Comparison of methods to identify individuals prescribed opioid analgesics for pain

Reem Farjo,Hsou-Mei Hu,Jennifer F Waljee,Michael J Englesbe,Chad M Brummett,Mark C Bicket
DOI: https://doi.org/10.1136/rapm-2023-105164
2024-01-25
Regional Anesthesia & Pain Medicine
Abstract:Introduction While identifying opioid prescriptions in claims data has been instrumental in informing best practises, studies have not evaluated whether certain methods of identifying opioid prescriptions yield better results. We compared three common approaches to identify opioid prescriptions in large, nationally representative databases. Methods We performed a retrospective cohort study, analyzing MarketScan, Optum, and Medicare claims to compare three methods of opioid classification: claims database-specific classifications, National Drug Codes (NDC) from the Centers for Disease Control and Prevention (CDC), or NDC from Overdose Prevention Engagement Network (OPEN). The primary outcome was discrimination by area under the curve (AUC), with secondary outcomes including the number of opioid prescriptions identified by experts but not identified by each method. Results All methods had high discrimination (AUC>0.99). For MarketScan (n=70,162,157), prescriptions that were not identified totalled 42,068 (0.06%) for the CDC list, 2,067,613 (2.9%) for database-specific categories, and 0 (0%) for the OPEN list. For Optum (n=61,554,852), opioid prescriptions not identified totalled 9,774 (0.02%) for the CDC list, 83,700 (0.14%) for database-specific categories, and 0 (0%) for the OPEN list. In Medicare claims (n=92,781,299), the number of opioid prescriptions not identified totalled 8,694 (0.01%) for the CDC file and 0 (0%) for the OPEN list. Discussion This analysis found that identifying opioid prescriptions using methods from CDC and OPEN were similar and superior to prespecified database-specific categories. Overall, this study shows the importance of carefully selecting the approach to identify opioid prescriptions when investigating claims data.
anesthesiology
What problem does this paper attempt to address?