The association between class of antipsychotic and rates of hospitalization: Results of a retrospective analysis of data from the 2005 medicare current beneficiary survey

Abstract Background: When second-generation antipsychotics (SGAs), also called atypical antipsychotics, were introduced in the 1990s, early research suggested that these drugs offered better tolerability and adherence than first-generation antipsychotics (FGAs), or typical antipsychotics. This presumably would reduce the need for hospital services. However, health research to test this hypothesis has focused mostly on psychiatric readmissions. Objective: The objective of this study was to compare rates of all-cause hospitalization among patients receiving different classes of antipsychotics (SGAs, FGAs, both, or neither) in a large, all-ages sample of both institutionalized and noninstitutionalized Medicare beneficiaries. Methods: We examined the 2005 Medicare Current Beneficiary Survey Cost and Use file for 11,236 survey participants. Antipsychotic utilization was characterized in terms of class: FGA (ie, chlorpromazine, fluphenazine, haloperidol, loxapine, perphenazine, thiothixene, thioridazine, or trifluoperazine) or SGA (ie, aripiprazole, clozapine, olanzapine, quetiapine, risperidone, or ziprasidone). Hospitalization was defined in terms of whether a Medicare beneficiary was admitted to the hospital for any reason in 2005, and was measured in terms of the number of hospital visits. In our final model, we included the following confounding variables: disability status (≥1 limitation in activities of daily living), Rosow-Breslau impairment score (difficulty with walking, stooping, crouching, kneeling, or doing heavy housework), cognitive impairment (diagnosis of Alzheimer's disease or memory loss that interfered with daily activity), and health behavior variables (body mass index and smoking status). Results: A total of 3.5% of Medicare beneficiaries (1.3 million) filled ≥1 prescription for an antipsychotic medication in 2005. Controlling for demographic, socioeconomic, health, and disability variables, SGA-only users were more than twice as likely (odds ratio [OR] = 2.2 [95% CI, 1.7–2.9]) and combination users were more than 6 times as likely (OR = 6.3 [95% CI, 2.4–16.2]) as nonusers to be hospitalized. The odds of FGA users being hospitalized were not significantly different from nonusers (OR = 1.4 [95% CI, 0.7–2.8]). Conclusions: This analysis yielded provocative, but by no means conclusive, evidence that SGAs as a class are not necessarily superior to FGAs in mitigating patient's use of hospital services under real-world conditions. Systematic analysis of this relationship with a large, multiple-year sample of Medicare beneficiaries is warranted.