Close
IntroductionKey FindingsNational InsightsPrevalence of Unhealthy BehaviorsPrevalence of Multiple Unhealthy BehaviorsPrevalence of Zero Unhealthy BehaviorsOdds of Reporting Fair or Poor Health StatusState InsightsMultiple Unhealthy Behaviors: Education-Based DifferencesConclusionsAppendix 1Appendix 2Appendix 3Appendix 4FootnotesAbout United Health Foundation
The Behavioral Risk Factor Surveillance System (BRFSS) is the largest and longest running state-based data collection tool in the United States. The telephone-administered surveys collect data on health-related risk behaviors, chronic health conditions, and use of preventive services in all 50 states, the District of Columbia, Guam, and Puerto Rico. [xiv] The validity and reliability of estimates produced by BRFSS data has been repeatedly confirmed through comparisons to other national household surveys. [xv]
Using BRFSS data for the five chosen unhealthy behaviors: smoking, obesity, physical inactivity, excessive drinking, and insufficient sleep, dichotomous variables were created representing the presence or absence of each behavior. A composite variable was then created to model the combination of the five unhealthy behaviors. The composite variable was defined as the total number of unhealthy behaviors, ranging from zero to five. The proportions of zero to five unhealthy behaviors were examined at the national level. At the state and subpopulations levels, the multiple unhealthy behaviors (MUBs) variable was collapsed into two categories, having three or more unhealthy behaviors or having less than three unhealthy behaviors, to allow for adequate sample size in the reporting of proportions by state and subpopulation.
The association between the number of unhealthy behaviors as an indicator variable and self-reported fair or poor health status was examined using logistic regression. The outcome variable, general health status, was dichotomized into 1=“low” (fair or poor health status) and 0=”high” (good, very good or excellent health status). The results are reported as prevalence odds ratios, which has been shown to overestimate the strength of the association. [xvi] For this reason, confidence intervals are included with estimates of association and sample sizes to improve the accuracy of interpretation.
The following covariates were included in the regression analysis to evaluate their effect on the association between MUBs and self-reported fair or poor health status: sex, age, race/ethnicity, education level, and income. All responses coded as “don’t know,” “not sure,” “refused,” or “missing” were excluded for all variables.