[13]. Fourth, the NVDRS, as noted, is limited to 17 states and may not be nationally representative [13]. Fifth, the time periods between NESARC and NVDRS differ. A separate analysis of National Survey on Drug Use and Health (2002-2011) data [30] showed that average drinking trends for men and women did not change over time (Appendix 5). Sixth, NESARC and NVDRS were not matched by dates of death or interview; however, we compared the prevalence of BAC+ and alcohol use by day of the week and month of the year and found parallel trends (Appendix 6). Lastly, results may be subject to residual confounding or potential effect modification due to unmeasured health conditions or other substance abuse; however, adjustment for alcohol problem in the models had a relatively small impact on the estimates.
