Validation of patient reports, automated pharmacy records, and pill countswith electronic monitoring of adherence to antihypertensive therapy

OBJECTIVES. To evaluate the validity of patient report, pharmacy dispensing records, and pill counts as measures of antihypertensive adherence using e lectronic monitoring as the validation standard. METHODS. The study was conducted among 286 members of Harvard Pilgrim Healt h Care, a managed care organization, who were at least 18 years of age, on monotherapy for hypertension, and had prescription drug coverage. Prescript ion refill adherence during the 12 months before enrollment was determined from their automated pharmacy dispensing records. Participants were intervi ewed about their medication adherence before and after a 3-month electronic monitoring period during which pill counts were also performed. Adherence to both recommended number and timing of doses was estimated from electroni c monitoring data. Data analysis was based on statistical correlation and a nalysis of variance. RESULTS. Electronic adherence monitoring revealed that the proportion of pr escribed doses consumed was higher (0.92) than the proportion of doses take n on time (0.63). The correlation between adherence to quantity and timing of doses was 0.32. Concurrent pill counts and earlier refilling patterns we re moderately correlated with electronic monitoring (pill count: r = .52 wi th quantity and r = .17 with timing; refill adherence r = .32 with quantity and r = .22 with timing). There was considerable misclassification of adhe rence reported by patients, although nonadherence was generally accurately reported. CONCLUSIONS. Deviation from recommended timing of doses appears to be great er than from prescribed number of doses. Pharmacy dispensing records demons trate predictive validity as measures of cumulative exposure and gaps in me dication supply. Adherence levels determined from pill counts and pharmacy dispensing records correlate more closely with quantity than with timing of doses. Nonadherence reported by patients can serve as a qualitative indica tor and predictor of reduced adherence.