continued to increase their daily walking from weeks 6 to 12 by 959 (SD=2131)
steps, and the PM+FF arm by 192 (SD=1988). Meanwhile, the control arm decreased
their steps by 454 (1733). For both time periods, the difference in the differences in daily
steps was not significant.
Body Mass Index

BMI within the control and PM arms was stable throughout the 12-week study. However,
BMI within the PM+FF arm significantly increased by 0.6 (SD=0.6, P=0.005) from weeks
6 to 12. The difference in the differences in BMI from weeks 6 to 12 between the PM and
PM+FF arms was statistically significant (P=0.0156).
96
Perceived FF Social Support

Perceived FF social support increased from weeks 0 to 6 within the control, PM, and
PM+FF arms by 0.2 (SD=0.6), 0.1 (SD=0.8), and 0.4 (SD=0.8), respectively. From
weeks 6 to 12, the perception of social support decreased for the PM arm (mean
difference=-0.1, SD=0.4), remained same within the control arm (mean difference=0.0,
SD=0.6), and increased for the PM+FF arm (mean difference=0.1, SD=0.5). For both
time periods, the difference in the differences in perceived FF social support was not
significant.
Self-Efficacy

Barriers self-efficacy increased within the control arm from weeks 0 to 6 (mean
difference=0.2, SD=0.5) and weeks 6 to 12 (mean difference=0.1, SD=0.2). For the PM
arm, it decreased from weeks 0 to 6 (mean difference=-0.1, SD=0.4), but then increased
by the same amount from weeks 6 to 12 (mean difference=0.1, SD=0.3). The PM+FF
arm experienced no change in barriers self-efficacy from weeks 0 to 6 (mean
difference=0.0, SD=0.6), and a decrease from weeks 6 to 12 (mean difference=-0.1,
SD=0.4).

From weeks 0 to 6, there was a decrease in exercise self-efficacy within the control
(mean difference=-0.5, SD=0.7), PM (mean difference=-0.2, SD=0.7), and PM+FF arms
(mean difference=-0.3, SD=0.8). It did not change for the PM arm from weeks 6 to 12,
but increased for the control (mean difference=0.3, SD=0.6) and the PM+FF arms (mean
difference=0.1, SD=0.7). From weeks 0 to 6 and 6 to 12, the difference in the differences
in both measures of self-efficacy was not statistically significant (P>.05).

97
Discussion
Process Feasibility
Our recruitment and retention was suitable and comparable to recent studies of text
messaging interventions targeting adults with diabetes [15], [26], [36], [37]. Given that
the target population is greatly underrepresented in related literature [11], [32], [38], [39],
our finding adds to the evidence base that participation by urban, low-income, Latino
adults with T2D in eHealth research can be effective [15].

Moreover, our results indicate that adjustments to the screening process may be needed
given that some participants perceived physical limitations precluding changes in PA and
others experienced injuries as a result of changes in PA. Participants were recruited
from a diabetes management program, which is reserved for patients with uncontrolled
diabetes. Based on our previous research studies conducted with the target population,
it would not be uncommon for these individuals to have comorbid chronic conditions
and/or advanced secondary complications of diabetes that would make them susceptible
to experiencing the aforementioned issues. While these types of individuals have tended
to be excluded from participating in studies of walking interventions [40], [41], we chose
not do so because they represent individuals who are in most of need of ongoing
diabetes self-management support. Future research is needed to develop a protocol that
will effectively identify individuals who are able to safely engage in unsupervised,
predominantly outdoor brisk walking. Once enrolled in the study, long-term daily step
goals can be individually tailored according to abilities. In addition, greater emphasis can
be placed on instructing participants to start slowly, based on their abilities and fitness
levels, and to gradually increase activity over several weeks.
98
Technical Feasibility

Engagement with messages requiring a response was high among participants who
opted to receive text messages, and was comparable to what was reported in studies
examining the use of text messaging to support self-management in a diabetes
population [36] and weight loss in a general population [42]. Our results thus indicate
that urban, low-income, Latino adults with T2D are willing to engage with interventions
leveraging this type of communication modality. We also found that engagement varied
by mode of message delivery: participants who opted to receive text messages were
better engaged than those who opted to receive voice messages. A study of a chronic
disease self-management intervention for Spanish-speaking patients delivered using IVR
phone calls reported high levels of participant engagement [16]. Participants in our study
who received voice messages may have been willing to reply to messages requiring a
response, but the platform we used to deliver the voice messages may have created
barriers that made it difficult for them to do so. Namely, important settings could not be
customized: time given for participants to reply and the language of the instructions for
replying to survey-type calls. If using this or a similar type of service for delivering voice
messages, our results suggest that text messaging is a better mode of delivery for
ensuring patient engagement. Given that most participants reported having a phone with
text messaging capability, a larger study can restrict the mode of delivery to text
messages and teach participants how to send and receive them.

In addition, we found that participants lost pedometers and were frustrated by poor
attachment clips. Others had trouble learning how to use pedometers or had trouble
reading the steps because of the font size. Some participants wore pedometers
99
regularly, but were afraid of touching the buttons. To facilitate pedometer wearing and
use, future studies can select a pedometer (or any activity tracker) with a better
attachment mechanism for preventing its loss and with a larger font size to
accommodate individuals who have vision impairments. Although guidelines for using
pedometers in research studies state that middle-aged and older adults have few
problems using pedometers when given simple instructions [43], our findings suggest
that urban, low-income, Latino adults with T2D may require a different, perhaps more
interactive, approach for learning to use the pedometers so that they feel comfortable
navigating them and using them without fear.
Perceived Usefulness

Our results suggest that guidance in self-regulation may be a useful mechanism for
supporting PA behavior change via ST/VM. The main themes of the interviews touched
on five of six self-regulatory processes [35]: self-monitoring, goal setting, self-instruction,
feedback, and social support. Pedometers and walking logs facilitated systematic
observation and recording of daily steps; daily step goals motivated individuals to
change their PA behavior in order to achieve the success they desired; information about
their PA behavior prompted individuals to “talk to themselves” about plans for increasing
their steps; and feedback provided via ST/VMs motivated them to continue their behavior
change efforts. ST/VMs and FF (for PM+FF) were perceived as sources of social
support. They had multiple functions that were perceived as useful for supporting the
behavior change process: 1) prompting participants to be physically active and being
joined by their FF in walking sessions (instrumental support), 2) feeling cared for
(emotional support), and 3) learning about the benefits of PA as well as where/how to be
active (informational support).
100
There is strong evidence that social support affects engagement in PA [44]–[50] and
chronic disease self-management [12]. There is an unmet need for studies to identify
disease-specific family behaviors that positively influence adult patient outcomes [51].
Our results identified FF behaviors that were perceived as useful and have the potential
to be effective at improving self-management and clinical outcomes among urban, low-
income, Latino adults with T2D. We also demonstrated that these FF behaviors could be
prompted using a ST/VM-based intervention. Future research can examine whether
these specific FF behaviors make a significant impact. Moreover, while much of the
discussion of social support for chronic disease self-management focuses on the role of
family and peers, our findings add to the growing body of evidence that technology-
based self-management interventions may also be perceived as sources of social
support. For example, one study found that participants perceived an automated text
message-based diabetes intervention as a “friend” that cared for them [52].

Although self-regulation theory was not developed with the direct intention of being used
with technology, our study has demonstrated that it can be successfully applied and
perceived as useful in a ST/VM-based intervention to support PA, and perhaps for other
health behavior changes as well. A recent systematic review examining the impact of
information technology on behavior change for various health conditions found that only
30 percent (14/40) of interventions delivered via phones explicitly reported having used a
behavior change theory as a guide for intervention design [38]. Among those that did
report using a theory, none used self-regulation. Nine studies did, however, report using
cognitive theory, which may have overlapping constructs with self-regulation.
Nonetheless, the possibility that self-regulation is not being leveraged to design phone-
101
based interventions is troubling given that meta-regression analyses of interventions to
promote PA and healthy eating have found that the use of self-regulation techniques
(i.e., goal setting, self-monitoring, and feedback) are associated with positive outcomes
[53], [54]. Future studies may use our intervention characteristics – which are aligned
with self-regulation theory and patient preferences – to inform the design of phone-based
interventions.
Potential Effectiveness

Evidence for potential effectiveness was mixed. We expected only PM and PM+FF to
improve their daily steps (the latter group showing greater improvements given the
enhanced support). We found that the control arm also increased their steps, but only in
the first 6 weeks of the study. The 10,000 steps per day recommendation coupled with
use of a pedometer and walking log may have been a sufficient intervention. There is
evidence that using a pedometer leads to increases in PA [55]. Alternatively, the care
received at the DMP may have unintentionally been part of the intervention. Given that
the DMP provides comprehensive care for preventing complications, it would not be
uncommon for clinicians to discuss and promote PA during routine clinic visits.

In regards to BMI, we did not expect to see changes in the 12 weeks. Walking
interventions for adults with diabetes have not reported significant changes in BMI within
this follow-up period [40], [41], [56]. We found a significant increase in BMI within the
PM+FF arm. We hypothesize that this increase could be a result of unintended adverse
types of social support, such as rewarding participants with unhealthy food, although
future research is needed to confirm this. In the meantime, ST/VMs can be modified so
that FF are prompted to encourage healthy eating and PA.
102

Furthermore, there were net improvements in perceived FF social support within the
control and PM+FF arms even though we only expected to observe improvements within
the latter. Nonetheless, the improvement in perceived FF social support within the
PM+FF arm was nearly two times greater than that of the control arm. The improvement
in perceived FF social support within the PM+FF arm did not appear to result in greater
daily steps. Family members or close friends are rarely included in technology-based
diabetes self-management interventions. To our knowledge, there are only two published
studies in which FF learned, via an IVR call or e-mail, what to do to support patients’
diabetes self-management [16], [26]. However, participants in these studies got to
choose whether they wanted to participate with a FF. In addition, neither study reported
measuring changes in perceived FF social support. One of them reported that FF
participation was significantly more common among individuals with lower income or
inadequate health literacy [26]. Our results indicate that using technology to prompt
social support behaviors has the potential to improve perceived FF social support among
urban, low-income, Latino adults with T2D. However, our results also suggest that this
perception may not necessarily translate to improved PA outcomes. Future research is
needed to investigate what specific behaviors by FF are effective at helping individuals in
the target population improve their PA.

Finally, although the self-regulation techniques used in the intervention may be similar to
those for increasing self-efficacy, we did not directly target and thus did not expect to
observe changes in self-efficacy. The control arm, however, increased their barriers self-
efficacy at each follow-up period. A study of individuals with T2D also found that self-
103
monitoring of PA using pedometers and walking logs enhanced barriers self-efficacy
[57]. In this study, it is not clear why we observed an improvement in self-efficacy only
within the control arm and not the intervention arms given that pedometers and walking
logs were distributed to everyone. Future research is needed to understand the reason
for the improvement. Moreover, all study arms experienced a net decrease in exercise
self-efficacy. In a study of a PA program, Speck and Looney [58] similarly observed a
decrease in self-efficacy in both control and intervention arms. We speculate that most
participants in our study did not have a regular PA routine prior to beginning the
program. Thus, without experience, they may have overestimated their confidence in
their ability to exercise regularly without quitting. As the study progressed, participants
may have developed a more realistic sense of their abilities, which would explain the
observed decrease in self-efficacy. Future research is needed to test this hypothesis.
Limitations

The first limitation of the study was the small sample size. Although the sample size was
based on a recommendation for pilot studies [27], we did not expect that the standard
deviations of the outcomes would be as great as what was observed. For this reason,
the findings cannot be generalized to the broader population of urban, low-income,
Latino adults with T2D based on this study alone. The second limitation is that the
randomization process resulted in the PM+FF arm having participants with more daily
steps at baseline compared to the other arms. The results of this imbalance may explain
the observation of a significant increase in daily steps within the control and PM arms,
but not the PM+FF arm. Another limitation is that our results of potential effectiveness
may have been negatively affected by the five participants in the interventions arms that
reported during follow-up interviews having physical limitations precluding changes in PA
104
or experiencing injuries as a result of changes in PA. A subgroup analysis excluding
these individuals (not presented) resulted in greater improvements in daily steps among
the interventions arms. A fourth limitation of the study is that the majority of participants’
perceptions about the usefulness of the program were positive, which may indicate
social desirability bias. However, the fact that between the PM and PM+FF arms, there
was only one participant that was lost to follow-up because they stopped responding to
our contact attempts (compared to three in the control arm) makes us confident that our
results may indeed represent participants’ perceptions. A final limitation is that we did
not interview FF to investigate their actual supportive behaviors, but instead relied on
perceptions of support by participants in the PM+FF arm. Nonetheless, others have
reported that perceived social support has better effects on health outcomes than actual
social support [59].
Conclusions

This study demonstrates the potential of using ST/VM to support PA behavior change
among urban, low-income Latino adults with T2D and to prompt social support from their
family members and close friends. The process of conducting the study is generally
feasible, however, adjustments to the screening methods are needed in order to avoid
enrolling participants who cannot safely engage in unsupervised, predominantly outdoor
brisk walking. Text messaging may be a better mode of message delivery than voice
messaging for ensuring participant receipt of and engagement with messages.
Pedometers can successfully be used by investigators for data collection purposes and
by participants for self-monitoring, although adjustments to instructions are needed so
that participants feel more comfortable using it without fear. Moreover, designing a
ST/VM intervention based on self-regulation techniques (i.e., self-monitoring, goal
105
setting, self-instruction, feedback, and social support) is feasible and perceived as useful
by participants. Finally, such an intervention may improve PA in terms of daily steps as
well as perceived social support from FF that participate in the intervention.

Acknowledgements

This project was supported with a Research Supplement to Promote Diversity in Health-
Related Research from the National Institute of Neurological Disorders and Stroke,
National Institutes of Health, U54NS081764. It was also supported with a grant from the
Daniel Epstein Institute at the University of Southern California. We acknowledge the
patients, staff, and providers at the Roybal Comprehensive Health Center who
participated in the study.
Abbreviations

BMI: body mass index
BSES: Barriers Self-Efficacy Scale
DCE: discrete choice experiment
DMP: diabetes management program
DSM: diabetes self-management
ESES: Exercise Self-Efficacy Scale
FF: family members and close friends
IVR: interactive voice response
PA: physical activity
PM: phone messaging
PM+SS: phone messaging plus social support from FF
106
RA: research assistant
SSES: Social Support and Exercise Survey
ST/VM: short text or voice message
T2D: type 2 diabetes
US: United States
References
[1] American Diabetes Association. “Standards of medical care in diabetes--2014,”
Diabetes Care, vol. 37, pp. S14–S80, 2014.
[2] K. Brunisholz, P. Briot, and S. Hamilton, “Diabetes self-management education
improves quality of care and clinical outcomes determined by a diabetes bundle
measure,” J Multidiscip Heal., vol. 7, pp. 533–542, 2014.
[3] R. Weaver, B. Hemmelgarn, and D. Rabi, “Association between participation in a
brief diabetes education programme and glycaemic control in adults with newly
diagnosed diabetes,” Diabet Med, vol. 31, no. 12, pp. 1610–1614, 2014.
[4] A. Steinsbekk, L. Rygg, M. Lisulo, M. Rise, and A. Fretheim, “A group based
diabetes self-management education compared to routine treatment for people with
type 2 diabetes mellitus. A systematic review with meta-analysis,” BMC Heal. Serv
Res, vol. 12, no. 1, 2012.
[5] I. Duncan, C. Birkmeyeter, S. Coughlin, Q. Li, D. Sherr, and S. Boren, “Assessing
the value of diabetes education,” Can J Diabetes, vol. 33, no. 5, pp. 752–760,
2009.
[6] L. Fan and S. Sidani, “Effectiveness of diabetes self-management education
intervention elements: a meta-analysis,” Can J Diabetes, vol. 33, no. 1, pp. 18–26,
2009.
[7] S. Ellis, T. Speroff, R. Dittus, A. Brown, J. Pichert, and T. Elasy, “Diabetes patient
education: a meta-analysis and meta-regression,” Patient Educ. Couns., vol. 52,
no. 1, pp. 97–105, 2004.
[8] S. L. Norris, J. Lau, S. J. Smith, C. H. Schmid, and M. M. Engelgau, “Self-
management education for adults with type 2 diabetes: a meta-analysis of the
effect on glycemic control,” Diabetes Care, vol. 25, no. 7, pp. 1159–71, Jul. 2002.
[9] S. L. Norris, M. Engelgau, and K. Venkat Narayan, “Effectiveness of Self-
Management Training in Type 2 Diabetes: A systematic review of randomized
controlled trials,” Diabetes Care, vol. 24, no. 3, pp. 561–587, 2001.
107
[10] J. Piette, “Interactive behavior change technology to support diabetes self-
management: where do we stand?,” Diabetes Care, vol. 30, no. 10, pp. 2425–32,
Oct. 2007.
[11] A. K. Hall, H. Cole-Lewis, and J. M. Bernhardt, “Mobile Text Messaging for Health:
A Systematic Review of Reviews,” Annu. Rev. Public Heal., vol. 36, pp. 393–415,
2015.
[12] M. P. Gallant, “The Influence of Social Support on Chronic Illness Self-
Management: A Review and Directions for Research,” Heal. Educ. Behav., vol. 30,
no. 2, pp. 170–195, Apr. 2003.
[13] CDC, “CDC Health Disparities and Inequalities Report - United States, 2011.”
[14] M. Menchine, A. Vishwanath, and S. Arora, “Prevalence, health and demographics
of emergency department patients with diabetes,” West J Emerg Med, vol. 11, no.
5, pp. 419–422, 2010.
[15] S. Arora, A. L. Peters, E. Burner, C. N. Lam, and M. Menchine, “Trial to Examine
Text Message-Based mHealth in Emergency Department Patients With Diabetes
(TExT-MED): A Randomized Controlled Trial,” Ann. Emerg. Med., vol. 63, no. 6,
pp. 745–754, 2013.
[16] J. Piette, N. Marinec, E. C. Gallegos-Cabriales, J. M. Gutierrez-Valverde, J.
Rodriguez-Saldaña, M. Mendoz-Alevares, and M. J. Silveira, “Spanish-speaking
patients’ engagement in interactive voice response (IVR) support calls for chronic
disease self-management: data from three countries.,” J. Telemed. Telecare, vol.
19, no. 2, pp. 89–94, Feb. 2013.
[17] J. Hu, K. Amirehsani, D. C. Wallace, and S. Letvak, “Perceptions of barriers in
managing diabetes: perspectives of Hispanic immigrant patients and family
members.,” Diabetes Educ., vol. 39, no. 4, pp. 494–503, 2013.
[18] K. M. Nelson, G. Reiber, and E. J. Boyko, “Diet and Exercise Among Adults With
Type 2 Diabetes,” Diabetes Care, vol. 25, no. 10, pp. 1722–1728, 2002.
[19] S. Zanuso, a Jimenez, G. Pugliese, G. Corigliano, and S. Balducci, “Exercise for
the management of type 2 diabetes: a review of the evidence.,” Acta Diabetol., vol.
47, no. 1, pp. 15–22, Mar. 2010.
[20] R. J. F. Manders, J.-W. M. Van Dijk, and L. J. C. van Loon, “Low-intensity exercise
reduces the prevalence of hyperglycemia in type 2 diabetes.,” Med. Sci. Sports
Exerc., vol. 42, no. 2, pp. 219–25, Feb. 2010.
[21] S. R. Colberg, R. J. Sigal, B. Fernhall, J. G. Regensteiner, B. J. Blissmer, R. R.
Rubin, L. Chasan-Taber, A. L. Albright, and B. Braun, “Exercise and type 2
diabetes: the American College of Sports Medicine and the American Diabetes
Association: Joint Position Statement,” Diabetes Care, vol. 33, no. 12, pp. e147–
67, Dec. 2010.
108
[22] D. Hansen, P. Dendale, R. A. M. Jonkers, M. Beelen, R. J. F. Manders, L. Corluy,
A. Mullens, J. Berger, R. Meeusen, and L. J. C. van Loon, “Continuous low- to
moderate-intensity exercise training is as effective as moderate- to high-intensity
exercise training at lowering blood HbA(1c) in obese type 2 diabetes patients,”
Diabetologia, vol. 52, no. 9, pp. 1789–97, Sep. 2009.
[23] S. Balducci, S. Zanuso, P. Cardelli, L. Salvi, A. Bazuro, L. Pugliese, C. Maccora, C.
Iacobini, F. G. Conti, A. Nicolucci, and G. Pugliese, “Effect of high- versus low-
intensity supervised aerobic and resistance training on modifiable cardiovascular
risk factors in type 2 diabetes; the Italian Diabetes and Exercise Study (IDES).,”
PLoS One, vol. 7, no. 11, p. e49297, Jan. 2012.
[24] D. Umpierre, C. K. Kramer, C. B. Leita, J. L. Gross, J. P. Ribeiro, and B. D.
Schaan, “Physical Activity Advice Only or Structured With HbA 1c Levels in Type 2
Diabetes,” JAMA, vol. 305, no. 17, pp. 1790–1799, 2011.
[25] C. O’Hagan, G. De Vito, and C. a G. Boreham, “Exercise prescription in the
treatment of type 2 diabetes mellitus : current practices, existing guidelines and
future directions.,” Sports Med., vol. 43, no. 1, pp. 39–49, Jan. 2013.
[26] J. E. Aikens, K. Zivin, R. Trivedi, and J. Piette, “Diabetes self-management support
using mHealth and enhanced informal caregiving,” J. Diabetes Complications, vol.
28, no. 2, pp. 171–176, 2014.
[27] S. A. Julious, “Sample size of 12 per group rule of thumb for a pilot study,” Pharm.
Stat., vol. 4, no. 4, pp. 287–291, Oct. 2005.
[28] T. L. Hart, A. M. Swartz, S. E. Cashin, and S. J. Strath, “How many days of
monitoring predict physical activity and sedentary behaviour in older adults?,” Int.
J. Behav. Nutr. Phys. Act., vol. 8, no. 1, p. 62, 2011.
[29] C. Tudor-Locke, R. Bell, A. Myers, S. Harris, N. Lauzon, and N. Rodger,
“Pedometer-determined ambulatory activity in individuals with type 2 diabetes,”
Diabetes Res Clin Pr., vol. 55, no. 3, pp. 191–199, 2002.
[30] D. X. Marquez and E. McAuley, “Social cognitive correlates of leisure time physical
activity among Latinos.,” J. Behav. Med., vol. 29, no. 3, pp. 281–9, Jun. 2006.
[31] J. Sallis, R. Grossman, R. Pinski, T. Patterson, and P. Nader, “The development of
scales to measure social support for diet and exercise behaviors,” Prev. Med.
(Baltim)., vol. 16, no. 6, pp. 825–836, 1987.
[32] K. R. Jones, N. Lekhak, and N. Kaewluang, “Using mobile phones and short
message service to deliver self-management interventions for chronic conditions: a
meta-review.,” Worldviews Evid. Based. Nurs., vol. 11, no. 2, pp. 81–8, 2014.
[33] E. McAuley, “The role of efficacy cognitions in the prediction of exercise behavior in
middle-aged adults,” J. Behav. Med., vol. 15, no. 1, pp. 65–88, 1992.
109
[34] E. McAuley, “Self-efficacy and the maintenance of exercise participation in older
adults,” J Behav Med, vol. 16, no. 1, pp. 103–113, 1993.
[35] K. Glanz, B. Rimer, and K. Viswanath, Eds., Health Behavior and Health
Education: Theory, Research, and Practice, 4th ed. Wiley, 2008.
[36] S. Nundy, J. J. Dick, C.-H. Chou, R. S. Nocon, M. H. Chin, and M. E. Peek, “Mobile
Phone Diabetes Project Led To Improved Glycemic Control And Net Savings For
Chicago Plan Participants,” Health Aff., vol. 33, no. 2, pp. 265–272, Feb. 2014.
[37] K. Capozza, S. Woolsey, M. Georgsson, J. Black, N. Bello, C. Lence, S. Oostema,
and C. North, “Going mobile with diabetes support: a randomized study of a text
message-based personalized behavioral intervention for type 2 diabetes self-
care.,” Diabetes Spectr., vol. 28, no. 2, pp. 83–91, 2015.
[38] S. Sawesi, M. Rashrash, K. Phalakornkule, J. Carpenter, and J. Jones, “The
impact of information technology on patient engagement and health behavior
change: a systematic review of the literature,” JMIR Med. Informatics, vol. 4, no. 1,
2016.
[39] M. Saffari, G. Ghanizadeh, and H. G. Koenig, “Health education via mobile text
messaging for glycemic control in adults with type 2 diabetes: A systematic review
and meta-analysis,” Prim. Care Diabetes, vol. 8, no. 4, pp. 275–285, 2014.
[40] P. Araiza, H. Hewes, C. Gashetewa, C. A. Vella, and M. R. Burge, “Efficacy of a
pedometer-based physical activity program on parameters of diabetes control in
type 2 diabetes mellitus,” Metabolism., vol. 55, no. 10, pp. 1382–1387, 2006.
[41] C. Tudor-Locke, R. C. Bell, A. M. Myers, S. B. Harris, N. A. Ecclestone, N. Lauzon,
and N. W. Rodger, “Controlled outcome evaluation of the First Step Program: a
daily physical activity intervention for individuals with type II diabetes,” Int. J. Obes.
Relat. Metab. Disord., vol. 28, no. 1, pp. 113–119, Jan. 2004.
[42] K. Patrick, F. Raab, M. A. Adams, L. Dillon, M. Zabinski, C. L. Rock, W. G.
Griswold, and G. J. Norman, “A text message-based intervention for weight loss:
randomized controlled trial,” J. Med. Internet Res., vol. 11, no. 1, p. e1, 2009.
[43] C. Tudor-Locke and A. Myers, “Methodological considerations for researchers and
practitioners using pedometers to measure physical (ambulatory) activity,” Res. Q.
Exerc. Sport, vol. 72, no. 1, pp. 1–12, 2001.
[44] M. Booth, N. Owen, A. Bauman, O. Clavisi, and E. Leslis, “Social-cognitive and
perceived environment influences associated with physical activity in older
Australians,” Prev. Med. (Baltim)., vol. 31, no. 1, pp. 15–22, 2000.
[45] B. Sternfeld, B. Ainsworth, and C. Quesenberry, “Physical activity patterns in a
diverse population of women,” Prev. Med. (Baltim)., vol. 28, no. 3, pp. 313–323,
1999.
110
[46] F. Treiber, T. Baranowski, D. Braden, and W. Strong, “Social support for exercise:
relationship to physical activity in young adults,” Prev. Med. (Baltim)., vol. 20, no. 6,
pp. 737–750, 1991.
[47] J. Sallis, M. Hovell, and C. Hofstetter, “Predictors of adoption and maintenance of
vigorous physical activity in men and women,” Prev. Med. (Baltim)., vol. 21, no. 2,
pp. 237–257, 1992.
[48] R. Dishman and J. Sallis, “Determinants and interventions for physical activity and
exercise,” in Physical Activity, Fitness, and Health, C. Bouchard, R. Shepard, T.
Stephens, J. Sutton, and B. McPherson, Eds. Champaign, IL: Human Kinetics,
1994.
[49] G. Felton and M. Parsons, “Factors influencing physical activity in average-weight
and overweight young women,” J. Community Health Nurs., vol. 11, no. 2, pp. 109–
119, 1994.
[50] A. A. Eyler, R. C. Brownson, R. J. Donatelle, A. C. King, D. Brown, and J. F. Sallis,
“Physical activity social support and middle- and older-aged minority women:
results from a US survey.,” Soc. Sci. Med., vol. 49, no. 6, pp. 781–9, Sep. 1999.
[51] A.-M. Rosland, M. Heisler, and J. Piette, “The Impact of Family Behaviors and
Communication Patterns on Chronic Illness Outcomes: A Systematic Review,” j
behav med, vol. 35, no. 2, pp. 221–239, 2012.
[52] S. Nundy, J. J. Dick, M. C. Solomon, and M. E. Peek, “Developing a behavioral
model for mobile phone-based diabetes interventions,” Patient Educ. Couns., vol.
90, no. 1, pp. 125–132, 2013.
[53] S. Dombrowski, F. Sniehotta, A. Avenell, M. Johnston, G. MacLennan, and V.
Araújo-Soares, “Identifying active ingredients in complex behavioural interventions
for obese adults with obesity-related co-morbidities or additional risk factors for co-
morbidities: a systematic review,” Health Psychol. Rev., vol. 6, no. 1, pp. 7–32,
Mar. 2012.
[54] S. Michie, C. Abraham, C. Whittington, J. McAteer, and S. Gupta, “Effective
techniques in healthy eating and physical activity interventions: a meta-
regression.,” Health Psychol., vol. 28, no. 6, pp. 690–701, Nov. 2009.
[55] D. Bravata, C. Smith-Spangler, V. Sundaram, A. Gienger, N. Lin, R. Lewis, C.
Stave, I. Olkin, and J. Sirard, “Using Pedometers to Increase Physical Activity and
Improve Health,” JAMA, vol. 298, no. 19, pp. 2296–2304, 2007.
[56] L. Engel and H. Lindner, “Impact of Using a Pedometer on Time Spent Walking in
Older Adults With Type 2 Diabetes,” Diabetes Educ., vol. 32, no. 1, pp. 98–107,
2006.
111
[57] Gleeson-Kreig, “Self-monitoring of physical activity: effects on self-efficacy and
behavior in people with type 2 diabetes,” Diabetes Educ., vol. 32, no. 1, p. p69,
2006.
[58] B. Speck and S. Looney, “Effects of a Minimal Intervention to Increase Physical
Activity in Women: Daily Activity Records,” Nurs. Res., vol. 50, no. 6, pp. 374–378,
2001.
[59] T. L. McDowell and J. M. Serovich, “The effect of perceived and actual social
support on the mental health of HIV-positive persons.,” AIDS Care, vol. 19, no. 10,
pp. 1223–9, Nov. 2007.

Appendix

Semi-Structured Interview Questions (PM and PM+FF only)

1. Over the past 6 weeks, you should have been receiving telephone calls/text
messages about walking as a form exercise. Have you been receiving these
telephone calls/text messages?
☐ Yes
☐ No

2. How many times per week would you say you receive telephone calls/text messages
from this study? _____ times per week

3. What are your thoughts on the number of messages you receive per week?

4. What are your thoughts on the times of the day or the days of the week you receive
the telephone calls/text messages?

5. What device do you use to receive these telephone calls/text messages?
☐ Landline telephone
☐ Cell phone
☐ Other: ______________________

6. Are you having any problems receiving/hearing/reading the telephone calls/text
messages? Please explain.

7. What do you do when you get a message? Do you answer the call (if voice)? Do you
read the message then, or later (if text)?

8. About once per week, you should have been receiving a message asking you to rate
how well you did with your walking goals in the last 7 days. Have you ever received
these messages?
112
☐ Yes
☐ No
☐ Other: _________________________

a. If Yes to #60: Do you usually reply to these messages?
☐ Yes
☐ No
☐ Other: ______________________________

a. If No to #60a: Why?

9. How do you feel about brisk walking as a form of exercise?

10. Do you think that participating in the study has affected the amount of walking that
you do? If so, how?

11. What are your thoughts on setting goals for walking?

12. How do you feel about keeping track of your walking with the pedometer and walking
log?

13. Overall, what do you think about the voice/text messages you have received?

14. What do you think about the messages you received each week after reporting how
you did with your goals?

15. Overall, what do you think about using telephone calls/text messages to
communicate with patients about their physical activity?

16. PM+SS patient only: I understand that your family member/friend also receives
telephone calls/text messages from this study as well. Do you know if your family
member/friend has indeed been receiving these?

17. PM+SS patient only: What has your family member/friend told you about the
telephone calls/text messages he/she receives from this study?

18. PM+SS patient only: Since the start of this study, how has the support you feel from
this person for improving your walking? What does this person do or say differently
to support you?

19. PM+SS patient only: Overall, what do you think about using telephone calls/text
messages to communicate with patients’ family members/friends about physical
activity?

20. PM+SS patient only: Is there anything else you would like to tell me about your family
member/friend’s involvement in the study?

113
21. How easy or hard was it to receive telephone calls/text messages from this study?

22. How easy or hard was it to reply to telephone calls/text messages?

23. Can you tell me about your experience with wearing a pedometer?

24. Overall, what do you like the most and least about this study? What do you like the
most and least about receiving telephone calls/text messages from this study?

25. If the program being tested in this study was offered to all patients as a part of their
diabetes care, would you willing to participate?

26. Would you recommend the program to other patients? Please explain.

27. If you could change one thing about the program, what would it be?

28. Is there anything else you would like to tell us about your experience in this program?
114
Chapter 5 Conclusion

Patient-facing communication and information technology (CIT) tools are increasingly
being developed to help patients manage their health and health care. Those designed
to support adults with diabetes have resulted in little to no effect on biological, cognitive,
behavioral, or emotional outcomes and have experienced high rates of user
discontinuation [1], [2]. The lack of significant impact may be attributed to the approach
used to design and evaluate the health CIT tools. Design approaches are, at best,
centered on behavioral theories, which do not consider the needs and preferences of the
intended user groups. Evaluation approaches focus on changes in patient outcomes,
largely ignoring important aspects of patients’ interaction with the technology that may
affect the performance of these tools.

The majority of health CIT research targeting individuals with diabetes comes from a
social and behavioral science perspective. This dissertation uses a human factors
perspective to design and evaluate health CIT tools that support vulnerable adults with
diabetes. Chapter 2 presented an evaluation of the design of an automated telephonic
assessment (ATA) system that was tested in a trial to accelerate the adoption of
collaborative depression care in a safety-net health care system. The ATA system
periodically called patients to assess depression symptoms, monitor treatment
adherence, prompt self-care behaviors, and inquire about patients’ needs for provider
contact. In the short term, patients provided high ratings of perceived usefulness, non-
intrusiveness, ease of use, and privacy/security. These high ratings decreased over time
115
as patients continued to receive ATA calls, but were no longer followed-up on by health
care providers. Factors predicting long-term technology acceptance are perceived
privacy/security and usefulness. The latter requires that the system help patients be
more aware of how they are feeling, remind them to take care of their health, and help
them stay better connected with their health care providers.

While the ATA system called patients periodically, Chapters 3 and 4 presented the
design and evaluation a user-centered health CIT tool that provided more frequent
ongoing support for a key aspect of diabetes self-care: engagement in physical activity
(PA). Patients were more likely to accept a short text/voice messaging (ST/VM) system
for PA behavior change support with the following features: recommends the message
frequency, educational content, and PA goals; provides feedback based on individual PA
performance; and prompts social support by family members. A pilot test of a ST/VM
system whose specifications were derived based on these findings revealed that text
messaging may be a better mode of message delivery than voice messaging for
ensuring patient receipt of and engagement with messages; self-monitoring, goal setting,
self-instruction, feedback, and social support are perceived as useful aspects of the
ST/VM system; and the ST/VM system has the potential to improve daily steps and
perceived social support from family members/friends.

This dissertation has generated preliminary findings for full trial to study health CIT for
improving PA among urban, low-income Latino adults with diabetes. The preliminary
findings contribute to the field of Industrial and Systems Engineering by elucidating an
important question that remains unanswered: How to design systems that facilitate
116
social support that is effective at changing health behaviors? The application of human
factors in health care technologies has focused primarily on medical devices and patient
safety. However, the growing recognition that patient-facing information technologies are
important components of health care services and that informal caregiver support is
crucial for self-management calls for a new area of research in human factors. Adequate
guidance does not exist for such health CIT tools, which require unique design criteria
relative to existing human factors application areas. Critical factors that will influence
design choices include users’ social network, cultural and health beliefs, education,
income, age, access to technology, chronic conditions, and health literacy [3]. Future
research is needed to understand first and foremost the mechanisms by which
families/close friends facilitate PA behavior change. And secondly, there is a need to
understand which technology design features and/or user participatory processes most
effectively leverage these factors to bring about such change. Having this knowledge
would help inform guidelines that would enable developers of health CIT tools to design
systems that are effective at helping individuals and their families to make lasting
behavior changes that improve health outcomes.

Moreover, technology acceptance models distill the critical factors predicting behavioral
intention to use a technology. Two of the most important factors are perceived ease of
use and usefulness [4] – both of which are largely under the control of designers. A
direction for future research that would extend the work of this dissertation involves
going beyond studying the usability of health CIT tools and users’ perception of their
usefulness. While a technology that is perceived as easy use and useful will have an
increased likelihood of being used, it is unknown whether or how this translates to
117
improvements in patient outcomes, which is, after all, the overarching goal of health CIT
tools. Future research is needed to examine the relationship between technology design
choices (e.g., frequency of alerts/messages/prompts) and patient outcomes (health
behavior change). An understanding of this relationship could lead to the development of
a model that predicts improvements in patient outcomes, and can also be used to guide
choices in the design of health CIT tools.

In conclusion, this dissertation adopted a human factors perspective by focusing the
design process on the user and evaluating aspects of the user-technology interaction
that may affect performance. The series of three studies have two major findings
regarding user-technology design. First, patient acceptance of remote monitoring
technology requires that the system be perceived as private/secure, and that it be
constantly useful for patients’ needs of awareness of feelings, self-care reminders, and
connectivity with health care providers. Second, a phone-based PA behavior change
support system designed based patient preferences is feasible, perceived as useful, and
may be potentially effective at improving daily steps and social support from FF. This
dissertation has shed light on a gap in the literature that we are lacking a model or
methods to design systems that facilitate effective social support to change health
behaviors. Future research should understand the relationship between technology
design features and/or user participatory processes and their effectiveness on leveraging
the factors that cause a person to make sustained health behavior changes and how
they affect patient outcomes. Such research could help inform guidelines for the design
of systems that are not just easy to use, but also effective at changing health behaviors
and improving patient outcomes.
118
References

[1] K. Pal, S. Eastwood, S. Michie, A. Farmer, M. Barnard, R. Peacock, B. Wood, J.
Inniss, and E. Murray, “Computer-based diabetes self-management interventions
for adults with type 2 diabetes mellitus (review),” Cochrane Libr., no. 3, 2013.

[2] J. Piette, “Interactive behavior change technology to support diabetes self-
management: where do we stand?,” Diabetes Care, vol. 30, no. 10, pp. 2425–32,
Oct. 2007.

[3] H. B. Jimison, M. Pavel, A. Parker, and K. Mainello, “The Role of Human
Computer Interaction in Consumer Health Applications: Current State, Challenges
and the Future,” pp. 163–187, 2015.

[4] F. D. Davis, “User acceptance of information technology: system characteristics,
user perceptions and behavioral impacts,” Int. J. Man. Mach. Stud., vol. 38, no. 3,
pp. 475–487, 1993.

Abstract (if available)

Abstract Communication and information technology (CIT) tools are increasingly being developed for use by patients to help them manage their health and health care. While it has been demonstrated that health CIT tools can improve health outcomes, those designed to support adults with diabetes have resulted in little to no effect on diabetes outcomes and have not been adopted long-term. The current health literature approach to design and evaluation of health CIT tools is a major problem that may be contributing to the observed lack of effectiveness and sustained user engagement. Rigorous design and evaluation approaches must be utilized in order to make health CIT tools that will have a significant impact. This dissertation used a human factors perspective to design and evaluate health CIT tools that support adults with diabetes. Such an approach to design and evaluation of health CIT may increase the likelihood that patients’ interaction with the tools is one that enhances performance, increases user satisfaction, and increases safety. The first part of this dissertation involved evaluating the design of an automated telephonic assessment (ATA) system that periodically called patients to assess depression symptoms, monitor treatment adherence, prompt self-care behaviors, and inquire about patients’ needs for provider contact. The manuscript in Chapter 2 presents secondary data analyses performed to answer the following research questions: What is patients’ acceptance of ATA calls over time? What factors predict patients’ long-term acceptance of ATA calls? While the ATA system called patients periodically, the second part of this dissertation involved designing and evaluating a user-centered health CIT tool that provided more frequent ongoing support for a key aspect of diabetes self-care: engagement in physical activity (PA). The communication modality used was short text/voice messaging (ST/VM). The manuscript in Chapter 3 presents results that answer the research question: What configuration of ST/VM system features for PA behavior change support do patients most prefer? Patient preferences were then translated into system specifications and a pilot test of the resulting ST/VM system conducted to gauge patients’ response. The manuscript in Chapter 4 presents results that answer the research question: What is the feasibility, acceptance, and potential effectiveness of a ST/VM system to support PA behavior change? The major findings and contributions of the three manuscripts that comprise this dissertation are summarized.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A series of longitudinal analyses of patient reported outcomes to further the understanding of care-management of comorbid diabetes and depression in a safety-net healthcare system

Asset Metadata

Creator Ramirez, Magaly (author)

Core Title Using a human factors engineering perspective to design and evaluate communication and information technology tools to support depression care and physical activity behavior change among low-inco...

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Industrial and Systems Engineering

Publication Date 09/21/2016

Defense Date 04/06/2016

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag clinical decision support systems,consumer health information,Depression,Diabetes Mellitus,disease management,Exercise,Health education,Hispanic Americans,Latino Americans,low income population,OAI-PMH Harvest,patient care management,pilot projects,reminder systems,safety-net clinics,self care,short message service,social support,technology assessment,Telecommunications,telemedicine,Telephone

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Wu, Shinyi (committee chair), Rahimi, Mansour (committee member), Spruijt-Metz, Donna (committee member)

Creator Email maggie.a.ramirez@gmail.com,ramirezma@ucla.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-303999

Unique identifier UC11281390

Identifier etd-RamirezMag-4789.pdf (filename),usctheses-c40-303999 (legacy record id)

Legacy Identifier etd-RamirezMag-4789.pdf

Dmrecord 303999

Document Type Dissertation

Format application/pdf (imt)

Rights Ramirez, Magaly

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA