RESUMO
BACKGROUND: The effectiveness of long-acting injectable antipsychotics (LAIAs) has been demonstrated in studies using prescription claims data. However, the validity of claims data for LAIAs has not been established. OBJECTIVE: We aimed to validate date dispensed, quantity dispensed and days supplied fields in prescription claims data, and to compare claims- and medical record-derived persistence estimates. METHODS: We evaluated LAIA dispensations in the Drug Programs Information Network prescription claims database from Manitoba, Canada against a random sample of medical records. Adults with one or more LAIA prescription between April 2015 and March 2016 were eligible. Results were stratified by LAIA type (first-generation LAIA, risperidone LAI or paliperidone LAI). Persistence estimates were assessed using Kaplan-Meier survival analysis and proportion of patients covered method. RESULTS: Claims data had high positive predictive value, ranging from 80.0% (95% CI 51.9-95.7) to 100.0% (95% CI 89.7-100.0), but low negative predictive value, ranging from 0.0% (95% CI 0.0-2.5) to 62.5% (95% CI 40.6-81.2). Quantity dispensed and days supplied exactly matched dose and dosing interval, respectively, for 99.7% and 97.1% of risperidone LAI doses, 100.0% and 76.6% of paliperidone doses, and 8.9% and 28.3% of first-generation LAIA doses. There were no significant differences in claims-derived versus medical record-derived persistence estimates. CONCLUSIONS: Quantity dispensed and days supplied provide valid estimates of dose and dosing interval for second-generation LAIAs, but underestimated these parameters for first-generation LAIAs. However, a large proportion of medical record-confirmed doses were missing from claims data, and dose and dosing interval are underestimated in claims data.
