Background:
Medication adherence is extremely important in the treatment of hypertension. Most medication adherence metrics are based on refill rates from pharmacy claims databases and require at least two drug dispensings for the calculations. Little is known about individuals who demonstrate early non-adherence to antihypertensive medications, either by never filling a prescription (primary non-adherence) or by filling a prescription only once (early non-persistence).
Methods:
We conducted a retrospective cohort study of 6,393 Kaiser Permanente Colorado enrollees with hypertension who received a first prescription for an antihypertensive medication between January 1, 2007 and June 30, 2008. We linked prescription orders to medication dispensings and then classified patients into primary non-adherent, early non-persistent, or ongoing dispensings groups. Multivariate logistic modeling was performed to investigate potential predictors of primary non-adherence compared to ongoing dispensings.
Results:
Five percent (331 of 6,393) of patients were primarily non-adherent, 26% (1,672 of 6,393) demonstrated early non-persistence, and 69% (4,390 of 6,393) received ongoing dispensings. In a multivariate model, the following characteristics were associated with a higher likelihood of exhibiting primary non-adherence compared to receiving ongoing dispensings: Hispanic (OR 1.74, 95% CI 1.20-2.52) or other racial/ethnic minority (1.48, 1.13-1.95), an enrollment period less than 10 years (1.28,1.00-1.62), more than four comorbidities compared to no comorbidities (1.76, 1.02-3.02), and fewer than five health care contacts in the six months after the medication was prescribed (1.31, 1.02-1.67). The copayment amount for office visits or medications was not associated with primary non-adherence. The c-statistic for the model containing demographic characteristics, comorbidities, and health care plan information was 0.63.
Conclusions:
A substantial number of individuals newly prescribed an antihypertensive medication exhibited early non-adherence. Our multivariate model had limited discrimination, indicating that further research is needed to better characterize these individuals and explore barriers to early adherence.