BACKGROUND: Drugs reimbursed through a single-party payer such as health maintenance organizations or provincial governments are generally captured in administrative data if they have full-benefit status on that payer's formulary. However, drugs subject to restrictive drug coverage policies are often not fully captured if patients receive these drugs through mechanisms other than the single-payer formulary. OBJECTIVE: The goal of this study was to estimate the association between restrictive drug coverage and drug exposure misclassification across the Canadian provinces of Manitoba and Saskatchewan, which provide universal coverage for formulary-approved drugs to all citizens regardless of age or socioeconomic status. METHODS: Monthly dispensations were compared for 75 drugs between 2005 and 2008 from Canada's National Prescription Drug Utilization System database, which captures provincial drug formulary claims only, versus the IMS Brogan CompuScript Database, which captures all drug dispensations irrespective of formulary status. The association between restrictive drug coverage and drug exposure misclassification was measured using generalized estimating equations and multivariable adjustment. RESULTS: On average, 84% of monthly retail drug dispensations were captured by provincial claims data: 100% of monthly dispensations were captured for drugs with full-benefit status but only 61% of dispensations for drugs with restrictive drug coverage (adjusted risk ratio = 0.65 [95% confidence interval, 0.56-0.75]). The direction and magnitude of the potential misclassification bias between full-benefit and restricted policy drugs were consistent across all drug classes examined: acid-reducing drugs (97% vs 66%), analgesics (89% vs 64%), central nervous system drugs (103% vs 61%), cardiovascular drugs (100% vs 57%), diabetes drugs (98% vs 61%), osteoporosis drugs (96% vs 57%), and respiratory drugs (112% vs 60%). CONCLUSIONS: Drugs subject to restrictive coverage policies are substantially under-captured in administrative databases, leading to potential drug exposure misclassification in pharmacoepidemiologic studies relying on administrative databases. Pharmacoepidemiologic studies should clearly describe whether evaluated drugs are available as full benefits or subject to restrictive coverage policies and the potential impact on their results.
BACKGROUND: Drugs reimbursed through a single-party payer such as health maintenance organizations or provincial governments are generally captured in administrative data if they have full-benefit status on that payer's formulary. However, drugs subject to restrictive drug coverage policies are often not fully captured if patients receive these drugs through mechanisms other than the single-payer formulary. OBJECTIVE: The goal of this study was to estimate the association between restrictive drug coverage and drug exposure misclassification across the Canadian provinces of Manitoba and Saskatchewan, which provide universal coverage for formulary-approved drugs to all citizens regardless of age or socioeconomic status. METHODS: Monthly dispensations were compared for 75 drugs between 2005 and 2008 from Canada's National Prescription Drug Utilization System database, which captures provincial drug formulary claims only, versus the IMS Brogan CompuScript Database, which captures all drug dispensations irrespective of formulary status. The association between restrictive drug coverage and drug exposure misclassification was measured using generalized estimating equations and multivariable adjustment. RESULTS: On average, 84% of monthly retail drug dispensations were captured by provincial claims data: 100% of monthly dispensations were captured for drugs with full-benefit status but only 61% of dispensations for drugs with restrictive drug coverage (adjusted risk ratio = 0.65 [95% confidence interval, 0.56-0.75]). The direction and magnitude of the potential misclassification bias between full-benefit and restricted policy drugs were consistent across all drug classes examined: acid-reducing drugs (97% vs 66%), analgesics (89% vs 64%), central nervous system drugs (103% vs 61%), cardiovascular drugs (100% vs 57%), diabetes drugs (98% vs 61%), osteoporosis drugs (96% vs 57%), and respiratory drugs (112% vs 60%). CONCLUSIONS: Drugs subject to restrictive coverage policies are substantially under-captured in administrative databases, leading to potential drug exposure misclassification in pharmacoepidemiologic studies relying on administrative databases. Pharmacoepidemiologic studies should clearly describe whether evaluated drugs are available as full benefits or subject to restrictive coverage policies and the potential impact on their results.
Authors: Vasily Giannakeas; Suzanne M Cadarette; Joann K Ban; Lorraine Lipscombe; Steven A Narod; Joanne Kotsopoulos Journal: Br J Cancer Date: 2018-11-13 Impact factor: 7.640