In Reply Dr Lozano suggests that in our article,1 we failed to take into account the variation in depression prevalence between the various medication user groups and matched nonusers and that our findings were therefore biased.
However, the multivariable logistic regression models we used to estimate the 180-day mortality risk included depression (see Tables 2 and 3 of our article).1 By including depression and other clinical variables listed in the table footnotes in the regression models, we controlled for the variation in characteristics between medication users and nonusers (eg, difference in depression prevalence), allowing us to isolate the association of a particular medication with mortality.
Maust DT, Schneider LS, Kales HC. Estimating Mortality Associated With Antipsychotics and Other Psychotropics—Reply. JAMA Psychiatry. 2015;72(10):1058. doi:10.1001/jamapsychiatry.2015.0675