Figure 3. Forest plot of results for key subgroup and sensitivity analyses evaluating the association between antipsychotic adherence and crash responsibility among drivers with schizophrenia in the adherence cohort. X-axis depicts the adjusted OR; squares, the point estimate, with size reflecting the inverse of the SE; horizontal lines, the 95% CI. We collapsed some age and crash-type categories because of low sample size and regression model non-convergence. *This sensitivity analysis sought to exclude individuals for whom antipsychotics were intentionally de-prescribed. **Antipsychotics prescribed in hospital do not appear in PharmaNet or any other accessible province-wide data source. This sensitivity analysis sought to evaluate whether this source of bias in calculating medication possession ratio (MPR) influenced our effect estimate. ***These sensitivity analyses suggest there is no ‘threshold MPR’ above which driving safety improves. ****This sensitivity analysis compared drivers who filled a long-acting injectable (LAI) antipsychotic but were not yet due to receive their next dose (implying perfect adherence at the time of crash) to drivers with an MPR of zero. Overall, the lack of association between antipsychotic adherence and crash responsibility appears to be consistent across subgroup and sensitivity analyses.