Propensity Score Matching
None of the matched cohort groups listed in Tables 3 and 4 (or in Table S1 in the data supplement that accompanies the online edition of this article) had any meaningful standardized differences across any of the continuous variables or any level of the categorical variables used in the propensity score matching process. Hence, we followed
the recommended approach and did not include any of the matching variables in the Cox modeling procedures(26).
Cox Regression
Matched drug and population proxy cohorts . The final Cox models demonstrated that individuals assigned to the methamphetamine group or to the other drug cohorts had a significantly greater risk of readmission with a schizophrenia diagnosis than individuals in the appendicitis or exclusion criteria cohorts (Table 3). The results generated from use of either of the population proxies demonstrated, in each drug group, similarly elevated risks of schizophrenia, similar patterns of statistical significance, and overlapping 95% confidence intervals for the corresponding hazard ratio estimates.
Matched meth amphetamine and other drug cohort analyses . In the final Cox models, individuals assigned to the methamphetamine group manifested a significantly greater risk of readmission with a schizophrenia diagnosis than individuals in the cocaine, alcohol, and opioid drug groups, but no difference in hazard in comparison with the matched cannabis group (Table 4).