Antihypertensive adherence and outcomes among community-dwelling Medicare beneficiaries: the Atherosclerosis Risk in Communities Study

Rationale, aims, and objectives: Despite proven benefits for reducing incidence of major cardiac events, antihypertensive drug therapy remains underutilized in the United States. This analysis assesses antihypertensive drug adherence, utilization predictors, and associations between adherence and outcomes (a composite of cardiovascular events, Medicare inpatient payments, and inpatient days). Methods: The sample consisted of Atherosclerosis Risk in Communities Study cohort participants reporting hypertension without prevalent cardiovascular disease during 2006 to 2007 annual follow-up calls. Atherosclerosis Risk in Communities records were linked to Medicare claims through 2012. Antihypertensive medication adherence was measured as more than 80% proportion days covered by using Medicare Part D claims. Standard and hierarchical regression models were used to evaluate adjusted associations between person characteristics and adherence and between adherence and outcomes. Results: Among 1826 hypertensive participants with Part D coverage, 31.5% had no antihypertensive class with more than 80% proportion days covered in the 3 months preceding the report of hypertension in 2006 to 2007. After adjustment for confounders, positive predictors of use included female gender and diabetes; negative predictors were African-American race and current smoking. Adjusted association between receiving no therapy and a composite endpoint of cardiovascular outcomes through 2012 was not statistically significant (hazard ratio: 0.93; 95% confidence interval: 0.72, 1.22) nor was the adjusted association with Medicare inpatient days or payments (incremental difference at 48 months in payments: $1217; 95% CI: −$2030, $4463). Conclusions: Despite having medical and prescription coverage, nearly a third of hypertensive participants were not adherent to antihypertensive drug therapy. Differences in clinical outcomes associated with nonadherence, though not statistically significant, were consistent with results from randomized trials. The approach provides a model framework for rigorous assessment of detailed data that are increasingly available through emerging sources.