Distinguishing between indications for use of medications in administrative data sources application to pregabalin and gabapentin A contribution from the ConcePTION project
Résumé
Introduction: For drug safety studies in pregnancy, it is important to distinguish the effect of drug from the effect of maternal disease. However, prescription data do not necessarily contain structured information on the indication for the prescription. This is problematic for drugs with multiple indications, such as antiepileptic agents and analgesics. Our objective was to describe an algorithm to disentangle the different indications. We tested it using the example of gabapentinoids (pregabalin/gabapentin), approved in Europe for epilepsy, neuropathic pain, and generalized anxiety disorder.
Material and Methods: The algorithm was developed in six European health care data sources. It identified data components used as markers of prescribing: indication, prescriber specialty, primary care and/or specialized care diagnoses, procedures, reimbursement status, and prescribing/dispensing data. Each component estimated one, several or no reason for prescribing. Results from all components were then aggregated to estimate the indication for each individual. Sensitivity analyses explored multiple assessment windows and values for the codes.
Results: We present results on data from EFEMERIS (a French cohort of pregnant women): 238 pregnancies in which the woman received a gabapentinoid prescription were included. The algorithm indicated that the reason for prescribing was pain in 34% of pregnancies, anxiety in 7%, and epilepsy in 2%. Multiple indications were identified in 18% of pregnancies, 80% of which were pain and anxiety. However, in almost 39% of pregnancies, no reason for prescribing could be detected.
Discussion/Conclusion: The results show it is possible to utilize the richness of large administrative healthcare data sources to improve information on reasons for prescribing. Our algorithm will be tested in five other large population data sources in the coming months.