Instrumental variable-based assessment of the effect of psychotherapy on suicide attempts, health, and economic outcomes in schizophrenia.

The literature on the effect of psychotherapy for schizophrenia is limited and characterized by small samples and possible bias from risk selection. We examined the effects of psychotherapy on suicide attempts and health and economic outcomes using an instrumental variable (IV) approach that exploits the variation in the propensities of hospital departments to prescribe psychotherapy. This was supplemented with naïve probit models as exogeneity could not be ruled out for all of the outcomes. The validity of the instrument was examined by distributional plots and various tests. The assumed randomness in referring patients to providers with high versus low propensities to psychotherapy appeared to be a critical aspect. Splitting the sample into homogeneous provider types did not substantially alter the results. Based on the IV results, we found no support for the effect of psychotherapy on suicide attempt, psychiatric readmission, assisted living, or labor market attachment. However, we cannot rule out smaller effects due to confidence intervals including the probit estimates. The main contribution of this study is new evidence on a broad range of outcomes and a large and representative population.