Body mass index identified as an independent predictor of psychiatric readmission

Background: Psychiatric hospital readmissions correlate with illness severity, drug selection, and compliance with treatment in the outpatient setting. The risk factors for psychiatric rehospitalization have been mainly assessed in databases lacking information regarding somatic comorbidity and anthropometric variables, such as body mass index (BMI), which are known to predict readmissions in nonpsychiatric settings. Objective: To determine independent predictors of 1-year readmission occurring among unselected adults consecutively admitted for treatment of severe mental illness to an academic, freestanding psychiatric hospital in New York City from August 2010 through January 2011. Method: After identifying univariate correlates of readmission, we used logistic regression with backward elimination to identify independent predictors of readmissions within 1 year after the index psychiatric hospitalization. Results: Among 224 (23.7%) of 945 readmitted patients, psychiatric readmission was significantly associated with age (P = .0029), length of stay (P = .036), schizophrenia/schizoaffective disorder (P < .0001), dementia (P = .027), major depressive disorder (P = .0006), treatment with atypical antipsychotic drugs (P = .0054), electroconvulsive therapy (P < .0001), and BMI (P = .0079), but not with physical comorbidities and routine laboratory data. The independent predictors of readmission were higher BMI (median = 28.5 kg/ m²; odds ratio [OR] = 3.6; CI, 1.2-10.6), a diagnosis of schizophrenia/schizoaffective disorder (OR = 2.2; CI, 1.5-3.4), clozapine treatment (OR = 2.8; CI, 1.1-6.9), no electroconvulsive therapy (OR = 0.13; CI, 0.02-0.45), and shorter length of stay (median = 18 days; OR = 0.08; CI, 0.01-0.42). Conclusions: Body mass index was identified, for the first time, as an independent predictor of psychiatric rehospitalization. Enhanced outpatient treatment programs for overweight and obese psychiatric patients might influence readmission rates and should be explored in prospective studies.