Estimating the risk of pneumonia in patients with schizophrenia newly receiving clozapine: A nationwide cohort study

Purpose/Background Patients with schizophrenia are vulnerable to pneumonia. Clozapine is associated with the greatest risk of pneumonia. We investigated the risk factors of pneumonia in patients with schizophrenia who use clozapine. Methods/Procedures We used a large cohort of patients with schizophrenia (N = 22,774) who newly use clozapine (baseline). We divided the data set into a training cohort (entry between 1998 and 2008, n = 18,496) and test cohort (entry between 2009 and 2012, n = 4278), where 483 and 168 patients developed pneumonia requiring hospitalization within 1 year after baseline, respectively. For prediction, we developed a static model using Cox proportional hazards regression and a dynamic model using Cox regression with time-dependent modeling. Areas under receiver operating curves (AUCs) for the predictive model were estimated in the training cohort and then in the test cohort for validation. Findings/Results Based on the baseline characteristics, the static model for predicting pneumonia in 3 periods (90, 180, and 365 days) was unsatisfactory (AUCs, 0.64, 0.64, and 0.65, respectively). The predictors were older age, male sex, history of nonpsychiatric hospitalization, dementia, asthma, and tuberculosis within 1 year before baseline. However, the results were improved (AUCs, 0.83, 0.79, and 0.77, respectively) after control for time-dependent variables, namely, duration of clozapine use and concomitant medications (ie, benzodiazepines, valproic acid, systemic corticosteroids). Implications/Conclusions Several risk factors for predicting subsequent pneumonia after initial use of clozapine were explored, including older age, male, history of nonpsychiatric hospitalization, dementia, asthma, tuberculosis, benzodiazepines, valproic acid, systemic corticosteroids, and the use duration of clozapine. Clinical staff can use the risk factors to administer evidence-based treatment.