Validation of a claims-based antipsychotic polypharmacy measure†

Antipsychotic polypharmacy is common,1–5 despite sparse evidence supporting the practice.6–8 Evidence of efficacy is limited to small randomized controlled clinical trials, case reports, and individual clinician experience.8–13 At the same time, antipsychotic polypharmacy has been associated with an increased risk of metabolic syndrome14–16 and increased healthcare costs,17–21 and may carry an increased risk of mortality.22 Although some individuals may benefit from ongoing antipsychotic polypharmacy, in one clinical trial, a majority of individuals on antipsychotic polypharmacy were able to successfully transition to a single antipsychotic medication.23 Given well-established metabolic and neurologic side effects of antipsychotics, concerns about the increased risks associated with multiple antipsychotics, and high costs of these agents, antipsychotic polypharmacy has been identified as a quality concern by states and national accrediting bodies.24,25 Wide variability has been reported in the prevalence of antipsychotic polypharmacy among those treated with antipsychotics, ranging from 7% to 50%,14,26–29 although most studies report rates between 10% and 30%.6,9 This variation is likely related to differences in the general demographic and clinical characteristics of the study populations, year of the study, variations in study method, treatment setting, and the duration of the study period. For example, antipsychotic polypharmacy was found to be 8% among adolescents in Florida's Medicaid program2, 29% among adults enrolled in a California Medicaid program,28 and over 50% among psychiatric inpatients.30 Some of the variability in reported prevalence estimates likely results from differences in the methods of measuring polypharmacy. Definitions that use narrow time frames of concurrent treatment with multiple antipsychotics may overestimate true prevalence, as they may include instances when a change in medications occurs before the previous prescription has been finished (e.g., within 30 days), or temporary medication overlaps such as those for patients switching medications using a cross-taper.31 Previous research has variously defined polypharmacy to be concomitant therapy lasting over 14,28 60,4 and 90 days.29 In administrative data, observed periods without medication may represent poor adherence, short hospital admissions, use of free medication samples, eligibility gaps, or actual discontinuation of medication. In previous claims-based antipsychotic polypharmacy measures, medication gaps have been operationalized by not allowing gaps,4,31 or allowing gaps of 143 or ≤31 days.28 When measuring long-term polypharmacy, such breaks may represent changes in the antipsychotic regimen, which occur in the context of long-term polypharmacy. A valid measure of antipsychotic polypharmacy is important not only to pharmacoepidemiologists but also to state Medicaid offices, healthcare plans, and pharmacy benefit managers interested in quality of care improvement initiatives. Pharmacy programs use measures to support interventions with individual patients and physicians, including prior authorization before filling prescriptions, retrospective drug utilization review, second opinion consultation requirements, and computer-based decision support tools, like electronic health records and prescribing applications that provide messages to prescribers. As measures may be used to intervene directly in clinical care pathways, it is important to establish that they are clinically meaningful and valid. With medical records as the criterion standard, we assessed the validity of an antipsychotic polypharmacy measure that was developed for a statewide quality improvement initiative. A claims-based measure was developed in an empirically derived and clinically guided manner to enhance the utility of the measure. We examine the demographic and operating characteristics of this measure, and test the impact of alternative definitions of antipsychotic polypharmacy on the prevalence, sensitivity, specificity, and positive predictive value (PPV) of the measure, specifically 14, 60, and 90 days current use, and a gap allowance in days’ supply of individual antipsychotics of 0, 14, and 32 days.