Many surveys relevant to researchers evaluating the effects of health reform have income measures that lack needed specificity to precisely define the relevant eligible populations, such as those with incomes that make them eligible for Medicaid. Nevertheless, because these surveys are often the best (or only) source for a rich set of measures of health out-comes and health behaviors, researchers have employed various methods in order to best utilize the available income measures to examine potentially eligible populations. The U.S. Centers for Disease Control and Prevention's (CDC) Behavioral Risk Factor Surveillance System (BRFSS) survey is one such resource; an annual household telephone survey of civilian non-institutionalized adults age 18 years or older that asks respondents about health behaviors, chronic health conditions, and the use of preventive health services. The BRFSS has a relatively large sample size, interviewing more than 400,000 respondents annually, and allows researchers to produce estimates for all 50 states and D.C. The BRFSS' income measure asks respondents to report their total annual household income within eight possible categories. Because these categories do not align with the federal poverty guideline (FPG) thresholds used to determine eligibility for programs such as Medicaid expansion (up to 138% FPG) or the Affordable Care Act's (ACA) cost sharing reductions (up to 250% FPG) or premium tax credits (up to 400% FPG), this creates a problem for researchers who want to use the BRFSS to study health reform. To deal with this issue, researchers have typically chosen to assign a continuous income to the respondent based on the categorical income measure, choosing either the lower bound of each category, the upper bound of each category, or the midpoint of each category. There is no clear consensus in the literature about which approach to assigning continuous income from categorical values is best, and we propose that the most appropriate method depends in part on the analytic issue at hand. In this brief, we first outline how each method impacts the income distribution in the BRFSS overall and by state. We then use the Current Population Survey (CPS) to evaluate the impact of using different methods to assign continuous income from a categorical income variable. We chose the CPS because the survey is used broadly to report on income, and contains both a categorical and a continuous income variable. As a result, we can compare the impacts of different strategies of assigning continuous income from a categorical variable to actual, continuous income from the same data source. We then summarize our findings from the CPS and their implications for evaluating the impact of different health reform policies (e.g., Medicaid expansion) on health outcomes in the BRFSS.
Copyright:
Reproduced with permission of the copyright holder. Further use of the material is subject to CC BY-NC-ND license. (More information)