Florida Behavioral Risk Factor Surveillance Research Articles

Associations of Content and Context of Communication with Prostate-Specific Antigen Testing.

There is limited research about the content and context of communication on prostate-specific antigen (PSA) testing among men in the state of Florida. The purpose of this study is to understand how the content communication (discussion of advantages and disadvantages of PSA testing between provider and patient; provider recommendations of PSA testing) and the context of communication (continuity of care denoted by the presence of a personal doctor) influence PSA testing. Data were drawn from the Florida Behavioral Risk Factor Surveillance System. Receipt of PSA testing was the primary outcome. Multiple logistic regression analyses were used to adjust for sociodemographic, clinical, healthcare access, and lifestyle characteristics when associating the content and context of communication with PSA testing. Discussions were classified into four mutually exclusive categories: discussions of advantages and disadvantages, only advantages, only disadvantages, and no discussion. The most significant predictors for PSA testing included physician recommendation, discussions including advantages, older age, non-smoking, and having a personal doctor. Individualized PSA screening may be a pathway to reducing racial disparities in screening for prostate cancer (PCa) and, by extension, lower incidence and mortality rates. Developing a bill to create an Office of Men's Health at Health & Human Services is recommended.

Abstract P527: Epidemiology of Adverse Childhood Experiences and Cardiometabolic Risks

Purpose: Adverse Childhood Experiences (ACEs) have been linked to a wide range of adverse health outcomes. The purpose of this research is to better understand this association and elucidate the mechanisms by which ACEs influence cardiometabolic (CM) risk factors (blood pressure, BMI, pre-diabetes). Racial/ethnic distribution of ACEs and potential health outcomes were also examined. Methods: This study utilized data from the Florida Behavioral Risk Factor Surveillance System for the 2019-2021 survey years to examine the racial/ethnic distribution of ACEs and cardiometabolic risks. Results: A total of 13,198 participants were included in this study with non-Hispanic Whites representing 76% of the study population. The highest percentage of each racial/ethnic group reported two or fewer ACEs. The Multiracial non-Hispanic group tended to have higher percentages of reported ACEs. The association between ACEs and CM risk factors significantly differed across racial/ethnic groups.

Constructing Statistical Intervals for Small Area Estimates Based on Generalized Linear Mixed Model in Health Surveys.

Generalized Linear Mixed Model (GLMM) has been widely used in small area estimation for health indicators. Bayesian estimation is usually used to construct statistical intervals, however, its computational intensity is a big challenge for large complex surveys. Frequentist approaches, such as bootstrapping, and Monte Carlo (MC) simulation, are also applied but not evaluated in terms of the interval magnitude, width, and the computational time consumed. The 2013 Florida Behavioral Risk Factor Surveillance System data was used as a case study. County-level estimated prevalence of three health-related outcomes was obtained through a GLMM; and their 95% confidence intervals (CIs) were generated from bootstrapping and MC simulation. The intervals were compared to 95% credential intervals through a hierarchial Bayesian model. The results showed that 95% CIs for county-level estimates of each outcome by using MC simulation were similar to the 95% credible intervals generated by Bayesian estimation and were the most computationally efficient. It could be a viable option for constructing statistical intervals for small area estimation in public health practice.

Community-Engaged Needs Assessment of Deaf American Sign Language Users in Florida, 2018.

Deaf American Sign Language (ASL) users comprise a linguistic and cultural minority group that is understudied and underserved in health education and health care research. We examined differences in health risk behaviors, concerns, and access to health care among Deaf ASL users and hearing English speakers living in Florida. We applied community-engaged research methods to develop and administer the first linguistically accessible and contextually tailored community health needs assessment to Deaf ASL users living in Florida. Deaf ASL users (n = 92) were recruited during a 3-month period in summer 2018 and compared with a subset of data on hearing English speakers from the 2018 Florida Behavioral Risk Factor Surveillance System (n = 12 589). We explored prevalence and adjusted odds of health behavior, including substance use and health care use. Mental health was the top health concern among Deaf participants; 15.5% of participants screened as likely having a depressive disorder. Deaf people were 1.8 times more likely than hearing people to engage in binge drinking during the past month. In addition, 37.2% of participants reported being denied an interpreter in a medical facility in the past 12 months. This study highlights the need to work with Deaf ASL users to develop context-specific health education and health promotion activities tailored to their linguistic and cultural needs and ensure that they receive accessible health care and health education.

Socioeconomic And Survival Differences Among Minorities With Hepatocellular Carcinoma In Florida.

BackgroundDisparities in hepatocellular carcinoma (HCC) have been partly attributed to low socioeconomic status among minorities. We investigated associations between race, socioeconomic characteristics, geographic characteristics and survival in HCC patients in Florida.MethodsUsing the Florida Cancer Data System (FCDS), we analyzed HCC cases diagnosed between 1/1/2004 and 12/31/2013. To ascertain population-level socioeconomic characteristics, we linked FCDS to the 2010–2014 US Census American Community Survey and the 2013 Florida Behavioral Risk Factor Surveillance System. We also estimated patient distance to liver transplant and academic cancer centers. Using Cox proportional hazards, we modeled the association between race and survival.ResultsOf 10,852 patients, 13.1% were Black, 67.1% White, 15.7% Hispanic, and 3.2% Asian. At diagnosis, Blacks were younger with more extensive disease, p <0.001. Transplants were performed in 9.3% of Hispanics, 7.5% of Whites, 5.8% of Asians and 4.2% of Blacks, p <0.001. Median survival was longest in Hispanics and shortest in Blacks, p<0.001 When adjusted for gender, age, payer, SEER stage, surgery type, and receipt of treatment, Blacks had a 17% increased risk of death [hazard ratio (HR) 1.17, 95% confidence interval (CI) 1.07–1.29] and Whites a 9% increased risk of death [HR 1.09, 95% CI 1.02–1.17] compared to Hispanics. As a group, Hispanics lived closest to any transplant or academic cancer center, p <0.001. Neighborhood poverty level was highest where Hispanic patients lived, p <0.001.ConclusionThough socioeconomic differences may contribute to disparities, Hispanics survived longer than Blacks and Whites in Florida despite living in the most socioeconomically depressed neighborhoods. Increased access to transplant likely contributed to improved survival. Additional research is needed to identify which individual socioeconomic and geographic determinants contribute most to disparities.

Demographics and Clinical Profiles of Patients Visiting a Free Clinic in Miami, Florida.

Background: Although the ranks of the uninsured in the United States have decreased in recent years, some states still lack Medicaid expansion programs, leaving many Americans, especially the indigent and homeless, without adequate healthcare coverage. Free-for-care clinics are oftentimes the last safety net for these vulnerable populations. Because these clinics have limited funding, a thorough understanding of the patients they serve is necessary to effectively direct their resources. The objective of the present study is to investigate the characteristics and clinical profiles of patients utilizing a free clinic in Miami, Florida.Methods: Aggregate EMR data reflecting consecutive adult patient visits to the Miami Rescue Mission Clinic in Miami, Florida between January 1st, 2018 to March 15th, 2019 (n = 846) were reviewed for sociodemographic characteristics and chronic disease prevalence. Prevalence rates were compared by sex and to county estimates from the Florida Behavioral Risk Factor Surveillance System.Results: The most common conditions were mental health (19.3%), circulatory system (14.7%), and musculoskeletal system disorders (13.9%). Males had a greater prevalence of depression (difference = 6.6%; 95% CI [1.5 to 10.7%]; χ2 = 6.2; p = 0.013) and overall mental illness (22.0 vs. 10.4%, difference = 11.6%; 95% CI [5.7 to 16.4%]; χ2 = 13.2; p = 0.0003) compared to females, and male sex was identified as an independent risk factor for mental illness on multivariate logistic regression analysis (OR = 2.8; 95% CI [1.7 to 4.7]; p < 0.001). There was also a higher prevalence of depression (difference = 6.41%; 95% CI [2.1 to 10.2%]; χ2 = 8.0; p = 0.0047) and HIV (difference = 1.4%; 95% CI [0.3 to 3.0%]; χ2 = 7.3; p = 0.007) in male patients compared to county estimates. Rates of hypertension, diabetes, elevated cholesterol, asthma, and COPD were lower in the clinic population compared to the surrounding county.Conclusion: There is an acute need for mental health services in this population. The lowered prevalence of other chronic conditions is due to underdiagnosis and loss to follow-up. Such analyses are important in guiding policy decisions for meeting the health needs of vulnerable, at risk populations.

Spatial smoothing models to deal with the complex sampling design and nonresponse in the Florida BRFSS survey

Public health and governmental organizations have acknowledged the importance of obtaining information of various characteristics for small areas, such as counties. Spatial smoothing models have been developed to gain reliable information on the geographical distribution of the outcome of interest. When the geographical analysis is based on survey data, two issues pose challenges: (1) the complex design of the survey and (2) the presence of missing data due to non-response. We investigate the influence of missing data and the adjustment thereof in the context of the 2013 Florida Behavioral Risk Factor Surveillance System (BRFSS) health survey. We focus on the application and comparison of the Hajek ratio estimator and two model-based approaches for estimation of the spatial trend of the prevalence of having no health insurance coverage. The model-based methods are compared using the Deviance Information Criterion which show the benefits of modeling the weights as flexibly as possible. Methods are extended towards subgroup analyses and the estimation of area-specific standardized rates, where household incomes was identified as an important factor to include in the analysis.

Florida Populations Most at Risk of Not Being Up to Date With Colorectal Cancer Screening.

IntroductionThe purpose of this study was to examine the characteristics of populations at risk of not being up to date on colorectal cancer screening in Florida.MethodsWe used Exhaustive Chi-squared Automatic Interaction Detection, a classification tree analysis, to identify subgroups not up to date with colorectal cancer screening using the 2013 Florida Behavioral Risk Factor Surveillance System. The data set was restricted to adults aged 50 to 75 years (n = 14,756).ResultsOnly 65.5% of the sample was up to date on colorectal cancer screening. Having no insurance and having a primary care provider were the most significant predictors of not being up to date on screening. The highest risk subgroups were 1) respondents with no insurance and no primary care provider, regardless of their employment status (screening rate, 12.1%–23.7%); 2) respondents with no insurance but had a primary care provider and were employed (screening rate, 32.3%); and 3) respondents with insurance, who were younger than 55 years, and who were current smokers (screening rate, 42.0%).ConclusionSome populations in Florida are at high risk for not being up to date on colorectal cancer screening. To achieve Healthy People 2020 goals, interventions may need to be further tailored to target these subgroups.

Association between race/ethnicity and disability status and receipt of vaccines among older adults in Florida

BackgroundIndividuals living with a disability or are a member of a certain racial/ethnic group may be at heightened risk for not receiving important vaccinations. ObjectiveThis study examined whether race/ethnicity and disability status are associated with the receipt of two vaccines (influenza and pneumococcal) among older adults living in Florida. MethodsUsing the 2011–2015 Florida Behavioral Risk Factor Surveillance System, a cross-sectional survey, we ran bivariate and multivariate analyses to determine the associations for race/ethnicity and disability status with receipt of vaccinations among individuals 65 years and older. Interactions between race/ethnicity and disability status were tested in each model. ResultsAmong our study sample, 68% received the pneumococcal vaccine in their lifetime and 54% of them received influenza vaccine in the past 12 months. Multivariate logistic regression indicated that Non-Hispanic Blacks and Hispanics were less likely to receive both vaccines compared to Non-Hispanic Whites. Older adults with a disability were more likely to receive influenza and pneumococcal vaccines compared to those without. A significant interaction was observed between race/ethnicity and disability status for predicting pneumococcal vaccination receipt. ConclusionsLarge proportions of older adults in Florida continue to go without needed vaccinations. Although race/ethnicity and disability status were shown to have some association with receipt of vaccines, having a regular source of care, employment and income also were shown to be important predictors.

Abstract A01: Socioeconomic and Survival Differences among Minorities With Hepatocellular Carcinoma in Florida

Abstract Purpose: Hepatocellular carcinoma (HCC) is a leading cause of cancer-related mortality with racial disparities in survival after its diagnosis. It has been suggested that racial disparities in survival reflect socioeconomic disparities. We aimed to determine associations between race, socioeconomic characteristics and survival in patients with HCC in Florida. Methods: We identified HCC cases diagnosed between 1/1/2004 - 12/31/2013 in the Florida Cancer Data System (FCDS). We linked FCDS to the 2010-2014 US Census American Community Survey by census tract and the 2013 Florida Behavioral Risk Factor Surveillance System by county to ascertain population-level characteristics. We performed univariate, bivariate and Cox proportional hazards regression analyses to describe the association between race and survival. Results: Of 10,852 patients with HCC, 77.4% were men, 13.4% were Black, 67.1% non-Hispanic White, 15.7% Hispanic, 3.2% Asian and &amp;lt;1% other. Blacks were younger at presentation, with a median age of 59 years compared to 64 years in Hispanics and 63 years in all other races, p&amp;lt;0.01. Median size of the largest tumor was greatest in Asians, 6.0 cm, followed by 5.4 cm in Blacks, 4.8 cm in Hispanics and 4.7 cm in Whites, p&amp;lt;0.01. Blacks were most likely to have distant disease, 18.1%, compared to 14.3% of Whites, 14.9% of Hispanics, 15.9% of Asians, and 10.4% of others, p&amp;lt;0.01. Only 16.3% of Blacks had surgery or radiofrequency ablation compared to 25.4% of Asians, 24.8% of Hispanics and 21.7% of Whites, p &amp;lt;0.01. Transplant was performed in 9.7% of Hispanics, 8% of Whites, 6.9% of Asians and 4.6% of Blacks, p&amp;lt;0.01. One-year survival was 45.6% in Blacks, 52.1% in Whites, 55.4% in Hispanics, 58.1% in Asians, and 57.3% in others, p&amp;lt;0.01. Five-year survival was 31.7% in Blacks, 36.5% in Whites, 36.9% in Hispanics 46.2% in Asians, and 51% in other, p&amp;lt;0.01.Median survival was 208 days in Hispanics, 171 days in Whites, 158 days in Asians and 136 days in Blacks, p&amp;lt;0.01. When adjusted for gender, age at diagnosis, payer, tumor size, stage, and surgical treatment, Hispanics had 16% reduced rate of death (hazard ratio (HR) 0.84, 95% confidence interval (CI) 0.77-0.92, p&amp;lt;0.01) and Whites had 7% reduced rate of death (HR 0.93, 95% CI 0.86-0.99, p 0.03) compared to Blacks. Hispanics had 9% adjusted reduced rate of death (HR 0. 91, 95% CI 0.85-0.97, p&amp;lt;0.01) compared to Whites. In our sample, Blacks and Asians were more frequently uninsured, 11.6%, compared to 8.1% of Hispanics and 6.3% of Whites, p&amp;lt;0.01. There were significant racial differences in population characteristics. For example, the census tracts where Hispanics live have the highest uninsured population and the highest percentage of participation in the supplemental nutrition assistance and/or food stamps program, compared to all other races, p &amp;lt;0.01. Also, Hispanics lived in census tracts where more families reported income below the federal poverty level and had lower mean household income, compared to Blacks, p&amp;lt;0.01 In regards to community health, in the counties where Blacks live, a higher percentage of adults have been diagnosed with diabetes, elevated cholesterol, cardiovascular disease and are obese than in the counties where Hispanic patients live, p&amp;lt;0.01. However, in the counties where Hispanics live, a higher percentage of adults have been told they are overweight, compared to Blacks, Whites or Others, p&amp;lt;0.01. In counties where Blacks lived, more adults reported having a colonoscopy, but fewer adults reported having a mammogram, compared to counties where Hispanic patients resided, p&amp;lt;0.01. Conclusions: Access to care and socioeconomic differences may help explain survival differences in some settings; however, despite living in areas with similar or worse socioeconomic features, Hispanics with HCC survive longer than Blacks and Whites in Florida. Although access to transplant may inform these findings, additional contributors must be identified. Citation Format: Patricia D. Jones, Valery Muenyi, Joselin Gonzalez-Diaz, Paul Martin, Erin Kobetz. Socioeconomic and Survival Differences among Minorities With Hepatocellular Carcinoma in Florida. [abstract]. In: Proceedings of the Ninth AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2016 Sep 25-28; Fort Lauderdale, FL. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2017;26(2 Suppl):Abstract nr A01.

Chronic Diseases and Use of Contraception Among Women at Risk of Unintended Pregnancy.

Women with chronic diseases are at increased risk of having unintended pregnancies. Little is known whether chronic diseases are associated with increased likelihood of effective/highly effective contraceptive use. We analyzed 2008-2010 Florida Behavioral Risk Factor Surveillance System data for women aged 18-44 years who were at risk of unintended pregnancy. Multivariable Poisson regression estimated adjusted prevalence ratios (aPRs) and 95% confidence intervals (CIs) for contraceptive use in relation to diabetes, cardiovascular disease (CVD), and current asthma. We assessed the association of chronic disease status with use of three different contraception outcomes: (1) any method versus none, (2) less effective methods (methods associated with ≥10 unintended pregnancies/100 women/year) versus none, and (3) effective/highly effective methods (<10 unintended pregnancies/100 women/year) versus none. Among 4473 women at risk for unintended pregnancy, 87% were using any method of contraception (22.5% less effective methods and 64.5% effective/highly effective methods). Women with CVD were more likely than those without CVD to use any contraception (aPR = 1.09, 95% CI: 1.04, 1.15), less effective (aPR = 1.39, 95% CI: 1.13, 1.70), and effective/highly effective (aPR = 1.10, 95% CI: 1.03, 1.19) contraception. Women with diabetes were more likely to use less effective methods than women without diabetes (aPR = 1.34, 95% CI: 1.05, 1.72). No significant associations were observed for asthma, regardless of contraceptive effectiveness. Self-reported use of effective/highly effective contraception was higher than nonuse or use of less effective methods among all women at risk of unintended pregnancy, but could be improved, especially among women with chronic diseases.

The cross-sectional association between severity of non-cognitive disability and self-reported worsening memory

BackgroundResearch has demonstrated a clear association between cognitive decline and non-cognitive disability; however, all of these studies focus on disability as a correlate or result of some level of cognitive impairment or dysfunction. The relationship between disability and cognition is likely a complex one, that is currently incompletely described in the literature. ObjectivesOur objective was to estimate the prevalence of long-term, non-cognitive disability using a population-representative sample of adults aged 18 and older, and then estimate the association between long-term, non-cognitive disability and self-reported worsening memory. MethodsUsing the 2009 Florida Behavioral Risk Factor Surveillance System (BRFSS), we measured the relationship between non-cognitive disability and worsening memory using multivariable logistic regression analysis weighted to account for the complex sampling design of the BRFSS. We also estimated the adjusted odds of worsening memory by disability severity, classified according to the types of assistance needed. ResultsApproximately 18% (95% confidence interval = (16%, 19%)) of Floridians were living with a long-term, non-cognitive disability in 2009. Among adults with no disability during or prior to the last year, only 5% reported worsening memory. The proportion of Floridians reporting worsening memory increases with increasing severity of disability-related limitations. In a multivariable logistic regression model, odds of worsening memory increased significantly with severity of disability-related limitations. ConclusionsThese results highlight the association between non-cognitive disability and subsequent increased odds of worsening memory, independent of several other known risk factors, and a dose-response association with disability-related limitations.

Care Utilization Patterns and Diabetes Self-Management Education Duration.

Objective. Previous studies have shown that receiving diabetes self-management education (DSME) is associated with increased care utilization. However, the relationship between DSME duration and care utilization patterns remains largely unexamined. Our purpose is to characterize DSME duration and examine the relationship between DSME duration and clinical- and self-care utilization patterns.Methods. The study sample included 1,446 adults who were ≥18 years of age, had diabetes, and had participated in the 2008 Florida Behavioral Risk Factor Surveillance System survey. Clinical- and self-care outcomes were derived using responses to the survey’s diabetes module and were based on minimum standards of care established by the American Diabetes Association. The outcomes examined included self-monitoring of blood glucose at least once per day; receiving at least one eye exam, one foot exam, A1C tests, and an influenza vaccination in the past year; and ever receiving a pneumococcal vaccination. DSME duration was categorized as no DSME, >0 to <4 hours, 4–10 hours, and >10 hours.Results. After adjusting for sociodemographic variables, compared to those who did not receive DSME, those who had 4–10 or 10+ hours of DSME were more likely to receive two A1C tests (odds ratio [95% CI] 2.69 [1.30–5.58] and 2.63 [1.10–6.31], respectively) and have a pneumococcal vaccination (1.98 [1.03–3.80] and 1.92 [1.01–3.64], respectively). Those receiving 10+ hours of DSME were 2.2 times (95% CI 1.18–4.09) as likely to have an influenza vaccination.Conclusion. These data reveal a positive relationship between DSME duration and utilization of some diabetes clinical care services.

Model-based standardization to adjust for unmeasured cluster-level confounders with complex survey data.

Model-based standardization uses a statistical model to estimate a standardized, or unconfounded, population-averaged effect. With it, one can compare groups had the distribution of confounders been identical in both groups to that of the standard population. We develop two methods for model-based standardization with complex survey data that accommodate a categorical confounder that clusters the individual observations into a very large number of subgroups. The first method combines a random-intercept generalized linear mixed model with a conditional pseudo-likelihood estimator of the fixed effects. The second method combines a between-within generalized linear mixed model with census data on the cluster-level means of the individual-level covariates. We conduct simulation studies to compare the two approaches. We apply the two methods to the 2008 Florida Behavioral Risk Factor Surveillance System survey data to estimate standardized proportions of people who drink alcohol, within age groups, adjusting for measured individual-level and unmeasured cluster-level confounders.

Association Between Living Alone and Physical Inactivity Among People With and Without Disability, Florida Behavioral Risk Factor Surveillance System, 2009

People with disability may be at risk of developing diseases due to physical inactivity; social support from family and friends is positively related to engaging in regular physical activity. We compared the association between living alone and engagement in physical activity among people with and without disability in Florida. We used multivariate logistical regression to analyze 2009 Florida Behavioral Risk Factor Surveillance System data (n = 10,902) to assess differences in physical activity in disability levels for respondents who lived alone versus those who did not. Respondents with a disability were less likely to engage in physical activity than were people without a disability, regardless of disability type, and the lowest rates of engaging in physical activity were found for people with disability who lived alone. Public health efforts should consider the role of household composition when targeting physical activity interventions among people with disability.

Visitability surveillance, prevalence, and correlates in Florida

BackgroundA primary means of social connection is visiting friends and families in their homes. Visitability is designing houses in a way that enables people to visit others' homes regardless of physical limitations or use of mobility assistive devices. ObjectiveThe goals of this study were to develop a set of questions about visitability that could be used for surveillance and to assess the prevalence and correlates of visitability features in Florida. MethodsWe added five questions to the 2011 Florida Behavioral Risk Factor Surveillance System (n = 12,399 respondents) and used complementary log–log regression models to estimate the prevalence ratio of each visitability feature. ResultsThe prevalence of visitability features in Florida homes was high for respondents with and without disabilities, though there was variation by visitability feature. A level entrance to the home and wide doorways were present in most respondents' homes (84.9% and 86.2%, respectively), while a main floor bathroom (59.6%) and a zero-step entrance (45.4%) were reported less commonly. People with a disability were less likely to report that their own home had doorways wide enough to accommodate a wheelchair compared to people without a disability (PR = 0.87, 95% CI: 0.80–0.95). Visitability features were less common in households with lower income and also in trailers or mobile homes than in detached single-family homes. ConclusionsThe survey questions used in this study could be implemented in other states to measure and track visitability and monitor progress toward the Healthy People 2020 goal. Building or retro-fitting homes to include visitability features could increase the participation and inclusion of people with disabilities in community life.

Methods of estimating or accounting for neighborhood associations with health using complex survey data.

Reasons for health disparities may include neighborhood-level factors, such as availability of health services, social norms, and environmental determinants, as well as individual-level factors. Investigating health inequalities using nationally or locally representative data often requires an approach that can accommodate a complex sampling design, in which individuals have unequal probabilities of selection into the study. The goal of the present article is to review and compare methods of estimating or accounting for neighborhood influences with complex survey data. We considered 3 types of methods, each generalized for use with complex survey data: ordinary regression, conditional likelihood regression, and generalized linear mixed-model regression. The relative strengths and weaknesses of each method differ from one study to another; we provide an overview of the advantages and disadvantages of each method theoretically, in terms of the nature of the estimable associations and the plausibility of the assumptions required for validity, and also practically, via a simulation study and 2 epidemiologic data analyses. The first analysis addresses determinants of repeat mammography screening use using data from the 2005 National Health Interview Survey. The second analysis addresses disparities in preventive oral health care using data from the 2008 Florida Behavioral Risk Factor Surveillance System Survey.

Adjusting for confounding by neighborhood using generalized linear mixed models and complex survey data

When investigating health disparities, it can be of interest to explore whether adjustment for socioeconomic factors at the neighborhood level can account for, or even reverse, an unadjusted difference. Recently, we proposed new methods to adjust the effect of an individual-level covariate for confounding by unmeasured neighborhood-level covariates using complex survey data and a generalization of conditional likelihood methods. Generalized linear mixed models (GLMMs) are a popular alternative to conditional likelihood methods in many circumstances. Therefore, in the present article, we propose and investigate a new adaptation of GLMMs for complex survey data that achieves the same goal of adjusting for confounding by unmeasured neighborhood-level covariates. With the new GLMM approach, one must correctly model the expectation of the unmeasured neighborhood-level effect as a function of the individual-level covariates. We demonstrate using simulations that even if that model is correct, census data on the individual-level covariates are sometimes required for consistent estimation of the effect of the individual-level covariate. We apply the new methods to investigate disparities in recency of dental cleaning, treated as an ordinal outcome, using data from the 2008 Florida Behavioral Risk Factor Surveillance System (BRFSS) survey. We operationalize neighborhood as zip code and merge the BRFSS data with census data on ZIP Code Tabulated Areas to incorporate census data on the individual-level covariates. We compare the new results to our previous analysis, which used conditional likelihood methods. We find that the results are qualitatively similar.

Conditional pseudolikelihood methods for clustered ordinal, multinomial, or count outcomes with complex survey data

In order to adjust individual-level covariate effects for confounding due to unmeasured neighborhood characteristics, we have recently developed conditional pseudolikelihood methods to estimate the parameters of a proportional odds model for clustered ordinal outcomes with complex survey data. The methods require sampling design joint probabilities for each within-neighborhood pair. In the present article, we develop a similar methodology for a baseline category logit model for clustered multinomial outcomes and for a loglinear model for clustered count outcomes. All of the estimators and asymptotic sampling distributions we present can be conveniently computed using standard logistic regression software for complex survey data, such as sas proc surveylogistic. We demonstrate validity of the methods theoretically and also empirically by using simulations. We apply the new method for clustered multinomial outcomes to data from the 2008 Florida Behavioral Risk Factor Surveillance System survey in order to investigate disparities in frequency of dental cleaning both unadjusted and adjusted for confounding by neighborhood.

Doubly robust testing and estimation of model‐adjusted effect‐measure modification with complex survey data

Model-based standardization enables adjustment for confounding of a population-averaged exposure effect on an outcome. It requires either a model for the probability of the exposure conditional on the confounders (an exposure model) or a model for the expectation of the outcome conditional on the exposure and the confounders (an outcome model). The methodology can also be applied to estimate averaged exposure effects within categories of an effect modifier and to test whether these effects differ or not. Recently, we extended that methodology for use with complex survey data, to estimate the effects of disability status on cost barriers to health care within three age categories and to test for differences. We applied the methodology to data from the 2007 Florida Behavioral Risk Factor Surveillance System Survey (BRFSS). The exposure modeling and outcome modeling approaches yielded two contrasting sets of results. In the present paper, we develop and apply to the BRFSS example two doubly robust approaches to testing and estimating effect modification with complex survey data; these approaches require that only one of these two models be correctly specified. Furthermore, assuming that at least one of the models is correctly specified, we can use the doubly robust approaches to develop and apply goodness-of-fit tests for the exposure and outcome models. We compare the exposure modeling, outcome modeling, and doubly robust approaches in terms of a simulation study and the BRFSS example.