Models for the main analyses are described in Table 1; Tables 2 and 3 summarize these results. These analyses cover the extended time period of 1990–2010, and the fully adjusted models include additional covariates (demographic factors, and additional state covariates). In the first partially adjusted model (Model 1), similar to the prior report (Anderson et al., 2014), the overall association between medical marijuana policy and suicide risk suggested a statistically significant protective effect (OR=0.956; 95% CI: 0.923, 0.992; p=.02), particularly among men (OR=0.956; 95% CI: 0.929, 0.984; p=.002). However, after addition of demographic covariates (Model 2), the magnitudes of the estimates were slightly reduced and no longer nominally significant. In the fully adjusted Model 3, which included the additional state characteristics, the odds ratio was exactly 1 (OR=1.000; 95% CI: 0.956, 1.045; p=.98). Likewise, the association was not statistically significant in the sex-stratified analyses after adjusting for demographic variables, and odds ratios were very close to 1 in the fully adjusted models (OR=0.996; 95% CI: 0.951, 1.043; p=.87 for men; OR=1.011; 95% CI: 0.948, 1.078; p=.74 for women; Table S43).
