Accuracy of a provincial prescription database for assessing medication adherence in heart failure patients

British Columbia's central prescription database, PharmaNet, is often used for both clinical and research applications. However, PharmaNet details prescription transactions, not actual medication consumption, resulting in many potential sources of inaccuracy when the information is assumed to reflect population or individual drug utilization.

To assess the accuracy of PharmaNet for adherence assessment in patients with heart failure who are taking beta-blockers.

A 6-month prospective, longitudinal assessment of adherence to the prescribed beta-blocker regimen was carried out using both PharmaNet data and the Medication Event Monitoring System (MEMS) for each patient enrolled. The limit of agreement between the 2 adherence assessment methods was assessed using the Bland-Altman approach.

Fifteen of 58 patients initially enrolled in the study were excluded, most due to misuse of MEMS or failure to return the MEMS vial despite thorough follow-up. For the 43 patients included in the final analysis, mean +/- SD adherence was 97.8 +/- 11.8% when assessed by PharmaNet and 97.1 +/- 7.3% when MEMS was used. However, the limit of agreement, reported as the mean of the differences +/- 2SD, was 6.8 +/- 18.5%, indicating a moderate-to-high level of agreement between the 2 methods when the confidence interval is taken into consideration.

These results suggest that PharmaNet data accurately reflect medication adherence for most patients. The MEMS system proved unreliable in several cases, illustrating the difficulty of identifying a gold standard for adherence assessment.