Abstract P385: Comparison of Different Algorithms to Measure Adherence to Multiple Anti-Hypertensive Medications Using Administrative Claims Database

Background: Adherence to antihypertensive medication is a key component of hypertension management. Insurance claims data provide an opportunity to examine and monitor patients’ adherence rate to prescribed medications. However, patients may take multiple drugs or drug classes simultaneously to control high blood pressure. Little is known regarding the differences between algorithms used to measure adherence to antihypertensive medications using claims database. Hypothesis: We hypothesize that different algorithms, number of drug classes and observation periods lead to different medication adherence estimates. Methods: Using the 2017-2019 IBM MarketScan Medicare supplement claims database, we conducted retrospective analysis of 107,194 insured members with hypertension diagnosis and were taking anti-hypertensive drugs (e.g., angiotensin II inhibitors, ACE inhibitors, Beta Blockers, Calcium Channel Blockers, Antihypertensive combinations, Alpha Blockers, Diuretics). We calculated medication adherence rates using the proportion of days covered (PDC) for 6 months and 12 months among polytherapy users who use more than one antihypertensive drug classes. Three PDC algorithms were used and compared, including (1) number of days covered by at least 1 antihypertensive medication class (“maximum”) in 6 months and 12 months; (2) number of days covered by all antihypertensive medication classes after they are prescribed in 6 months and 12 months (“minimum”); (3) the average number of days covered by each medication class measured separately in 6 months and 12 months (“average”). Results: The proportion of patients who took two, three, or four or more antihypertensive drug classes was 26.96%, 9.44%, and 3.21% for the 6-month observation period, and was 26.73%, 9.37%, and 3.21% for the 12-month observation period. The mean PDC estimate and standard deviation (SD) using the “maximum” algorithm was 89% (0.2) for 6-month and 88% (0.2) for 12-month observation period, respectively. The mean PDC estimate and SD using the “minimum” algorithm was 45% (0.33) and 45% (0.33) for 6- and 12-month observation period, respectively. And the mean PDC estimate and SD using the “average” algorithm was 68% (0.23) and 67% (0.23) for 6- and 12-month observation period, respectively. Conclusions: There is a lack of consensus on the algorithms to measure adherence to anti-hypertensive medications based on pharmacy claims data. The PDC estimates based on the “average” algorithm was less dependent on the number of antihypertensive drug classes, and the 6- and 12-month observation period did not significantly influence PDC estimates.