Evaluation of the Michigan Behavioral Risk Factor Surveillance System for Changes in Alcohol Consumption Behavior Estimates Due to Methodological Changes *

BACKGROUND:   In 2011, the Behavioral Risk Factor Surveillance System (BRFSS) began sampling from cell phone-only households and weighting by raking to adjust for the recent increase in the proportion of U.S. households that rely on cell phones.  People who use cell phones exclusively tend to be younger, less educated, and have lower incomes, and these demographics also describe a large proportion of excessive drinkers.  By including cell phone-only respondents, it is expected that heavy drinking and binge drinking prevalence estimates will appear to increase, and provide a more accurate representation of the excessive drinking population.  

METHODS:   Data from the Michigan BRFSS (MiBRFSS) were analyzed in three separate methodology groups: 2008-2010, landline-only, poststratified weighting (old); 2011-2012, landline & cell, raked weighting (new); and 2011-2012, landline-only, raked weighting (new landline-only).  Demographic characteristics studied included sex, age, household income, race/ethnicity, education, region type, and marital status.  Prevalence estimates (percent or mean) and asymmetric 95% confidence intervals (CI) were calculated for each type of alcohol consumption (current, heavy, binge, binge frequency, binge intensity, driving after drinking) by methodology group overall and in cross-tabulations with each demographic factor.  Differences were significant at p<0.05 if confidence intervals did not overlap.

RESULTS:   Overall, binge drinking prevalence estimates among Michigan adults increased from 16.6% (95% CI: 15.9-17.3) to 19.4% (95% CI: 18.6-20.3) under the new methodology.  The prevalence of heavy drinking also increased, from 5.4% (95% CI: 5.0-5.8) to 6.6% (95% CI: 6.1-7.1).  Within the 2011-2012 time period, the landline-only binge drinking prevalence estimate (17.4%, 95% CI: 16.4-18.5) was significantly lower than that of the new (landline & cell phone) methodology (19.4%, 95% CI: 18.6-20.3), but not different from the 2008-2010 (landline only) estimate (16.6%, 95% CI: 15.9-17.3).  The same pattern was seen for both current drinking and driving after drinking.

CONCLUSIONS:   With the inclusion of cell phone-only respondents, the reported prevalence of binge and heavy alcohol consumption in Michigan adults increased significantly overall and within demographic groups common among cell phone-only households.  Furthermore, there were no differences in any overall consumption estimates between the 2008-2010 and 2011-2012 landline-only samples, suggesting that the differences between the old and new estimates are not due to increases in these behaviors over time, but to the inclusion of cell phone respondents. Given these data, we can conclude that the increased prevalence estimates reflect a more representative view of the population, and not necessarily a real increase in consumption behaviors.