Statistical Analyses

Age adjustment is a standard analytical technique used to compare estimates between populations with different age distributions (e.g., between states) and over time. In this report, prevalence estimates were directly age adjusted so that the reader can compare estimates across states and MMSAs with different age distributions. Age adjusted prevalence estimates were standardized to the 2000 projected U.S. population, which is consistent with recommendations from the CDC National Center for Health Statistics (12).

For prevalence estimates among adults aged ≥18 years, three age adjustment categories were used: 18–44 years (standardized proportion: 0.5305), 45–64 years (standardized proportion: 0.2992), and ≥65 years (standardized proportion: 0.1703). For prevalence estimates among adults aged 18–64 years, two age adjustment categories were used: 18–44 years (standardized proportion: 0.6394) and 45–64 years (standardized proportion: 0.3606). For prevalence estimates among adults aged ≥45 years, four age adjustment categories were used: 45–54 years (standardized proportion: 0.3869), 55–64 years (standardized proportion: 0.2504), 65–74 years (standardized proportion: 0.1895), and ≥75 years (standardized proportion: 0.1732). Age-adjusted prevalence estimates are taken from direct responses and are not the results of modeling. Age was imputed for the limited number of persons who were missing data on age. To account for BRFSS’s complex sampling design, all prevalence estimates in this report were calculated using weights and strata in SAS version 9.3 (SAS Institute Inc., Cary, North Carolina) or SAS-callable SUDAAN Version 11 (RTI International, Research Triangle Park, North Carolina). Crude (unadjusted) estimates for each state and MMSA are available on the BRFSS website (13). Most prior BRFSS reports (i.e., those reporting on 2012 survey data and earlier) included crude prevalence estimates rather than age-adjusted prevalence estimates. The age-adjusted prevalence estimates in this BRFSS report should not be directly compared with crude prevalence estimates in most prior BRFSS reports.

This report presents unweighted sample sizes; age-adjusted, weighted prevalence estimates with standard errors; and 95% confidence intervals for the prevalence of chronic health conditions, health-risk behaviors, and use of preventive health care services by state, territory, and MMSA using 2015 BRFSS data. Only MMSAs with ≥500 respondents are included in this report. County-level estimates are not presented in this report. Modeled small area estimates at the county level will be released at a future date.

If the unweighted sample size of any jurisdiction or subpopulation was <50 or if the relative standard error was >30%, the findings were suppressed to avoid unstable estimates. Relative standard error was calculated by dividing the standard error of the estimated prevalence by the estimated prevalence and multiplying by 100 (for percent). Responses coded as “refused” or “do not know” were excluded from the given analysis.