Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Healthcare provider recommendations: a panacea to improving influenza vaccination rates?
(USC Thesis Other)
Healthcare provider recommendations: a panacea to improving influenza vaccination rates?
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
HEALTHCARE PROVIDER RECOMMENDATIONS: A PANACEA TO IMPROVING INFLUENZA VACCINATION RATES? by Reginald Villacorta A Dissertation Presented to the Faculty of the University of Southern California Graduate School University of Southern California In Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy (Pharmaceutical Economics and Policy) May 2015 Copyright 2015 Reginald Villacorta ! ii! Table of Contents List of Tables v List of Figures vi Preface vii References xv Chapter 1 Determinants of healthcare provider recommendations for influenza vaccinations 1 Abstract 1 1.0 Introduction 2 2.0 Methods 4 2.1 Data source 4 2.2 Study population 4 2.3 Primary outcome measures 5 2.4 Key variables 6 3.0 Statistical analysis 7 4.0 Results 8 4.1 Unadjusted analyses of sample characteristics 8 4.2 Unadjusted analyses of opinions about influenza vaccines 9 4.3 Adjusted analyses of ACIP priority groups and HCP recommendation receipt 10 4.4 Adjusted analyses of determinants of HCP recommendations for flu vaccinations 10 4.5 Sensitivity analyses of adjusted models 11 5.0 Discussion 12 6.0 Conclusions 16 7.0 References 17 8.0 Tables 20 Chapter 2 What Explains Racial/Ethnic Disparities in Influenza Vaccination Rates? 32 Abstract 32 1.0 Introduction 34 2.0 Methods 35 2.1 Data 35 2.2 Explanatory variables 37 2.3 Statistical analyses 37 2.4 Decomposition model 38 3.0 Results 40 4.0 Discussion 44 ! iii! 5.0 Conclusions 46 6.0 References 47 7.0 Tables 49 Chapter 3 Healthcare provider recommendations and influenza vaccinations: A causal effects estimate 64 Abstract 64 1.0 Introduction 65 2.0 Theoretical framework 67 2.1 Framework of omitted variable bias 69 3.0 Methods 70 3.1 Study participants 70 3.2 Study design 71 3.2.1 Study 1 71 3.2.2 Study 2 72 3.3 Statistical analysis plan 73 3.3.1 Study 1 73 3.3.2 Study 2 74 3.4 Hypotheses 74 4.0 Results 76 4.1 Study participants 76 4.1.1 Study 1 results 76 4.1.2 Study 2 results 78 4.2 Empirical evidence of omitted variable bias 80 4.3 Can the predisposition effect from Equation (4) be estimated? 81 4.4 Sensitivity analyses on flu vaccine intent and actual flu vaccination 82 5.0 Discussion 83 5.1 Predisposition has broad implications 87 5.2 Limitations 89 5.3 Directions for future research 90 6.0 Conclusion 91 7.0 References 92 8.0 Figures 95 9.0 Tables 97 Bibliography 110 Appendices Appendix 1A. Sensitivity of logit model outcomes 115 Appendix 1B. Multinomial logit model of healthcare provider recommendations 122 ! iv! Appendix 1C. Predicted probabilities for healthcare provider recommendations and flu vaccines 130 Appendix 2A. Full descriptive statistics 137 Appendix 2B. Full decomposition of differences in flu vaccination rates between Whites and Blacks 143 Appendix 2C. Full decomposition of differences in flu vaccination rates between Whites and Hispanics 149 Appendix 2D. Select decomposition of differences in flu vaccination rates between Whites and Hispanics using the 2013 Behavioral Risk Factor Surveillance System (BRFSS) dataset 154 Appendix 2E. Select decomposition of differences in flu vaccination rates between Whites and Blacks using the 2013 Behavioral Risk Factor Surveillance System (BRFSS) dataset 155 Appendix 2F. Sub-aggregate decomposition of differences in flu vaccination rates between Whites and Hispanics (omitting flu vaccine beliefs and provide recommendations) 156 Appendix 2G. Sub-aggregate decomposition of differences in flu vaccination rates between Whites and Blacks (omitting flu vaccine beliefs and provide recommendations) 157 Appendix 3A. Provider recommendations and influenza vaccinations for the 2013-2014 season and H1N1 flu vaccines by healthcare worker status 158 Appendix 3B. Random assignment of healthcare provider recommendations 162 Appendix 3C. Subpopulation descriptive statistics 168 Appendix 3D: Subpopulation regression analysis 169 ! v! List of Tables Table 1-1 Descriptive statistics of healthcare provider recommendations 20 Table 1-2 Descriptive statistics of healthcare provider recommendations (continued) 25 Table 1-3 Multivariate logistic regression for healthcare provider recommendations 27 Table 2-1 Vaccination type by race/ethnicity 49 Table 2-2 Select descriptive statistics 50 Table 2-3 Logit regression 55 Table 2-4 Sub-aggregate decomposition of differences in flu vaccination rates between Whites and Blacks 62 Table 2-5 Sub-aggregate decomposition of differences in flu vaccination rates between Whites and Hispanics 63 Table 3-1 Study 2 experimental conditions 97 Table 3-2 Descriptive statistics for Study 1 respondents 98 Table 3-3 Effect of provider recommendations on flu vaccine receipt 101 Table 3-4 Descriptive statistics for Study 2 respondents 102 Table 3-5 Effect of provider recommendations on flu vaccine intent 106 Table 3-6 Agreement with seasonal flu vaccination effectiveness 107 Table 3-7 Measures of predisposition 108 Table 3-8 Predicted probability of flu vaccination 109 ! vi! List of Figures Figure 3-1 Study 1 sample flow diagram 95 Figure 3-2 Study 2 sample flow diagram 96 !vii! PREFACE The burden of influenza disease in the United States (US) accounts for an estimated annual average of 36,000 deaths and 226,000 hospitalizations [Thompson, 2003; Thompson, 2004; Thompson, 2010]. Influenza is the greatest vaccine-preventable burden of disease in the US, where for the 2013-2014 season, the influenza vaccine prevented 7.2 million illness, 3.1 million medically attended illness, and 90,000 influenza-associated hospitalizations. Yet, if vaccination levels met the HealthyPeople 2020 target of 70% for the adult population, an approximate additional 5.9 million illness, 2.3, million medically attended illnesses, and 42,000 influenza-associated hospitalizations might have been averted [HealthyPeople2020, 2014; Reed, 2014]. This underscores the need for influenza vaccination coverage to increase and meet HealthyPeople 2020 targets to reduce the US burden of influenza disease. The demand for influenza vaccines can be bolstered by healthcare providers through patient education on the consequences of influenza infections and benefits of the influenza vaccine [Orenstein and Schaffner, 2008]. This will likely lead to increases in vaccine demand because in a health system with a choice of physician (i.e. a healthcare provider) such as in the US, patients typically choose their physicians based on their own assessment on the physician’s judgment. Thus, patients become disposed to accepting physician recommendations since they have located a physician whose judgment they trust. At the same time, physicians are encouraged to recommend the influenza vaccine to their patients since influenza vaccination is considered the best defense against influenza-related morbidity and mortality. !viii! Recently, the National Vaccine Advisory Committee (NVAC) considers healthcare provider recommendations for recommended vaccinations (e.g., the annual influenza vaccine) as a standard of practice [Bhatt et al, 2014]. This standard of practice comes from prior work demonstrating a positive, strong and statistically significant correlation between provider recommendations and vaccine uptake [Bhatt et al, 2014]. Yet, complicating this potential to increase influenza vaccination coverage through healthcare provider recommendations is whether the patient actually believes the vaccine is effective. Researchers have argued that vaccine side effects deter patients from obtaining the vaccine [Fiscella, 2005; Hebert et al., 2005; Mirza et al, 2008]. This suggests a provider recommendation may not be salient to patients in the presence of these negative vaccination beliefs. The notion of vaccine side effects and limited effectiveness is credible. For example, common side effects can include sore throat and headaches. And, influenza vaccine effectiveness vary year-to-year for certain patient groups where, for example, the vaccination was associated with only a 71% reduction in flu-related hospitalizations among adults of all ages [Centers for Disease Control and Prevention (CDC), 2014]. These influenza vaccine effectiveness characteristics provide a reasonable basis for why vaccination coverage is below HealthyPeople2020 targets. Barriers to influenza vaccination are numerous and tend to involve major stakeholders in the influenza vaccination program [Bhatt et al, 2014; Orenstein and Schaffner, 2008]. In ! ix! the US, these stakeholders (e.g., state, federal, and local policymakers; vaccine producers and distributors; clinicians; and, workplace immunization programs) participate in the supply and distribution of influenza vaccines. Yet, the NVAC focuses on the role of providers in ensuring coverage rates improve because they consider that the fundamental responsibility to ensure that all patients are up-to-date with their recommended immunizations resides with the healthcare provider. Additionally, with patients coming to their own conclusions with influenza vaccine effectiveness, two important questions regarding recommendation effectiveness arises. Will provider recommendations be the best method to improving vaccination rates in the presence of how patients perceive influenza vaccine effectiveness? Will the recommendation effect be overshadowed by underlying patient intent towards obtaining the influenza vaccine? This dissertation evaluates these questions regarding the effect of healthcare provider recommendation for influenza vaccinations by: measuring the patient groups receiving a healthcare provider recommendation; evaluating the extent to which healthcare provider recommendations reduce racial/ethnic disparities in influenza vaccination rates; and, assessing causal effects estimate between healthcare provider recommendations and patient flu vaccinations. Who gets a provider recommendation for influenza vaccinations? Similar questions have been asked in provider recommendations for human papillomavirus vaccines and breast cancer screenings. Yet, no study evaluates this question for influenza vaccinations. The literature in influenza vaccinations is predominantly focused on correlations between ! x! patient characteristics (i.e., demographics, socioeconomic status, whether they received a provider recommendation) and influenza vaccinations [Ding et al, 2011; Pandolfi et al, 2012; Santibanez et al, 2010]. Still, researchers have yet to identify who received a provider recommendation for influenza vaccinations. This is an important policy question because the majority of this work was published prior to the NVAC standard. Furthermore, these patient populations were predominantly high-risk patients (i.e., chronically ill, elderly, or infant children) with limited focus on the general population. Identifying the patient populations likely to receive a recommendation provides insight into which patients groups received recommendations prior to the policy change. This dissertation adds to the literature by demonstrating healthcare provider recommendations for influenza vaccinations play an important role in improving vaccination rates, especially among Advisory Committee on Immunization Practices (ACIP) priority groups. This study demonstrates these priority groups were more likely to report healthcare provider recommendations for influenza vaccinations during the 2009-2010 flu season when compared to non-priority groups. Unlike similar studies in HPV vaccine recommendations, racial/ethnic minority groups were more likely to receive recommendations compared to Non-Hispanic, White adults. Racial/ethnic disparities exist in flu vaccine uptake and are widely recognized in the literature [Fiscella, 2005; Hebert et al., 2005]. Moreover, researchers have offered explanations to the existence of these disparities. Yet, with healthcare provider recommendations a standard of care in the US, to what extent do these recommendations ! xi! reduce racial/ethnic disparities? Do recommendations reduce racial/ethnic disparities at higher rates when compared to patient vaccination beliefs? These are important policy questions because numerous barriers exist in explaining racial/ethnic disparities; yet, no research explores the extent to which recommendations potentially reduce racial/ethnic disparities in the presence of other barriers. This dissertation shows the majority of the share in influenza vaccination disparities between Non-Hispanic Whites (Whites), Non-Hispanic Blacks (Blacks), and Hispanics can be explained by differences in observable characteristics. Whites report higher influenza vaccination rates by 13.7 and 12.0 percentage points relative to Blacks and Hispanics, respectively. Differences in socioeconomic levels and beliefs about influenza vaccines mainly explain these disparities in influenza vaccinations. 29.2% of the difference in influenza vaccination rates between Whites and Blacks can be explained by differences in beliefs about influenza vaccines. 14.0% of the same difference can be explained by differences in socioeconomic status. 6.1% of the difference in influenza vaccination rates between Whites and Hispanics can be explained by differences in beliefs about influenza vaccines. 22.6% of the same difference can be explained by differences in socioeconomic status. Health policies incorporating healthcare provider recommendations to reduce flu vaccination disparities should expect a modest change (<1 percentage point) in disparities that can vary depending on the comparator racial/ethnic groups. !xii! What is the causal effect of provider recommendations on flu vaccinations? Strong, positive and statistically significant correlations exist between healthcare provider recommendations and influenza vaccinations. Yet, recommendations are likely non- random because vaccinated providers are likely to recommend the vaccine to their patients [Santibanez et al, 2010]. Furthermore, patients may be disposed to obtaining the flu vaccine without a provider recommendation because of their belief that the vaccine is effective [Santibanez et al, 2010]. Given the non-random nature of recommendations from providers and patient predisposition to the vaccine without a recommendation, these prior positive correlations may be overestimated. To determine this upward bias, a causal estimate is warranted. This dissertation hypothesizes prior estimates in the effect of provider recommendations are biased upwards. The results demonstrate provider recommendations generate an unadjusted 39% (p<0.001) increased likelihood of obtaining a flu vaccine. A subsequent unadjusted provider recommendation causal effect on flu vaccination intent was 16% (p<0.001). Therefore, by construction, there exists a 23-percentage point bias in the recommendation effect. These findings suggest prevailing measures of the effect of provider recommendations on flu vaccinations are overestimated. Further research is needed to describe the sensitivities of this bias in different treatment scenarios. The potential for provider recommendations to improve influenza vaccination rates is underscored by the standard of care policy initiated by NVAC [Bhatt et al, 2014]. This dissertation generates important recommendation effects that inform the NVAC policy. !xiii! By measuring the patient groups likely to receive a recommendation, policy makers can be sure that recommendations are reaching high-risk patient groups. Yet, recommendation disparities exist among racial/ethnic minority groups. Further, among these groups, recommendations may not reduce disparities at greater rates when compared to patient beliefs regarding the vaccine. Lastly, the causal effect measured in this dissertation suggests prior estimates are biased upwards. In addition, by controlling for patient predisposition defined as their behavior in acquiring the vaccine from their provider, the recommendation effect is reduced by one half. One can come away with two competing outcomes from this dissertation. First, similar to prior estimates, the recommendation effect on influenza vaccination is positively and statistically significant. Second, when compared to patient predisposition on vaccine effectiveness, the recommendation effect may not be as high as previously measured. By setting recommendations as a standard of practice, the NVAC recognizes the importance of recommendations in improving influenza vaccinations. Yet, whether it is due to the belief that the vaccine is effective or ineffective, these recommendations may not strongly influencing patient behavior. In a broader context, patient-provider communication practices are widely researched to the point where patient values and preferences are becoming important elements in healthcare decision-making. A consistent question from this research often arises: Which is more salient to the patients in consuming health services, their own beliefs or provider council? In the present study, patient predisposition towards the flu vaccine is a !xiv! stronger predictor of obtaining flu vaccinations than provider recommendations. In this case, patients have a basis for the flu vaccine and providers can only add to the demand of flu vaccine with council. Directions for future research include examining how predisposition arises. This study captures the behavior of requesting the flu vaccine from their provider as a relatively simple description of predisposition. Including similar behavioral questions in immunization surveys (e.g., National Immunization Survey, National Health Interview Survey, and Behavioral Risk Factor Surveillance System) provides for a closer representation of a causal recommendation effect on influenza vaccinations can be attained. Further applications of this dissertation can be extended to similar work in recommendations for prophylactic medical care (e.g., human papillomavirus vaccinations, breast cancer screenings). ! xv! REFERENCES Bhatt A, Bridges C, Donoghue K, Fernandez C, Gehring R, Hall LL, et al. Recommendations from the National Vaccine Advisory committee: standards for adult immunization practice. Public Health Rep. 2014; 129(2): 115-23. Centers for Disease Control and Prevention. Vaccine Effectiveness – How well does the flu vaccine work? 2014. http://www.cdc.gov/flu/about/qa/vaccineeffect.htm Ding H, Santibanez TA, Jamieson DJ, et al. Influenza vaccination coverage among pregnant women – National 2009 H1N1 Flu Survey (NHFS). Am J Obstet Gynecol 2011;204(6):S96-S106. Fiscella K. Commentary – Anatomy of racial disparity in influenza vaccination. HSR: Health Services Research 2005;40(2): 539-550. HealthyPeople2020. Immunization and Infections Diseases. 2014. https://www.healthypeople.gov/2020/topics-objectives/topic/immunization-and- infectious-diseases/objectives. Hebert PL, Frick KD, Kane RL, McBean AM. The causes of racial and ethnic differences in influenza vaccination rates among elderly Medicare beneficiaries. HSR: Health Services Research 2005;40(2): 517-538. Mirza A, Subedar A, Fowler SL, et al. Influenza vaccine: awareness and barriers to immunization in families of children with chronic medical conditions other than asthma. South Med J 2008;101(11):1101-1105. Orenstein WA, Schaffner W. Lessons learned: Role of influenza vaccine production, distribution, supply, and demand – What it means for the provider The American Journal of Medicine. 2008;121: S21-S27. Pandolfi E, Marino MG, Carloni E, et al. The effect of physician’s recommendation on seasonal influenza immunization in children with chronic diseases. BMC Public Health 2012;12(1):984. Reed C, Kim IK, Singleton JA, et al. Estimated influenza illnesses and hospitalizations averted by vaccinations – United States, 2013-14 influenza season. MMWR 2014; 63(49): 1151-1154. Santibanez TA, Mootrey GT, Euler GL, Janssen AP. Behavior and beliefs about influenza vaccine among adults aged 50-64 years. Am J Health Behav 2010;34(1):77-89. Thompson WW, Shay DK, Weintraub E, et al. Mortality associated with influenza and respiratory syncytial virus in the United States. JAMA. 2003;289:179-186. !xvi! Thompson WW, Shay DK, Weintraub E, et al. Influenza-associated hospitalizations in the United States. JAMA. 2004;292:1333-1340. Thompson MG, Shay DK, Zhou H, et al. Updated Estimates of Mortality Associated with Seasonal Influenza through the 2006-2007 Influenza Season. MMWR 2010; 59(33): 1057-1062. ! 1! Chapter 1. Determinants of healthcare provider recommendations for influenza vaccinations Abstract Objective: Investigate determinants of receiving healthcare provider (HCP) recommendations for seasonal and H1N1 influenza vaccinations. Methods: Using a United States national sample of adults 18+ from the National 2009 H1N1 Flu Survey, multivariate regression models estimated the likelihood of receiving a HCP recommendation. Covariates included demographics, socioeconomic status, and Advisory Committee on Immunization Practices (ACIP) priority groups. Results: Adults age 65+ were more likely to report a HCP recommendation when compared to adults age 18-34 (OR:1.738, 95%CI:1.427-2.116). Chronically ill adults had 58.0% (95%CI:1.414-1.765) higher odds of reporting a HCP recommendation than non- chronically ill adults. Patients visiting a doctor once and twice had 28.7% (95%CI:0.618- 0.821) and 17.1% (95%CI:0.721-0.952) lower odds of reporting a HCP recommendation when compared to adults visiting their doctor at least four times. And, racial/ethnic minorities were more likely to receive a recommendation. Conclusions: ACIP priority groups experienced higher rates of recommendations compared to non-ACIP groups. Racial/ethnic minority groups have an increased likelihood of receiving a recommendation compared to non-Hispanic Whites. Further efforts to increase recommendation rates for all patient groups can improve influenza vaccination rates. ! 2! 1.0 Introduction In the United States (US), influenza (i.e., the flu) infections result in >200,000 hospitalizations and 24,000 deaths on average [Groshkpof et al., 2013]. Seasonal influenza vaccination is an important method for preventing the transmission of the influenza virus. Despite this recognition that flu vaccinations can reduce the spread of influenza infections, gaps in vaccination coverage exist. Disparities in adult US influenza vaccination coverage exist between the elderly and non-elderly; populations at high-risk for influenza-related complications compared to otherwise; and, racial/ethnic minority groups compared to White, non-Hispanic groups [Centers for Disease Control and Prevention, 2011; Fiscella, 2005; Hebert, Frick, Kane, and McBean, 2005; Lu et al., 2013; O’Malley and Forrest, 2006]. Receipt of a physician recommendation for an influenza vaccination has been studied based on patient [Armstrong et al., 2001; Ding et al., 2011; Fiebach and Viscoli, 1991; Gnanasekaran et al., 2006; Hemingway and Poehling, 2004; Lyn-Cook, Halm, and Wisnivesky, 2007; Mirza et al., 2008; Nichol, Lofgren, and Gapinski, 1992; Pandolfi et al., 2012; Poehling et al., 2001; Santibanez, Mootrey, Euler, and Janssen , 2010] or physician [Dominguez and Daum, 2005; Jessop, Dumas, and Moser, 2013; Levy, Ambrose, Oleka, and Lewing, 2009; Nichol and Zimmerman, 2001] self-reports. These studies find a strong association between physician recommendation and the likelihood of obtaining an influenza vaccination for various patient groups. However, these studies predominantly focus on groups at high-risk for influenza-related complications (i.e., ! 3! asthmatics, elderly adults) and racial/minority groups that have relatively low flu vaccine uptake. Therefore, there is limited generalizability to the general population. Other studies demonstrate disparities in influenza vaccination rates for racial or ethnic minorities and those with lower socio-economic status [Annunziata et al., 2012; Gu and Sood, 2011; Takayama, Wetmore, and Mokdad, 2012; Singleton, Santibanez, and Wortley, 2005]. However, it is not known the extent to which these patient groups received flu vaccine recommendations from their provider. Examining the patient populations likely to report a physician recommendation can influence policy initiatives with the goal of reducing disparities in vaccination rates. Similar work related to factors associated with recommendations for human papillomavirus (HPV) vaccines were recently assessed from patient [Ylitalo, Lee, and Mehta, 2013] and provider [Vadaparampil et al., 2014] perspectives. These studies find disparities in HPV vaccine recommendations amongst racial/ethnic groups. This study has two objectives. The first objective is to investigate the association between healthcare provider recommendations for influenza vaccinations and patient demographic, socioeconomic, and health access characteristics from a US population. The second objective is to determine whether Advisory Committee on Immunization Practices (ACIP) priority groups experienced flu vaccine recommendations from their healthcare provider at higher rates than non-ACIP priority groups. ! 4! 2.0 Methods 2.1 Data source Data came from the public-use National 2009 H1N1 Flu Survey (NHFS) by the Centers of Disease Control and Prevention [US Department of Health and Human Services, 2012]; which was reviewed by the National Center for Health Statistics Disclosure Review Board to protect participant privacy and data confidentiality. Households were identified from all 50 US states and the District of Columbia; where the monitoring of both H1N1 and seasonal influenza vaccination coverage rates, at national and state levels, was evaluated from persons age ≥6 months. The content of NHFS household interviews consisted of survey-respondent history of chronic conditions and respiratory illness; H1N1 and seasonal flu vaccination history; demographics and socioeconomic information; household characteristics; and, for adults, questions about knowledge, attitudes, and practices related to 2009 H1N1 and seasonal influenza. The reported Council of American Survey Research Organizations response rate range was 33.4% for landline telephones and 26.1% for cell phones [US Department of Health and Human Services, 2012]. 2.2 Study population This study identified adult survey-respondents age 18+ that were interviewed from January through June 2010. We identified adults that visited a doctor’s office, hospital, or clinic since August 2009 up to the interview date. We focused on adults because important respondent characteristics were only captured from adults (i.e., chronic medical condition status, work status, and opinions about the seasonal and H1N1 influenza ! 5! vaccine). The analysis was restricted to interviews conducted between January 2010 and June 2010 to allow for a sufficient time window to observe provider recommendations [28]. This restriction of interviews was consistent with the recommendation of analyzing factors associated with influenza vaccinations by the NHFS. 2.3 Primary outcome measures The primary outcome of this study was adults’ self-report of a doctor or other health professional personal recommendation for the H1N1 or seasonal flu vaccination since August 2009. Posted signs, newsletters, pamphlets, or television and radio ads were not considered a recommendation. Thus, the primary outcome captures patients experiencing face-to-face flu vaccine recommendations that were likely tailored to the individual patient. For this study, we considered all recommendations to be healthcare provider (HCP) recommendations. Survey-respondents were given the following choices of HCP recommendations: (1) H1N1 flu vaccination; (2) seasonal flu vaccination; (3) both vaccinations; (4) neither vaccination; (5) don’t know; and, (6) refused. Respondents reporting don’t know and refused were grouped with neither vaccination response to create a four choice framework. For our primary outcome, reporting a HCP recommendation was defined as a binary variable equal to one when the respondent indicated they received recommendations for H1N1 flu vaccination only, seasonal flu vaccination only, or both vaccinations. The 2009-2010 flu season was unique in providing both seasonal and H1N1 flu vaccinations, and there may be differences in HCP recommendations between these two ! 6! vaccinations related to disease severity or infectiousness. Therefore, we conducted the following sensitivity analyses on the classification of HCP recommendations: defining the outcome as any seasonal (season flu vaccine only and both vaccinations) or any H1N1 (H1N1 flu vaccine only and both vaccinations) flu vaccine recommendations (Appendix 1A); relative risk ratios from a multinomial logit (MNL) model analyzing the polychotomous outcome of: no recommendation, receipt of H1N1 recommendation only, receipt of seasonal vaccination only, and receipt of both seasonal and H1N1 recommendations (Appendix 1B). To generalize our study to the prior literature, we estimated marginal effects where the primary outcome was HCP recommendation and compared them to marginal effects where the primary outcome was flu vaccinations (Appendix 1C). 2.4 Key variables Prior studies have limited information on predictors of healthcare provider recommendations for influenza vaccines. Therefore, we utilize previously studied determinants of influenza vaccination to inform the variables in the adjusted models. Demographic characteristics such as males and non-White race are significantly associated with a lower likelihood of influenza vaccinations compared to females and White race groups, respectively. Further, compared to younger adults, older adults experience higher rates of influenza vaccinations [Annunziata et al., 2012; Gu and Sood, 2011; Takayama, Wetmore, and Mokdad, 2012; Singleton, Santibanez, and Wortley, 2005]. In particular, adults age 65+ is an ACIP priority group because of their high risk ! 7! of infection compared to adults <65 [Centers for Disease Control and Prevention, 2009 and 2009]. Adults with health insurance, higher incomes, and those who visit their doctor more frequently are significantly more likely to receive an influenza vaccination [Annunziata et al., 2012; Gu and Sood, 2011; Takayama, Wetmore, and Mokdad, 2012; Singleton, Santibanez, and Wortley, 2005]. Furthermore, chronically ill adults (i.e., asthmatics, diabetics) have higher likelihoods of receiving an influenza vaccination when compared to non-chronically ill adults [Annunziata et al., 2012; Takayama, Wetmore, and Mokdad, 2012]. These patient characteristics are also considered key variables in our analysis. Lastly, negative beliefs and opinions about vaccine effectiveness (e.g., vaccine side effects) create significant barriers to vaccination that contribute to disparities in vaccination rates [Armstrong et al., 2001; Fiscella, 2005; O’Malley et al., 2006; Santibanez, Mootrey, Euler, and Janssen, 2010; Singleton, Santibanez, and Wortley, 2005]. Therefore, we examined whether HCP recommendations reach patients reporting similar barriers to vaccinations. In summary, these sample characteristics are grouped into demographic, socioeconomic, health, and access variables (Table 1A) as well as opinion about flu vaccine variables (Table 1B). 3.0 Statistical analysis All analyses were weighted to make the sample representative of the US adult population. We compared the rate of HCP recommendations by respondent characteristics to ! 8! investigate associations between demographic, socioeconomic, health, access variables, and HCP recommendation. Next, after adjusting for Table 1A and B variables, multivariate logistic regression models examined significant determinants of HCP recommendations. Sensitivity analyses on the model specifications included alternate definitions of the primary outcome (see Appendix). All analyses were conducted with Stata 11 (Stata Corp, College Station, TX). 4.0 Results 4.1 Unadjusted analyses of sample characteristics About one fifth of the population was greater than 65 years old (95%CI: 19.9-21.5) and about a quarter of the population was in the 18-34 year old age group (95% CI: 25.1- 27.4). The study population was 69.9% (95%CI: 68.6-71.1) Non-Hispanic, White only and 54.9% female (95%CI: 53.7-56.1). About a third were college graduates (95% CI: 34.9-37.1) and roughly a quarter had incomes greater than $75,000 per year (95% CI: 25.4-27.5). More than 80% had health insurance coverage (95% CI: 82.1-84.2) and about 30% reported having a chronic condition (95% CI: 29.3-31.4). Unadjusted analysis of the study population revealed older age was positively associated with receiving HCP recommendations (Table 1-1). For example, 51.8% (95% CI: 49.8- 53.9) of adults aged 65+ years received a HCP recommendation while 36.2% (95% CI: 33.6-38.8) of adults 18-34 years old received a HCP recommendation. Other important positive predictors of receiving HCP recommendations included being married and female gender. The chronically ill and those with health insurance were more likely to ! 9! report receiving HCP recommendation. For example, 43.1% (95% CI: 41.8-44.3) of insured adults received a HCP recommendation while 30.7% (95% CI: 26.6-35.0) of uninsured adults received a HCP recommendation. And, approximately one half (95%CI: 46.6-51.0) and a third (95%CI: 31.2-35.3) of patients with ≥4 and one doctor’s visit received a HCP recommendation, respectively. 4.2 Unadjusted analyses of opinions about influenza vaccines Table 1-2 reports the opinions about vaccine effectiveness, risk of getting sick with the flu without the vaccine, and worry about getting sick from the vaccine. First, the majority of our study sample considered the seasonal and H1N1 influenza vaccine as somewhat and very effective. These patient groups were more likely to have received a HCP recommendation. For example, 51.9% (95%CI: 50.0-53.9) of patients that considered the seasonal vaccine as very effective received a HCP recommendation compared to 26.9% (95%CI: 22.5-31.9) of patients that considered the seasonal vaccine as not at all effective. Second, about two thirds of the study population thought that they had a “very low” or “somewhat low” risk of getting sick with flu without either vaccine. However, the perception of getting sick with the flu, without either vaccine, was positively associated with receiving a HCP recommendation. Lastly, the study population was not predominantly worried about getting sick from either the seasonal or H1N1 flu vaccine; where patients with high levels of worry were more likely to report having a HCP recommendation. !10! 4.3 Adjusted analyses of ACIP priority groups and HCP recommendation receipt The logistic regression model for the primary outcome of this study (Table 1-3) demonstrates ACIP priority groups such as adults aged 65+ and those reporting a chronic medical condition were more likely to report a HCP recommendation compared to their non-ACIP counterparts. Compared to 18-34 year olds, adults 65+ were 73.8% more likely to receive a recommendation (95%CI: 1.427-2.116, Table 1-3). Adults with a chronic medical condition were 58.0% (95%CI: 1.414-1.765, Table 1-3) more likely to report a recommendation versus adults with no chronic medical condition. And, healthcare and non-healthcare workers did not experience differences in reporting recommendations. 4.4 Adjusted analyses of determinants of HCP recommendations for flu vaccinations Similar to previous studies on predictors of influenza vaccinations, our adjusted logistic model demonstrates patients with health insurance were more likely to receive a HCP recommendation compared to patients with no health insurance (OR: 1.448, 95%CI: 1.165-1.801, Table 1-3). Also, high number of physician visits contributed to an increased likelihood of receiving a recommendation. For example, compared to adults visiting a doctor at least 4 times, patients visiting a doctor once were 28.7% less likely to receive a recommendation (OR: 0.713, 95%CI: 0.618-0.821, Table 1-3). Lastly, there were racial/ethnic differences in HCP recommendations; where Non-Hispanic Black only adults were more likely to receive a HCP recommendation when compared to Non- Hispanic White only adults (OR: 1.284, 95%CI: 1.064-1.549). Differences in income levels were not associated with likelihood of receiving HCP recommendations. !11! 4.5 Sensitivity analyses of adjusted models The sensitivity analysis of the primary outcome can be found in Appendix 1A-1C. The results from these model specifications were similar to the Table 2 results with few exceptions. For example, the results for any H1N1 vaccine recommendation outcome suggest no differences in HCP recommendation rates between racial/minorities and Non- Hispanic, White adults (Model 2 – Appendix 1A). There were no significant correlations associated with H1N1 vaccine only recommendations and age, race/ethnicity, and frequency of doctor’s visits (Model 1C – Appendix 1B). Combined, these results suggest recipients of H1N1 flu vaccine recommendations were evenly distributed amongst age, race/ethnicity, and frequency of doctor’s visit when compared to seasonal flu vaccine recommendations. The MNL model results (Appendix 1B) demonstrate differences in recommendation rates in key variables (i.e., age, race/ethnicity, and chronic illness status) were driven by seasonal flu vaccine recommendations rather than H1N1 flu vaccine recommendations. For example, Table 1-3 demonstrates adults age 65+ and Non-Hispanic Black groups were more likely to receive recommendations compared to adults age 18-34 years and Non-Hispanic Whites, respectively. From the MNL model, these differences do not exist when the outcome is an H1N1 vaccine only recommendation (Model 1C – Appendix 1B). A similar trend can be ascertained when comparing the MNL model results to any H1N1 flu vaccine recommendation (Model 2 – Appendix 1A), where receiving a !12! recommendation for both seasonal and H1N1 flu vaccines does not contribute to differences in recommendations for age and race/ethnicity groups. Lastly, similar to previous research, a HCP recommendation was significantly associated with obtaining seasonal and H1N1 flu vaccinations (Model 2 - Appendix 1C). Furthermore, this analysis reveals patients with low levels of opinion about seasonal flu vaccine effectiveness were less likely to receive a recommendation and either vaccine when compared to patients with high opinions about vaccine effectiveness (Models 1 and 2 – Appendix 1C). Patients with low risk perceptions of getting sick with seasonal flu without the vaccine experienced similar negative associations with recommendations and vaccinations. 5.0 Discussion Previous research on healthcare provider recommendations for influenza vaccination considers its effect on vaccination uptake. Our study adds to the literature by characterizing the patient groups reporting a HCP recommendation from the general population. This study provides two important findings about determinants of HCP recommendations for flu vaccines. First, our study demonstrates that Non-Hispanic, Black adults (a racial/ethnic group typically less likely to obtain a flu vaccine [Centers for Disease Control and Prevention, 2011; Fiscella, 2005; Hebert, Frick, Kane, and McBean, 2005; Lu et al., 2013; O’Malley and Forrest, 2006]) were more likely to receive a !13! recommendation compared to Non-Hispanic, White adults (Table 1-3). This contrasts with the HPV literature where Non-Hispanic, Black adolescents are less likely receive recommendations for HPV vaccines from their providers compared to Non-Hispanic, Whites [Ylitalo, Lee, and Mehta, 2013]. These findings suggest that differences in HCP recommendation rates by race/ethnicity are not a likely explanation for disparities in flu vaccination rates by race/ethnicity. This naturally raises the question: Why do Non- Hispanic Black adults have lower vaccination rates despite receiving higher rates of HCP recommendations? One potential reason is that Non-Hispanic Black adults might be less receptive to advice from healthcare providers. For example, this demographic group may be resistant to vaccinations [Hebert, Frick, Kane, and McBean, 2005] or more concerned about being experimented upon by physicians without consent [Fiscella, 2005]. Another explanation might be that racial/ethnic minority groups experience healthcare discrimination that may influence interactions within the healthcare setting leading to low patient adherence [MacIntosh, Desai, Lewis, Jones, and Nunez-Smith, 2013]. Finally, other differences between racial/ethnic minority groups and Non-Hispanic Whites such as socio-economic status and trust in modern health care might explain the disparities in vaccination rates [O’Malley and Forrest, 2006]. Future research should carefully evaluate the importance of each of the above explanations to identify potential interventions for improving vaccination rates among minority racial/ethnic groups. Second, ACIP priority groups are more likely to receive recommendations compared to non-ACIP groups and recommendations can contribute, in large part, towards obtaining a flu vaccination. For example, from Models 1 and 2 in Appendix 1C, our sample !14! population has a 33.1% increased probability of obtaining flu vaccination given a vaccine recommendation from their provider. Chronically ill adults have a 4.9% higher chance of obtaining a flu vaccine compared to non-chronically ill adults. Moreover, chronically ill adults have a 10.9% higher chance of receiving a recommendation for flu vaccines compared to non-chronically ill adults. Taken together, receiving flu vaccine recommendations from providers explains 73.6% (i.e., 33.1% times 10.9% and divided by 4.9%) of the difference in flu vaccination rates between chronically ill and non- chronically ill adults. The HealthyPeople 2020 influenza vaccination goals suggest further research is needed to improve vaccination rates for all patient groups [HealthyPeople2020, 2013]. For example, the baseline vaccination rate for noninstitutionalized adults aged 65+ is 66.6%; where the HealthyPeople 2020 target is 90.0%. With evidence that HCP recommendations for flu vaccinations are significant determinants of a person obtaining a flu vaccination, this study demonstrates certain patient groups did not experience flu vaccine recommendations from their provider during the 2009-2010 flu season. The National Vaccine Advisory Committee has recently outlined recommendations as a standard for providers [Bhatt et al., 2014]; which is a promising step towards ensuring recommendations reach all patient groups. It is also important to consider how provider recommendations reach patients and how providers respond when patients voice resistance to vaccine recommendations [Opel et al., 2013]. !15! The ACIP has identified strategies for implementing vaccination recommendations in healthcare settings [Fiore et al., 2009]. These strategies include the offering of the vaccine during visits throughout the influenza season and a plan for identifying persons recommended for vaccination. Our study suggests not every visit incurs a recommendation for the flu vaccine. The likelihood of a recommendation during one or two visits to the doctor was significantly less when compared to at least four visits. However, poor patient health may be associated with more visits to the healthcare provider; thus creating greater opportunities to receive a recommendation. While we control for health in our regression analyses using a chronic illness indicator, there might be other differences in health status or other factors (such as vaccine supply/availability) that are correlated with number of physician visits. There are several limitations to this study. First, it is likely that some doctor’s visits were to non-primary care physicians or healthcare providers who are less likely to recommend seasonal and H1N1 flu vaccination. Ideally, we would like to distinguish between visits to primary care physicians versus other providers but we did not have data to make this distinction. Second, there may be recall bias related to recommendations for influenza vaccinations. Further, the distinction between seasonal and H1N1 flu vaccinations, unique to the 2009-2010 flu season, may not be fully understood by survey-respondents; however, our sensitivity analyses related to this distinction suggest generally robust results. Third, the 2009-2010 flu season experienced the H1N1 flu pandemic. Consequently, these findings may not generalize to other flu seasons. Although, these results are relevant to future influenza pandemics because policies related to ensuring !16! influenza vaccination coverage will benefit from our study conclusions on HCP recommendations. 6.0 Conclusions Healthcare provider recommendations for influenza vaccinations play an important role in improving vaccination rates, especially among ACIP priority groups. This study demonstrates these priority groups were more likely to report healthcare provider recommendations for influenza vaccinations during the 2009-2010 flu season when compared to non-priority groups. Patients with more doctors’ visits have an increased likelihood of receiving a recommendation. Unlike similar studies in HPV vaccine recommendations, racial/ethnic minority groups were more likely to receive recommendations compared to Non-Hispanic, White adults. !17! 7.0 References Annunziata K, Rak A, Del Buono H, et al. Vaccination rates among the general adult population and high-risk groups in the United States. PLoS One 2012;7(11):e50553. Armstrong K, Merlin M, Schwartz JS, et al. Barriers to influenza immunization in a low- income urban population. Am J Prev Med 2001;20(1):21-25. Bhatt A, Bridges C, Donoghue K, Fernandez C, Gehring R, Hall LL, et al. Recommendations from the National Vaccine Advisory committee: standards for adult immunization practice. Public Health Rep. 2014; 129(2): 115-23. Centers for Disease Control and Prevention. Prevention and control of seasonal influenza with vaccines: recommendations of the Advisory Committee on Immunization Practices (ACIP), 2009. MMWR Recommendations and Reports 2009;58(RR08):1-52. Centers for Disease Control and Prevention. Use of influenza A (H1N1) 2009 monovalent vaccine: recommendations of the Advisory Committee on Immunization Practices (ACIP), 2009. MMWR Recommendations and Reports 2009;58(RR10):1-8. Centers for Disease Control and Prevention. CDC Health Disparities and Inequalities Report – United States, 2011. Health-care access and preventive health services: Influenza vaccination coverage – United States, 2000-2010. MMWR 2011;60(Suppl):38-42. Ding H, Santibanez TA, Jamieson DJ, et al. Influenza vaccination coverage among pregnant women – National 2009 H1N1 Flu Survey (NHFS). Am J Obstet Gynecol 2011;204(6):S96-S106. Dominguez SR, Daum RS. Physician knowledge and perspectives regarding influenza and influenza vaccination. Human Vaccines 2005;1(2):74-79. Fiore AE, Shay DK, Broder K, et al. Prevention and control of influenza with vaccines: recommendations of the Advisory Committee on Immunization Practices (ACIP), 2009. MMWR Recomm Rep 2009, 58(RR08):1-52. Fiebach NH, Viscoli CM. Patient acceptance of influenza vaccination. Am J Med 1991;91(4):393-400. Fiscella K. Commentary – Anatomy of racial disparity in influenza vaccination. HSR: Health Services Research 2005;40(2): 539-550. !18! Gnanasekaran SK, Finkelstein JA, Hohman K, et al. Parental perspective on influenza vaccination among children with asthma. Public Health Rep 2006;121(2):181- 188. Groshkopf LA, Shay DK, Shimabukuro TT, et al. Prevention and control of seasonal influenza with vaccines: recommendations of the Advisory Committee on Immunization Practices (ACIP) – United States, 2013-2014. MMWR 2013;62(RR07):1-43. Gu Q, Sood N. Do people taking flu vaccines need them the most? PLoS One 2011;6(12):e26347. HealthyPeople2020. Immunization and Infections Diseases. 2013. http://www.healthypeople.gov/2020/topicsobjectives2020/objectiveslist.aspx?topi cId=23. Hebert PL, Frick KD, Kane RL, McBean AM. The causes of racial and ethnic differences in influenza vaccination rates among elderly Medicare beneficiaries. HSR: Health Services Research 2005;40(2): 517-538. Hemingway CO, Poehling KA. Change in recommendation affects influenza vaccinations among children 6 to 59 months of age. Pediatrics 2004;114(4):948- 952. Jessop AB, Dumas H, Moser CA. Delivering influenza vaccine to high-risk adults: subspecialty physician practices. American Journal of Medical Quality 2013;28:232-237. Levy DJ, Ambrose CS, Oleka N, Lewin EB. A survey of pediatricians’ attitudes regarding influenza immunization in children. BMC Pediatrics 2009;9(8):1-5. Lu PJ, Singleton JA, Euler GL, et al. Seasonal influenza vaccination coverage among adult populations in the United States, 2005-2011. Am J Epidemiol. 2013;178(9):1478-1487. Lyn-Cook R, Halm EA, Wisnivesky JP. Determinants of adherence to influenza vaccination among inner-city adults with persistent asthma. Prim Care Resp 2007;16(4):229-235. MacIntosh T, Desai MM, Lewis TT, Jones BA, Nunez-Smith M. Socially-assigned race, healthcare discrimination and preventive healthcare services. PLoS One. 2013;8(5): 1-7 Mirza A, Subedar A, Fowler SL, et al. Influenza vaccine: awareness and barriers to immunization in families of children with chronic medical conditions other than asthma. South Med J 2008;101(11):1101-1105. !19! Nichol KL, Lofgren RP, Gapinski J. Influenza vaccination: knowledge, attitudes, and behavior among high-risk outpatients. Arch Intern Med 1992;152(1):106-110. Nichol KL, Zimmerman R. Generalist and subspecialist physicians’ knowledge, attitudes, and practices regarding influenza and pneumococcal vaccinations for elderly and other high-risk patients: a nationwide survey. Arch Intern Med 2001;161:2702-2708. O’Malley AS, Forrest CB. Immunization disparities in older Americans: determinants and future research needs. Am J Prev Med 2006;31(2):150-158. Opel DJ, Heritage J, Taylor JA, Mangione-Smith R, Salas HS, DeVere V, et al. The architecture of provider-parent vaccine discussion at health supervision visits. Pediatrics 2013; 132(6): 1037-1046. Pandolfi E, Marino MG, Carloni E, et al. The effect of physician’s recommendation on seasonal influenza immunization in children with chronic diseases. BMC Public Health 2012;12(1):984. Poehling KA, Speroff T, Dittus RS, et al. Predictors of influenza virus vaccination status in hospitalized children. Pediatrics 2001;108(6):1-6. Santibanez TA, Mootrey GT, Euler GL, Janssen AP. Behavior and beliefs about influenza vaccine among adults aged 50-64 years. Am J Health Behav 2010;34(1):77-89. Singleton J, Santibanez T, Wortley P. Influenza and pneumococcal vaccination of adults aged ≥65 racial/ethnic differences. Am J Prev Med 2005;29(5):412-420. Takayama M, Wetmore CM, Mokdad AH. Characteristics associated with the uptake of influenza vaccination among adults in the United States. Prev Med 2012;54(5):358-362. U.S. Department of Health and Human Servces (DHHS). National Center for Health Statistics. The National 2009 H1N1 Flu Survey, Hyattsville, MD: Centers for Disease Control and Prevention, 2012. Information about the NHFS is located at http://www.cdc.gov/nchs/nis/about_nis.htm#h1n1. Vadaparampil ST, Malo TL, Kahn JA, Salmon DA, Lee JH, Quinn GP, et al. Physicians’ human papillomavirus vaccine recommendations, 2009 and 2011. Am J Prev Med 2014;46(1):80-4. Ylitalo KR, Lee H, Mehta NK. Health care provider recommendation, human papillomavirus vaccination, and race/ethnicity in the US National Immunization Survey. Am J Public Health 2013;103(1):164-9. !20! 8.0 Tables Table 1-1. Descriptive statistics of healthcare provider recommendations N 23358 90507676 Healthcare provider recommendation b 95% Confidence Interval 95% Confidence Interval Variable Unweighted Weighted, % Lower Limit Upper Limit Weighted % Lower Limit Upper Limit DEMOGRAPHIC Age group 18-34 3833 26.2 25.1 27.4 36.2 33.6 38.8 35-44 2969 16.5 15.5 17.5 37.3 34.2 40.4 45-54 4396 20.1 19.2 21.1 35.2 32.6 37.8 55-64 5064 16.6 15.8 17.4 47.1 44.7 49.6 65+ 7096 20.7 19.9 21.5 51.8 49.8 53.9 Race/ethnicity Hispanic 1244 11.6 10.6 12.7 41.5 36.5 46.6 Non-Hispanic, Black Only 2083 12.6 11.7 13.5 43.0 d 39.2 46.8 Non-Hispanic, White Only 18713 69.9 68.6 71.1 41.0 39.8 42.2 Non-Hispanic, Other or Multiple Race 1318 5.9 5.4 6.6 38.9 34.1 44.0 Gender Male 8708 45.1 43.9 46.3 37.8 36.0 39.7 Female 14650 54.9 53.7 56.1 44.0 42.5 45.5 Married Yes 11697 52.7 51.5 53.9 42.9 41.3 44.5 !21! No 10559 41.5 40.3 42.7 39.8 38.0 41.6 Missing 1102 5.8 5.2 6.5 35.5 30.3 41.1 Number of children 0 17176 64.4 63.2 65.6 41.8 40.5 43.2 1 2598 14.6 13.7 15.5 40.5 37.2 44.0 2 2121 12.3 11.5 13.2 37.5 34.0 41.2 3 1330 7.8 7.1 8.6 43.1 38.0 48.3 Missing 133 1.0 0.7 1.3 40.1 e 25.9 54.3 Number of people in household 1 6721 16.8 16.0 17.6 41.1 38.9 43.4 2 8846 34.6 33.5 35.7 43.0 41.3 44.8 3 3194 17.7 16.8 18.7 40.3 37.6 43.2 4 2788 18.0 17.0 19.1 38.7 35.6 41.9 5 1432 9.3 8.5 10.2 40.4 35.7 45.2 6 276 2.3 1.9 2.9 43.2 d 32.8 53.7 7 101 1.2 0.9 1.7 42.8 f 25.9 59.7 3-category MSA status MSA, principle city 6864 31.9 30.8 33.2 41.0 38.7 43.3 MSA, not principle city 10403 51.4 50.2 52.6 41.3 39.7 42.9 Non-MSA 6091 16.6 15.9 17.4 41.3 39.0 43.6 Census region of residence a Region 1 4030 19.0 18.4 19.6 47.2 44.5 49.9 Region 2 5055 21.9 21.3 22.5 41.3 39.1 43.4 Region 3 8562 37.2 36.5 38.0 39.4 37.6 41.2 Region 4 5711 21.9 21.2 22.6 39.0 36.1 42.0 Interview date Jan-10 1279 4.6 4.2 5.0 40.3 35.8 45.1 !22! Feb-10 3654 17.8 16.9 18.7 41.9 39.2 44.8 Mar-10 4124 18.6 17.7 19.6 40.4 37.7 43.1 Apr-10 4380 19.4 18.4 20.4 40.0 37.4 42.7 May-10 5424 19.6 18.7 20.5 42.5 40.1 45.0 Jun-10 4497 20.1 19.1 21.1 41.4 38.6 44.2 SOCIOECONOMIC Self-report education level <12 years 1970 9.9 9.1 10.8 42.6 38.2 47.0 12 years 5152 21.1 20.1 22.0 44.3 41.8 46.8 Some college 6301 27.3 26.2 28.4 40.5 38.3 42.9 College graduate 8863 35.9 34.9 37.1 40.5 38.8 42.3 Missing 1072 5.8 5.2 6.4 34.9 29.7 40.5 Income poverty status Above poverty threshold, >=$75,000 income 5865 26.5 25.4 27.5 39.6 37.5 41.7 Above poverty threshold, <$75,000 income 11383 44.9 43.7 46.1 42.2 40.4 43.9 Below poverty threshold 2324 11.8 10.9 12.7 44.1 40.1 48.2 Poverty status unknown 3786 16.9 16.0 17.8 39.2 36.4 42.0 Work status Employed 10960 50.4 49.2 51.6 37.8 36.1 39.4 Unemployed 1170 6.5 5.9 7.2 37.6 32.4 43.1 Not in labor force 10088 36.9 35.8 38.1 47.6 45.7 49.4 Don't know/Refused/Missing 1140 6.2 6.2 5.5 34.9 30.0 40.3 Works in health care field No 20338 86.2 85.3 87.0 40.9 39.7 42.2 Yes 2428 11.2 10.4 12.0 44.9 41.2 48.7 !23! Missing 592 2.6 2.3 3.1 34.7 28.2 41.8 Home rented or owned Home is owned 16654 65.2 64.0 66.4 42.4 41.0 43.7 Home is rented or other arrangement 5062 26.6 25.4 27.8 39.9 37.3 42.6 Don't know/Refused/Mising 1642 8.2 7.5 9.0 36.4 32.2 40.7 HEALTH Chronic medical condition c No 14710 66.6 65.5 67.7 36.1 34.6 37.5 Yes 7914 30.3 29.3 31.4 53.2 51.1 55.2 Missing 734 3.0 2.7 3.5 34.9 29.2 41.1 Health status Sick with fever and cough or sore throat in past month No 21746 92.6 91.9 93.2 40.9 39.7 42.1 Yes 1161 5.5 5.0 6.1 49.1 44.0 54.3 Missing 451 1.9 1.6 2.3 32.9 25.5 41.2 Other people in house with fever and cough or sore throat No 20127 81.6 80.5 82.6 41.1 39.8 42.3 Yes 2893 16.7 15.7 17.7 42.6 39.3 46.1 Missing 338 1.8 1.5 2.1 33.5 25.6 42.5 ACCESS Has health insurance coverage Yes 20530 83.2 82.1 84.2 43.1 41.8 44.3 No 1800 11.0 10.1 12.0 30.7 26.6 35.0 Don't 1028 5.8 5.2 6.5 34.8 29.6 40.4 !24! know/Refused/Missing Number of times seen doctor since August 2009 >=4 6633 28.8 27.7 29.9 48.8 46.6 51.0 3 5980 14.7 13.8 15.6 44.4 41.1 47.7 2 3481 27.1 26.0 28.2 39.6 37.4 41.9 1 6780 27.8 26.7 28.8 33.2 31.2 35.3 Missing 484 1.7 1.5 2.0 40.8 33.7 48.3 a Region 1: CT, ME, MA, NH, VT, RI, NJ, NY, PA; Region 2: IL, IN, MI, OH, WI, IA, KS, MN, MO, NE, ND, SD; Region 3: DE, DC, FL, GA MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK, TX; Region 4: AZ, CO, ID, MT, NV, NM, UT, WY, AK, CA, HI, OR, WA b Healthcare provider recommendation is a binary variable where the value of one is equal to receiving a recommendation for the seasonal flu vaccination only, H1N1 flu vaccination only, or both vaccinations. c This indicates whether the person has any of the following chronic medical conditions: asthma or an other lung condition, diabetes, a heart condition, a kidney condition, sickle cell anemia or other anemia, a neurological or neuromuscular condition, a liver condition, or a weakened immune system caused by a chronic illness or by medicines taken for a chronic illness. d 1 stratum omitted because it contains no subpopulation members. e 4 strata omitted because it contains no subpopulation members. f 10 strata omitted because it contains no subpopulation members. !25! Table 1-2. Descriptive statistics of healthcare provider recommendations (continued) N 23358 90507676 Healthcare provider recommendation a 95% Confidence Interval 95% Confidence Interval Variable Unweighted Weighted, % Lower Limit Upper Limit Weighted % Lower Limit Upper Limit OPINIONS ABOUT FLU VACCINE Opinion: Effectiveness of H1N1 vaccine Very effective 6977 30.3 29.2 31.5 49.9 47.6 52.2 Somewhat effective 10378 44.2 43.0 45.4 39.5 37.8 41.2 Not very effective 1575 7.1 6.5 7.8 32.1 28.3 36.2 Not at all effective 717 3.5 3.1 3.9 28.8 23.5 34.7 Don't know/Refused/Missing 3711 14.9 14.1 15.7 35.8 33.1 38.7 Opinion: Risk of getting sick with H1N1 flu without vaccine Very high 1469 6.9 6.3 7.7 60.9 55.6 66.0 Somewhat high 4581 19.5 18.5 20.5 53.1 50.3 55.9 Somewhat low 8604 35.9 34.8 37.0 39.8 37.9 41.8 Very low 7502 32.9 31.9 34.1 31.8 29.9 33.6 Don't know/Refused/Missing 1202 4.7 4.2 5.2 39.2 34.1 44.6 Opinion: Worry about getting sick from the H1N1 vaccine Very worried 1904 9.7 8.9 10.5 46.1 41.7 50.6 Somewhat worried 5053 22.7 21.7 23.8 46.6 44.0 49.2 Not very worried 8009 33.6 32.5 34.8 39.5 37.6 41.5 !26! Not at all worried 8065 32.6 31.5 33.7 37.9 36.0 39.7 Don't know/Refused/Missing 327 1.3 1.1 1.7 39.2 34.1 44.6 Opinion: Effectiveness of seasonal vaccine Very effective 9414 37.4 36.3 38.6 51.9 50.0 53.9 Somewhat effective 9781 43.8 42.6 45.0 37.4 35.7 39.2 Not very effective 1899 8.7 8.1 9.4 30.7 27.0 34.5 Not at all effective 968 4.7 4.2 5.3 26.9 22.5 31.9 Don't know/Refused/Missing 1296 5.3 4.9 5.9 27.0 22.9 31.5 Opinion: Risk of getting sick with seasonal flu without vaccine Very high 2697 12.2 11.3 13.1 56.1 52.1 60.0 Somewhat high 6895 28.1 27.1 29.2 53.1 50.8 55.4 Somewhat low 7754 33.4 32.3 34.5 35.0 33.2 36.9 Very low 5038 22.6 21.6 23.7 27.7 25.5 30.0 Don't know/Refused/Missing 974 3.7 3.3 4.1 39.9 34.4 45.7 Opinion: Worry about getting sick from the seasonal vaccine Very worried 1577 7.7 7.0 8.5 44.3 39.4 49.3 Somewhat worried 4336 19.9 19.0 20.9 45.9 43.2 48.6 Not very worried 6704 29.7 28.6 30.8 40.1 37.9 42.4 Not at all worried 10313 40.9 39.8 42.2 39.5 37.8 41.1 Don't know/Refused/Missing 428 1.7 1.5 2.1 33.7 26.0 42.4 a Healthcare provider recommendation is a binary variable where the value of one is equal to receiving a recommendation for the seasonal flu vaccination only, H1N1 flu vaccination only, or both vaccinations. !27! Table 1-3. Multivariate logistic regression for healthcare provider recommendations a 95% Confidence Interval Variable Odds Ratio Lower Limit Upper Limit DEMOGRAPHIC Age group 18-34 Reference 35-44 1.001 0.828 1.211 45-54 0.952 0.798 1.136 55-64 1.483 1.237 1.778 65+ 1.738 1.427 2.116 Race/ethnicity Hispanic 1.165 0.928 1.462 Non-Hispanic, Black Only 1.284 1.064 1.549 Non-Hispanic, White Only Reference Non-Hispanic, Other or Multiple Race 1.041 0.830 1.304 Gender Male Reference Female 1.140 1.027 1.266 Married Yes 1.091 0.941 1.264 No Reference Missing 1.302 0.622 2.724 Number of children 0 Reference 1 1.175 0.958 1.440 2 1.104 0.840 1.453 !28! 3 1.454 0.983 2.151 Missing 2.016 0.937 4.338 Number of people in household 1 1.053 0.897 1.236 2 1.067 0.865 1.317 3 1.057 0.811 1.377 4 0.995 0.685 1.445 5 1.057 0.608 1.839 6 0.816 0.358 1.857 7 Reference 3-category MSA status MSA, principle city 1.097 0.946 1.271 MSA, not principle city 1.060 0.933 1.205 Non-MSA Reference Census region of residence b Region 1 Region 2 0.752 0.648 0.872 Region 3 0.666 0.578 0.767 Region 4 0.662 0.559 0.784 Interview date Jan-10 Reference Feb-10 1.089 0.857 1.383 Mar-10 0.975 0.769 1.237 Apr-10 0.958 0.757 1.212 May-10 1.084 0.859 1.368 Jun-10 1.026 0.805 1.308 !29! SOCIOECONOMIC Self-report education level <12 years Reference 12 years 1.180 0.946 1.473 Some college 1.092 0.870 1.370 College graduate 1.061 0.848 1.326 Missing 0.903 0.474 1.719 Income poverty status Above poverty threshold, >=$75,000 income Reference Above poverty threshold, <$75,000 income 1.049 0.921 1.194 Below poverty threshold 1.065 0.841 1.348 Poverty status unknown 1.049 0.868 1.269 Work status Employed Reference Unemployed 1.021 0.791 1.318 Not in labor force 1.066 0.935 1.216 Don't know/Refused/Missing 0.921 0.512 1.659 Works in health care field No Reference Yes 1.101 0.931 1.301 Missing 1.193 0.620 2.298 Home rented or owned Home is owned Reference Home is rented or other arrangement 0.940 0.812 1.089 Don't know/Refused/Missing 0.960 0.695 1.326 !30! HEALTH Chronic medical condition c No Reference Yes 1.580 1.414 1.765 Missing 1.113 0.731 1.694 Health status Sick with fever and cough or sore throat in past month No Reference Yes 1.115 0.882 1.408 Missing 0.867 0.399 1.886 Other people in house with fever and cough or sore throat No Reference Yes 0.966 0.821 1.137 Missing 0.778 0.413 1.465 ACCESS Has health insurance coverage Yes 1.448 1.165 1.801 No Reference Don't know/Refused/Missing 1.154 0.522 2.548 Number of times seen doctor since August 2009 >=4 Reference 3 0.915 0.778 1.076 2 0.829 0.721 0.952 1 0.713 0.618 0.821 Missing 0.746 0.531 1.048 !31! a The regression model controls for variables reported in Table 1-2. The relationships between the Table 1-2 variables and recommendations can be found in the Appendix tables. The outcome for this model was defined as a binary variable equal to one when the respondent indicated they received recommendations for H1N1 flu vaccination only, seasonal flu vaccination only, or both vaccinations. b Region 1: CT, ME, MA, NH, VT, RI, NJ, NY, PA; Region 2: IL, IN, MI, OH, WI, IA, KS, MN, MO, NE, ND, SD; Region 3: DE, DC, FL, GA MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK, TX; Region 4: AZ, CO, ID, MT, NV, NM, UT, WY, AK, CA, HI, OR, WA c This indicates whether the person has any of the following chronic medical conditions: asthma or an other lung condition, diabetes, a heart condition, a kidney condition, sickle cell anemia or other anemia, a neurological or neuromuscular condition, a liver condition, or a weakened immune system caused by a chronic illness or by medicines taken for a chronic illness. !32! Chapter 2. What Explains Racial/Ethnic Disparities in Influenza Vaccination Rates? Abstract Objectives: Explain causes of racial/ethnic differences in influenza vaccination rates in the United States (US) general population. Methods: Racial/ethnic vaccination disparities were decomposed into differences in observable and unobservable characteristics. Multivariate decomposition of logit models for influenza vaccinations using a US national sample of adults 18+ from the National 2009 H1N1 Flu Survey. Covariates included demographics, socioeconomic status, beliefs about influenza vaccine effectiveness and Advisory Committee on Immunization Practices (ACIP) priority groups. Results: The majority of the share in influenza vaccination disparities between Non- Hispanic Whites (Whites), Non-Hispanic Blacks (Blacks), and Hispanics can be explained by differences in observable characteristics. Whites report higher influenza vaccination rates by 13.7 and 12.0 percentage points relative to Blacks and Hispanics, respectively. 29.2% of the difference in influenza vaccination rates between Whites and Blacks can be explained by differences in beliefs about influenza vaccines. 14.0% of the same difference can be explained by differences in socioeconomic status. 6.1% of the difference in influenza vaccination rates between Whites and Hispanics can be explained by differences in beliefs about influenza vaccines. 22.6% of the same difference can be explained by differences in socioeconomic status. Healthcare provider recommendations can modestly decrease influenza vaccination disparities between these racial/ethnic groups. !33! Conclusion: A majority of the differences in influenza vaccination rates between Whites and Blacks and Whites and Hispanics can be explained by differences in observable characteristics. In particular, influenza vaccination beliefs are significant contributors to explaining influenza vaccination differences. !34! 1.0 Introduction Influenza is a contagious respiratory illness and influenza (i.e., the flu) infections generate >200,000 hospitalizations and 24,000 deaths, on average, in the United States (US). Annual influenza vaccinations can significantly reduce influenza-related mortality and morbidity [Groshkpof et al., 2013]. Since 2009, influenza vaccination coverage among adults aged 18 or older remain below the HealthyPeople2020 target of 70% [Centers for Disease Control and Prevention (CDC), 2014; HealthPeople2020, 2014]. In particular, significant racial disparities in influenza vaccination coverage persist among US adults [CDC, 2008; CDC, 2011]. Recent coverage estimates report Non- Hispanic White adults achieve significantly greater coverage rates than Non-Hispanic Black, Hispanic, and Other race adults [CDC, 2014; Lu et al., 2013]. The National Vaccine Advisory Committee (NVAC) considers racial/ethnic disparities a significant barrier in immunization receipt; thus underscoring the need to further evaluate causes of the disparity. Prior work has discussed racial/ethnic disparities in influenza vaccinations can arise from differences in access to care, comorbidity rates, attitudes and beliefs about the vaccine, and provider discrimination [Hebert et al., 2005]. These causes are difficult to measure because not all disparities are racial/ethnic in nature and can be a function of socioeconomic status, vaccine supply and availability issues, and attitudes toward health care and healthcare providers [Fiscella, 2005; Musa et al., 2009; NVAC, 2012; O’Malley et al., 2005]. !35! Evidence-based interventions to reduce racial/ethnic gaps in healthcare predominantly involve delivering education and training followed closely by care team restructuring and providing reminders [Clarke AR, 2013]. Currently, there exist inadequate examination in the literature on how healthcare provider recommendations can reduce disparities, especially among racial/ethnic groups. These recommendations are standards of practice endorsed by NVAC to ensure patients acquire needed vaccinations; with the intent of reducing disparities in flu vaccination rates [Bhatt et al., 2014]. Given the multi-factorial causes of disparities in healthcare, how recommendations reduce disparities when compared to, for example, resistant attitudes to preventive medicine can inform new strategies to interventions improving minority health. Therefore, this study broadens the focus of possible causes of racial/ethnic disparities in influenza vaccinations among the US general population by incorporating differences in healthcare provider recommendations for influenza vaccination and attitudes and beliefs about the vaccination. 2.0 Methods 2.1 Data Data came from the public-use National 2009 H1N1 Flu Survey (NHFS) by the Centers of Disease Control and Prevention [CDC, 2012]; which was reviewed by the National Center for Health Statistics Disclosure Review Board to protect participant privacy and data confidentiality. Households were identified from all 50 US states and the District of Columbia; where both H1N1 and seasonal influenza vaccination coverage rates were evaluated at national and state levels. !36! This study identified adult survey-respondents age 18+ that were interviewed from January through June 2010. We identified adults that visited a doctor’s office, hospital, or clinic since August 2009 up to the interview date. This sample captures patients experiencing face-to-face flu vaccine recommendations that were likely tailored to the individual patient. We focused on adults because important respondent characteristics were only captured from adults (i.e., chronic medical condition status, work status, and beliefs about influenza vaccine effectiveness). The analysis was restricted to interviews conducted between January 2010 and June 2010 to allow for a sufficient time window to observe provider recommendations. This was consistent with the NHFS recommendation of analyzing factors associated with influenza vaccinations. We compare our NHFS findings to recent 2013 Behavioral Risk Factor Surveillance System (BRFSS) for temporal trends [CDC, 2014]. The 2013 BRFSS was chosen as the comparator dataset because the H1N1 flu vaccination did not exist as a separate vaccination for the observed period, which may limit the 2009 NHFS findings. And, the benefit of the 2009 NHFS is it’s reporting of survey-respondent vaccination beliefs and recommendations from a nationally representative sample. We evaluate the tradeoff between the 2009 NHFS with these respondent characteristics and the more current 2013 BRFSS without these characteristics. !37! 2.2 Explanatory variables This study contains individual characteristics that have shown to explain disparities in influenza vaccinations [Fiscella, 2005; Hebert et al., 2005]. First, access barriers such as less frequent use of care could contribute to disparities. We evaluate this barrier by including number of doctor’s visits. Second, minorities experience higher rates of comorbidities than whites. We investigate this effect by incorporating a dichotomous variable if survey respondents experience a chronic illness. Third, this study contains patient attitudes and beliefs by race/ethnicity because differing attitudes and beliefs can lead to disparities in vaccination rates. Lastly, provider characteristics such as discrimination in delivery of care in favor of white patients over minority patients along with minority patients potentially seeing providers less likely to administer vaccinations can lead to disparities in coverage. While these provider characteristics are difficult to measure, as a close proxy to these characteristics this study incorporates healthcare provider recommendations to explore whether differing rates in recommendations (i.e., delivery of care) contribute to vaccination disparities. Additional characteristics are also considered: age, gender, income, education, geographic region, and healthcare worker status. 2.3 Statistical analyses Descriptive statistics by race/ethnicity groups (Non-Hispanic Whites defined as Whites; Non-Hispanic Blacks defined as Blacks; and, Hispanics) estimated differences in demographics, socioeconomic, health and access, and beliefs about flu vaccines characteristics. Multivariable logistic regression models controlled for effects of these !38! confounding variables. Next, Blinder-Oaxaca decomposition estimated disparities in influenza vaccination rates between racial/ethnic groups due to: observed characteristics (i.e. the characteristic effect) and unobserved characteristics (i.e., the coefficients effect). In order to compare 2009 NHFS decompositions with the 2013 BRFSS decomposition, we conduct the NHFS decomposition without vaccination beliefs and recommendations to evaluate similarities with the BRFSS decomposition. All analyses were conducted with Stata 12 (Stata Corp, College Station, TX). 2.4 Decomposition model The decomposition technique utilizes separate influenza vaccination regression estimates for each racial/ethnic cohort. Using Non-Hispanic White adults (W) as the reference group, this decomposition compares Non-Hispanic Black adults (B) and Hispanic adults (H) with the reference group. The regressions contain the outcome variable (Y) as a dichotomous measure of influenza vaccination and a series of explanatory variables (X) that were described above. We employ the mvdcmp program developed for Stata 12.0 in measuring the decomposition effects [Powers et al., 2011]. This program overcomes path dependence and identification issues associated with nonlinear response models by generating weights from a first-order Taylor linearization of (1) and averaging coefficient effects of a set of dummy variables while changing the order of the reference groups, respectively. We describe the main empirical question in the following using W and B cohorts: ! ! −! ! =!+!= !(! ! ! ! )−!(! ! ! ! ) + !(! ! ! ! )−!(! ! ! ! ) (1) !39! The primary outcome (Y) of this study was a binary variable equal to one when adults self-report receiving both H1N1 and seasonal flu vaccinations and either H1N1 or seasonal flu vaccine since August 2009. F() is a logit regression model where X contains the explanatory variables. The mean difference in Y between W and B is decomposed into a characteristics effect (C) and coefficients effect (E). This allows us to account for how much of the mean difference in Y is comprised of the observed (C) and unobserved (E). The outcome differential contained in C explains group differences in levels of the explanatory variables (i.e. provider recommendations) across W and B. The differential accounted for by E reflects the unobserved heterogeneity between W and B. The Blinder-Oaxaca decomposition has been applied to health services research investigating racial/ethnic disparities in health insurance coverage [Pylypchuk and Selden, 2008; Wehby et al., 2012], drug therapy utilization [Bowblis and Yun, 2010], and preventive services [Bustamante et al., 2009; Bustamante et al., 2010; Laden, 2012]. However, this technique has not been extensively used to study influenza vaccination rates; especially in the presence of healthcare provider recommendations. A related study examines the contribution of observed and unobserved factors in disparities in a bundle of preventive care services that includes influenza vaccinations [Bustamante et al., 2010]. However, this bundling does not focus solely on influenza vaccinations. Therefore, this study adds to the literature by highlighting disparities in influenza vaccinations featuring a US general population. Given the recent NVAC standards of practice for provider recommendations, this study measures the extent healthcare provider recommendations !40! and beliefs about the flu vaccine explain racial/ethnic disparities in influenza vaccinations. 3.0 Results The sample shows Blacks and Hispanics have lower rates of seasonal and H1N1 flu vaccinations compared Whites (Table 2-1). Whites experienced lower rates of receiving H1N1 flu only vaccinations relative to Blacks and Hispanics (Table 2-1). This suggests the aggregate seasonal and H1N1 flu vaccination disparity is mainly comprised of differences in seasonal flu vaccinations. The sample of Whites was predominantly near the 65+ age range while Hispanics were predominantly 18-34 years old (Table 2-2). Compared to Whites, a greater proportion of Blacks and Hispanics were below the poverty threshold and did not have health insurance coverage (Table 2-2). All three race/ethnicities received the same rates of healthcare provider recommendations for seasonal and H1N1 flu vaccines (Table 2-2). Lastly, relative to Whites, Blacks and Hispanics worry about getting sick from either the seasonal or H1N1 vaccines (Table 2- 2). See Appendix Table 2A for additional explanatory variables by race/ethnicity. Table 2-3 reports pooled logit regression marginal effects controlling for observable factors and marginal effects by race/ethnicity. The percentage of Blacks that receive a flu vaccination is 7.6 percentage points lower than Whites. Hispanics report a 0.6 percentage point decrease in vaccination rates compared to Whites. The findings reveal the existence of disparities in flu vaccination rates between Blacks and Whites, but not between Hispanics and Whites. !41! Comparing average characteristics (Table 2-2) and coefficient estimates (Table 2-3) by race/ethnicity reveal significant variations in magnitude and direction of factors contributing to flu vaccination rates. To further characterize these results, Table 2-4 reports select decomposition results between Whites and Blacks. 61.9% of the disparity in flu vaccinations between Whites and Blacks can be explained by differences in characteristics while 38.1% can be explained by differences in coefficients. If Blacks had the same characteristics as Whites, then 8.5 percentage points of the 13.7 point difference in Black and White flu vaccination rates would not exist. Thus, more than half the vaccination differential constitutes differences in characteristics rather than how the characteristics affect the rate of flu vaccinations (i.e., coefficient differences). This is an important finding since Blacks on average have lower education levels, income, and health insurance coverage relative to Whites. Further, on average, Blacks worry more about getting sick from the seasonal and H1N1 vaccines relative to Whites (Table 2-2). To identify which variables drive variations in flu vaccinations, the aggregate effect is broken down into sub-aggregate effects (Table 2-4). Select sub-aggregate effects are shown and Appendix Table 2B reports the all sub-aggregate effects. Significant sub- aggregate characteristics effects for income and education are positive (Appendix Table 2B). This suggests if Blacks have similar incomes and education as Whites then the disparity between the two groups would decrease. For example, 0.0192 is the sum of all socioeconomic characteristic effects in Table 2-4; which means if Blacks had similar characteristics as Whites, then flu vaccination disparities would decrease by 2.32 !42! percentage points (e.g., 17% of the disparity share). The associated coefficients effect reveals if Blacks had similar coefficients as Whites, then flu vaccination disparities would decrease by 3.6% percentage points (e.g., 26% of the difference). Table 2-5 reports select decomposition results between Whites and Hispanics (see Appendix Table 2C for further sub-aggregate effects). Similar to Table 2-4, flu vaccination disparities between Whites and Hispanics are attributed to differences in characteristics effects. The sum of the select demographic and socioeconomic sub- aggregate effects constitutes a 7.4 point decrease (i.e., 62% of the disparity share) in flu vaccination disparities if Hispanics had similar characteristics as Whites. When compared to Table 2-4 decomposition results, differences between Whites and Blacks and between Whites and Hispanics can significantly explain the characteristics effect. This suggests the coefficients effect has little impact on explaining disparities in flu vaccinations rates for this sample. Comparisons to the 2013 BRFSS data reveal omitting beliefs and recommendations in the 2009 NHFS decomposition can recover comparable aggregate characteristic and coefficient effects as the 2013 BRFSS decomposition. For example, Appendix 2G reports sub-aggregate 2009 NHFS decompositions between Whites and Blacks (omitting flu vaccine beliefs and provider recommendations). This table is compared to Appendix 2E reporting decompositions between Whites and Blacks using the 2013 BRFSS. Omitting beliefs and recommendations from the 2009 decomposition reveal the characteristics effect share as 37% (compared to 41% from the BRFSS) and coefficient !43! effects share as 63% (compared to 59% from the 2013 decomposition). Sub-aggregate demographic, socioeconomics, and health and access effects contribute different percent shares between the 2009 and 2013 time periods. This is likely due to differences in the variables captured between the BRFSS and NHFS. Furthermore, Appendix 2F reports sub-aggregate NHFS decompositions between Whites and Hispanics (omitting flu vaccine beliefs and provider recommendations). This table is compared to Appendix 2D reporting BRFSS decompositions between Whites and Hispanics. The findings reveal the characteristics effect from 2009 data constitute 121% of the difference in vaccination rates (Appendix 2F) while the 2013 data characteristics effect constitute 93% of the difference (Appendix 2D). Lastly, by omitting beliefs from the decomposition, the resulting lack of their coefficient estimates may provide inaccurate coefficient differences between the racial/ethnic groups. Consequently, the characteristics effect tends to explain more of the difference in vaccination rates between the groups because beliefs are likely correlated to the other characteristic differences between the racial/ethnic groups (i.e., chronically ill patients may be more likely believe the vaccine is effective). These 2009 and 2013 time period trends between Whites and minorities reveal several important findings. First, omitting vaccination beliefs and provider recommendations from the 2009 decomposition recovers similar aggregate characteristics and coefficient effects as the 2013 decomposition. This suggests, despite differences in time periods, beliefs and provider recommendations are important factors in explaining disparities in !44! vaccination rates. Second, the tradeoff for using the 2009 NHFS for the relatively recent 2013 BRFSS is the ability to reveal the extent flu vaccine beliefs and recommendations explain racial/ethnic vaccination disparities (which is otherwise missing in national surveys such as the BRFSS). Although, it is important to note that influenza infections and vaccine effectiveness change year-to-year and any differences across time periods may reflect differences in influenza infectivity rather than differences in aggregate effects. 4.0 Discussion Exploring the factors driving racial/ethnic disparities in flu vaccination rates can inform appropriate public health policies. The decomposition of disparities in flu vaccination rates by race/ethnicity suggests Blacks and Hispanics are significantly different from Whites through the characteristics effect (i.e., observable differences). In particular, the relatively large sub-aggregate effects generated from health and access and beliefs about flu vaccines variables highlight the importance of these characteristics in explaining racial/ethnic disparities. Public health policies with the goal of reducing flu vaccination disparities by race/ethnicity should focus on why Blacks and Hispanics have different characteristics than Whites. This study shows the majority of flu vaccination disparities are not explained by behavioral differences when Blacks and Hispanics have similar characteristics as Whites (i.e., coefficient effects). Consequently, policy-makers and researchers should focus on improving socioeconomic and access levels between Whites and Blacks/Hispanics. In particular, flu vaccination disparities can be reduced by four percentage points if Blacks expressed similar beliefs about flu vaccines as Whites. !45! National health policies consider reducing racial/ethnic disparities in preventive treatments a priority; thus underscoring the need to examine factors associated with variations in flu vaccinations across racial/ethnic groups. Recently, healthcare provider recommendations for flu vaccinations became a standard of care for providers. This study demonstrates healthcare provider recommendations can modestly reduce disparities in flu vaccination rates when compared against other factors. For example, Blacks experiencing similar rates of recommendations as Whites behave differently when receiving flu vaccines. This study demonstrates if Blacks have similar coefficient estimates as Whites for provider recommendations, then 2.5 percentage points of the 13.6 point difference in flu vaccination rates would not exist. Yet, disparities would significantly increase, although by less than 1 point, if Blacks experienced the same rates of recommendations as Whites. This means there is potential for recommendations to reduce racial/ethnic disparities in flu vaccination rates. However, it may need to be combined with, for example, ensuring Blacks have similar beliefs about flu vaccines as Whites. A particularly interesting finding is the relatively lower share that differences in unobservable characteristics explain racial/ethnic flu vaccine disparities. This suggests, for example, Whites and Blacks experience a relatively similar likelihood in obtaining the flu vaccine despite experiencing differences in observable characteristics. Further characterizing this notion is how beliefs about flu vaccines affect differences in flu vaccination rates between Whites and Blacks. For example, 29% of vaccine disparities !46! arise from differences in observable beliefs while the 2% share is due to differences in unobservable beliefs. Similar to prior work, this study shows Blacks, relative to Whites, consider the flu vaccine as less effective. Yet, despite this difference, each cohort experiences similar tendencies (i.e., coefficients) for the vaccine. This provides an explanation to the 29% versus 2% difference in explaining flu vaccine disparities. A limitation of this study is the presence of recall bias related to flu vaccinations and healthcare provider recommends for influenza vaccinations. Further, the distinction between seasonal and H1N1 flu vaccinations, unique to the 2009-2010 flu season, may not be fully understood by survey-respondents. And, the 2009-2010 flu season experienced the H1N1 flu pandemic. While we compare the NHFS data to a more recent BRFSS sample, slight differences observed in what constitutes disparities in racial/ethnic vaccination rates may be due to year-to-year differences in influenza infectivity. Yet, these results are relevant to future influenza pandemics because policies related to ensuring influenza vaccination coverage will benefit from our study conclusions. 5.0 Conclusions Reducing flu vaccination disparities between racial/ethnic groups should focus on improving socioeconomic levels and ensuring similar positive beliefs about flu vaccines. Health policies incorporating healthcare provider recommendations to reduce flu vaccination disparities should expect a modest reduction that can vary depending on the specific racial/ethnic groups in comparison. !47! 6.0 References Bhatt A, Bridges C, Donoghue K, Fernandez C, Gehring R, Hall LL, et al. Recommendations from the National Vaccine Advisory committee: standards for adult immunization practice. Public Health Rep. 2014; 129(2): 115-23. Bowblis JR, Yun MS. Racial and ethnic disparities in the use of drug therapy. Social Science Research 2010;39:674-684. Bustamante AV, Fang H, Rizzo JA, Ortega AN. Understanding observed and unobserved health care access and utilization disparities among U.S. Latino adults. Med Care Res Rev. 2009;66(5):561-577. Bustamante AV, Chen J, Rodriguez HP, Rizzo JA, Ortega AN. Use of preventive care services among Latino subgroups. Am J Prev Med 2010;38(6): 610-619. Centers for Disease Control and Prevention. Behavioral Risk Factor Surveillance System. 2013. http://www.cdc.gov/brfss. Centers for Disease Control and Prevention. Self-reported influenza vaccination coverage trends 1989–2008 among adults by age group, risk group, race/ethnicity, health-care worker status, and pregnancy status, United States, National Health Interview Survey (NHIS). 2008.! http://www.cdc.gov/flu/pdf/professionals/nhis89_08fluvaxtrendtab.pdf. Centers for Disease Control and Prevention. CDC Health Disparities and Inequalities Report – United States, 2011. Health-care access and preventive health services: Influenza vaccination coverage – United States, 2000-2010. MMWR 2011;60(Suppl):38-42. Centers for Disease Control and Prevention. U.S. Department of Health and Human Services, National Center for Health Statistics. The National 2009 H1N1 Flu Survey (NHFS), Hyattsville, MD: Centers for Disease Control and Prevention, 2012. Information about the NHFS is located at http://www.cdc.gov/nchs/nis/about_nis.htm#h1n1. Centers for Disease Control and Prevention. Flu vaccination coverage, United States, 2013-14 influenza season. 2014. http://www.cdc.gov/flu/fluvaxview/coverage- 1314estimates.htm. Clarke AR, Goddu AP, Nocon RS, …, Chin MH. Thirty years of disparities intervention research: What are we doing to close racial and ethnic gaps in healthcare? Med Care. 2013;51(11):1-14. Fiscella K. Commentary – Anatomy of racial disparity in influenza vaccination. HSR: Health Services Research 2005;40(2): 539-550. !48! Groshkopf LA, Shay DK, Shimabukuro TT, et al. Prevention and control of seasonal influenza with vaccines: recommendations of the Advisory Committee on Immunization Practices (ACIP) – United States, 2013-2014. MMWR 2013;62(RR07):1-43. HealthyPeople2020. Immunization and Infections Diseases. 2014. https://www.healthypeople.gov/2020/topics-objectives/topic/immunization-and- infectious-diseases/objectives. Hebert PL, Frick KD, Kane RL, McBean AM. The causes of racial and ethnic differences in influenza vaccination rates among elderly Medicare beneficiaries. HSR: Health Services Research 2005;40(2): 517-538. Ladin K. Decomposing differences in utilization of health services between depressed and non-depressed elders in Europe. Eur J Ageing. 2012;9(1): 51-64. Lu PJ, Singleton JA, Euler GL, et al. Seasonal influenza vaccination coverage among adult populations in the United States, 2005-2011. Am J Epidemiol. 2013;178(9):1478-1487. Musa D, Schulz R, Harris R, Silverman M, Thomas SB. Trust in the health care system and the use of preventive health services by older black and white adults. Am J Public Health. 2009;99:1293-1299. National Vaccine Advisory Committee. A pathway to leadership for adult immunization: recommendations of the National Vaccine Advisory Committee. Public Health Rep. 2012; 127(1): 1-41. O’Malley AS, Forrest CB. Immunization disparities in older Americans: determinants and future research needs. Am J Prev Med 2006;31(2):150-158. Powers DA, Yoshioka H, Yun MS. Mvdcmp: Multivariate decomposition for nonlinear response models. The Stata Journal 2011;11(4): 556-576. Pylypchuk, Y., Selden, T.M., 2008. A discrete choice decomposition analysis of racial and ethnic differences in children’s health insurance coverage. Journal of Health Economics 27, 1109–1128. Wehby GL, Murray JC, McCarthy AM, Castilla EE. Racial gaps in child health insurance coverage in four South American countries: The role of wealth, human capital, and other household characteristics. HSR: Health Services Research. 2011;46(6): 2119-2138. !49! 7.0 Tables Table 2-1. Vaccination type by race/ethnicity N 18713 2083 1244 Difference Vaccination type Whites (%) Blacks (%) Hispanics (%) Whites minus Blacks (%) Whites minus Hispanics (%) Seasonal and H1N1 flu 60.52 46.81 48.55 13.72 11.97 Both seasonal and H1N1 flu 28.04 14.88 20.66 13.16 7.39 Seasonal flu only 28.03 26.12 19.61 1.92 8.42 H1N1 flu only 4.45 5.81 8.28 -1.36 -3.83 None 39.48 53.19 51.45 -13.72 -11.97 Abbreviations: Whites, Non-Hispanic Whites; Blacks, Non-Hispanic Blacks !50! Table 2-2. Select descriptive statistics Whites Blacks Hispanics N=18713 N=2083 N=1244 Variables Mean SE Mean SE Mean SE DEMOGRAPHIC Age group (years) 18-34 0.1429 0.0026 0.1959 0.0087 0.3312 0.0133 35-44 0.1162 0.0023 0.1507 0.0078 0.2195 0.0117 45-54 0.1867 0.0028 0.2079 0.0089 0.1841 0.0110 55-64 0.2266 0.0031 0.2026 0.0088 0.1310 0.0096 65+ 0.3276 0.0034 0.2429 0.0094 0.1342 0.0097 Gender Male 0.3734 0.0035 0.3265 0.0103 0.3826 0.0138 Female 0.6266 0.0035 0.6735 0.0103 0.6174 0.0138 3-category MSA status MSA, principle city 0.2604 0.0032 0.5098 0.0110 0.3883 0.0138 MSA, not principle city 0.4579 0.0036 0.3447 0.0104 0.4622 0.0141 Non-MSA 0.2817 0.0033 0.1455 0.0077 0.1495 0.0101 Census region of residence a Region 1 0.1841 0.0028 0.1018 0.0066 0.1600 0.0104 Region 2 0.2405 0.0031 0.1277 0.0073 0.1021 0.0086 Region 3 0.3399 0.0035 0.6923 0.0101 0.2990 0.0130 Region 4 0.2355 0.0031 0.0783 0.0059 0.4389 0.0141 Interview month 10-Jan 0.0550 0.0017 0.0499 0.0048 0.0571 0.0066 10-Feb 0.1562 0.0027 0.1531 0.0079 0.1608 0.0104 !51! 10-Mar 0.1751 0.0028 0.1949 0.0087 0.1849 0.0110 10-Apr 0.1889 0.0029 0.1767 0.0084 0.1897 0.0111 10-May 0.2318 0.0031 0.2352 0.0093 0.2243 0.0118 10-Jun 0.1930 0.0029 0.1901 0.0086 0.1833 0.0110 SOCIOECONOMIC Self-report education level <12 years 0.0639 0.0018 0.1603 0.0080 0.2484 0.0123 12 years 0.2211 0.0030 0.2391 0.0093 0.2186 0.0117 Some college 0.2682 0.0032 0.2885 0.0099 0.2339 0.0120 College graduate 0.4048 0.0036 0.2650 0.0097 0.2299 0.0119 Missing 0.0421 0.0015 0.0470 0.0046 0.0691 0.0072 Income poverty status Above poverty threshold, >=$75,000 income 0.2699 0.0032 0.1421 0.0077 0.1672 0.0106 Above poverty threshold, <$75,000 income 0.5010 0.0037 0.4633 0.0109 0.4043 0.0139 Below poverty threshold 0.0722 0.0019 0.2108 0.0089 0.2484 0.0123 Poverty status unknown 0.1569 0.0027 0.1839 0.0085 0.1801 0.0109 HEALTH AND ACCESS Chronic medical condition b No 0.6277 0.0035 0.5982 0.0107 0.7058 0.0129 Yes 0.3413 0.0035 0.3658 0.0106 0.2781 0.0127 Missing 0.0309 0.0013 0.0360 0.0041 0.0161 0.0036 Has health insurance coverage Yes 0.9001 0.0022 0.8262 0.0083 0.7098 0.0129 !52! No 0.0593 0.0017 0.1296 0.0074 0.2203 0.0118 Don't know/Refused/Missing 0.0407 0.0014 0.0442 0.0045 0.0699 0.0072 Number of times seen doctor since August 2009 >=4 0.2842 0.0033 0.3447 0.0104 0.2653 0.0125 3 0.1465 0.0026 0.1699 0.0082 0.1680 0.0106 2 0.2561 0.0032 0.2376 0.0093 0.2781 0.0127 1 0.2938 0.0033 0.2132 0.0090 0.2733 0.0126 Missing 0.0193 0.0010 0.0346 0.0040 0.0153 0.0035 Healthcare provider recommendation c Seasonal and H1N1 flu vaccine 0.4340 0.0036 0.4455 0.0109 0.4236 0.0140 BELIEFS ABOUT FLU VACCINES Opinion: Effectiveness of H1N1 vaccine Very/somewhat effective 0.7515 0.0032 0.6942 0.0101 0.7195 0.0127 Not very/not at all effect/Don't know/Refused/Missing 0.2485 0.0032 0.3058 0.0101 0.2805 0.0127 Opinion: Risk of getting sick with H1N1 flu without vaccine Very/somewhat high 0.2518 0.0032 0.2597 0.0096 0.3585 0.0136 Somewhat/very low/Don't know/Refused/Missing 0.7482 0.0032 0.7403 0.0096 0.6415 0.0136 Opinion: Worry about getting sick from the H1N1 vaccine Very/somewhat worried 0.2778 0.0033 0.3615 0.0105 0.4325 0.0141 Not very/not at all 0.7222 0.0033 0.6385 0.0105 0.5675 0.0141 !53! worried/Don't know/Refused/Missing Opinion: Effectiveness of seasonal vaccine Very/somewhat effective 0.8331 0.0028 0.7657 0.0093 0.7838 0.0117 Not very/not at all effect/Don't know/Refused/Missing 0.1669 0.0028 0.2343 0.0093 0.2162 0.0117 Opinion: Risk of getting sick with seasonal flu without vaccine Very/somewhat high 0.4147 0.0036 0.3533 0.0105 0.4510 0.0141 Somewhat/very low/Don't know/Refused/Missing 0.5853 0.0036 0.6467 0.0105 0.5490 0.0141 Opinion: Worry about getting sick from the seasonal vaccine Very/somewhat worried 0.2325 0.0031 0.3341 0.0103 0.3883 0.0138 Not very/not at all worried/Don't know/Refused/Missing 0.7675 0.0031 0.6659 0.0103 0.6117 0.0138 Abbreviations: Whites, Non-Hispanic Whites; Blacks, Non-Hispanic Blacks; SE, standard error; MSA, metropolitan statistical area a Region 1: CT, ME, MA, NH, VT, RI, NJ, NY, PA; Region 2: IL, IN, MI, OH, WI, IA, KS, MN, MO, NE, ND, SD; Region 3: DE, DC, FL, GA MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK, TX; Region 4: AZ, CO, ID, MT, NV, NM, UT, WY, AK, CA, HI, OR, WA b This indicates whether the person has any of the following chronic medical conditions: asthma or an other lung condition, diabetes, a heart condition, a kidney condition, sickle cell anemia or other anemia, a neurological or neuromuscular condition, a liver condition, or a weakened immune system caused by a chronic illness or by medicines taken for a chronic illness. c Reporting a HCP recommendation was defined as a binary variable equal to one when the respondent indicated !54! they received recommendations for H1N1 flu vaccination only, seasonal flu vaccination only, or both vaccinations. !55! Table 2-3. Logit regression Pooled Whites Blacks Hispanics Variables ME SE ME SE ME SE ME SE DEMOGRAPHIC Race/ethnicity Hispanic -0.006 0.018 NA NA NA Non-Hispanic, Black Only -0.076 *** 0.014 NA NA NA Non-Hispanic, White Only Ref NA NA NA Age group (years) 18-34 Ref Ref Ref Ref 35-44 0.028 * 0.015 0.025 0.016 -0.002 0.045 0.068 0.051 45-54 0.083 *** 0.014 0.079 *** 0.015 0.011 0.043 0.203 *** 0.070 55-64 0.162 *** 0.015 0.167 *** 0.016 0.033 0.045 0.179 ** 0.074 65+ 0.331 *** 0.016 0.329 *** 0.017 0.249 *** 0.049 0.328 *** 0.103 Gender Male Ref Ref Ref Ref Female -0.024 *** 0.008 -0.025 *** 0.009 -0.003 0.029 -0.045 0.038 Married No Ref Ref Ref Ref Yes 0.029 ** 0.011 0.027 ** 0.013 0.084 ** 0.035 -0.025 0.043 Missing -0.045 0.046 -0.015 0.050 0.008 0.134 -0.362 0.222 Number of people in household !56! 1 Ref Ref Ref Ref 2 -0.029 ** 0.013 -0.035 ** 0.014 -0.030 0.036 0.061 0.060 3 -0.031 * 0.016 -0.040 ** 0.017 -0.048 0.043 0.169 ** 0.077 4 -0.045 ** 0.018 -0.053 *** 0.020 -0.006 0.049 0.016 0.067 5 -0.054 ** 0.021 -0.070 *** 0.024 -0.030 0.067 0.145 0.082 6 -0.102 *** 0.039 -0.130 *** 0.047 -0.038 0.100 0.089 0.106 7 -0.014 0.063 0.007 0.083 0.093 0.151 0.009 0.139 3-category MSA status MSA, principle city 0.026 ** 0.011 0.036 *** 0.012 -0.007 0.040 -0.144 ** 0.062 MSA, not principle city 0.011 0.010 0.017 * 0.010 -0.003 0.042 -0.155 ** 0.064 Non-MSA Ref Ref Ref Ref Census region of residence a Region 1 Ref Ref Ref Ref Region 2 0.005 0.013 0.009 0.013 -0.102 * 0.055 0.065 0.072 Region 3 -0.008 0.012 -0.004 0.012 -0.051 0.044 0.019 0.054 Region 4 0.000 0.013 -0.005 0.013 0.036 0.061 0.070 0.054 Interview month 10-Jan Ref Ref Ref Ref 10-Feb 0.018 0.020 0.013 0.021 -0.019 0.067 0.112 0.090 10-Mar 0.010 0.019 -0.005 0.020 0.068 0.066 0.068 0.086 10-Apr 0.015 0.019 0.000 0.020 0.103 0.066 0.056 0.085 10-May 0.026 0.019 0.013 0.020 0.037 0.064 0.160 0.090 10-Jun 0.031 0.019 0.018 0.020 0.088 0.066 0.088 0.087 SOCIOECONOMIC Self-report education level <12 years Ref Ref Ref Ref !57! 12 years 0.021 0.016 0.032 * 0.019 0.015 0.043 -0.022 0.052 Some college 0.036 ** 0.016 0.041 ** 0.019 0.037 0.044 0.038 0.055 College graduate 0.096 *** 0.017 0.102 *** 0.019 0.030 0.048 0.144 ** 0.068 Missing 0.099 ** 0.049 0.108 * 0.055 0.002 0.137 0.144 0.182 Income poverty status Above poverty threshold, >=$75,000 income Ref Ref Ref Ref Above poverty threshold, <$75,000 income -0.042 *** 0.011 -0.045 *** 0.011 -0.015 0.043 0.005 0.055 Below poverty threshold -0.067 *** 0.018 -0.082 *** 0.020 -0.007 0.056 -0.037 0.070 Poverty status unknown -0.019 0.016 -0.014 0.017 -0.070 0.058 -0.003 0.077 Work status Employed Ref Ref Ref Ref Unemployed 0.006 0.019 -0.004 0.021 0.030 0.051 0.091 0.073 Not in labor force 0.044 *** 0.010 0.036 *** 0.011 0.098 *** 0.034 0.079 * 0.048 Don't know/Refused/Missing 0.122 *** 0.045 0.051 0.050 0.411 *** 0.128 0.235 0.177 Works in health care field No Ref Ref Ref Ref Yes 0.172 *** 0.014 0.171 *** 0.015 0.155 *** 0.043 0.163 ** 0.074 Missing -0.041 0.054 -0.050 0.059 0.012 0.192 -0.126 0.261 Home rented or owned Home is owned Ref Ref Ref Ref Home is rented or other arrangement -0.023 ** 0.011 -0.019 0.012 -0.013 0.031 -0.032 0.043 Don't know/Refused/Missing -0.021 0.026 -0.038 0.028 0.104 0.075 -0.010 0.107 !58! HEALTH AND ACCESS Chronic medical condition b No Ref Ref Ref Ref Yes 0.055 *** 0.009 0.053 ** 0.010 0.068 ** 0.029 0.041 0.041 Missing 0.098 *** 0.037 0.120 ** 0.039 0.062 0.116 -3.468 94.60 5 Sick with fever and cough or sore throat in past month No Ref Ref Ref Ref Yes -0.026 0.019 -0.034 ** 0.021 0.045 0.060 -0.030 0.068 Missing 0.080 0.056 0.069 * 0.059 -0.048 0.194 -3.383 158.7 84 Other people in house with fever and cough or sore throat No Ref Ref Ref Ref Yes -0.001 0.013 0.011 ** 0.014 -0.070 0.044 -0.030 0.046 Missing -0.018 0.048 -0.014 * 0.051 -0.064 0.133 7.110 184.2 85 Has health insurance coverage No Ref Ref Ref Ref Yes 0.139 *** 0.016 0.145 ** 0.019 0.067 0.041 0.196 *** 0.064 Don't know/Refused/Missing 0.105 * 0.055 0.153 * 0.061 -0.090 0.168 0.208 0.206 Number of times seen doctor since August 2009 >=4 Ref Ref Ref Ref 3 0.021 * 0.013 0.021 0.014 -0.019 0.039 0.098 * 0.058 !59! 2 0.012 0.011 0.009 0.012 0.012 0.035 0.054 0.048 1 0.006 0.011 0.003 0.012 0.048 0.038 -0.008 0.049 Missing 0.079 *** 0.029 0.080 ** 0.032 0.005 0.076 0.203 0.146 Healthcare provider recommendation c Seasonal and H1N1 flu vaccine 0.311 *** 0.008 0.314 *** 0.009 0.266 *** 0.027 0.284 *** 0.073 BELIEFS ABOUT FLU VACCINES Opinion: Effectiveness of H1N1 vaccine Very/somewhat effective 0.027 *** 0.010 0.032 *** 0.011 -0.036 0.032 0.055 0.044 Not very/not at all effect/Don't know/Refused/Missing Ref Ref Ref Ref Opinion: Risk of getting sick with H1N1 flu without vaccine Very/somewhat high 0.133 *** 0.011 0.150 *** 0.013 0.014 0.035 0.091 * 0.049 Somewhat/very low/Don't know/Refused/Missing Ref Ref Ref Ref Opinion: Worry about getting sick from the H1N1 vaccine Very/somewhat worried 0.003 0.010 0.006 0.011 -0.044 0.032 0.048 0.044 Not very/not at all worried/Don't know/Refused/Missing Ref Ref Ref Ref !60! Opinion: Effectiveness of seasonal vaccine Very/somewhat effective 0.252 *** 0.012 0.251 *** 0.013 0.244 *** 0.036 0.202 *** 0.066 Not very/not at all effect/Don't know/Refused/Missing Ref Ref Ref Ref Opinion: Risk of getting sick with seasonal flu without vaccine Very/somewhat high 0.346 *** 0.010 0.346 *** 0.010 0.332 *** 0.033 0.268 *** 0.074 Somewhat/very low/Don't know/Refused/Missing Ref Ref Ref Ref Opinion: Worry about getting sick from the seasonal vaccine Very/somewhat worried -0.157 *** 0.011 -0.169 *** 0.011 -0.070 ** 0.032 -0.045 0.045 Not very/not at all worried/Don't know/Refused/Missing Ref Ref Ref Ref Sample size 22040 18713 2083 1244 Pseudo R 2 28.97 30.66 21.32 23.67 Significance level: *** 1% , ** 5%, * 10% Abbreviations: Whites, Non-Hispanic Whites; Blacks, Non-Hispanic Blacks; SE, standard error; MSA, metropolitan statistical area; ME, marginal effect a Region 1: CT, ME, MA, NH, VT, RI, NJ, NY, PA; Region 2: IL, IN, MI, OH, WI, IA, KS, MN, MO, NE, ND, SD; Region 3: DE, DC, FL, GA MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK, TX; Region 4: AZ, CO, ID, MT, NV, NM, UT, WY, AK, CA, HI, OR, WA b This indicates whether the person has any of the following chronic medical conditions: asthma or an other lung condition, !61! diabetes, a heart condition, a kidney condition, sickle cell anemia or other anemia, a neurological or neuromuscular condition, a liver condition, or a weakened immune system caused by a chronic illness or by medicines taken for a chronic illness. c Reporting a HCP recommendation was defined as a binary variable equal to one when the respondent indicated they received recommendations for H1N1 flu vaccination only, seasonal flu vaccination only, or both vaccinations. !62! Table 2-4. Sub-aggregate decomposition of differences in flu vaccination rates between Whites and Blacks Difference between Whites and Blacks Characteristics effect Coefficients effect Estimate SE Share Estimate SE Share Aggregate effect 0.0849 *** 0.0040 61.91 0.0522 *** 0.0106 38.09 DEMOGRAPHIC Sub-aggregate effect a 0.0232 NA 16.95 0.0360 NA 26.24 SOCIOECONOMIC Sub-aggregate effect 0.0192 NA 14.01 0.0639 NA 46.59 HEALTH AND ACCESS Sub-aggregate effect 0.0024 NA 1.74 -0.0509 NA -37.11 BELIEFS ABOUT FLU VACCINES Sub-aggregate effect 0.04008 NA 29.22 0.0033 NA 2.37 Significance level: *** 1% , ** 5%, * 10% Abbreviations: Whites, Non-Hispanic Whites; Blacks, Non-Hispanic Blacks; SE, standard error; NA, not applicable a The sub-aggregate effect estimate and share is calculated from the summing all sub-aggregate effects from Appendix Table 2B. Statistics are reported for each variable constituting each sub-aggregate effect in Appendix Table 2B. !63! Table 2-5. Sub-aggregate decomposition of differences in flu vaccination rates between Whites and Hispanics Difference between Whites and Hispanics Characteristics effect Coefficients effect Estimate SE Share Estimate SE Share Aggregate effect 0.1207 *** 0.0052 100.78 -0.0009 0.0134 -0.78 DEMOGRAPHIC Sub-aggregate effect a 0.0636 NA 53.14 0.0029 NA 2.39 SOCIOECONOMIC Sub-aggregate effect 0.0271 NA 22.62 -0.0009 NA -0.71 HEALTH AND ACCESS Sub-aggregate effect 0.0227 NA 18.98 -0.0038 NA -3.19 BELIEFS ABOUT FLU VACCINES Sub-aggregate effect 0.0072 NA 6.05 0.0009 NA 0.72 Significance level: *** 1% , ** 5%, * 10% Abbreviations: Whites, Non-Hispanic Whites; Blacks, Non-Hispanic Blacks; SE, standard error; NA, not applicable a The sub-aggregate effect estimate and share is calculated from the summing all sub-aggregate effects from Appendix Table 2C. Statistics are reported for each variable constituting each sub-aggregate effect in Appendix Table 2C. !64! Chapter 3. Healthcare provider recommendations and influenza vaccinations: A causal effects estimate Abstract Prior studies have thus far demonstrated a significant positive correlation between healthcare provider recommendations for the flu vaccine and patient adherence to the recommendation. These existing studies employ regression techniques that do not consider a deeper question of whether the regression of provider recommendations on flu vaccination has a causal interpretation. The objective of this study is to provide a measure of bias associated with prior estimates of provider recommendations for flu vaccines and patient adherence to these recommendations. This study hypothesizes these prior estimates of the effect of provider recommendations are biased upwards. The results demonstrate provider recommendations generate an unadjusted 39% (p<0.001) increased likelihood of obtaining a flu vaccine when measured from an Amazon Mechanical Turk (MTurk) sample of the United States general population. A causal effect of provider recommendations on flu vaccination intent was measured by randomly assigning provider recommendations to treatment scenarios from a similar, yet independent, MTurk sample. The resulting unadjusted provider recommendation effect on flu vaccination intent was 16% (p<0.001). Therefore, by construction, there exists a 23-percentage point bias in the recommendation effect. These findings suggest prevailing measures of the effect of provider recommendations on flu vaccinations are overestimated. Further research is needed to describe the sensitivities of this bias in different treatment scenarios. !65! 1.0 Introduction Healthcare provider recommendations for influenza (i.e., the flu) vaccines are positively correlated with an individual’s likelihood to obtain an influenza vaccine [Ding et al., 2011; Pandolfi et al., 2012]. Based on this evidence, the National Vaccine Advisory Committee (NVAC) considers strongly recommending all immunizations that patients need as a standard of practice for all providers [Bhatt et al., 2014]. Other approaches such as use of standing orders, reducing patient out-of-pocket costs, and worksite promotion of vaccination services can increase vaccination coverage. However, the NVAC’s emphasis on healthcare provider recommendations relies on their position that every healthcare provider has a fundamental responsibility to ensure that all patients are up-to-date with their recommended vaccinations. While provider recommendations can improve vaccination rates, especially for influenza vaccinations, an underlying issue exists in whether they can be the best mechanism to induce vaccination uptake in the presence of other barriers. For example, a recommendation may have little influence on a patient’s decision to adhere to the recommendation if they are more concerned about potential side effects (i.e., flu infections) from the flu vaccine; which would underestimate the recommendation effect on patients [Mirza et al, 2008]. Alternatively, patients may be predisposed to obtaining the flu vaccine because they are aware that getting the flu shot every year is recommended for persons their age; thus, overestimating (i.e., biasing) the recommendation effect on patient adherence [Santibanez et al., 2010]. This suggests that provider recommendations for flu vaccines may not be salient to the patient. !66! From the provider perspective, a strong determinant of vaccine use and recommending vaccinations is the perception that the vaccine is effective [Nichol et al., 2001; Rickert et al., 2006]. Localized studies show a majority of physician practices in Philadelphia, Pennsylvania offer the vaccine to patients [Jessop et al., 2013] while pediatricians from Maryland regularly vaccinate their patients [Levy et al., 2009]. Yet, it is uncertain whether providers are consistently recommending the influenza vaccination to all patients as proposed by the NVAC. Adding to this notion are studies where physicians in Chicago, Illinois were not aware of the severity of flu infections for children [Dominguez and Daum, 2005]. And, a study from a nationally representative sample of United States (US) physicians suggest the perceived responsibility of providing vaccinations to high- risk children vary among physician specialties [Rickert et al., 2006]. From this evidence, variations in physicians’ awareness of flu vaccine effectiveness and the related practice of recommending the vaccine to their patients adds doubt to whether recommendations will consistently improve vaccination rates. With studies suggesting patient vaccination beliefs can confound the recommendation effect on flu vaccinations, health policy initiatives utilizing recommendations to improve vaccination rates can be misinformed due to these confounding effects. The goal of this current study is to develop an unbiased estimate for the effect of provider recommendations on influenza vaccinations. With recommendations a current standard of care for healthcare providers set by the NVAC, exploring a causal recommendation effect can better inform health policy. The primary objective of this study is to provide a !67! measure of bias associated with prior estimates of provider recommendations for flu vaccines and patient adherence to recommendations. This study hypothesizes prior estimates of the effect of provider recommendations are biased upwards. The rest of this study is organized as follows. The next section describes the theoretical framework of the bias potentially associated with provider recommendations. Then, the paper describes the study methods and statistical analysis plan generating the data that measures the potential bias. A formal outline of the study hypotheses is proposed. Finally, this paper summarizes the conclusions and offers potential implications to health policies. 2.0 Theoretical framework Consider the potential-outcomes notation of two potential vaccination outcomes: !"#$%#&'(!!"#$!%&= ! !! , if!! ! = 1 ! !! , if!! ! = 0 ! (1) For this case, ! !! is i’s vaccination status without a recommendation and ! !! is i’s vaccination status with a recommendation. ! ! is a dummy variable indicating receipt of a healthcare provider recommendation. Prior research is interested in the difference between ! !! and ! !! (i.e., the causal effect of recommendations on individual i); where the observed outcome can be written as: !"!!"#!$!!"#$!%&=! ! =! !! + (! !! −! !" )! ! (2) !68! And since neither potential outcomes are observed at the same time per i, previous work intends to estimate the average of ! !! −! !! (i.e., ![! !! −! !! |! ! = 1], the average causal effect of recommendations on those who received a recommendation. However, this is a naïve comparison of averages by recommendation status. It is possible that those who receive a recommendation would have obtained a vaccination without a provider recommendation; therefore, the naïve comparison would exaggerate the benefits of a recommendation. For example, individuals may consider flu vaccinations as an effective measure to protecting against influenza infections, thus a provider recommendation would do little to convince the individual to obtain the vaccine since they are positively predisposed to the vaccine. A formal treatment of this notion is [Green et al., 2008; Angrist et al. 2008]: ! ! !! −! !! 1 !!!!!!!!!!!= !!!!!!!!!!! ! ! ! ! = 1 −![! ! |! ! = 0] Average causal effect Observed difference in vaccine uptake =! ! !! −! !! ! ! = 1 !!!!!!!!!!!!!!+ !!!!!!!!!!!!!!! ! !! ! ! = 1 −![! !! |! ! = 0] (3) Average treatment effect on the treated Selection bias Therefore, the ‘average causal effect’ can overestimate the benefits of recommendations due to a positive selection bias. A positive selection bias exists because those who received a recommendation may have obtained a vaccination without a recommendation 2 (overestimating the benefits of a recommendation). !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 1 !The!potential!outcome!is!a!binary!variable!equal!to!one!denoting!vaccination!receipt.!!The!average!causal!effect! is!not!affected!by!whether!the!outcome!is!binary!since!the!outcome!follows!a!Bernoulli!trial,!! ! !! −! !! = ! ! !! = 1 −![! !! = 1].! 2 Prior work has assumed a selection-on-observables approach by controlling for individual demographic and socioeconomic characteristics; which assumes the residual in the linear causal model are mean- independent. This reliance of selection-on-observables may not be guaranteed if unobservable traits affect vaccination receipt. !69! 2.1 Framework of omitted variable bias Suppose the healthcare provider recommendation regression controls for individual variables measuring current and past motivation for obtaining the vaccine or an individual’s behavior in seeking healthcare provider recommendations. These attributes can be denoted by a vector ! ! defined here as ‘predisposition.’ The regression of vaccination (! ! ) on recommendations (! ! ) controlling for predisposition becomes: ! ! =! ! +! ! ! ! +! ! ! ! +! ! (4) where ! ! , ! ! (average causal effect), and ! ! are population regression coefficients and ! ! is a residual uncorrelated with all regressors by definition. If the selection-on- observables assumption applies given ! ! , then ! ! is unbiased (i.e., the average causal effect is not overestimated given (3)). If predisposition is omitted from (4) then the model becomes: ! ! = ! ! +! ! ! ! +! ! (5) and ![! ! ]=! ! +! ! ! ! (6) with ! ! equal to the slope of the regression of ! ! on ! ! [Green et al., 2008]. Note that (6) is equal to the notation in (3) where ! ! ! ! (the omitted variable bias) represents the selection bias. !70! In summary, the notion of predisposition (i.e., intrinsic motivation) is difficult to measure (and likely unobservable) and its omission from regression analysis creates a likely bias in the causal effect of recommendations on vaccine receipt. Prior literature proxy’s for this variable through individual opinions about the vaccine; relying on the selection-on- observables assumption to provide a possible basis for empirical work. But these measures are imperfect because they are likely to be influenced by or measured after the treatment (! ! ) [Green et al., 2008; Angrist et al. 2008]. Therefore, the best-case scenario for the selection-on-observables assumption is random assignment of ! ! , conditional on ! ! (a vector of individual characteristics), in an experiment. The next section describes two independent studies where the first study estimates regression (5) using methods similar to prior research and the second study re-estimates regression (5) after random assignment of ! ! for a similar outcome (i.e., flu vaccine intent instead of self-reported flu vaccination). 3.0 Methods 3.1 Study participants This research conducted two separate surveys that were both approved by the University of Southern California Institutional Review Board. Survey respondents were recruited from Amazon’s Mechanical Turk (MTurk). MTurk is an online crowdsourcing tool that allows ‘workers’ to complete online tasks (i.e., human intelligence tasks, HITs) for relatively small amounts of compensation (Boynton et al., 2014). Respondents were eligible to participate in either survey if they were adults (at least 18 years of age) and residents of the United States (US). !71! For each study, HITs were posted on MTurk that invited interested participants to complete a survey related to flu vaccinations. Upon completion of the survey, the worker was provided a unique completion code consisting of a random character string. Each study respondent was asked to copy and paste the code into MTurk to confirm his or her actual participation of the study. 3.2 Study design 3.2.1 Study 1 The first study (Study 1) was conducted from March 17, 2014 through March 18, 2014. This study design was similar to prior literature where respondents were asked their flu vaccination status and whether their provider recommended the flu vaccine [Ding et al., 2011]. Study 1 was a HIT posted on MTurk that invited interested participants to complete a survey related to their flu vaccination status and whether their healthcare provider recommended the flu vaccination to them for the 2013-2014 flu season. Similar to prior studies, all participants were asked whether they received at least one seasonal flu vaccination at the time of the survey and whether their healthcare provider personally recommended they receive the flu vaccine. Questions that capture predisposition were the following: 1. Have you ever requested a seasonal flu vaccination from your usual care doctor or healthcare provider? a. Responses: Yes; No. !72! 2. I would have obtained the seasonal flu vaccination had I not received a recommendation from my doctor or healthcare provider to get the flu vaccination. 3 a. Responses: Strongly disagree; Disagree; Somewhat disagree; Somewhat agree; Agree; Strongly agree. Respondents were also asked demographic (i.e. race/ethnicity, gender) and socioeconomic (i.e., education, household income) questions. The Study 1 survey questions can be found in Appendix 3A. Each participant was compensated for completing the survey ($US 0.20). 3.2.2 Study 2 The second study (Study 2) was conducted from March 23, 2014 through March 27, 2014. Study 2 was a HIT posted on MTurk inviting participants to complete a survey related to their flu vaccination intent after random assignment of conditions. This study randomly assigned participants (in a 1:1:1:1 ratio) to one of four conditions (Table 3-1). The experimental condition is described below: The information below describes a fictional scenario between you and your usual care doctor or healthcare provider. This scenario is designed to mimic the events that may occur during a routine doctor's visit during a typical flu season. Scenario: You are currently not vaccinated against the flu. You have recently visited your regular doctor for a routine physical. As a result of your doctor's visit, [you learn that your doctor received the flu vaccine OR it is unknown to you whether your doctor obtained a flu vaccination]. [Your doctor recommends that you get the flu vaccine OR You receive no recommendation from your doctor to get the flu vaccine]. If you had to make a decision now, would you get the flu vaccine? Yes or No !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 3 This question was asked if participants received a flu vaccine and a healthcare provider recommendation at the time of the survey. !73! Prior to random assignment, survey respondents were asked three questions related to their strength of agreement (e.g., strongly agree to strongly disagree) with, for example, whether the seasonal flu vaccination is effective. These questions are the following: 1. The seasonal flu vaccination is effective in preventing the seasonal flu. 2. If I didn’t get a seasonal flu vaccination, I have a high chance of getting sick with the seasonal flu. 3. I am worried about getting sick from the seasonal flu vaccine. As in Study 1, respondents were also asked demographic (i.e. race/ethnicity, gender) and socioeconomic (i.e., education, household income) questions. The remaining Study 2 survey questions can be found in Appendix 3B. Each participant was compensated for completing the survey ($US 0.20). 3.3 Statistical analysis plan 3.3.1 Study 1 The primary outcomes were the proportion of respondents who received a flu vaccination (e.g., Since September 2013, have you received at least one seasonal flu vaccination?) and whether they received recommendation to obtain the flu vaccination from their healthcare provider (e.g., Since September 2013, did your usual care doctor or healthcare provider personally recommend that you receive a seasonal flu vaccination?). Both outcomes were dichotomous variables equal to one if the respondent answered yes. A total enrollment of 800 patients was planned to provide at least 80% power, at the two- sided alpha=0.05 significant level, to detect a 0.10 between two independent proportion (i.e., 40% versus 30%) for the primary outcomes (Faul et al., 2009). A P value of 0.05 defined statistical significance. All variables were analyzed using descriptive statistics to evaluate data distributions. Ordinary chi-square tests for categorical variables and !74! independent sample t tests for continuous variables evaluated bivariate associations with vaccination and recommendation status. Ordinary least squares models were used for multivariate adjustment. 3.3.2 Study 2 The primary outcome was the proportion of respondents indicating intent to become vaccinated after assignment to the experimental condition (e.g., If you had to make a decision now, would you get the flu vaccine?). The results of a power analysis indicates randomly assigning 350 respondents to each of the four experimental conditions will provide at least 80% power, at the two-sided alpha=0.05 significant level, to detect a 0.10 difference between two independent proportion (i.e., 40% versus 30%) (Faul et al., 2009). All variables were analyzed similar to Study 1. 3.4 Hypotheses The study hypotheses are formalized as follows: Hypothesis 1: Provider recommendations will increase vaccination rates. Prior work suggests the provider recommendation effect is potentially confounded with whether the patient has an underlying positive attraction to the vaccine (i.e., the belief that the vaccine can prevent future infections) that can cause a positive bias to Hypothesis 1. In this case, respondents would have received a flu vaccine without a recommendation. A competing argument is a negative barrier to the vaccine (i.e., the !75! vaccine can cause adverse effects) that can cause a negative bias to Hypothesis 1. Few studies explore this predisposition (i.e., intrinsic motivation) that ultimately affects the strength of the recommendation in the presence of vaccination perceptions. A formal statement of the hypothesis is: Hypothesis 2: After adjusting for intrinsic motivation, provider recommendations will slightly increase vaccination rates. According to this premise, study participants receiving an extrinsic intervention via a provider recommendation adds to a positive intrinsic motivation to become vaccinated. However, provider recommendations do not likely occur by chance since providers recommending the flu vaccine are likely to have been vaccinated [Santibanez et al, 2010]. From the patient perspective, it is possible the combined knowledge that their vaccinated provider recommends the flu vaccine may contribute to their intrinsic likelihood to become vaccinated. The hypothesis is formalized in the following: Hypothesis 3: Provider recommendations coming from a vaccinated provider will considerably increase vaccination rates. Studies show compulsory vaccination policies for healthcare workers improve healthcare worker vaccination rates [Maurer et al., 2012]. Whether these workplace policies affect patient vaccination behavior is unknown. If provider vaccination status results in increased patient vaccination rates, compulsory healthcare worker vaccination policies !76! can indirectly improve vaccination rates; especially in the presence of a provider recommendation. 4.0 Results 4.1 Study participants 4.1.1 Study 1 results Eight hundred and sixteen participants initiated the survey. From these survey initiators, twenty-two were removed because the confirmation code generated to ensure the respondent completed the survey was incorrect. The final sample comprised of 794 adults. Figure 3-1 presents the sample flow diagram for this study. Study 1 respondent characteristics are reported in Table 3-1 where 284 (35.8%) respondents received a flu vaccination since September 2013. A majority of respondents (N=505, 63.6%) did not receive a recommendation from their provider to obtain the flu vaccine. The respondent sample experienced ‘Some college’ or greater education (86.0%), are single (68.5%), aged 25-44 years old (78.7%), and White/Caucasian (78.9%). A comparison with US national estimates from the 2013 American Community Survey shows the respondent sample reported ‘Some college’ or greater education by approximately more than 27 points [US American Community Survey, 2013]. The sample was predominantly younger because 25-44 years constitute approximately 26 percent of the US total population (e.g., 66.8% of the US population are 25+ years old). Survey respondents were predominantly never married (i.e., single) when compared to !77! national estimates of 33.1%. However, the sample reporting 78.7% White/Caucasian was similar to US national estimates of approximately 74%. Table 3-2 shows unadjusted and adjusted ordinary least square regression estimates measuring the effect of recommendations on flu vaccination. This study demonstrates the unadjusted effect of recommendations on flu vaccinations is 38.9% (t-statistic: 11.96). The notion of predisposition can be represented by whether respondents requested the flu vaccination from their provider. From Table 3-2, omitting this question from the adjusted regressions shows the effect of recommendations on flu vaccinations is 33.3% (t-statistic: 9.87). Controlling for this variable reduces the recommendation effect on flu vaccinations by more than half to 13.7% (t-statistic: 4.71); thus, controlling for this measure of predisposition is a strong predictor of flu vaccinations. A subpopulation analysis for respondents indicating that they did not request the flu vaccination from their healthcare provider (N=619) measured the extent the population without this predisposition is influenced by provider recommendations. The results show the recommendation effect from this subpopulation is about 50% of the value when compared to the total population (see Table 3-2). The recommendation effect on flu vaccination is 16.0% (t-statistic: 4.45) after controlling for respondent characteristics. Appendix 3C and 3D report the subpopulation characteristics and model regression results, respectively. !78! 4.1.2 Study 2 results One thousand four hundred and thirty-one participants initiated the survey. Of these, sixteen were removed prior to randomization because they did not consent to participate nor continue after consent. Figure 3-2 presents the sample flow for this study. Table 3-3 displays the sample characteristics for respondents across the four experimental scenarios. Across experimental scenarios there were no significant differences in demographic and socioeconomic variables except for age. For age, the proportion of respondents between 25-44 years old ranged from 71.6% (Experiment 1) to 85.4% (Experiment 3) across scenarios. Respondents across experimental scenarios agreed and strongly agreed with the statement ‘The seasonal flu vaccination is effective in preventing the seasonal flu’ ranging from 82.1-84.2%. Survey respondents did not consider themselves at high risk of getting sick with the seasonal flu without a flu vaccination. Lastly, these respondents were not worried about getting sick from the flu vaccine. Across experimental scenarios, the primary outcome ‘flu vaccine intent’ ranged from 44.4% (Scenario 4) to 61.4% (Scenario 1). Holding recommendation status constant and varying whether the respondent was aware of their doctor’s flu vaccination status (i.e., Experiment 1 versus 2 and Experiment 3 versus 4) did not measurably affect a respondent’s intent to acquire the flu vaccine. However, holding whether the respondent !79! was aware of their doctor’s flu vaccination constant and varying recommendation status (i.e., Experiment 1 versus 3 and Experiment 2 versus 4) had a significant impact on a respondent’s intent to obtain the flu vaccine. Table 3-4 shows unadjusted and adjusted ordinary least square regression estimates for the effect of recommendations on flu vaccination intent after combining respondents in experimental conditions 2 and 4 (i.e., the recommendation effect) as well as experimental conditions 1 and 3 (i.e., the provider vaccination status effect). With the doctor’s vaccination status unknown (experiments 2 and 4), this study demonstrates the unadjusted effect of recommendations on flu vaccine intent is 15.9% (t-statistic: 4.28) in Model 1 Table 3-4. 4 Knowing their doctor obtained the flu vaccine resulted in a 12.2% (t-statistic: 3.24) unadjusted recommendation effect on flu vaccine intent (Model 4 Table 3-4). However, unlike experimental conditions 2 and 4, adjusting for respondent characteristics in Table 3-3 did not dramatically alter the recommendation effect from the unadjusted model (Models 5 and 6 Table 3-4). The effect of opinions about flu vaccine effectiveness and risk of infection on flu vaccine were positive and significant. However, worry about getting sick from the flu vaccine did not affect flu vaccine intent. This trend of opinions was relatively stable when respondents were aware that their provider received the flu vaccine (Models 3 and 6 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 4 !Adjusting for respondent characteristics in Models 2 and 3 (Table 3-4) moderately increases the effect to 17.1% (t-statistic: 4.43) and 20.2% (t-statistic: 6.21), respectively. ! !80! Table 3-4). Moreover, including these opinions in the model improved the model fit defined by the adjusted R-square (e.g., Model 3 versus Model 2 in Table 3-4). 4.2 Empirical evidence of omitted variable bias Comparing the respondent characteristics between Studies 1 and 2 show similar distributions of demographic and socioeconomic characteristics; which allows for a comparison of model estimates. For example, similar to Study 1, the respondent sample in Study 2 (for all experimental conditions) experienced ‘Some college’ or greater education (85.3%), single (65.4%), aged 25-44 years old (77.5%), and White/Caucasian (78.2%). Recall from Model 1 Table 3-2 (Study 1) that the difference between vaccine uptake rates, 38.9% (t-statistic: 11.96), is an estimate of the average causal effect as described in (3). And, since obtaining a flu vaccine was a dummy variable then the average causal effect is the causal effect on vaccine uptake rates. Similarly, Model 1 Table 3-4 (Study 2) demonstrates the difference between vaccine intent rates was 15.9% (t-statistic: 4.28). Taken together, Study 2 demonstrates the effect of recommendations on influenza vaccine receipt is likely to be biased, despite slight differences in study outcomes. It is possible to measure the likely bias from the theoretical framework described in this study. From (6), ! ! ! =39% (Study 1) and ! ! =16% (Study 2), suggesting the omitted variable bias, ! ! ! ! , is positive and equal to 23% by construction. The causal estimate from Study 2 can also be recovered from Study 1.by controlling for the response !81! to whether the respondent requested the flu vaccine from their provider (67.7%, t- statistic: 19.71, Table 3-2 Model 3). Table 3-2 Model 3 demonstrates the recommendation effect experiences a 50% reduction from 33.4% (t-statistic: 9.87) to 13.6% (t-statistic: 4.71); which is reasonably close to the 13.5% (t-statistic: 4.07, Table 3-4 Model 6) and 20.3% (t-statistic: 6.37, Table 3-4 Model 3) causal effects estimates from Study 2. Evaluating how respondents agree to the Study 1 statement, “I would have obtained the seasonal flu vaccination had I not received a recommendation from my doctor or healthcare provider to get the flu vaccination” further establishes the presence of a positive omitted variable. Recall from (3) that selection bias is ! ! !! ! ! = 1 − ![! !! |! ! = 0]. ![! !! |! ! = 0] can be measured from Study 1 as 21.6% (Table 3-1). ! ! !! ! ! = 1 is unobservable; yet a proxy to this measure comes from the proportion of respondents that Somewhat Agree, Agree, and Strongly Agree to whether they would obtain the flu vaccine without a recommendation (Study 1). This is estimated to be 76.6% from Study 1 (Table 3-1). ! ! !! ! ! = 1 −![! !! |! ! = 0] is equal to 54.9% by construction. Comparing this value to the 23% ! ! ! ! calculation and 67.8% (t-statistic: 19.71, Table 3-2 Model 3) effect from requesting flu vaccinations from their provider suggests the omitted variable bias is likely to be positive for this sample. 4.3 Can the predisposition effect from Equation (4) be estimated? The measured omitted variable bias resulting from Study 1 and 2 can be decomposed into ! ! and ! ! . This presentation primarily focuses on Experiments 2 and 4 to reduce the !82! potential correlation with knowing the doctor obtained the flu vaccine in Experiments 1 and 3. From the Equation (4) model specification, ! ! can be represented by Study 2 survey respondents’ response to questions 1-3 (Table 3-5). Equation (4) was estimated with ! ! equal to the individual’s response to whether the influenza vaccine is effective (e.g., Question 1 response ranging from 0-4). ! ! and ! ! was 18.3% (t=5.47) and 31.9% (t=13.24), respectively. Thus, from equation (6) and since ! ! ! =39% (from Study 1, Model 1, Table 3-2), ! ! (the slope of ! ! on ! ! ) is 64.9% by construction. Thus, the correlation between ! ! on ! ! is positive. Table 3-6 shows ! ! , ! ! and ! ! estimates when ! ! is defined from the other pretreatment questions (i.e., Questions 2 and 3). 4.4 Sensitivity analysis on flu vaccine intent and actual flu vaccination Not all individuals declaring their intent to acquire the flu vaccine in Study 2 will actually obtain the flu vaccine [Harris et al., 2010]. This prior work demonstrates approximately 50% of individuals will actually get the flu vaccine at the end of the flu season after declaring their intent to get the flu vaccine mid flu season. It also demonstrates at least 95% of individuals with no intention to get the vaccine mid season will be unvaccinated by end-of-season. Therefore, using the results from Experiments 2 and 4, Table 3-7 assumes 50% of Survey 2 respondents intending to obtain the flu vaccine will actually get the flu vaccine while 100% of respondents remain unvaccinated if they do not intend to get the vaccine. The difference between vaccine uptake rates becomes approximately !83! 8%. Subsequently, the omitted variable bias, by construction, increases to approximately 31%. 5.0 Discussion Many of us develop an understanding of flu vaccine effectiveness. Yet, how does an extrinsic intervention such as a provider recommendation influence someone’s decision to become vaccinated in the presence of intrinsic incentives or barriers (related to reducing the spread of influenza infections)? This study presents a significant, strong and positive correlation with provider recommendations for flu vaccines and flu vaccine uptake. The unadjusted recommendation effect on flu vaccinations is 38.9% (t-statistic: 11.96); that is, receiving a recommendation increases the chance of obtaining the flu vaccine by approximately 39%. Similar to prior work in the recommendation effect for flu vaccines, controlling for respondent characteristics (e.g., demographic, socioeconomic, health status) reduces the recommendation effect on flu vaccination to 33.3% (t-statistic: 9.87). These findings support the hypothesis that provider recommendations will increase vaccination rates (Hypothesis 1). However, the estimated 33% recommendation effect is potentially biased if predisposition is omitted from regression analysis. If this intrinsic motivation exists and is correlated with flu vaccinations at higher levels than provider recommendations, then provider recommendations are overestimated (Hypothesis 2). This study proposes and !84! evaluates two novel and simple methods of measuring predisposition: (a) whether the respondent requested the flu vaccination from their provider; and, (b) whether the respondent would have obtained the flu vaccine without a recommendation. The act of requesting can be a representation of someone’s attraction to the vaccine whether it be due to availability of the vaccine or reduced costs from their provider. Controlling for whether the respondent requested the flu vaccine from their provider reduces the recommendation effect on flu vaccinations to 13.7% (t-statistic: 4.71); thus, the findings support the hypothesis that provider recommendations will slightly increase vaccination rates after adjusting for a measure of intrinsic motivation (Hypothesis 2). In other words, the omitted variable bias (resulting from omitting the vaccine seeking behavior from providers) is positive. Further supporting Hypothesis 2 is the measure of selection bias (i.e., omitted variable bias) from evaluating whether the respondent would have obtained the flu vaccine without a recommendation. This question represents the counterfactual scenario (! ! !! ! ! = 1 ) that is difficult to measure; yet, this study estimates it from the proportion of respondents agreeing with the counterfactual question. And, since ! ! !! ! ! = 0 is also estimated from the study, the constructed selection bias is 54.9%; which suggests a positive omitted variable bias. While the findings are exploratory, it is definitive in demonstrating provider recommendations are overestimated if researchers do not consider a predisposition effect from respondents. !85! By controlling for predisposition (in the form of vaccine-seeking behavior) in regression or subpopulation analyses, the recommendation effect is reduced by at least 50%. This provides early indication that recommendations are not as strong of a motivator to someone’s vaccination receipt in the presence of a behavior. The two hypotheses evaluated from Study 1 are validated in the randomization of recommendations in Study 2. The unadjusted recommendation effect after randomization is 15.9% (t-statistic: 4.28) and adjusting for patient characteristics (e.g., demographics, socioeconomics, beliefs regarding flu vaccine effectiveness) marginally increases the effect to 20.3% (t-statistic: 6.37). The recommendation effect remains relatively stable after including the conditions of provider vaccination status; therefore, the estimated unbiased range of the causal recommendation effect is 12.2% to 20.3% (Table 3-4). When compared to the adjusted Study 1 recommendation effect of 13.7%, the findings support the notion that the recommendation effect is biased upwards if predisposition is omitted from regression analysis. Through this sequence of evaluating and validating the existence of predisposition (in the context of flu vaccinations) as an omitted variable when measuring a recommendation effect, one can conclude that recommendations are still important for improving vaccination rates. But, it may not be as important to improving vaccination rates after considering intrinsic motives towards the vaccine. Furthermore, the reliance of the selection-on-observables assumptions exhibited in prior work should be reconsidered given the existence of predisposition. To ensure a more accurate measure of the recommendation effect, future work should strongly consider evaluating respondent !86! disposition towards the flu vaccine. This can be achieved by including simple questions incorporating vaccine-seeking behavior to reduce the bias in regression analyses. Further observations resulting from including predisposition into regression analyses are worth noting. First, the goodness of fit of the regression model improves two-fold to 0.49 (Table 3-2 Model 3) from 0.23 (Table 3-2 Model 2) after adjusting for predisposition (i.e., requested flu vaccination from their provider). This suggests predisposition (the omitted variable) is strongly correlated to flu vaccinations even after controlling for provider recommendations and patient characteristics. Second, it is also possible to characterize predisposition through patient beliefs in vaccine effectiveness. Controlling for these patient beliefs significantly improves the model goodness of fit (Table 3-4 Models 3 versus 2 and Models 6 versus 5). And with prior work relying on the selection- on-observables assumption in controlling for imperfect measures of patient beliefs, this study improves on prior study designs through the randomization procedure; which ensures patient beliefs are independent from any influence recommendations may have on opinions about vaccine effectiveness. Lastly, further attempts to characterize predisposition were conducted from Study 2 (Table 3-6). By defining predisposition as a patient belief (another form of intrinsic motivation and measured independently from provider recommendations), it is possible to recover a significant and positive omitted variable (Question 1 and 2 Table 3-6). This similarity in direction with patient behavior (i.e., requesting the flu vaccine from the !87! provider) further acknowledges the importance of controlling for predisposition when evaluating a recommendation effect. This study further extends the literature on provider recommendations for flu vaccinations by including the vaccination status of the provider. This has potential towards promoting compulsory vaccination policies for healthcare workers because recommendations from a vaccinated provider can improve patient vaccination rates (Hypothesis 3). Thus, these compulsory vaccination policies can indirectly improve vaccination rates. The findings from Study 2 suggests knowing that their provider was vaccinated was not salient to the patient in the presence of a provider recommendation. The recommendation effect remains relatively stable (ranging from 12.2% to 13.5%, Table 3-4) with or without the knowledge that their provider received the flu vaccine. This does not support the notion that recommendations coming from a vaccinated provider will considerably increase vaccination rates (Hypothesis 3). And, while compulsory vaccination policies reduce the spread of influenza infections, this study adds doubt to whether it can indirectly improve patient vaccination rates. 5.1 Predisposition has broad implications Predisposition is a novel description of an individual’s underlying motivation (i.e., intrinsic motivation) to obtain the flu vaccine. This description is similar in theory to research on returns to education where ability describes an individual’s underlying relationship with their future income [Angrist et al. 2008]. For example, ‘returns to recommendation’ is correlated with how the individual perceives the flu vaccine in the !88! presence of a provider recommendation. This correlation can be in two directions. First, a negative correlation exists if the strength of the recommendation cannot overcome other barriers to flu vaccination such as fear of vaccine side effects. Second, a positive correlation exists if patients exhibit vaccine-seeking behavior that strengthens the recommendation effect on flu vaccine receipt. In this paper, predisposition is revealed to be a likely reason for the overestimation reported from these two studies. In a broader context, the findings extend research in empirical health economics. The consumer demand function for health-related goods is often studied when investigating the role of health information on consumer demand for medical care [Culyer and Newhouse, 2000; Hsieh and Lin, 1997; Kenkel, 1990; Parente et al., 2005]. This health information can come from multiple sources (i.e., advice nurses and computers) that influence the consumer’s opinion about medical care effectiveness; potentially decreasing reliance on health professionals for information [Wagner et al., 2001]. The developing predisposition towards a medical intervention can influence the demand for that intervention. The implications of this study reveal health information from a health professional may not be salient to patients with a positive predisposition towards influenza vaccines. Patient-provider communication practices are widely researched to the point where patient values and preferences are becoming important elements in healthcare decision- making [Legare and Witteman, 2013; Opel et al., 2013; Street and Haidet, 2013]. A prominent theme in this area of work is the notion that providers with an accurate !89! understanding of their patients’ health beliefs would be better equipped to reconcile differences in how the patient or provider perceives the patient’s illness; thus, allowing for a unique and personal solution to patient treatment. Yet, where do these patient health beliefs originate and how do they affect the demand for health services? A number of sources include printed health, computer-based resources, and healthcare organizations [Dwyer and Liu, 2013; Nicholson et al., 2005]. The question then becomes which is more salient to the patient when they consume health services: their own beliefs or provider council? In the present study, patient predisposition towards the flu vaccine is a stronger predictor of obtaining flu vaccinations than provider recommendations. In this case, patients have a basis for the flu vaccine and providers can only add to the demand of flu vaccine with council. 5.2 Limitations There are several limitations to this work. First, the hypothetical scenarios in Study 2 may not be realistic to all patients. The scenarios assume survey respondents experienced routine physicals from their regular doctor. Despite this restriction in purpose for the doctor’s visit, patients likely experience variations in recommendations and doctor’s flu vaccination status that affect their decision to obtain the flu vaccine. There is limited information in describing the type of doctor’s visits that would inform these experimental conditions. These findings should be placed in the context of this limitation in experimental conditions. !90! Second, the sample population for Studies 1 and 2 was not representative of the general US population. Prior research demonstrates minority racial/ethnic groups such as Non- Hispanic, Black adults exhibit resistance to vaccinations [Hebert, Frick, Kane, and McBean, 2005; Fiscella, 2005]. The results from this study may not be consistent with different demographic groups according to race/ethnicity. More generally, there may be other demographic groups (e.g., parents deciding to vaccinate their children) resistant to the flu vaccine for reasons related to vaccine side effects. Third, not all Study 2 respondents with intent to obtain the flu vaccine will actually acquire the vaccine. Based on prior work, if this study assumes half of the respondents intending to get the flu vaccine actually obtain the flu vaccine, the estimated recommendation effect on flu vaccination is reduced. The subsequent overestimation increases as a result. 5.3 Direction for future research Capturing the behavior of requesting the flu vaccine from their provider (i.e., Question 2 in Study 1) is relatively simple description of predisposition. By controlling for this question in the regression analysis the in Study 1, the causal effects estimate in Study 2 can be approximately recovered. This type of behavioral question is missing in national immunization surveys (e.g., National Immunization Survey, National Health Interview Survey, and Behavioral Risk Factor Surveillance System). Incorporating behavioral questions towards the vaccine can accurately capture the provider recommendation effect on flu vaccinations. !91! Additional implications of this study involve the growing body of work related to provider vaccine communication practices to improve vaccination behavior [Opel et al., 2013]. The Study 2 experimental condition that patients know their doctor’s flu vaccination status add to this literature by informing communication practices. The findings from this experimental condition suggest knowing that the provider was vaccinated (in the absence of a recommendation) modestly adds approximately 5- percentage points to a patient’s flu vaccination intent (Experiment 3 versus 4 in Table 3- 3, Study 2). However, in the presence of a recommendation, a provider’s vaccination status is overshadowed by the recommendation effect on flu vaccine intent (Experiment 1 versus 2 in Table 3-3, Study 2). The proportion of respondents intending to get the flu vaccine remains stable in the presence of a recommendation and knowing whether the doctor obtained the flu vaccine. 6.0 Conclusion The effects of healthcare provider recommendations on flu vaccination are likely overestimated. This upward bias can be characterized from an individual’s underlying opinion about vaccine effectiveness or vaccine-seeking behavior. !92! 7.0 References Angrist JD and Pischke JS. Most harmless econometrics: an empiricist’s companion. (2008) Princeton University Press, Princeton, New Jersey. Bhatt A, Bridges C, Donoghue K, Fernandez C, Gehring R, Hall LL, et al. Recommendations from the National Vaccine Advisory Committee: standards for adult immunization practice. Public Health Rep. 2014; 129(2): 115-23. Boynton MH, Richman LS. An online daily diary study of alcohol use using Amazon’s Mechanical Turk. Drug and Alcohol Review. 2014;33(4):456-461. Charness G and Gneezy U. Incentives to exercise. Journal of the Econometric Society 2009;77(3):909-931. Culyer AJ and Newhouse JP. (2000) Handbook of Health Economics, Volume 1A. Amsterdam: New Holland. Ding H, Santibanez TA, Jamieson DJ, et al. Influenza vaccination coverage among pregnant women – National 2009 H1N1 Flu Survey (NHFS). Am J Obstet Gynecol 2011;204(6):S96-S106. Dominguez SR, Daum RS. Physician knowledge and perspectives regarding influenza and influenza vaccination. Human Vaccines 2005;1(2):74-79. Dwyer DS, Liu H. The impact of consumer health information on the demand for health services. The Quarterly Review of Economics and Finance 2013;53:1-11. Faul F, Erdfelder E, Buchner A, Lang AG. Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods. 2009;41:1149-1160. Fiscella K. Commentary – Anatomy of racial disparity in influenza vaccination. HSR: Health Services Research 2005;40(2): 539-550. Green WH. Econometric analysis. (2008, 6 th edition) Pearson Education, Inc., Upper Saddle River, New Jersey. Harris KM, Maurer J, Lurie N. Do people who intend to get a flu shot actually get one? J Gen Intern Med. 2009;24(12):1311-1313. Hebert PL, Frick KD, Kane RL, McBean AM. The causes of racial and ethnic differences in influenza vaccination rates among elderly Medicare beneficiaries. HSR: Health Services Research 2005;40(2): 517-538. !93! Hsieh CR and Lin SJ. Health information and the demand for preventive care among the elderly in Taiwain. 1997;32(2):308-333. Jessop AB, Dumas H, Moser CA. Delivering influenza vaccine to high-risk adults: subspecialty physician practices. American Journal of Medical Quality 2013;28:232-237. Kenkel D. Consumer health information and the demand for medical care. The Review of Economics and Statistics 1999;72(4):587-595. Legare F, Witteman HO. Shared decision making: Examining key elements and barriers to adoption into routine clinical practice. Health Affairs 2013;32(2):276-284. Levy DJ, Ambrose CS, Oleka N, Lewin EB. A survey of pediatricians’ attitudes regarding influenza immunization in children. BMC Pediatrics 2009;9(8):1-5. Maurer J, Harris KM, Black CL, Euler GL. Support for seasonal influenza vaccination requirements among US healthcare personnel. Infect Control Hosp Epidemiol 2012;33(3):213-221. Nichol KL, Zimmerman R. Generalist and subspecialist physicians’ knowledge, attitudes, and practices regarding influenza and pneumococcal vaccinations for elderly and other high-risk patients: a nationwide survey. Arch Intern Med 2001;161:2702-2708. Nicholson W, Gardner B, Grason HA, Power NR. The association between women’s health information use and health care visits. Women’s Health Issues 2005;15:240-248. Opel DJ, Heritage J, Taylor JA, Mangione-Smith R, Salas HS, DeVere V, et al. The architecture of provider-parent vaccine discussion at health supervision visits. Pediatrics 2013; 132(6): 1037-1046. Pandolfi E, Marino MG, Carloni E, et al. The effect of physician’s recommendation on seasonal influenza immunization in children with chronic diseases. BMC Public Health 2012;12(1):984. Parente ST, Salkever DS, DaVanzo J. The role of consumer knowledge of insurance benefits in the demand for preventive health care among the elderly. 2005;14(1):25-38. Rickert D, Santoli J, Shefer A, Myrick A, Yusuf H. Influenza vaccination of high-risk children: what the providers say. Am J Prev Med. 2006;30:111–118. doi: 10.1016/j.amepre.2005.10.016. !94! Santibanez TA, Mootrey GT, Euler GL, Janssen AP. Behavior and beliefs about influenza vaccine among adults aged 50-64 years. Am J Health Behav 2010;34(1):77-89. Street RL, Haidet P. How well do doctors know their patients? Factors affecting physician understanding of patients’ health beliefs. J Gen Intern Med 2010;26(1):21-27. United States Census Bureau; American Community Survey, 2013 American Community Survey 1-Year Estimates, Tables S0201 and DP05; generated by Reginad Villacorta; using American FactFinder; <http://factfinder2.census.gov>; (29 December 2014). Wagner T, Hu TW, Hibbard JH. The demand for consumer health information. 2001;20(6):1059-1075. !95! 8.0 Figures ! ! ! Figure 3-1. Study 1 sample flow diagram ! ! ! Enrollment period: 16 March 2014 to 17 March 2014 Assessed for eligibility (n=816) Excluded (n=22) - No confirmation code response signifying survey completion (n= 17) - Confirmation code was not auto-generated (n= 5) Analyzed (n=794) !96! ! ! Figure 3-2. Study 2 sample flow diagram. Code mismatch is an incorrect survey confirmation code for completion. Enrollment period: 23 March 2014 to 27 March 2014 Assessed for eligibility (n=1431) Randomized (n=1415) Scenario 1 (n=354) Scenario 2 (n= 354) Analyzed (n=345) Analyzed (n=351) Analyzed (n=349) Analyzed (n=349) Excluded (n=9) - Code mismatch (n=9) Excluded (n=3) - Code mismatch (n=3) Excluded (n=16) - Did not consent to participate (n=3) - Did not continue after consent (n=13) Scenario 3 (n=354) Excluded (n=5) - Code mismatch (n=5) Scenario 4 (n=353) Excluded (n=4) - Code mismatch (n=4) !97! 9.0 Tables Table 3-1: Study 2 experimental scenarios Experimental scenario HCP vaccination status HCP recommendation 1 Vaccinated Recommendation 2 Unknown Recommendation 3 Vaccinated No recommendation 4 Unknown No recommendation Abbreviations: HCP, healthcare provider !98! Table 3-2. Descriptive statistics for Study 1 respondents! Flu vaccination status All respondents Not vaccinated Vaccinated (N=794) (N=510) (N=284) p-value HCP recommendation No 505 396 109 <0.001 Yes 289 114 175 Requested flu vaccination from HCP No 619 505 114 <0.001 Yes 175 5 170 Would obtain flu vaccine without recommendation Strongly disagree NA NA a 11 NA Disagree 16 Somewhat disagree 13 Somewhat agree 27 Agree 64 Strongly agree 43 Unknown 1 Healthcare worker No 729 498 231 <0.001 Yes 65 12 53 Direct patient care if healthcare worker b No 31 10 21 0.006 Yes 34 2 32 Health insurance past 12 months No 203 152 51 <0.001 Yes 588 355 233 Unknown 3 3 0 Self-reported current health status Excellent 146 91 55 0.262 Very good 333 202 131 Good 217 148 69 Fair 83 58 25 Poor 15 11 4 !99! Chronic condition in the past 2 years c No 651 441 210 <0.001 Yes 143 69 74 Region d 1 169 102 67 0.764 2 182 119 63 3 232 148 84 4 208 139 69 Unknown 3 2 1 Education Grammar school 2 1 1 0.024 High School or equivalent 84 66 18 Vocational/Technical School (2 year) 24 14 10 Some college 330 212 118 College graduate (4 year) 277 179 98 Master's Degree (MS) 57 26 31 Doctoral Degree (PhD) 8 5 3 Professional Degree (MD, JD, etc.) 11 6 5 Other 1 1 0 Gender Female 273 163 110 0.149 Male 516 344 172 Unknown 5 3 2 Current marital status Divorced 32 22 10 0.009 Married 210 117 93 Separated 5 1 4 Single 544 367 177 Widowed 1 1 0 Unknown 2 2 0 Household income <$10,000 81 64 17 0.001 $10,000 - $19,999 84 66 18 $20,000 - $29,999 106 69 37 ! 100! $30,000 - $39,999 102 65 37 $40,000 - $49,999 96 62 34 $50,000 - $74,999 159 93 66 $75,000 - $99,999 71 34 37 $100,000 - $150,000 67 40 27 >$150,000 24 14 10 Missing 4 3 1 Age 18-24 5 2 3 0.511 25-34 279 175 104 35-44 346 234 112 45-54 102 64 38 55-64 42 24 18 65+ 17 9 8 Unknown 3 2 1 Race White/Caucasian 627 407 220 0.773 African American 36 22 14 Hispanic 42 29 13 Asian 73 44 29 Native American 4 2 2 Pacific Islander 3 2 1 Other 8 4 4 Unknown 1 0 1 Abbreviations: NA, not applicable; HCP, healthcare provider a This question was asked to respondents with a flu vaccination and recommendation b This was asked to respondents that were healthcare workers c Chronic medical condition is dummy variable indicating whether the respondent was advised by their healthcare provider that they had asthma, a lung condition other than asthma, a heart condition, diabetes, a neurological or neuromuscular condition, a liver condition, and a weakened immune system due to chronic illness in the past 2 years. d Region 1: Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont, New Jersey, New York, and Pennsylvania; Region 2: Illinois, Indiana, Michigan, Ohio, Wisconsin, Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, and South Dakota; Region 3: Delaware, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia, Washington D.C., West Virginia, Alabama, Kentucky, Mississippi, Tennessee, Arkansas, Louisiana, Oklahoma, and Texas; Region 4: Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, Wyoming, Alaska, California, Hawaii, Oregon, and Washington ! ! ! ! 101! ! Table 3-3. Effect of provider recommendations on flu vaccine receipt! Model 1 a Model 2 b Model 3 c Coefficient Coefficient Coefficient HCP recommendation d 0.3897 0.3337 0.1374 (11.96) (9.87) (4.71) Requested flu vaccination from their provider - - 0.6777 (19.71) Sample size 794 794 794 Adjusted R-square 0.1520 0.2306 0.4943 Abbreviation: HCP, healthcare provider a Ordinary least squares regression unadjusted for respondent characteristics in Table 3-1 b Ordinary least squares regression adjusted for respondent characteristics in Table 3-1. This model omits the question: Respondent requested flu vaccination from HCP. Response to questions described in Table 3-1 Footnote ‘a’ and ‘b’ were omitted from the regressions because not all respondents were asked these questions. c Ordinary least squares regression adjusted for respondent characteristics in Table 3-1. Response to questions described in Table 3-1 Footnote ‘a’ and ‘b’ were omitted from the regressions because not all respondents were asked these questions. d t-statistics in parentheses ! ! ! ! ! ! 102! ! Table 3-4. Descriptive statistics for Study 2 respondents Experimental scenario 1 2 3 4 Doctor obtained the flu vaccine Known Unknown Known Unknown Doctor recommended the flu vaccine Yes Yes No No p-value N (Total: 1394) 345 351 349 349 a Flu vaccine intent No 133 139 177 194 <0.001 Yes 212 212 172 155 The seasonal flu vaccination is effective in preventing the seasonal flu. Strongly disagree 9 12 7 5 0.448 Disagree 46 51 55 48 Agree 209 219 213 211 Strongly agree 81 69 74 83 Missing 0 0 0 2 If you didn't get a seasonal flu vaccination, I have a high chance of getting sick with the seasonal flu. Strongly disagree 44 37 42 35 0.149 Disagree 169 193 163 168 Agree 111 95 124 109 Strongly agree 21 26 19 36 Missing 0 0 1 1 I am worried about getting sick from the seasonal flu vaccine Strongly disagree 115 109 102 108 0.299 Disagree 121 141 146 134 Agree 79 68 84 81 Strongly agree 30 33 17 26 Missing 0 0 0 0 Currently works in a healthcare facility such as a hospital, medical clinic, doctor's office, or nursing home. This include part-time and unpaid work in a healthcare facility and home nursing care No 321 315 322 326 0.487 Yes 24 35 26 23 Missing 0 1 1 0 ! 103! Provides direct patient care such as physical or hands-on contact with patients b No 12 16 9 15 0.195 Yes 12 19 17 8 Had health care coverage in past 12 months (i.e., health insurance, prepaid plans such as HMOs, or government plans such as Medicaid) No 82 95 84 102 0.457 Yes 262 256 265 246 Missing 1 0 0 1 Rate of current health status Excellent 64 71 58 53 0.372 Very good 148 152 157 141 Good 104 97 99 106 Fair 24 27 30 36 Poor 5 4 4 12 Missing 0 0 1 1 Chronic condition in past 2 years c No 291 293 284 282 0.562 Yes 54 58 65 67 Region d 1 69 69 54 80 0.143 2 85 76 86 70 3 120 114 123 109 4 71 92 86 88 Unknown 0 0 0 2 Education Grammar school 2 1 1 1 0.809 High School or equivalent 44 29 34 33 Vocational/Tech nical School (2 year) 11 15 15 17 Some college 127 143 151 140 College graduate (4 year) 122 128 119 117 Master's Degree (MS) 33 27 19 28 Doctoral Degree (PhD) 2 2 3 4 ! 104! Professional Degree (MD, JD, etc.) 3 6 6 9 Other 0 0 0 0 Missing 1 0 1 0 Gender Female 109 126 132 130 0.277 Male 235 223 215 214 Unknown 1 2 2 5 Current marital status Divorced 16 15 14 12 0.936 Married 102 92 99 110 Separated 4 4 5 3 Single 222 238 229 222 Widowed 0 1 2 2 Unknown 1 1 0 0 Household income <$10,000 35 21 37 37 0.296 $10,000 - $19,999 44 38 34 34 $20,000 - $29,999 55 61 49 42 $30,000 - $39,999 38 46 42 50 $40,000 - $49,999 41 36 44 33 $50,000 - $74,999 55 68 80 73 $75,000 - $99,999 31 38 35 35 $100,000 - $150,000 29 29 12 25 >$150,000 16 14 15 19 Missing 1 0 1 1 Age 18-24 0 5 1 0 <0.001 25-34 107 123 122 126 35-44 140 153 176 133 45-54 66 49 36 46 55-64 22 14 9 31 ! 105! 65+ 9 7 3 13 Unknown 1 0 2 0 Race White/Caucasian 270 284 272 264 0.930 African American 17 16 17 27 Hispanic 18 17 20 14 Asian 31 26 30 35 Native American 2 1 2 1 Pacific Islander 0 2 2 1 Other 5 5 5 6 Unknown 2 0 1 1 a Two respondents did not provide an answer to flu vaccine intent. These two were included as No responses. b This was asked to respondents that were healthcare workers c Chronic medical condition is dummy variable indicating whether the respondent was advised by their healthcare provider that they had asthma, a lung condition other than asthma, a heart condition, diabetes, a neurological or neuromuscular condition, a liver condition, and a weakened immune system due to chronic illness in the past 2 years. d Region 1: Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont, New Jersey, New York, and Pennsylvania; Region 2: Illinois, Indiana, Michigan, Ohio, Wisconsin, Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, and South Dakota; Region 3: Delaware, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia, Washington D.C., West Virginia, Alabama, Kentucky, Mississippi, Tennessee, Arkansas, Louisiana, Oklahoma, and Texas; Region 4: Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, Wyoming, Alaska, California, Hawaii, Oregon, and Washington ! ! ! ! 106! ! Table 3-5. Effect of provider recommendations on flu vaccine intent! Experimental conditions 2 and 4 1 and 3 Model 1 a 2 b 3 b 4 a 5 b 6 b Coef Coef Coef Coef Coef Coef HCP recommendation c 0.1599 0.1708 0.2027 0.1217 0.1361 0.1353 (4.28) (4.43) (6.37) (3.24) (3.51) (4.07) Flu vaccine is effective Strongly agree/Agree d - - 0.3549 - - 0.4039 (7.67) (8.54) High chance of getting flu without a flu vaccine Strongly agree/Agree d - - 0.4319 - - 0.3607 (12.49) (10.34) Worried about getting sick from flu vaccine Strongly agree/Agree d - - -0.0478 - - -0.0419 (-1.31) (-1.13) Sample size 700 700 700 694 694 694 Adjusted R-square 0.0242 0.0387 0.3480 0.0135 0.0483 0.3034 Abbreviation: HCP, healthcare provider; Coef, coefficient a Uncontrolled for respondent characteristics in Table 3-3. b Controlling for respondent characteristics in Table 3-3. c t-statistics in parentheses d Compared to Strongly Disagree/Disagree/No response ! ! ! ! ! 107! ! Table 3-6. Agreement with seasonal flu vaccination effectiveness! Question a Obs Mean Std Dev Min Max 1 b 700 3.019 0.694 0 4 2 c 700 2.363 0.790 0 4 3 d 700 2.071 0.925 1 4 a Response categories for each question: 0, no response; 1, strongly disagree; 2, disagree; 3, agree; 4, strongly agree b Question 1: The seasonal flu vaccination is effective in preventing the seasonal flu c Question 2: If I didn't get a seasonal flu vaccination, I have a high chance of getting sick with the seasonal flu d Question 3: I am worried about getting sick from the seasonal flu vaccine Abbreviations: Obs, observations; Std Dev, standard deviation; Min, minimum; Max, maximum ! ! ! ! ! ! 108! ! Table 3-7. Measures of predisposition! Question a 1 2 3 ! ! ! b 0.3897 0.3897 0.3897 (11.96) (11.96) (11.96) ! ! c 0.1827 0.1902 0.1598 (5.47) (5.82) (4.33) ! ! c 0.3191 0.3055 -0.0822 (13.24) (14.76) (-4.12) ! ! d 0.6487 0.6532 -2.7958 a t-statistics in parentheses b Estimated from Study 1 with no adjustment c Estimated from Study 2 with no adjustment d Constructed from ![! ! ]=! ! +! ! ! ! (6) ! ! ! ! ! 109! ! Table 3-8. Predicted probability of flu vaccination! Recommendation status Predicted no flu vaccine Predicted flu vaccine a Total Probability flu vaccine intent Predicted probability flu vaccine c Received 245 106 351 60.40% 30.20% Did not receive 271.5 77.5 349 44.41% 22.21% Total 516.5 183.5 700 15.99% b 7.99% d a Predicted estimates assume 50% of respondents intending to get the vaccine in Experiment 2 and 4 actually get the vaccine. The remaining respondents do not become vaccinated. b This estimate is based on Model 1 Table 3-5 by taking the difference between 60.40% and 44.41%. c The revised estimate on predicted vaccination status by taking the difference between 30.20% and 22.21%. ! ! 110! BIBLIOGRAPHY Angrist JD and Pischke JS. Most harmless econometrics: an empiricist’s companion. (2008) Princeton University Press, Princeton, New Jersey. Annunziata K, Rak A, Del Buono H, et al. Vaccination rates among the general adult population and high-risk groups in the United States. PLoS One 2012;7(11):e50553. Armstrong K, Merlin M, Schwartz JS, et al. Barriers to influenza immunization in a low- income urban population. Am J Prev Med 2001;20(1):21-25. Bhatt A, Bridges C, Donoghue K, Fernandez C, Gehring R, Hall LL, et al. Recommendations from the National Vaccine Advisory Committee: standards for adult immunization practice. Public Health Rep. 2014; 129(2): 115-23. Bowblis JR, Yun MS. Racial and ethnic disparities in the use of drug therapy. Social Science Research 2010;39:674-684. Boynton MH, Richman LS. An online daily diary study of alcohol use using Amazon’s Mechanical Turk. Drug and Alcohol Review. 2014;33(4):456-461. Bustamante AV, Chen J, Rodriguez HP, Rizzo JA, Ortega AN. Use of preventive care services among Latino subgroups. Am J Prev Med 2010;38(6): 610-619. Bustamante AV, Fang H, Rizzo JA, Ortega AN. Understanding observed and unobserved health care access and utilization disparities among U.S. Latino adults. Med Care Res Rev. 2009;66(5):561-577. Centers for Disease Control and Prevention. Behavioral Risk Factor Surveillance System. 2013. http://www.cdc.gov/brfss. Centers for Disease Control and Prevention. CDC Health Disparities and Inequalities Report – United States, 2011. Health-care access and preventive health services: Influenza vaccination coverage – United States, 2000-2010. MMWR 2011;60(Suppl):38-42. Centers for Disease Control and Prevention. Flu vaccination coverage, United States, 2013-14 influenza season. 2014. http://www.cdc.gov/flu/fluvaxview/coverage- 1314estimates.htm. Centers for Disease Control and Prevention. Prevention and control of seasonal influenza with vaccines: recommendations of the Advisory Committee on Immunization Practices (ACIP), 2009. MMWR Recommendations and Reports 2009;58(RR08):1-52. Centers for Disease Control and Prevention. Self-reported influenza vaccination coverage trends 1989–2008 among adults by age group, risk group, race/ethnicity, health-care worker status, and pregnancy status, United States, National Health Interview Survey (NHIS). 2008.! http://www.cdc.gov/flu/pdf/professionals/nhis89_08fluvaxtrendtab.pdf. Centers for Disease Control and Prevention. U.S. Department of Health and Human Services, National Center for Health Statistics. The National 2009 H1N1 Flu Survey (NHFS), Hyattsville, MD: Centers for Disease Control and Prevention, 2012. Information about the NHFS is located at http://www.cdc.gov/nchs/nis/about_nis.htm#h1n1. Centers for Disease Control and Prevention. Use of influenza A (H1N1) 2009 monovalent vaccine: recommendations of the Advisory Committee on ! 111! Immunization Practices (ACIP), 2009. MMWR Recommendations and Reports 2009;58(RR10):1-8. Centers for Disease Control and Prevention. Vaccine Effectiveness – How well does the flu vaccine work? 2014. http://www.cdc.gov/flu/about/qa/vaccineeffect.htm Charness G and Gneezy U. Incentives to exercise. Journal of the Econometric Society 2009;77(3):909-931. Clarke AR, Goddu AP, Nocon RS, …, Chin MH. Thirty years of disparities intervention research: What are we doing to close racial and ethnic gaps in healthcare? Med Care. 2013;51(11):1-14. Culyer AJ and Newhouse JP. (2000) Handbook of Health Economics, Volume 1A. Amsterdam: New Holland. Ding H, Santibanez TA, Jamieson DJ, et al. Influenza vaccination coverage among pregnant women – National 2009 H1N1 Flu Survey (NHFS). Am J Obstet Gynecol 2011;204(6):S96-S106. Dominguez SR, Daum RS. Physician knowledge and perspectives regarding influenza and influenza vaccination. Human Vaccines 2005;1(2):74-79. Dwyer DS, Liu H. The impact of consumer health information on the demand for health services. The Quarterly Review of Economics and Finance 2013;53:1-11. Faul F, Erdfelder E, Buchner A, Lang AG. Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods. 2009;41:1149-1160. Fiebach NH, Viscoli CM. Patient acceptance of influenza vaccination. Am J Med 1991;91(4):393-400. Fiore AE, Shay DK, Broder K, et al. Prevention and control of influenza with vaccines: recommendations of the Advisory Committee on Immunization Practices (ACIP), 2009. MMWR Recomm Rep 2009, 58(RR08):1-52. Fiscella K. Commentary – Anatomy of racial disparity in influenza vaccination. HSR: Health Services Research 2005;40(2): 539-550. Gnanasekaran SK, Finkelstein JA, Hohman K, et al. Parental perspective on influenza vaccination among children with asthma. Public Health Rep 2006;121(2):181- 188. Green WH. Econometric analysis. (2008, 6 th edition) Pearson Education, Inc., Upper Saddle River, New Jersey. Groshkopf LA, Shay DK, Shimabukuro TT, et al. Prevention and control of seasonal influenza with vaccines: recommendations of the Advisory Committee on Immunization Practices (ACIP) – United States, 2013-2014. MMWR 2013;62(RR07):1-43. Gu Q, Sood N. Do people taking flu vaccines need them the most? PLoS One 2011;6(12):e26347. Harris KM, Maurer J, Lurie N. Do people who intend to get a flu shot actually get one? J Gen Intern Med. 2009;24(12):1311-1313. HealthyPeople2020. Immunization and Infections Diseases. 2013. http://www.healthypeople.gov/2020/topicsobjectives2020/objectiveslist.aspx?topi cId=23. ! 112! HealthyPeople2020. Immunization and Infections Diseases. 2014. https://www.healthypeople.gov/2020/topics-objectives/topic/immunization-and- infectious-diseases/objectives. Hebert PL, Frick KD, Kane RL, McBean AM. The causes of racial and ethnic differences in influenza vaccination rates among elderly Medicare beneficiaries. HSR: Health Services Research 2005;40(2): 517-538. Hemingway CO, Poehling KA. Change in recommendation affects influenza vaccinations among children 6 to 59 months of age. Pediatrics 2004;114(4):948- 952. Hsieh CR and Lin SJ. Health information and the demand for preventive care among the elderly in Taiwain. 1997;32(2):308-333. Jessop AB, Dumas H, Moser CA. Delivering influenza vaccine to high-risk adults: subspecialty physician practices. American Journal of Medical Quality 2013;28:232-237. Kenkel D. Consumer health information and the demand for medical care. The Review of Economics and Statistics 1999;72(4):587-595. Ladin K. Decomposing differences in utilization of health services between depressed and non-depressed elders in Europe. Eur J Ageing. 2012;9(1): 51-64. Legare F, Witteman HO. Shared decision making: Examining key elements and barriers to adoption into routine clinical practice. Health Affairs 2013;32(2):276-284. Levy DJ, Ambrose CS, Oleka N, Lewin EB. A survey of pediatricians’ attitudes regarding influenza immunization in children. BMC Pediatrics 2009;9(8):1-5. Lu PJ, Singleton JA, Euler GL, et al. Seasonal influenza vaccination coverage among adult populations in the United States, 2005-2011. Am J Epidemiol. 2013;178(9):1478-1487. Lyn-Cook R, Halm EA, Wisnivesky JP. Determinants of adherence to influenza vaccination among inner-city adults with persistent asthma. Prim Care Resp 2007;16(4):229-235. MacIntosh T, Desai MM, Lewis TT, Jones BA, Nunez-Smith M. Socially-assigned race, healthcare discrimination and preventive healthcare services. PLoS One. 2013;8(5): 1-7 Maurer J, Harris KM, Black CL, Euler GL. Support for seasonal influenza vaccination requirements among US healthcare personnel. Infect Control Hosp Epidemiol 2012;33(3):213-221. Mirza A, Subedar A, Fowler SL, et al. Influenza vaccine: awareness and barriers to immunization in families of children with chronic medical conditions other than asthma. South Med J 2008;101(11):1101-1105. Musa D, Schulz R, Harris R, Silverman M, Thomas SB. Trust in the health care system and the use of preventive health services by older black and white adults. Am J Public Health. 2009;99:1293-1299. National Vaccine Advisory Committee. A pathway to leadership for adult immunization: recommendations of the National Vaccine Advisory Committee. Public Health Rep. 2012; 127(1): 1-41. Nichol KL, Lofgren RP, Gapinski J. Influenza vaccination: knowledge, attitudes, and behavior among high-risk outpatients. Arch Intern Med 1992;152(1):106-110. ! 113! Nichol KL, Zimmerman R. Generalist and subspecialist physicians’ knowledge, attitudes, and practices regarding influenza and pneumococcal vaccinations for elderly and other high-risk patients: a nationwide survey. Arch Intern Med 2001;161:2702-2708. Nicholson W, Gardner B, Grason HA, Power NR. The association between women’s health information use and health care visits. Women’s Health Issues 2005;15:240-248. O’Malley AS, Forrest CB. Immunization disparities in older Americans: determinants and future research needs. Am J Prev Med 2006;31(2):150-158. Opel DJ, Heritage J, Taylor JA, Mangione-Smith R, Salas HS, DeVere V, et al. The architecture of provider-parent vaccine discussion at health supervision visits. Pediatrics 2013; 132(6): 1037-1046. Orenstein WA, Schaffner W. Lessons learned: Role of influenza vaccine production, distribution, supply, and demand – What it means for the provider The American Journal of Medicine. 2008;121: S21-S27. Pandolfi E, Marino MG, Carloni E, et al. The effect of physician’s recommendation on seasonal influenza immunization in children with chronic diseases. BMC Public Health 2012;12(1):984. Parente ST, Salkever DS, DaVanzo J. The role of consumer knowledge of insurance benefits in the demand for preventive health care among the elderly. 2005;14(1):25-38. Poehling KA, Speroff T, Dittus RS, et al. Predictors of influenza virus vaccination status in hospitalized children. Pediatrics 2001;108(6):1-6. Powers DA, Yoshioka H, Yun MS. Mvdcmp: Multivariate decomposition for nonlinear response models. The Stata Journal 2011;11(4): 556-576. Pylypchuk, Y., Selden, T.M., 2008. A discrete choice decomposition analysis of racial and ethnic differences in children’s health insurance coverage. Journal of Health Economics 27, 1109–1128. Reed C, Kim IK, Singleton JA, et al. Estimated influenza illnesses and hospitalizations averted by vaccinations – United States, 2013-14 influenza season. MMWR 2014; 63(49): 1151-1154. Rickert D, Santoli J, Shefer A, Myrick A, Yusuf H. Influenza vaccination of high-risk children: what the providers say. Am J Prev Med. 2006;30:111–118. doi: 10.1016/j.amepre.2005.10.016. Santibanez TA, Mootrey GT, Euler GL, Janssen AP. Behavior and beliefs about influenza vaccine among adults aged 50-64 years. Am J Health Behav 2010;34(1):77-89. Singleton J, Santibanez T, Wortley P. Influenza and pneumococcal vaccination of adults aged ≥65 racial/ethnic differences. Am J Prev Med 2005;29(5):412-420. Street RL, Haidet P. How well do doctors know their patients? Factors affecting physician understanding of patients’ health beliefs. J Gen Intern Med 2010;26(1):21-27. Takayama M, Wetmore CM, Mokdad AH. Characteristics associated with the uptake of influenza vaccination among adults in the United States. Prev Med 2012;54(5):358-362. ! 114! Thompson MG, Shay DK, Zhou H, et al. Updated Estimates of Mortality Associated with Seasonal Influenza through the 2006-2007 Influenza Season. MMWR 2010; 59(33): 1057-1062. Thompson WW, Shay DK, Weintraub E, et al. Influenza-associated hospitalizations in the United States. JAMA. 2004;292:1333-1340. Thompson WW, Shay DK, Weintraub E, et al. Mortality associated with influenza and respiratory syncytial virus in the United States. JAMA. 2003;289:179-186. U.S. Department of Health and Human Servces (DHHS). National Center for Health Statistics. The National 2009 H1N1 Flu Survey, Hyattsville, MD: Centers for Disease Control and Prevention, 2012. Information about the NHFS is located at http://www.cdc.gov/nchs/nis/about_nis.htm#h1n1. United States Census Bureau; American Community Survey, 2013 American Community Survey 1-Year Estimates, Tables S0201 and DP05; generated by Reginad Villacorta; using American FactFinder; <http://factfinder2.census.gov>; (29 December 2014). Vadaparampil ST, Malo TL, Kahn JA, Salmon DA, Lee JH, Quinn GP, et al. Physicians’ human papillomavirus vaccine recommendations, 2009 and 2011. Am J Prev Med 2014;46(1):80-4. Wagner T, Hu TW, Hibbard JH. The demand for consumer health information. 2001;20(6):1059-1075. Wehby GL, Murray JC, McCarthy AM, Castilla EE. Racial gaps in child health insurance coverage in four South American countries: The role of wealth, human capital, and other household characteristics. HSR: Health Services Research. 2011;46(6): 2119-2138. Ylitalo KR, Lee H, Mehta NK. Health care provider recommendation, human papillomavirus vaccination, and race/ethnicity in the US National Immunization Survey. Am J Public Health 2013;103(1):164-9. ! 115! APPENDICES Appendix 1A. Sensitivity of logit model outcomes a Recommendation Any seasonal flu vaccine Any H1N1 flu vaccine Model 1 Model 2 95% Confidence Interval 95% Confidence Interval Variable Odds Ratio Lower Limit Upper Limit Odds Ratio Lower Limit Upper Limit DEMOGRAPHIC Age group 18-34 Reference 35-44 1.026 0.848 1.240 0.854 0.691 1.055 45-54 1.054 0.880 1.262 0.742 0.607 0.907 55-64 1.648 1.372 1.978 0.994 0.814 1.215 65+ 2.030 1.671 2.466 0.887 0.712 1.104 Race/ethnicity Hispanic 1.080 0.857 1.362 1.055 0.826 1.349 Non-Hispanic, Black Only 1.263 1.049 1.522 1.099 0.896 1.349 Non-Hispanic, White Only Reference Non-Hispanic, Other or Multiple Race 1.018 0.807 1.285 0.833 0.651 1.066 Gender Male Reference Female 1.140 1.027 1.266 1.061 0.943 1.194 Married Yes 1.045 0.906 1.206 1.131 0.955 1.340 ! 116! No Reference Missing 1.195 0.529 2.703 1.064 0.600 1.885 Number of children 0 Reference 1 1.153 0.939 1.414 1.250 0.996 1.569 2 1.094 0.825 1.451 1.224 0.908 1.651 3 1.279 0.853 1.919 1.497 0.966 2.318 Missing 1.934 0.868 4.306 1.533 0.617 3.809 Number of people in household 1 Reference 2 1.098 0.936 1.288 1.042 0.867 1.253 3 1.101 0.895 1.354 1.037 0.813 1.323 4 1.210 0.925 1.584 0.970 0.724 1.299 5 1.125 0.770 1.645 1.010 0.666 1.531 6 0.835 0.504 1.384 1.227 0.671 2.245 7 0.867 0.384 1.958 0.666 0.267 1.662 3-category Metropolitan Statistical Area (MSA) status Reference MSA, principle city 1.140 0.985 1.320 1.101 0.932 1.299 MSA, not principle city 1.057 0.930 1.202 0.977 0.846 1.129 Non-MSA Census region of residence b Region 1 Reference Region 2 0.736 0.634 0.855 0.817 0.693 0.963 Region 3 0.662 0.574 0.764 0.729 0.624 0.851 Region 4 0.644 0.542 0.766 0.798 0.663 0.961 ! 117! Interview date 10-Jan Reference 10-Feb 1.002 0.792 1.267 1.059 0.818 1.372 10-Mar 0.948 0.749 1.199 0.929 0.722 1.195 10-Apr 0.948 0.751 1.198 0.921 0.717 1.183 10-May 1.006 0.799 1.267 1.118 0.873 1.431 10-Jun 0.959 0.755 1.219 1.025 0.790 1.329 SOCIOECONOMIC Self-report education level <12 years Reference 12 years 1.249 0.998 1.563 1.192 0.925 1.537 Some college 1.078 0.860 1.351 1.253 0.967 1.624 College graduate 1.094 0.875 1.369 1.249 0.965 1.616 Missing 0.936 0.465 1.884 2.196 1.151 4.190 Income poverty status Above poverty threshold, >=$75,000 income Reference Above poverty threshold, <$75,000 income 1.001 0.879 1.140 0.982 0.851 1.133 Below poverty threshold 0.956 0.751 1.218 1.067 0.821 1.386 Poverty status unknown 1.047 0.864 1.269 0.988 0.800 1.220 Work status Employed Reference Unemployed 0.893 0.701 1.137 1.093 0.823 1.452 Not in labor force 1.064 0.933 1.213 1.003 0.869 1.159 Don't know/Refused/Missing 1.033 0.556 1.918 0.552 0.303 1.006 Works in health care field ! 118! No Reference Yes 1.155 0.977 1.365 1.263 1.055 1.512 Missing 1.200 0.609 2.363 0.995 0.445 2.225 Home rented or owned Home is owned Reference Home is rented or other arrangement 0.875 0.754 1.016 0.958 0.813 1.130 Don't know/Refused/Missing 0.893 0.643 1.241 1.118 0.789 1.583 HEALTH Chronic medical condition c No Reference Yes 1.620 1.449 1.811 1.607 1.423 1.814 Missing 1.201 0.792 1.821 0.897 0.548 1.468 Health status Sick with fever and cough or sore throat in past month No Reference Yes 1.127 0.893 1.423 1.057 0.819 1.365 Missing 0.865 0.397 1.884 1.210 0.538 2.717 Other people in house with fever and cough or sore throat No Reference Yes 0.888 0.755 1.044 0.948 0.797 1.127 Missing 0.706 0.357 1.395 0.684 0.308 1.517 ! 119! ACCESS Has health insurance coverage Yes 1.555 1.244 1.945 1.188 0.933 1.513 No Reference Don't know/Refused/Missing 1.308 0.588 2.906 0.848 0.398 1.808 Number of times seen doctor since August 2009 >=4 Reference 3 0.906 0.770 1.066 0.944 0.792 1.125 2 0.845 0.735 0.970 0.878 0.753 1.025 1 0.694 0.603 0.799 0.797 0.680 0.935 Missing 0.770 0.547 1.085 0.724 0.502 1.045 OPINIONS ABOUT FLU VACCINE Opinion: Effectiveness of H1N1 vaccine Very effective Reference Somewhat effective 0.963 0.845 1.098 0.608 0.531 0.696 Not very effective 1.065 0.847 1.340 0.449 0.346 0.582 Not at all effective 0.934 0.657 1.327 0.675 0.462 0.986 Don't know/Refused/Missing 0.964 0.809 1.150 0.498 0.403 0.615 Opinion: Risk of getting sick with H1N1 flu without vaccine Very high Reference Somewhat high 0.879 0.684 1.131 0.692 0.541 0.885 Somewhat low 0.835 0.644 1.083 0.386 0.297 0.501 ! 120! Very low 0.712 0.543 0.935 0.262 0.199 0.345 Don't know/Refused/Missing 0.875 0.598 1.279 0.459 0.314 0.670 Opinion: Worry about getting sick from the H1N1 vaccine Very worried Reference Somewhat worried 1.052 0.824 1.344 1.093 0.847 1.411 Not very worried 0.962 0.751 1.233 0.738 0.570 0.956 Not at all worried 1.012 0.785 1.305 0.915 0.703 1.191 Don't know/Refused/Missing 1.027 0.500 2.108 1.038 0.515 2.092 Opinion: Effectiveness of seasonal vaccine Very effective Reference Somewhat effective 0.709 0.626 0.802 0.951 0.827 1.094 Not very effective 0.603 0.483 0.753 1.113 0.872 1.421 Not at all effective 0.554 0.411 0.747 0.915 0.664 1.261 Don't know/Refused/Missing 0.416 0.303 0.571 0.926 0.655 1.310 Opinion: Risk of getting sick with seasonal flu without vaccine Very high Reference Somewhat high 0.941 0.781 1.134 1.163 0.959 1.411 Somewhat low 0.539 0.441 0.659 0.945 0.764 1.168 Very low 0.421 0.333 0.532 1.003 0.781 1.289 Don't know/Refused/Missing 0.742 0.501 1.098 1.219 0.810 1.834 Opinion: Worry about getting sick from the seasonal vaccine ! 121! Very worried Reference Somewhat worried 1.247 0.948 1.639 1.082 0.809 1.446 Not very worried 1.096 0.830 1.446 1.184 0.883 1.589 Not at all worried 1.143 0.871 1.501 1.085 0.815 1.444 Don't know/Refused/Missing 1.377 0.676 2.806 1.157 0.585 2.290 a The outcome from Model 1 was defined as a binary variable equal to one when the respondent indicated they received recommendations for seasonal flu vaccination only or both seasonal and H1N1 vaccinations. The outcome from Model 2 was defined as a binary variable equal to one when the respondent indicated they received recommendations for H1N1 flu vaccination only or both seasonal and H1N1 vaccinations. b Region 1: CT, ME, MA, NH, VT, RI, NJ, NY, PA; Region 2: IL, IN, MI, OH, WI, IA, KS, MN, MO, NE, ND, SD; Region 3: DE, DC, FL, GA MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK, TX; Region 4: AZ, CO, ID, MT, NV, NM, UT, WY, AK, CA, HI, OR, WA c This indicates whether the person has any of the following chronic medical conditions: asthma or an other lung condition, diabetes, a heart condition, a kidney condition, sickle cell anemia or other anemia, a neurological or neuromuscular condition, a liver condition, or a weakened immune system caused by a chronic illness or by medicines taken for a chronic illness. ! 122! Appendix 1B. Multinomial logit model of healthcare provider recommendations a Recommendation Both seasonal and H1N1 flu vaccines Seasonal flu vaccine only H1N1 flu vaccine only Model 1A Model 1B Model 1C 95% Confidence Interval 95% Confidence Interval 95% Confidence Interval Variable RRR Lower Limit Upper Limit RRR Lower Limit Upper Limit RRR Lower Limit Upper Limit DEMOGRAPHIC Age group 18-34 Reference 35-44 0.890 0.715 1.107 1.344 0.989 1.826 0.885 0.590 1.328 45-54 0.827 0.666 1.028 1.512 1.160 1.972 0.663 0.443 0.993 55-64 1.276 1.028 1.584 2.535 1.935 3.321 0.843 0.560 1.270 65+ 1.328 1.050 1.679 3.431 2.598 4.530 0.604 0.386 0.944 Race/ethnicity Hispanic 1.027 0.779 1.355 1.296 0.943 1.783 1.410 0.902 2.205 Non-Hispanic, Black Only 1.178 0.945 1.468 1.444 1.128 1.849 1.197 0.813 1.764 Non-Hispanic, White Only Reference Non-Hispanic, Other or Multiple Race 0.838 0.642 1.095 1.383 0.988 1.936 1.077 0.646 1.796 Gender Male Reference Female 1.103 0.971 1.253 1.202 1.044 1.384 1.106 0.865 1.416 Married ! 123! Yes 1.102 0.926 1.312 0.979 0.808 1.187 1.199 0.829 1.734 No Reference Missing 1.050 0.529 2.081 1.544 0.415 5.747 1.874 0.785 4.470 Number of children 0 Reference 1 1.269 0.997 1.615 1.010 0.750 1.362 1.171 0.707 1.940 2 1.223 0.879 1.702 0.875 0.583 1.314 1.191 0.713 1.990 3 1.440 0.878 2.361 1.212 0.701 2.094 1.807 0.899 3.629 Missing 1.783 0.588 5.405 2.464 1.041 5.829 2.149 0.645 7.161 Number of people in household 1 Reference 2 1.104 0.906 1.344 1.099 0.891 1.356 0.898 0.597 1.350 3 1.090 0.841 1.411 1.140 0.871 1.493 0.970 0.561 1.676 4 1.143 0.828 1.576 1.235 0.851 1.792 0.607 0.345 1.065 5 1.142 0.723 1.805 1.061 0.639 1.760 0.689 0.309 1.534 6 0.997 0.547 1.816 0.797 0.379 1.677 1.388 0.542 3.559 7 0.683 0.257 1.811 1.197 0.406 3.528 0.708 0.146 3.424 3-category Metropolitan Statistical Area (MSA) status MSA, principle city 1.182 0.987 1.415 1.056 0.876 1.273 0.899 0.620 1.302 MSA, not principle city 1.002 0.856 1.173 1.165 0.990 1.372 1.028 0.751 1.408 Non-MSA Reference Census region of residence b Region 1 Reference Region 2 0.729 0.610 0.872 0.725 0.594 0.884 0.894 0.618 1.294 ! 124! Region 3 0.640 0.540 0.757 0.672 0.556 0.812 0.802 0.554 1.162 Region 4 0.680 0.556 0.832 0.557 0.437 0.711 0.859 0.564 1.310 Interview date 10-Jan Reference 10-Feb 1.000 0.761 1.314 1.065 0.765 1.482 1.466 0.760 2.826 10-Mar 0.904 0.689 1.186 1.021 0.737 1.415 1.100 0.583 2.076 10-Apr 0.908 0.693 1.191 1.005 0.727 1.389 0.992 0.528 1.861 10-May 1.055 0.808 1.377 1.004 0.729 1.381 1.467 0.788 2.730 10-Jun 0.966 0.732 1.275 1.010 0.724 1.409 1.345 0.711 2.544 SOCIOECONOMIC Self-report education level <12 years Reference 12 years 1.328 1.002 1.761 1.101 0.832 1.455 0.907 0.560 1.468 Some college 1.243 0.937 1.650 0.914 0.688 1.214 1.157 0.725 1.847 College graduate 1.282 0.966 1.702 0.869 0.657 1.150 0.953 0.569 1.599 Missing 2.005 0.957 4.199 0.255 0.081 0.796 0.723 0.225 2.320 Income poverty status Above poverty threshold, >=$75,000 income Reference Above poverty threshold, <$75,000 income 0.963 0.828 1.120 1.127 0.939 1.353 1.270 0.885 1.824 Below poverty threshold 0.983 0.731 1.322 1.014 0.737 1.396 1.441 0.910 2.283 Poverty status unknown 1.014 0.806 1.276 1.133 0.874 1.467 1.028 0.614 1.721 ! 125! Work status Employed Reference Unemployed 0.945 0.709 1.261 0.919 0.648 1.305 1.446 0.883 2.366 Not in labor force 1.026 0.880 1.196 1.131 0.942 1.357 1.057 0.765 1.460 Don't know/Refused/Missing 0.642 0.318 1.297 1.680 0.768 3.677 0.513 0.175 1.506 Works in health care field No Reference Yes 1.303 1.082 1.571 0.841 0.641 1.103 0.906 0.571 1.437 Missing 1.091 0.432 2.751 1.653 0.781 3.496 1.008 0.265 3.841 Home rented or owned Home is owned Reference Home is rented or other arrangement 0.876 0.730 1.051 0.934 0.769 1.135 1.218 0.875 1.697 Don't know/Refused/Missing 1.002 0.673 1.492 0.807 0.533 1.222 1.409 0.719 2.759 HEALTH Chronic medical condition c No Reference Yes 1.862 1.628 2.129 1.372 1.187 1.586 1.204 0.930 1.558 Missing 1.033 0.601 1.773 1.357 0.802 2.297 0.614 0.202 1.861 Health status Sick with fever and cough or sore throat in past month No Reference ! 126! Yes 1.120 0.843 1.486 1.182 0.870 1.605 1.053 0.644 1.722 Missing 1.147 0.461 2.856 0.654 0.228 1.878 1.018 0.292 3.548 Other people in house with fever and cough or sore throat No Reference Yes 0.865 0.717 1.044 0.994 0.785 1.258 1.279 0.919 1.779 Missing 0.572 0.224 1.458 0.957 0.438 2.091 1.312 0.538 3.194 ACCESS Has health insurance coverage Yes 1.400 1.074 1.825 1.888 1.356 2.628 1.122 0.765 1.647 No Reference Don't know/Refused/Missing 1.027 0.461 2.288 1.870 0.536 6.523 0.855 0.207 3.537 Number of times seen doctor since August 2009 >=4 Reference 3 0.908 0.748 1.101 0.903 0.733 1.111 1.005 0.682 1.482 2 0.843 0.713 0.997 0.811 0.672 0.978 0.832 0.592 1.171 1 0.702 0.593 0.832 0.665 0.549 0.806 0.888 0.626 1.260 Missing 0.688 0.457 1.036 0.838 0.535 1.313 0.660 0.314 1.386 OPINIONS ABOUT FLU VACCINE Opinion: Effectiveness of H1N1 vaccine Very effective Reference Somewhat effective 0.687 0.591 0.799 1.548 1.292 1.854 0.591 0.439 0.795 ! 127! Not very effective 0.575 0.437 0.757 2.202 1.596 3.039 0.365 0.200 0.665 Not at all effective 0.729 0.466 1.140 1.344 0.854 2.114 0.665 0.349 1.267 Don't know/Refused/Missing 0.600 0.481 0.750 1.659 1.326 2.076 0.421 0.235 0.754 Opinion: Risk of getting sick with H1N1 flu without vaccine Very high Reference Somewhat high 0.754 0.574 0.991 1.199 0.738 1.949 0.571 0.370 0.881 Somewhat low 0.494 0.370 0.660 1.894 1.170 3.064 0.269 0.169 0.428 Very low 0.336 0.248 0.456 1.956 1.187 3.224 0.195 0.117 0.324 Don't know/Refused/Missing 0.608 0.394 0.938 1.509 0.845 2.694 0.179 0.090 0.357 Opinion: Worry about getting sick from the H1N1 vaccine Very worried Reference Somewhat worried 1.059 0.803 1.397 1.177 0.810 1.711 1.362 0.828 2.242 Not very worried 0.749 0.564 0.995 1.453 1.000 2.110 0.994 0.598 1.653 Not at all worried 0.932 0.695 1.248 1.212 0.836 1.758 0.978 0.579 1.653 Don't know/Refused/Missing 0.918 0.395 2.133 1.520 0.632 3.658 2.649 0.892 7.869 Opinion: Effectiveness of seasonal vaccine Very effective Reference Somewhat effective 0.795 0.684 0.923 0.634 0.536 0.750 1.233 0.882 1.724 Not very effective 0.848 0.648 1.111 0.382 0.266 0.548 1.494 0.927 2.408 Not at all effective 0.711 0.496 1.019 0.415 0.268 0.643 1.181 0.643 2.169 Don't know/Refused/Missing 0.608 0.407 0.908 0.275 0.183 0.413 1.485 0.745 2.960 ! 128! Opinion: Risk of getting sick with seasonal flu without vaccine Very high Reference Somewhat high 1.050 0.850 1.296 0.862 0.658 1.128 1.622 1.023 2.572 Somewhat low 0.664 0.528 0.835 0.448 0.332 0.605 1.847 1.142 2.988 Very low 0.618 0.472 0.810 0.296 0.213 0.411 2.232 1.267 3.930 Don't know/Refused/Missing 0.912 0.573 1.453 0.653 0.394 1.082 2.850 1.291 6.292 Opinion: Worry about getting sick from the seasonal vaccine Very worried Reference Somewhat worried 1.223 0.889 1.681 1.199 0.816 1.762 0.757 0.413 1.388 Not very worried 1.178 0.849 1.636 1.014 0.691 1.488 1.124 0.639 1.979 Not at all worried 1.151 0.834 1.589 1.092 0.754 1.582 0.898 0.512 1.573 Don't know/Refused/Missing 1.371 0.651 2.889 1.332 0.466 3.809 0.634 0.159 2.524 a Model 1A, 1B, and 1C outcomes were binary variables equal to one when the respondent indicated they received recommendations for both seasonal and H1N1 flu vaccinations, seasonal flu only vaccination, and H1N1 flu only vaccination, respectively. The comparator group for these models was whether the respondent replied with neither, don't know, and refused for whether they receive any seasonal and H1N1 flu vaccination recommendations. b Region 1: CT, ME, MA, NH, VT, RI, NJ, NY, PA; Region 2: IL, IN, MI, OH, WI, IA, KS, MN, MO, NE, ND, SD; Region 3: DE, DC, FL, GA MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK, TX; Region 4: AZ, CO, ID, MT, NV, NM, UT, WY, AK, CA, HI, OR, WA ! 129! c This indicates whether the person has any of the following chronic medical conditions: asthma or an other lung condition, diabetes, a heart condition, a kidney condition, sickle cell anemia or other anemia, a neurological or neuromuscular condition, a liver condition, or a weakened immune system caused by a chronic illness or by medicines taken for a chronic illness. Abbreviations: RRR, relative risk ratio ! 130! Appendix 1C. Predicted probabilities for healthcare provider recommendations and flu vaccines a Outcome Any seasonal and H1N1 flu vaccine recommendation Any seasonal and H1N1 flu vaccine Model 1 Model 2 95% Confidence Interval 95% Confidence Interval Variable dy/dx Lower Limit Upper Limit dy/dx Lower Limit Upper Limit Provider recommendation for seasonal and H1N1 flu vaccines Yes 0.33135 0.29945 0.36324 No Reference DEMOGRAPHIC Age group 18-34 Reference 35-44 0.00034 -0.04534 0.04601 -0.01681 -0.07072 0.03710 45-54 -0.01174 -0.05421 0.03073 0.08610 0.03679 0.13541 55-64 0.09474 0.05133 0.13815 0.18342 0.13019 0.23666 65+ 0.13282 0.08566 0.17999 0.31243 0.25295 0.37192 Race/ethnicity Hispanic 0.03664 -0.01798 0.09127 -0.05534 -0.12232 0.01164 Non-Hispanic, Black Only 0.06010 0.01493 0.10527 -0.04546 -0.10057 0.00965 Non-Hispanic, White Only Reference Non-Hispanic, Other or Multiple Race 0.00957 -0.04471 0.06385 0.03571 -0.03004 0.10147 Gender ! 131! Male Reference Female 0.03156 0.00642 0.05670 -0.00810 -0.03914 0.02294 Married Yes 0.02085 -0.01461 0.05631 0.01462 -0.02671 0.05594 No Reference Missing 0.06338 -0.11412 0.24089 -0.00898 -0.22514 0.20718 Number of children 0 Reference 1 0.03867 -0.01034 0.08769 0.05012 -0.00609 0.10633 2 0.02387 -0.04201 0.08976 0.10915 0.03523 0.18307 3 0.09002 -0.00392 0.18396 0.13397 0.02337 0.24456 Missing 0.16854 -0.01563 0.35270 0.17772 0.01493 0.34050 Number of people in household 1 Reference 2 0.01242 -0.02615 0.05098 0.01481 -0.03402 0.06364 3 0.01567 -0.03487 0.06621 -0.02041 -0.08318 0.04235 4 0.01329 -0.05029 0.07686 -0.07087 -0.14665 0.00490 5 -0.00116 -0.09087 0.08855 -0.09581 -0.20018 0.00857 6 0.01344 -0.11952 0.14639 -0.15431 -0.31045 0.00183 7 -0.04890 -0.24650 0.14869 -0.26502 -0.53923 0.00920 3-category Metropolitan Statistical Area (MSA) status MSA, principle city 0.02221 -0.01330 0.05772 0.00385 -0.04084 0.04855 MSA, not principle city 0.01405 -0.01673 0.04484 0.00365 -0.03557 0.04287 Non-MSA Reference Census region of residence b Region 1 Reference Region 2 -0.06862 -0.10446 -0.03278 0.01968 -0.02249 0.06185 ! 132! Region 3 -0.09775 -0.13181 -0.06369 -0.00187 -0.04311 0.03938 Region 4 -0.09922 -0.13991 -0.05852 0.01730 -0.03212 0.06672 Interview date 10-Jan Reference 10-Feb 0.02049 -0.03700 0.07799 0.04111 -0.03025 0.11246 10-Mar -0.00602 -0.06306 0.05103 0.04239 -0.02836 0.11314 10-Apr -0.01034 -0.06695 0.04628 0.07908 0.00912 0.14905 10-May 0.01946 -0.03644 0.07536 0.06865 0.00045 0.13685 10-Jun 0.00620 -0.05210 0.06451 0.06182 -0.00860 0.13223 SOCIOECONOMIC Self-report education level <12 years Reference 12 years 0.03987 -0.01342 0.09315 0.00759 -0.05984 0.07503 Some college 0.02114 -0.03347 0.07574 0.03750 -0.02862 0.10361 College graduate 0.01418 -0.03954 0.06789 0.07806 0.01141 0.14470 Missing -0.02451 -0.17928 0.13027 -0.11259 -0.28792 0.06274 Income poverty status Above poverty threshold, >=$75,000 income Reference Above poverty threshold, <$75,000 income 0.01144 -0.01970 0.04257 -0.03005 -0.06840 0.00830 Below poverty threshold 0.01511 -0.04150 0.07172 -0.06393 -0.13110 0.00324 Poverty status unknown 0.01155 -0.03413 0.05724 0.00246 -0.05352 0.05844 Work status Employed Reference Unemployed 0.00497 -0.05637 0.06632 -0.02508 -0.10063 0.05047 Not in labor force 0.01539 -0.01616 0.04693 0.05328 0.01389 0.09267 Don't know/Refused/Missing -0.01967 -0.16107 0.12173 0.10507 -0.05223 0.26238 ! 133! Works in health care field No Reference Yes 0.02311 -0.01707 0.06328 0.17220 0.12013 0.22428 Missing 0.04251 -0.11488 0.19990 0.09312 -0.09413 0.28037 Home rented or owned Home is owned Reference Home is rented or other arrangement -0.01477 -0.05001 0.02047 -0.03115 -0.07374 0.01145 Don't know/Refused/Missing -0.00988 -0.08761 0.06786 -0.05016 -0.13514 0.03481 HEALTH Chronic medical condition c No Reference Yes 0.10994 0.08334 0.13655 0.04864 0.01486 0.08242 Missing 0.02572 -0.07522 0.12667 0.04613 -0.07002 0.16228 Health status Sick with fever and cough or sore throat in past month No Reference Yes 0.02608 -0.03003 0.08220 -0.01470 -0.09947 0.07008 Missing -0.03417 -0.22087 0.15253 -0.05916 -0.25769 0.13938 Other people in house with fever and cough or sore throat No Reference Yes -0.00830 -0.04736 0.03076 -0.04568 -0.09132 -0.00004 Missing -0.06024 -0.21226 0.09178 0.02922 -0.12960 0.18803 ! 134! ACCESS Has health insurance coverage Yes 0.08904 0.03681 0.14128 0.18723 0.12396 0.25051 No Reference Don't know/Refused/Missing 0.03434 -0.15614 0.22482 0.20069 0.00251 0.39888 Number of times seen doctor since August 2009 >=4 Reference 3 -0.02145 -0.06046 0.01755 0.02907 -0.01895 0.07708 2 -0.04519 -0.07868 -0.01169 0.02852 -0.01487 0.07190 1 -0.08141 -0.11552 -0.04729 0.04367 0.00197 0.08537 Missing -0.07048 -0.15227 0.01131 0.07388 -0.03279 0.18055 OPINIONS ABOUT FLU VACCINE Opinion: Effectiveness of H1N1 vaccine Very effective Reference Somewhat effective -0.03212 -0.06306 -0.00118 -0.06678 -0.10655 -0.02700 Not very effective -0.03226 -0.08627 0.02175 -0.10731 -0.17428 -0.04035 Not at all effective -0.03225 -0.11214 0.04763 -0.12163 -0.22350 -0.01975 Don't know/Refused/Missing -0.04416 -0.08659 -0.00174 -0.07755 -0.13356 -0.02154 Opinion: Risk of getting sick with H1N1 flu without vaccine Very high Reference Somewhat high -0.06746 -0.13133 -0.00360 0.03292 -0.05443 0.12027 Somewhat low -0.11416 -0.18007 -0.04825 -0.05388 -0.14338 0.03563 Very low -0.16047 -0.22879 -0.09214 -0.10685 -0.19756 -0.01615 Don't know/Refused/Missing -0.11684 -0.20809 -0.02560 -0.00640 -0.12354 0.11074 ! 135! Opinion: Worry about getting sick from the H1N1 vaccine Very worried Reference Somewhat worried 0.03136 -0.02750 0.09021 -0.00013 -0.08014 0.07988 Not very worried -0.00356 -0.06198 0.05487 -0.00674 -0.08659 0.07311 Not at all worried 0.00534 -0.05438 0.06506 -0.04521 -0.12648 0.03607 Don't know/Refused/Missing 0.05160 -0.12170 0.22490 -0.20509 -0.37048 -0.03970 Opinion: Effectiveness of seasonal vaccine Very effective Reference Somewhat effective -0.06609 -0.09584 -0.03635 -0.20392 -0.23941 -0.16844 Not very effective -0.08841 -0.13942 -0.03741 -0.39179 -0.45387 -0.32971 Not at all effective -0.12190 -0.19122 -0.05257 -0.31560 -0.41074 -0.22046 Don't know/Refused/Missing -0.17230 -0.24508 -0.09952 -0.19661 -0.28753 -0.10569 Opinion: Risk of getting sick with seasonal flu without vaccine Very high Reference Somewhat high 0.01425 -0.03117 0.05967 -0.13962 -0.20615 -0.07309 Somewhat low -0.10085 -0.14908 -0.05262 -0.36195 -0.42959 -0.29431 Very low -0.14228 -0.19870 -0.08586 -0.50661 -0.58244 -0.43077 Don't know/Refused/Missing -0.01531 -0.10887 0.07824 -0.23195 -0.36424 -0.09965 Opinion: Worry about getting sick from the seasonal vaccine Very worried Reference Somewhat worried 0.03264 -0.03159 0.09687 0.09909 0.01578 0.18239 Not very worried 0.02599 -0.03804 0.09002 0.14587 0.06351 0.22822 Not at all worried 0.02342 -0.03921 0.08605 0.26570 0.18468 0.34671 Don't know/Refused/Missing 0.04705 -0.12975 0.22385 0.20288 0.01409 0.39167 ! 136! a The outcome for Model 1 was defined as a binary variable equal to one when the respondent indicated they received recommendations for H1N1 flu vaccination only, seasonal flu vaccination only, or both vaccinations. The Model 2 outcome was defined as a binary variable equal to one when the respondent indicated they received the H1N1 flu vaccination only, seasonal flu vaccination only, or both vaccinations. b Region 1: CT, ME, MA, NH, VT, RI, NJ, NY, PA; Region 2: IL, IN, MI, OH, WI, IA, KS, MN, MO, NE, ND, SD; Region 3: DE, DC, FL, GA MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK, TX; Region 4: AZ, CO, ID, MT, NV, NM, UT, WY, AK, CA, HI, OR, WA c This indicates whether the person has any of the following chronic medical conditions: asthma or an other lung condition, diabetes, a heart condition, a kidney condition, sickle cell anemia or other anemia, a neurological or neuromuscular condition, a liver condition, or a weakened immune system caused by a chronic illness or by medicines taken for a chronic illness. ! 137! Appendix Table 2A. Full descriptive statistics Whites Blacks Hispanics N=18713 N=2083 N=1244 Variables Mean SE Mean SE Mean SE DEMOGRAPHIC Age group (years) 18-34 0.1429 0.0026 0.1959 0.0087 0.3312 0.0133 35-44 0.1162 0.0023 0.1507 0.0078 0.2195 0.0117 45-54 0.1867 0.0028 0.2079 0.0089 0.1841 0.0110 55-64 0.2266 0.0031 0.2026 0.0088 0.1310 0.0096 65+ 0.3276 0.0034 0.2429 0.0094 0.1342 0.0097 Gender Male 0.3734 0.0035 0.3265 0.0103 0.3826 0.0138 Female 0.6266 0.0035 0.6735 0.0103 0.6174 0.0138 Married Yes 0.5321 0.0036 0.2784 0.0098 0.4437 0.0141 No 0.4246 0.0036 0.6697 0.0103 0.4871 0.0142 Missing 0.0433 0.0015 0.0518 0.0049 0.0691 0.0072 Number of people in household 1 0.2925 0.0033 0.3413 0.0104 0.1809 0.0109 2 0.4032 0.0036 0.2804 0.0098 0.2468 0.0122 3 0.1287 0.0024 0.1671 0.0082 0.1543 0.0102 4 0.1091 0.0023 0.1330 0.0074 0.1961 0.0113 5 0.0553 0.0017 0.0514 0.0048 0.1608 0.0104 6 0.0087 0.0007 0.0197 0.0030 0.0402 0.0056 7 0.0025 0.0004 0.0072 0.0019 0.0209 0.0041 ! 138! 3-category MSA status MSA, principle city 0.2604 0.0032 0.5098 0.0110 0.3883 0.0138 MSA, not principle city 0.4579 0.0036 0.3447 0.0104 0.4622 0.0141 Non-MSA 0.2817 0.0033 0.1455 0.0077 0.1495 0.0101 Census region of residence a Region 1 0.1841 0.0028 0.1018 0.0066 0.1600 0.0104 Region 2 0.2405 0.0031 0.1277 0.0073 0.1021 0.0086 Region 3 0.3399 0.0035 0.6923 0.0101 0.2990 0.0130 Region 4 0.2355 0.0031 0.0783 0.0059 0.4389 0.0141 Interview month 10-Jan 0.0550 0.0017 0.0499 0.0048 0.0571 0.0066 10-Feb 0.1562 0.0027 0.1531 0.0079 0.1608 0.0104 10-Mar 0.1751 0.0028 0.1949 0.0087 0.1849 0.0110 10-Apr 0.1889 0.0029 0.1767 0.0084 0.1897 0.0111 10-May 0.2318 0.0031 0.2352 0.0093 0.2243 0.0118 10-Jun 0.1930 0.0029 0.1901 0.0086 0.1833 0.0110 SOCIOECONOMIC Self-report education level <12 years 0.0639 0.0018 0.1603 0.0080 0.2484 0.0123 12 years 0.2211 0.0030 0.2391 0.0093 0.2186 0.0117 Some college 0.2682 0.0032 0.2885 0.0099 0.2339 0.0120 College graduate 0.4048 0.0036 0.2650 0.0097 0.2299 0.0119 Missing 0.0421 0.0015 0.0470 0.0046 0.0691 0.0072 Income poverty status Above poverty threshold, >=$75,000 0.2699 0.0032 0.1421 0.0077 0.1672 0.0106 ! 139! income Above poverty threshold, <$75,000 income 0.5010 0.0037 0.4633 0.0109 0.4043 0.0139 Below poverty threshold 0.0722 0.0019 0.2108 0.0089 0.2484 0.0123 Poverty status unknown 0.1569 0.0027 0.1839 0.0085 0.1801 0.0109 Work status Employed 0.4696 0.0036 0.4489 0.0109 0.5185 0.0142 Unemployed 0.0430 0.0015 0.0888 0.0062 0.0707 0.0073 Not in labor force 0.4427 0.0036 0.4100 0.0108 0.3352 0.0134 Don't know/Refused/Missing 0.0448 0.0015 0.0523 0.0049 0.0756 0.0075 Works in health care field No 0.8716 0.0024 0.8589 0.0076 0.8971 0.0086 Yes 0.1039 0.0022 0.1138 0.0070 0.0860 0.0080 Missing 0.0245 0.0011 0.0274 0.0036 0.0169 0.0037 Home rented or owned Home is owned 0.7587 0.0031 0.5295 0.0109 0.4815 0.0142 Home is rented or other arrangement 0.1762 0.0028 0.3889 0.0107 0.4196 0.0140 Don't know/Refused/Missing 0.0651 0.0018 0.0816 0.0060 0.0989 0.0085 HEALTH AND ACCESS Chronic medical condition b No 0.6277 0.0035 0.5982 0.0107 0.7058 0.0129 ! 140! Yes 0.3413 0.0035 0.3658 0.0106 0.2781 0.0127 Missing 0.0309 0.0013 0.0360 0.0041 0.0161 0.0036 Sick with fever and cough or sore throat in past month No 0.9340 0.0018 0.9323 0.0055 0.9172 0.0078 Yes 0.0475 0.0016 0.0466 0.0046 0.0732 0.0074 Missing 0.0185 0.0010 0.0211 0.0032 0.0096 0.0028 Other people in house with fever and cough or sore throat No 0.8692 0.0025 0.8699 0.0074 0.7878 0.0116 Yes 0.1172 0.0024 0.1119 0.0069 0.2034 0.0114 Missing 0.0136 0.0008 0.0182 0.0029 0.0088 0.0027 Has health insurance coverage Yes 0.9001 0.0022 0.8262 0.0083 0.7098 0.0129 No 0.0593 0.0017 0.1296 0.0074 0.2203 0.0118 Don't know/Refused/Missing 0.0407 0.0014 0.0442 0.0045 0.0699 0.0072 Number of times seen doctor since August 2009 >=4 0.2842 0.0033 0.3447 0.0104 0.2653 0.0125 3 0.1465 0.0026 0.1699 0.0082 0.1680 0.0106 2 0.2561 0.0032 0.2376 0.0093 0.2781 0.0127 1 0.2938 0.0033 0.2132 0.0090 0.2733 0.0126 Missing 0.0193 0.0010 0.0346 0.0040 0.0153 0.0035 Healthcare provider recommendation c ! 141! Seasonal and H1N1 flu vaccine 0.4340 0.0036 0.4455 0.0109 0.4236 0.0140 BELIEFS ABOUT FLU VACCINES Opinion: Effectiveness of H1N1 vaccine Very/somewhat effective 0.7515 0.0032 0.6942 0.0101 0.7195 0.0127 Not very/not at all effect/Don't know/Refused/Missing 0.2485 0.0032 0.3058 0.0101 0.2805 0.0127 Opinion: Risk of getting sick with H1N1 flu without vaccine Very/somewhat high 0.2518 0.0032 0.2597 0.0096 0.3585 0.0136 Somewhat/very low/Don't know/Refused/Missing 0.7482 0.0032 0.7403 0.0096 0.6415 0.0136 Opinion: Worry about getting sick from the H1N1 vaccine Very/somewhat worried 0.2778 0.0033 0.3615 0.0105 0.4325 0.0141 Not very/not at all worried/Don't know/Refused/Missing 0.7222 0.0033 0.6385 0.0105 0.5675 0.0141 Opinion: Effectiveness of seasonal vaccine Very/somewhat effective 0.8331 0.0028 0.7657 0.0093 0.7838 0.0117 ! 142! Not very/not at all effect/Don't know/Refused/Missing 0.1669 0.0028 0.2343 0.0093 0.2162 0.0117 Opinion: Risk of getting sick with seasonal flu without vaccine Very/somewhat high 0.4147 0.0036 0.3533 0.0105 0.4510 0.0141 Somewhat/very low/Don't know/Refused/Missing 0.5853 0.0036 0.6467 0.0105 0.5490 0.0141 Opinion: Worry about getting sick from the seasonal vaccine Very/somewhat worried 0.2325 0.0031 0.3341 0.0103 0.3883 0.0138 Not very/not at all worried/Don't know/Refused/Missing 0.7675 0.0031 0.6659 0.0103 0.6117 0.0138 Abbreviations: Whites, Non-Hispanic Whites; Blacks, Non-Hispanic Blacks; SE, standard error; MSA, metropolitan statistical area a Region 1: CT, ME, MA, NH, VT, RI, NJ, NY, PA; Region 2: IL, IN, MI, OH, WI, IA, KS, MN, MO, NE, ND, SD; Region 3: DE, DC, FL, GA MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK, TX; Region 4: AZ, CO, ID, MT, NV, NM, UT, WY, AK, CA, HI, OR, WA b This indicates whether the person has any of the following chronic medical conditions: asthma or an other lung condition, diabetes, a heart condition, a kidney condition, sickle cell anemia or other anemia, a neurological or neuromuscular condition, a liver condition, or a weakened immune system caused by a chronic illness or by medicines taken for a chronic illness. c Reporting a HCP recommendation was defined as a binary variable equal to one when the respondent indicated they received recommendations for H1N1 flu vaccination only, seasonal flu vaccination only, or both vaccinations. ! 143! ! Appendix 2B. Full decomposition of differences in flu vaccination rates between Whites and Blacks Difference between Whites and Blacks Characteristics effect Coefficients effect Estimate SE Share P- value Estimate SE Share P- value Aggregate effect 0.0849 0.0040 61.91 0.0000 0.0522 0.0106 38.09 0.0000 DEMOGRAPHIC Sub-aggregate effect 0.0232 16.95 0.0360 26.24 Age group 18-34 0.0046 0.0004 3.35 0.000 -0.0098 0.0041 -7.16 0.016 35-44 0.0024 0.0003 1.74 0.000 -0.0046 0.0033 -3.35 0.161 45-54 0.0006 0.0001 0.46 0.000 0.0002 0.0038 0.14 0.959 55-64 0.0008 0.0002 0.60 0.000 0.0104 0.0038 7.59 0.006 65+ 0.0128 0.0006 9.34 0.000 0.0069 0.0052 5.01 0.184 Gender Male 0.0004 0.0002 0.31 0.004 0.0027 0.0033 1.96 0.412 Female 0.0004 0.0002 0.31 0.004 -0.0055 0.0068 -4.04 0.412 Married Yes 0.0042 0.0032 3.07 0.195 -0.0051 0.0095 -3.72 0.592 No 0.0007 0.0031 0.53 0.813 0.0115 0.0227 8.40 0.612 Missing 0.0001 0.0002 0.08 0.568 0.0001 0.0033 0.04 0.986 Number of people in household 1 -0.0016 0.0007 -1.18 0.013 0.0096 0.0097 7.01 0.324 2 0.0010 0.0014 0.71 0.484 0.0062 0.0073 4.52 0.396 3 -0.0002 0.0005 -0.12 0.720 0.0051 0.0047 3.74 0.280 4 0.0001 0.0003 0.09 0.690 -0.0009 0.0040 -0.63 0.829 ! 144! 5 -0.0001 0.0001 -0.05 0.237 -0.0002 0.0021 -0.14 0.928 6 0.0007 0.0003 0.49 0.036 -0.0008 0.0012 -0.61 0.500 7 -0.0002 0.0002 -0.13 0.453 -0.0002 0.0007 -0.15 0.777 3-category MSA status MSA, principle city -0.0033 0.0012 -2.37 0.006 0.0081 0.0067 5.93 0.227 MSA, not principle city 0.0000 0.0005 -0.02 0.950 -0.0002 0.0048 -0.16 0.964 Non-MSA -0.0017 0.0006 -1.27 0.006 -0.0022 0.0025 -1.63 0.377 Census region of residence a Region 1 0.0000 0.0005 0.01 0.983 -0.0020 0.0024 -1.43 0.403 Region 2 0.0007 0.0006 0.54 0.222 0.0070 0.0028 5.12 0.012 Region 3 0.0011 0.0017 0.80 0.522 0.0078 0.0107 5.70 0.463 Region 4 -0.0006 0.0009 -0.42 0.503 -0.0037 0.0020 -2.68 0.064 Interview month 10-Jan 0.0000 0.0001 -0.02 0.669 0.0013 0.0017 0.94 0.462 10-Feb 0.0000 0.0000 0.01 0.524 0.0074 0.0033 5.36 0.027 10-Mar 0.0002 0.0001 0.12 0.235 -0.0044 0.0039 -3.23 0.256 10-Apr -0.0001 0.0001 -0.04 0.478 -0.0075 0.0036 -5.46 0.040 10-May 0.0000 0.0000 -0.01 0.423 0.0026 0.0044 1.89 0.553 10-Jun 0.0000 0.0000 0.02 0.225 -0.0037 0.0038 -2.73 0.328 SOCIOECONOMIC Sub-aggregate effect 0.0192 14.01 0.0639 46.59 Self-report education level <12 years 0.0040 0.0012 2.88 0.001 -0.0049 0.0047 -3.59 0.295 12 years 0.0003 0.0002 0.24 0.064 -0.0042 0.0062 -3.04 0.498 Some college 0.0002 0.0002 0.17 0.227 -0.0072 0.0073 -5.22 0.325 College graduate 0.0046 0.0013 3.38 0.000 0.0066 0.0069 4.85 0.335 Missing -0.0002 0.0002 -0.14 0.224 0.0023 0.0036 1.64 0.536 ! 145! Income poverty status Above poverty threshold, >=$75,000 income 0.0032 0.0009 2.37 0.000 0.0015 0.0034 1.12 0.651 Above poverty threshold, <$75,000 income -0.0003 0.0002 -0.19 0.182 -0.0058 0.0069 -4.23 0.399 Below poverty threshold 0.0047 0.0014 3.44 0.000 -0.0095 0.0046 -6.95 0.038 Poverty status unknown -0.0004 0.0002 -0.31 0.051 0.0086 0.0042 6.29 0.041 Work status** Employed -0.0003 0.0002 -0.22 0.155 0.0331 0.0124 24.17 0.007 Unemployed 0.0008 0.0007 0.61 0.203 0.0045 0.0030 3.29 0.138 Not in labor force 0.0004 0.0003 0.26 0.304 0.0144 0.0114 10.52 0.204 Don't know/Refused/Missing -0.0002 0.0002 -0.12 0.410 -0.0084 0.0036 -6.10 0.020 Works in health care field No -0.0004 0.0002 -0.27 0.048 0.0061 0.0395 4.46 0.877 Yes -0.0009 0.0002 -0.68 0.000 0.0035 0.0055 2.56 0.527 Missing 0.0002 0.0001 0.14 0.021 -0.0010 0.0025 -0.76 0.672 Home rented or owned Home is owned 0.0032 0.0018 2.31 0.071 0.0181 0.0108 13.17 0.096 Home is rented or other arrangement 0.0000 0.0018 -0.01 0.993 0.0112 0.0083 8.15 0.176 Don't know/Refused/Missing 0.0002 0.0002 0.17 0.309 -0.0051 0.0029 -3.74 0.073 HEALTH AND ACCESS Sub-aggregate effect 0.0024 1.74 -0.0509 -37.11 Chronic medical condition b No -0.0012 0.0003 -0.90 0.000 -0.0082 0.0172 -5.98 0.633 Yes 0.0001 0.0002 0.06 0.746 -0.0072 0.0108 -5.28 0.502 Missing -0.0002 0.0001 -0.17 0.016 0.0012 0.0020 0.88 0.539 Sick with fever and cough or sore throat in past month ! 146! No 0.0000 0.0000 -0.01 0.582 -0.0085 0.0443 -6.19 0.848 Yes 0.0000 0.0000 -0.02 0.054 -0.0030 0.0025 -2.19 0.224 Missing -0.0001 0.0001 -0.08 0.149 0.0016 0.0019 1.13 0.420 Other people in house with fever and cough or sore throat No 0.0000 0.0000 0.00 0.956 -0.0252 0.0297 - 18.34 0.397 Yes 0.0000 0.0001 0.03 0.539 0.0028 0.0041 2.07 0.492 Missing 0.0000 0.0001 0.03 0.711 0.0001 0.0012 0.05 0.955 Has health insurance coverage Yes 0.0025 0.0011 1.79 0.027 -0.0129 0.0337 -9.39 0.702 No 0.0051 0.0012 3.68 0.000 -0.0101 0.0056 -7.40 0.072 Don't know/Refused/Missing -0.0001 0.0001 -0.10 0.178 0.0041 0.0035 3.02 0.232 Number of times seen doctor since August 2009 >=4 0.0010 0.0004 0.73 0.017 -0.0037 0.0062 -2.69 0.553 3 0.0000 0.0002 0.02 0.885 0.0030 0.0037 2.19 0.418 2 -0.0002 0.0001 -0.13 0.153 -0.0028 0.0047 -2.07 0.544 1 -0.0011 0.0006 -0.83 0.041 -0.0085 0.0044 -6.21 0.054 Missing -0.0006 0.0003 -0.46 0.022 0.0016 0.0015 1.13 0.301 Healthcare provider recommendation c Seasonal and H1N1 flu vaccine -0.0026 0.0001 -1.91 0.000 0.0249 0.0084 18.15 0.003 BELIEFS ABOUT FLU VACCINES Sub-aggregate effect 0.04008 29.22 0.0033 2.37 Opinion: Effectiveness of H1N1 vaccine Very/somewhat effective 0.0007 0.0002 0.48 0.003 0.0165 0.0077 12.00 0.034 Not very/not at all effect/Don't know/Refused/Missing 0.0007 0.0002 0.48 0.003 -0.0072 0.0034 -5.29 0.034 ! 147! Opinion: Risk of getting sick with H1N1 flu without vaccine Very/somewhat high -0.0004 0.0000 -0.31 0.000 0.0131 0.0033 9.57 0.000 Somewhat/very low/Don't know/Refused/Missing -0.0004 0.0000 -0.31 0.000 -0.0374 0.0093 - 27.29 0.000 Opinion: Worry about getting sick from the H1N1 vaccine Very/somewhat worried -0.0002 0.0003 -0.14 0.553 0.0061 0.0041 4.45 0.138 Not very/not at all worried/Don't know/Refused/Missing -0.0002 0.0003 -0.14 0.553 -0.0108 0.0073 -7.86 0.138 Opinion: Effectiveness of seasonal vaccine Very/somewhat effective 0.0061 0.0003 4.45 0.000 0.0090 0.0100 6.58 0.365 Not very/not at all effect/Don't know/Refused/Missing 0.0061 0.0003 4.45 0.000 -0.0028 0.0030 -2.01 0.365 Opinion: Risk of getting sick with seasonal flu without vaccine Very/somewhat high 0.0077 0.0002 5.61 0.000 0.0064 0.0040 4.66 0.114 Somewhat/very low/Don't know/Refused/Missing 0.0077 0.0002 5.61 0.000 -0.0117 0.0074 -8.53 0.114 Opinion: Worry about getting sick from the seasonal vaccine Very/somewhat worried 0.0062 0.0004 4.52 0.000 -0.0131 0.0039 -9.52 0.001 Not very/not at all worried/Don't know/Refused/Missing 0.0062 0.0004 4.52 0.000 0.0260 0.0077 18.96 0.001 Constant 0.0091 0.0439 6.64 0.836 Abbreviations: Whites, Non-Hispanic Whites; Blacks, Non-Hispanic Blacks; SE, standard error; MSA, metropolitan statistical area ! 148! a Region 1: CT, ME, MA, NH, VT, RI, NJ, NY, PA; Region 2: IL, IN, MI, OH, WI, IA, KS, MN, MO, NE, ND, SD; Region 3: DE, DC, FL, GA MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK, TX; Region 4: AZ, CO, ID, MT, NV, NM, UT, WY, AK, CA, HI, OR, WA b This indicates whether the person has any of the following chronic medical conditions: asthma or an other lung condition, diabetes, a heart condition, a kidney condition, sickle cell anemia or other anemia, a neurological or neuromuscular condition, a liver condition, or a weakened immune system caused by a chronic illness or by medicines taken for a chronic illness. c Reporting a HCP recommendation was defined as a binary variable equal to one when the respondent indicated they received recommendations for H1N1 flu vaccination only, seasonal flu vaccination only, or both vaccinations. ! ! 149! ! Appendix 2C. Full decomposition of differences in flu vaccination rates between Whites and Hispanics Difference between Whites and Hispanics Characteristics effect Coefficients effect Estimate SE Share P-value Estimate SE Share P-value Aggregate effect 0.1207 0.0052 100.78 0.0000 -0.0009 0.0134 -0.78 0.9440 DEMOGRAPHIC Sub-aggregate effect 0.0636 53.14 0.0029 2.39 Age group 18-34 0.0161 0.0014 13.48 0.000 -0.0003 0.0104 -0.25 0.977 35-44 0.0070 0.0008 5.87 0.000 0.0002 0.0061 0.15 0.977 45-54 -0.0001 0.0000 -0.06 0.000 0.0007 0.0250 0.60 0.977 55-64 0.0032 0.0006 2.70 0.000 -0.0002 0.0057 -0.14 0.977 65+ 0.0289 0.0013 24.11 0.000 -0.0003 0.0119 -0.29 0.977 Gender Male -0.0001 0.0000 -0.07 0.004 0.0001 0.0046 0.11 0.977 Female -0.0001 0.0000 -0.07 0.004 -0.0002 0.0074 -0.18 0.977 Married Yes 0.0014 0.0011 1.21 0.195 0.0015 0.0505 1.21 0.977 No 0.0002 0.0008 0.15 0.813 0.0027 0.0950 2.28 0.977 Missing 0.0003 0.0006 0.29 0.568 -0.0006 0.0213 -0.51 0.977 Number of people in household 1 0.0036 0.0015 3.05 0.014 -0.0009 0.0319 -0.77 0.977 2 0.0012 0.0018 1.03 0.484 -0.0002 0.0077 -0.18 0.977 3 -0.0001 0.0003 -0.09 0.720 0.0006 0.0209 0.50 0.977 4 0.0004 0.0011 0.36 0.690 -0.0004 0.0131 -0.31 0.977 5 0.0018 0.0015 1.52 0.237 0.0007 0.0239 0.57 0.977 ! 150! 6 0.0019 0.0009 1.59 0.036 0.0002 0.0067 0.16 0.977 7 -0.0007 0.0009 -0.58 0.453 -0.0001 0.0037 -0.09 0.977 3-category MSA status MSA, principle city -0.0016 0.0006 -1.38 0.006 -0.0010 0.0363 -0.87 0.977 MSA, not principle city 0.0000 0.0000 0.00 0.950 -0.0011 0.0370 -0.89 0.977 Non-MSA -0.0017 0.0006 -1.39 0.006 0.0007 0.0260 0.62 0.977 Census region of residence a Region 1 0.0000 0.0001 0.00 0.983 -0.0003 0.0091 -0.22 0.977 Region 2 0.0009 0.0007 0.75 0.222 0.0001 0.0025 0.06 0.977 Region 3 -0.0001 0.0002 -0.10 0.522 -0.0002 0.0066 -0.16 0.977 Region 4 0.0007 0.0011 0.61 0.503 0.0007 0.0238 0.57 0.977 Interview month 10-Jan 0.0000 0.0000 0.01 0.669 -0.0002 0.0061 -0.15 0.977 10-Feb 0.0000 0.0000 -0.02 0.524 0.0002 0.0058 0.14 0.977 10-Mar 0.0001 0.0001 0.07 0.235 0.0000 0.0003 0.00 1.000 10-Apr 0.0000 0.0000 0.00 0.478 -0.0001 0.0049 -0.12 0.977 10-May 0.0000 0.0000 0.03 0.424 0.0007 0.0235 0.56 0.977 10-Jun 0.0001 0.0001 0.06 0.224 0.0000 0.0015 -0.03 0.978 SOCIOECONOMIC Sub-aggregate effect 0.0271 22.62 -0.0009 -0.71 Self-report education level <12 years 0.0075 0.0023 6.23 0.001 0.0000 0.0010 0.02 0.981 12 years 0.0000 0.0000 -0.04 0.063 -0.0005 0.0176 -0.42 0.977 Some college -0.0004 0.0003 -0.32 0.227 -0.0001 0.0018 -0.04 0.978 College graduate 0.0057 0.0016 4.78 0.000 0.0003 0.0108 0.26 0.977 Missing -0.0010 0.0008 -0.83 0.224 0.0001 0.0026 0.06 0.977 Income poverty status ! 151! Above poverty threshold, >=$75,000 income 0.0026 0.0007 2.15 0.000 -0.0002 0.0074 -0.18 0.977 Above poverty threshold, <$75,000 income -0.0007 0.0005 -0.55 0.182 0.0004 0.0144 0.34 0.977 Below poverty threshold 0.0059 0.0017 4.95 0.000 0.0003 0.0090 0.22 0.977 Poverty status unknown -0.0004 0.0002 -0.30 0.051 -0.0001 0.0049 -0.12 0.977 Work status Employed 0.0007 0.0005 0.60 0.155 -0.0017 0.0592 -1.42 0.977 Unemployed 0.0005 0.0004 0.41 0.203 0.0001 0.0018 0.04 0.977 Not in labor force 0.0012 0.0011 0.96 0.304 -0.0006 0.0192 -0.46 0.977 Don't know/Refused/Missing -0.0007 0.0008 -0.56 0.410 0.0003 0.0110 0.26 0.977 Works in health care field No 0.0007 0.0004 0.61 0.048 0.0012 0.0430 1.03 0.977 Yes 0.0017 0.0003 1.40 0.000 0.0000 0.0007 0.01 0.980 Missing -0.0005 0.0002 -0.41 0.022 0.0000 0.0009 -0.02 0.977 Home rented or owned Home is owned 0.0038 0.0021 3.15 0.071 -0.0001 0.0051 -0.12 0.977 Home is rented or other arrangement 0.0000 0.0021 -0.02 0.993 -0.0003 0.0111 -0.27 0.977 Don't know/Refused/Missing 0.0005 0.0005 0.39 0.309 0.0001 0.0036 0.09 0.977 HEALTH AND ACCESS Sub-aggregate effect 0.0227 18.98 -0.0038 -3.19 Chronic medical condition b No 0.0032 0.0008 2.68 0.000 0.0357 0.5973 29.82 0.952 Yes -0.0002 0.0006 -0.17 0.746 0.0139 0.2305 11.58 0.952 Missing 0.0007 0.0003 0.55 0.017 -0.0016 0.0269 -1.35 0.952 Sick with fever and cough or sore throat in past month No -0.0001 0.0003 -0.12 0.582 0.0442 2.3421 36.95 0.985 Yes 0.0008 0.0004 0.70 0.054 0.0036 0.1873 2.97 0.985 Missing 0.0004 0.0003 0.30 0.150 -0.0009 0.0493 -0.78 0.985 ! 152! Other people in house with fever and cough or sore throat No 0.0001 0.0010 0.05 0.956 -0.0780 2.2163 -65.12 0.972 Yes -0.0007 0.0012 -0.60 0.539 -0.0205 0.5806 -17.11 0.972 Missing 0.0000 0.0001 -0.04 0.711 0.0018 0.0501 1.47 0.972 Has health insurance coverage Yes 0.0062 0.0028 5.22 0.027 0.0003 0.0108 0.25 0.977 No 0.0114 0.0026 9.53 0.000 -0.0002 0.0076 -0.18 0.977 Don't know/Refused/Missing -0.0011 0.0008 -0.93 0.177 0.0000 0.0014 0.03 0.978 Number of times seen doctor since August 2009 >=4 -0.0003 0.0001 -0.26 0.017 -0.0005 0.0170 -0.41 0.977 3 0.0000 0.0002 0.02 0.885 0.0002 0.0074 0.18 0.977 2 0.0002 0.0002 0.18 0.153 0.0000 0.0005 0.00 0.991 1 -0.0003 0.0001 -0.24 0.041 -0.0006 0.0222 -0.53 0.977 Missing 0.0002 0.0001 0.14 0.021 0.0000 0.0016 0.04 0.977 Healthcare provider recommendation c Seasonal and H1N1 flu vaccine 0.0023 0.0001 1.94 0.000 -0.0012 0.0411 -0.99 0.977 BELIEFS ABOUT FLU VACCINES Sub-aggregate effect 0.0072 6.05 0.0009 0.72 Opinion: Effectiveness of H1N1 vaccine Very/somewhat effective 0.0004 0.0001 0.30 0.003 0.0003 0.0105 0.25 0.977 Not very/not at all effect/Don't know/Refused/Missing 0.0004 0.0001 0.30 0.003 -0.0001 0.0041 -0.10 0.977 Opinion: Risk of getting sick with H1N1 flu without vaccine Very/somewhat high -0.0057 0.0005 -4.77 0.000 -0.0006 0.0198 -0.48 0.977 Somewhat/very low/Don't know/Refused/Missing -0.0057 0.0005 -4.77 0.000 0.0010 0.0354 0.85 0.977 Opinion: Worry about getting sick from the H1N1 vaccine Very/somewhat worried -0.0004 0.0006 -0.30 0.553 0.0004 0.0129 0.31 0.977 ! 153! Not very/not at all worried/Don't know/Refused/Missing -0.0004 0.0006 -0.30 0.553 -0.0005 0.0170 -0.41 0.977 Opinion: Effectiveness of seasonal vaccine Very/somewhat effective 0.0044 0.0002 3.69 0.000 -0.0013 0.0447 -1.07 0.977 Not very/not at all effect/Don't know/Refused/Missing 0.0044 0.0002 3.69 0.000 0.0004 0.0123 0.30 0.977 Opinion: Risk of getting sick with seasonal flu without vaccine Very/somewhat high -0.0045 0.0001 -3.74 0.000 -0.0011 0.0389 -0.94 0.977 Somewhat/very low/Don't know/Refused/Missing -0.0045 0.0001 -3.74 0.000 0.0014 0.0474 1.14 0.977 Opinion: Worry about getting sick from the seasonal vaccine Very/somewhat worried 0.0094 0.0006 7.84 0.000 0.0012 0.0404 0.97 0.977 Not very/not at all worried/Don't know/Refused/Missing 0.0094 0.0006 7.84 0.000 -0.0018 0.0637 -1.53 0.977 Constant 0.0017 0.0598 1.43 0.977 Abbreviations: Whites, Non-Hispanic Whites; Blacks, Non-Hispanic Blacks; SE, standard error; MSA, metropolitan statistical area a Region 1: CT, ME, MA, NH, VT, RI, NJ, NY, PA; Region 2: IL, IN, MI, OH, WI, IA, KS, MN, MO, NE, ND, SD; Region 3: DE, DC, FL, GA MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK, TX; Region 4: AZ, CO, ID, MT, NV, NM, UT, WY, AK, CA, HI, OR, WA b This indicates whether the person has any of the following chronic medical conditions: asthma or an other lung condition, diabetes, a heart condition, a kidney condition, sickle cell anemia or other anemia, a neurological or neuromuscular condition, a liver condition, or a weakened immune system caused by a chronic illness or by medicines taken for a chronic illness. c Reporting a HCP recommendation was defined as a binary variable equal to one when the respondent indicated they received recommendations for H1N1 flu vaccination only, seasonal flu vaccination only, or both vaccinations. ! ! ! ! 154! ! Appendix 2D. Select decomposition of differences in flu vaccination rates between Whites and Hispanics using the 2013 Behavioral Risk Factor Surveillance System (BRFSS) dataset Difference between Whites and Hispanics Characteristics effect Coefficients effect Estimate SE Share Estimate SE Share Aggregate effect 0.1359 *** 0.0053 92.94 0.0103 * 0.0058 7.06 DEMOGRAPHIC Sub-aggregate effect a 0.0495 NA 33.84 0.0656 NA 44.84 b SOCIOECONOMIC Sub-aggregate effect 0.0220 NA 15.08 -0.3562 NA -243.56 b HEALTH AND ACCESS Sub-aggregate effect 0.0644 NA 44.01 0.3009 NA 205.78 b Significance level: *** 1% , ** 5%, * 10% Abbreviations: Whites, Non-Hispanic Whites; Blacks, Non-Hispanic Blacks; SE, standard error; NA, not applicable. a The sub-aggregate effect estimate and share is calculated from summing all sub-aggregate effects in the following: demographic (age, gender, marital status, number of adults in household, metropolitan statistical area, Census region of residence, interview quarter); socioeconomic (education, income, employment status, home rented or owned); health and access (chronic medical condition, health insurance, health status, has one person as personal healthcare provider, did not take medications due to cost, did not see doctor because of cost, length of time since last routine checkup, currently pregnant, participated in physical exercise, smoking status, body mass index, daily alcoholic beverages, number of times seen doctor, nurse, or other health professional). Statistics are not shown for each variable constituting each sub-aggregate effect. b! All!specific!sub6aggregate!effects!constituting!these!sum!values!are!not!significant!(not!shown).!!This!explains,!in!large!part,! the!relatively!high!shares!of!the!socioeconomic!and!health!and!access!variables.!!Further!support!of!this!observation!is!due!to! the!low!portion!of!the!aggregate!coefficient!effect!that!constitutes!the!overall!difference!in!vaccination!rates.!! ! ! 155! ! Appendix 2E. Select decomposition of differences in flu vaccination rates between Whites and Blacks using the 2013 Behavioral Risk Factor Surveillance System (BRFSS) dataset Difference between Whites and Blacks Characteristics effect Coefficients effect Estimate SE Share Estimate SE Share Aggregate effect 0.0453 *** 0.0010 41.13 0.0648 *** 0.0031 58.87 DEMOGRAPHIC Sub-aggregate effect a 0.0102 NA 9.29 0.0763 NA 69.34 SOCIOECONOMIC Sub-aggregate effect 0.0092 NA 8.37 -0.0066 NA -6.03 HEALTH AND ACCESS Sub-aggregate effect 0.0258 NA 23.47 0.0368 NA -4.45 Significance level: *** 1% , ** 5%, * 10% Abbreviations: Whites, Non-Hispanic Whites; SE, standard error; NA, not applicable a The sub-aggregate effect estimate and share is calculated from summing all sub-aggregate effects in the following: demographic (age, gender, marital status, number of adults in household, metropolitan statistical area, Census region of residence, interview quarter); socioeconomic (education, income, employment status, home rented or owned); health and access (chronic medical condition, health insurance, health status, has one person as personal healthcare provider, did not take medications due to cost, did not see doctor because of cost, length of time since last routine checkup, currently pregnant, participated in physical exercise, smoking status, body mass index, daily alcoholic beverages, number of times seen doctor, nurse, or other health professional). Statistics are not shown for each variable constituting each sub-aggregate effect. ! ! ! ! 156! Appendix 2F. Sub-aggregate decomposition of differences in flu vaccination rates between Whites and Hispanics (omitting flu vaccine beliefs and provide recommendations) Difference between Whites and Hispanics Characteristics effect Coefficients effect Estimate SE Share Estimate SE Share Aggregate effect 0.1452 *** 0.0052 121.31 -0.0255 * 0.015 -21.32 DEMOGRAPHIC Sub-aggregate effect a 0.0800 NA 66.82 -0.3188 NA -266.25 c SOCIOECONOMIC Sub-aggregate effect 0.0379 NA 31.69 0.2306 NA -192.63 c HEALTH AND ACCESS b Sub-aggregate effect 0.0273 NA 22.81 0.4087 NA 341.66 c BELIEFS ABOUT FLU VACCINES b Sub-aggregate effect NA NA NA NA NA NA Significance level: *** 1% , ** 5%, * 10% Abbreviations: Whites, Non-Hispanic Whites; Blacks, Non-Hispanic Blacks; SE, standard error; NA, not applicable a The sub-aggregate effect estimate and share is calculated from the summing all sub-aggregate similar to Table 2-5. Sub-aggregate effects for each demographic, socioeconomic, health and access, and beliefs about flu vaccines were estimated similar to Table 2-5. b Unlike the 2009 NHFS dataset, the 2013 BRFSS data does not contain provider recommendations and beliefs about flu vaccines. To allow for a comparison to the 2013 BRFSS findings, these variables were removed and the decomposition analysis was repeated. c! All!specific!sub6aggregate!effects!constituting!these!sum!values!are!not!significant!(not!shown).!!This!explains,! in!large!part,!the!relatively!high!shares!of!the!socioeconomic!and!health!and!access!variables.!!Further!support!of! this!observation!is!due!to!the!low!portion!of!the!aggregate!coefficient!effect!that!constitutes!the!overall! difference!in!vaccination!rates. ! 157! Appendix 2G. Sub-aggregate decomposition of differences in flu vaccination rates between Whites and Blacks (omitting flu vaccine beliefs and provide recommendations) Difference between Whites and Blacks Characteristics effect Coefficients effect Estimate SE Share Estimate SE Share Aggregate effect 0.0510 *** 0.0046 37.20 0.0861 *** 0.0043 62.80 DEMOGRAPHIC Sub-aggregate effect a 0.0221 NA 16.09 0.0133 NA 9.70 SOCIOECONOMIC Sub-aggregate effect 0.0253 NA 18.41 0.0603 NA 43.98 HEALTH AND ACCESS b Sub-aggregate effect 0.0037 NA 2.69 -0.0566 NA -41.27 BELIEFS ABOUT FLU VACCINES b Sub-aggregate effect NA NA NA NA NA NA Significance level: *** 1% , ** 5%, * 10% Abbreviations: Whites, Non-Hispanic Whites; Blacks, Non-Hispanic Blacks; SE, standard error; NA, not applicable a The sub-aggregate effect estimate and share is calculated from the summing all sub-aggregate effects similar to Table 2-4. Sub- aggregate effects for each demographic, socioeconomic, health and access, and beliefs about flu vaccines were estimated similar to Table 2-4. b Unlike the 2009 NHFS dataset, the 2013 BRFSS data does not contain provider recommendations and beliefs about flu vaccines. To allow for a comparison to the 2013 BRFSS findings, these variables were removed and the decomposition analysis was repeated. ! ! 158! Appendix 3A: Provider recommendations and influenza vaccinations for the 2013-2014 season Instructions. Please answer the following questions to the best of your ability. There are no right or wrong answers. If you do not know the answers to certain questions, you can skip them to indicate "Don't know" or "No opinion." Question 1. Since September 2013, have you received at least one seasonal flu vaccination? Seasonal flu vaccinations come in two types. One type is a shot and the other type is a spray, mist, or drip in the nose. Responses: Yes No Question 2. Since September 2013, have you ever requested a seasonal flu vaccination from your usual care doctor or healthcare provider? Responses: Yes No Question 3. Since September 2013, did your usual care doctor or healthcare provider personally recommend that you receive a seasonal flu vaccination? Posted signs, newsletters, pamphlets, or television and radio ads are not considered a recommendation. Responses: Yes No IF QUESTION 1 AND 3 RESPONSE IS YES THEN ASK QUESTION 3a ELSE SKIP TO QUESTION 4 Question 3a. I would have obtained the seasonal flu vaccination had I not received a recommendation from my doctor or healthcare provider to get the flu vaccination. Responses: Strongly disagree Disagree Somewhat disagree Somewhat agree Agree Strongly agree ! 159! Question 4. Do you currently work in a healthcare facility such as a hospital, medical clinic, doctor's office, or nursing home? This includes part-time and unpaid work in a healthcare facility as well as professional nursing care provided in the home. Responses: Yes No IF QUESTION 4 RESPONSE IS YES THEN ASK QUESTION 4a ELSE SKIP TO QUESTION 5 Question 4a As part of your routine work, do you provide direct patient care such as physical or hands-on contact with patients? Responses: Yes No Question 5. In the past 12 months, have you had any kind of health care coverage, including health insurance, prepaid plans such as HMOs, or government plans such as Medicaid? Responses: Yes No Question 6. In general, how would you rate your current health status? Responses: Excellent Very good Good Fair Poor Question 7. In the past 2 years has a doctor or other health professional ever said that you have any of the following health conditions? By “other health professional” we mean a nurse practitioner, a physician’s assistant, or some other license professional. Select all that apply. Responses: Asthma A lung condition other than asthma A heart condition Diabetes A neurological or neuromuscular condition ! 160! A liver condition A weakened immune system caused by a chronic illness or by medicines taken for a chronic illness Question 8. In what state do you currently reside? Responses: [50 states] District of Columbia Puerto Rico I do not reside in the United States Question 9. Please indicate the highest level of education completed. Responses: Grammar school High School or equivalent Vocational/Technical School (2 year) Some College College Graduate (4 year) Master's Degree (MS) Doctoral Degree (PhD) Professional Degree (MD, JD, etc.) Other Question 10. What is your gender? Responses: Female Male Question 11. What is your current marital status? Responses: Divorced Married Separated Single Widowed Question 12. Please indicate your current household income in U.S. dollars ! 161! Responses: <$10,000 $10,000 - $19,999 $20,000 - $29,999 $30,000 - $39,999 $40,000 - $49,999 $50,000 - $74,999 $75,000 - $99,999 $100,000 - $150,000 >$150,000 Question 13. How old are you? Responses: 18-24 25-34 35-44 45-54 55-64 65+ Question 14. What is your race? Responses: White/Caucasion African American Hispanic Asian Native American Pacific Islander Other ! 162! Appendix 3B: Random assignment of healthcare provider recommendations Instructions. All respondents will read the preface and then respond to the three items that measure their agreement with the effectiveness of the flu vaccine and their risk of becoming infected with the flu. Please answer the following questions to the best of your ability. There are no right or wrong answers. If you do not know the answers to certain questions, you can skip them to indicate "Don't know" or "No opinion." Preface Commonly called "the flu," seasonal influenza is caused by influenza viruses that typically infect the respiratory system (i.e., the nose, throat, lungs). Influenza viruses are mainly thought to spread from person to person in respiratory droplets of coughs and sneezes. Each year, in the United States, an average 5-20% of the population gets the flu and hospitalizations with seasonal flu-related complications occur in at least 200,000 people. Furthermore, the influenza virus widely circulates the United States from the late from through early spring. On a scale of 1 to 4 where 1 means strongly disagree and 4 means strongly agree, how much do you agree or disagree with the following statements regarding flu vaccinations? Question 1, Statement 1. The seasonal flu vaccination is effective in preventing the seasonal flu. Responses: Strongly disagree Disagree Agree Strongly agree Question 2, Statement 2. If I didn't get a seasonal flu vaccination, I have a high chance of getting sick with the seasonal flu. Responses: Strongly disagree Disagree Agree Strongly agree Question 3, Statement 3. I am worried about getting sick from the seasonal flu vaccine. Responses: Strongly disagree Disagree Agree Strongly agree ! 163! Experimental conditions randomly assigned to each survey respondent Instructions. Please answer the following question to the best of your ability. There are no right or wrong answers. If you do not know the answers to certain questions, you can skip them to indicate "Don't know" or "No opinion." Question 4, Condition 1. The information below describes a fictional scenario between you and your usual care doctor or healthcare provider. This scenario is designed to mimic the events that may occur during a routine doctor's visit during a typical flu season. Scenario: You are currently not vaccinated against the flu. You have recently visited your regular doctor for a routine physical. As a result of your doctor's visit, you learn that your doctor received the flu vaccine. Your doctor recommends that you get the flu vaccine. If you had to make a decision now, would you get the flu vaccine? Responses: Yes No OR Question 4, Condition 2. The information below describes a fictional scenario between you and your usual care doctor or healthcare provider. This scenario is designed to mimic the events that may occur during a routine doctor's visit during a typical flu season. Scenario: You are currently not vaccinated against the flu. You have recently visited your regular doctor for a routine physical. As a result of your doctor's visit, it is unknown to you whether your doctor obtained a flu vaccination. Your doctor recommends that you get the flu vaccine. If you had to make a decision now, would you get the flu vaccine? Responses: Yes No OR Question 4, Condition 3. The information below describes a fictional scenario between you and your usual care doctor or healthcare provider. This scenario is designed to mimic the events that may occur during a routine doctor's visit during a typical flu season. ! 164! Scenario: You are currently not vaccinated against the flu. You have recently visited your regular doctor for a routine physical. As a result of your doctor's visit, you learn that your doctor received the flu vaccine. You receive no recommendation from your doctor to get the flu vaccine. If you had to make a decision now, would you get the flu vaccine? Responses: Yes No OR Question 4, Condition 4. The information below describes a fictional scenario between you and your usual care doctor or healthcare provider. This scenario is designed to mimic the events that may occur during a routine doctor's visit during a typical flu season. Scenario: You are currently not vaccinated against the flu. You have recently visited your regular doctor for a routine physical. As a result of your doctor's visit, it is unknown to you whether your doctor obtained a flu vaccination. You receive no recommendation from your doctor to get the flu vaccine. If you had to make a decision now, would you get the flu vaccine? Responses: Yes No Question 5. Do you currently work in a healthcare facility such as a hospital, medical clinic, doctor's office, or nursing home? This includes part-time and unpaid work in a healthcare facility as well as professional nursing care provided in the home. Responses: Yes No IF QUESTION 5 RESPONSE IS YES THEN ASK QUESTION 5a ELSE SKIP TO QUESTION 6 Question 5a. As part of your routine work, do you provide direct patient care such as physical or hands-on contact with patients? Responses: Yes No Question 6. In the past 12 months, have you had any kind of health care coverage, including health insurance, prepaid plans such as HMOs, or government plans such as Medicaid? ! 165! Responses: Yes No Question 7. In general, how would you rate your current health status? Responses: Excellent Very good Good Fair Poor Question 8. In the past 2 years has a doctor or other health professional ever said that you have any of the following health conditions? By “other health professional” we mean a nurse practitioner, a physician’s assistant, or some other license professional. Select all that apply. Responses: Asthma A lung condition other than asthma A heart condition Diabetes A neurological or neuromuscular condition A liver condition A weakened immune system caused by a chronic illness or by medicines taken for a chronic illness Question 9. In what state do you currently reside? Responses: [50 states] District of Columbia Puerto Rico I do not reside in the United States Question 10. Please indicate the highest level of education completed. Responses: Grammar school High School or equivalent Vocational/Technical School (2 year) Some College College Graduate (4 year) Master's Degree (MS) ! 166! Doctoral Degree (PhD) Professional Degree (MD, JD, etc.) Other Question 11. What is your gender? Responses: Female Male Question 12. What is your current marital status? Responses: Divorced Married Separated Single Widowed Question 13. Please indicate your current household income in U.S. dollars Responses: <$10,000 $10,000 - $19,999 $20,000 - $29,999 $30,000 - $39,999 $40,000 - $49,999 $50,000 - $74,999 $75,000 - $99,999 $100,000 - $150,000 >$150,000 Question 14. How old are you? Responses: 18-24 25-34 35-44 45-54 55-64 65+ Question 15. What is your race? ! 167! Responses: White/Caucasion African American Hispanic Asian Native American Pacific Islander Other ! 168! Appendix 3C: Subpopulation descriptive statistics Flu vaccination status All respondents Not vaccinated Vaccinated (N=619) (N=505) (N=114) p-value HCP recommendation No 452 393 59 <0.001 Yes 167 112 55 Requested flu vaccination from HCP No 619 505 114 NA Yes 0 0 0 Would obtain flu vaccine without recommendation Strongly disagree NA NA a 7 NA Disagree 9 Somewhat disagree 6 Somewhat agree 8 Agree 14 Strongly agree 11 Unknown 0 Abbreviations: NA, not applicable; HCP, healthcare provider a This question was asked to respondents with a flu vaccination and recommendation ! 169! Appendix 3D: Subpopulation regression analysis Model 1 a Model 2 b Model 3 b Coefficient Coefficient Coefficient HCP recommendation d 19.9 0.1603 0.3337 (5.81) (4.45) (9.87) Requested flu vaccination from their provider - - - Sample size 619 619 794 Adjusted R-square 0.0503 0.0879 0.2306 Abbreviation: HCP, healthcare provider a Ordinary least squares regression unadjusted for respondent characteristics in Table 3-1 b Ordinary least squares regression adjusted for respondent characteristics in Table 3-1. This model omits the question: Respondent requested flu vaccination from HCP. Response to questions described in Table 3-1 Footnote ‘a’ and ‘b’ were omitted from the regressions because not all respondents were asked these questions. c Ordinary least squares regression adjusted for respondent characteristics in Table 3-1. Response to questions described in Table 3-1 Footnote ‘a’ and ‘b’ were omitted from the regressions because not all respondents were asked these questions. d t-statistics in parentheses !
Abstract (if available)
Abstract
Chapter 1. Determinants of healthcare provider recommendations for influenza vaccinations. Abstract. Objective: Investigate determinants of receiving healthcare provider (HCP) recommendations for seasonal and H1N1 influenza vaccinations. Methods: Using a United States national sample of adults 18+ from the National 2009 H1N1 Flu Survey, multivariate regression models estimated the likelihood of receiving a HCP recommendation. Covariates included demographics, socioeconomic status, and Advisory Committee on Immunization Practices (ACIP) priority groups. Results: Adults age 65+ were more likely to report a HCP recommendation when compared to adults age 18-34 (OR:1.738, 95%CI:1.427-2.116). Chronically ill adults had 58.0% (95%CI:1.414-1.765) higher odds of reporting a HCP recommendation than non-chronically ill adults. Patients visiting a doctor once and twice had 28.7% (95%CI:0.618-0.821) and 17.1% (95%CI:0.721-0.952) lower odds of reporting a HCP recommendation when compared to adults visiting their doctor at least four times. And, racial/ethnic minorities were more likely to receive a recommendation. Conclusions: ACIP priority groups experienced higher rates of recommendations compared to non-ACIP groups. Racial/ethnic minority groups have an increased likelihood of receiving a recommendation compared to non-Hispanic Whites. Further efforts to increase recommendation rates for all patient groups can improve influenza vaccination rates. ❧ Chapter 2. What Explains Racial/Ethnic Disparities in Influenza Vaccination Rates? Abstract. Objectives: Explain causes of racial/ethnic differences in influenza vaccination rates in the United States (US) general population. Methods: Racial/ethnic vaccination disparities were decomposed into differences in observable and unobservable characteristics. Multivariate decomposition of logit models for influenza vaccinations using a US national sample of adults 18+ from the National 2009 H1N1 Flu Survey. Covariates included demographics, socioeconomic status, beliefs about influenza vaccine effectiveness and Advisory Committee on Immunization Practices (ACIP) priority groups. Results: The majority of the share in influenza vaccination disparities between Non-Hispanic Whites (Whites), Non-Hispanic Blacks (Blacks), and Hispanics can be explained by differences in observable characteristics. Whites report higher influenza vaccination rates by 13.7 and 12.0 percentage points relative to Blacks and Hispanics, respectively. 29.2% of the difference in influenza vaccination rates between Whites and Blacks can be explained by differences in beliefs about influenza vaccines. 14.0% of the same difference can be explained by differences in socioeconomic status. 6.1% of the difference in influenza vaccination rates between Whites and Hispanics can be explained by differences in beliefs about influenza vaccines. 22.6% of the same difference can be explained by differences in socioeconomic status. Healthcare provider recommendations can modestly decrease influenza vaccination disparities between these racial/ethnic groups. Conclusion: A majority of the differences in influenza vaccination rates between Whites and Blacks and Whites and Hispanics can be explained by differences in observable characteristics. In particular, influenza vaccination beliefs are significant contributors to explaining influenza vaccination differences. ❧ Chapter 3. Healthcare provider recommendations and influenza vaccinations: a causal effects estimate. Abstract. Prior studies have thus far demonstrated a significant positive correlation between healthcare provider recommendations for the flu vaccine and patient adherence to the recommendation. These existing studies employ regression techniques that do not consider a deeper question of whether the regression of provider recommendations on flu vaccination has a causal interpretation. The objective of this study is to provide a measure of bias associated with prior estimates of provider recommendations for flu vaccines and patient adherence to these recommendations. This study hypothesizes these prior estimates of the effect of provider recommendations are biased upwards. The results demonstrate provider recommendations generate an unadjusted 39% (p<0.001) increased likelihood of obtaining a flu vaccine when measured from an Amazon Mechanical Turk (MTurk) sample of the United States general population. A causal effect of provider recommendations on flu vaccination intent was measured by randomly assigning provider recommendations to treatment scenarios from a similar, yet independent, MTurk sample. The resulting unadjusted provider recommendation effect on flu vaccination intent was 16% (p<0.001). Therefore, by construction, there exists a 23-percentage point bias in the recommendation effect. These findings suggest prevailing measures of the effect of provider recommendations on flu vaccinations are overestimated. Further research is needed to describe the sensitivities of this bias in different treatment scenarios.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Understanding primary nonadherence to medications and its associated healthcare outcomes: a retrospective analysis of electronic medical records in an integrated healthcare setting
PDF
The causal-effect of childhood obesity on asthma in young and adolescent children
PDF
Essays on the economics of infectious diseases
PDF
Discriminating changes in health using patient-reported outcomes
PDF
Economic aspects of obesity
PDF
Advances and applications for economic evaluation methods in health technology assessment (HTA)
PDF
Selected papers on the evaluation of healthcare costs of prematurity and necrotizing enterocolitis using large retrospective databases
PDF
Racial/ethnic differences in patient experiences with health care in association with the Gleason’s score level at prostate cancer diagnosis: findings from the SEER-CAHPS data
PDF
Developing and testing a heuristic-systematic model of health decision making: the role of affect, trust, confidence and media influence
PDF
Understanding the role of population and place in the dynamics of seasonal influenza outbreaks in the United States
PDF
Impact of pharmacy-based transitional care on healthcare utilization and costs
PDF
The impact of Patient-Centered Medical Home on a managed Medicaid plan
PDF
Multilevel sociodemographic correlates of the health and healthcare utilization of childhood cancer survivors
PDF
The impact of treatment decisions and adherence on outcomes in small hereditary disease populations
PDF
Selected papers on methods for evaluating the impact of marijuana use on BMI and other risk factors for metabolic syndrome
PDF
A biodemographic approach to understanding sociodemographic disparities in kidney functioning on three dimensions: individual, population, and cross-national
PDF
Racial/ethnic variation in care preferences and care outcomes among United States hospice enrollees
PDF
Three essays in health economics
PDF
The determinants and measurement of human capital
PDF
Exploring geography and multi-level factors associated with disparities in HPV vaccination among Latino adolescents in Los Angeles County
Asset Metadata
Creator
Villacorta, Reginald B.
(author)
Core Title
Healthcare provider recommendations: a panacea to improving influenza vaccination rates?
School
School of Pharmacy
Degree
Doctor of Philosophy
Degree Program
Pharmaceutical Economics and Policy
Publication Date
04/17/2015
Defense Date
02/23/2015
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
causal,disparity,influenza,OAI-PMH Harvest,provider,recommendations,vaccinations
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Sood, Neeraj (
committee chair
), Doctor, Jason N. (
committee member
), Huesch, Marco (
committee member
)
Creator Email
rbvillacorta@gmail.com,rvillaco@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-552760
Unique identifier
UC11297516
Identifier
etd-Villacorta-3326.pdf (filename),usctheses-c3-552760 (legacy record id)
Legacy Identifier
etd-Villacorta-3326.pdf
Dmrecord
552760
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Villacorta, Reginald B.
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
Tags
causal
disparity
influenza
provider
recommendations
vaccinations