2017 年 48 巻 5 号 p. 173-175
Background: Although pill counts can objectively measure medication adherence, they are often labor-intensive and time-consuming, especially in patients taking numerous medications. We hypothesized that counting pills for not all but only some medications, which would undoubtedly spare time and labor, could provide reliable information on adherence to all medications. The aim of the present study was to evaluate how accurately medication adherence could be assessed using partial pill counts.
Methods: We retrospectively analyzed pill count data from 158 consecutive Japanese type 2 diabetic outpatients treated with polypharmacy. Adherence was defined as an 80% or higher medication consumption rate, whereas non-adherence was defined as <80%. We assessed how accurately a patient's adherence to one medication could be predicted based on information on his/her adherence to another medication.
Results: The positive likelihood ratio for adherence (i.e.,reciprocal of the negative likelihood ratio of non-adherence) was 2.4 (95% confidence interval: 1.7 to 3.3) , whereas that for non-adherence (i.e., reciprocal of the negative likelihood ratio of adherence) was 9.9 (95% confidence interval: 7.1 to 13.8) (both p<0.05). The accuracy was dependent on the consistency of administration times. Adherence (or non-adherence) to a medication could be more accurately predicted based on information regarding another medication with the same administration schedule, compared with another with a different administration schedule.
Conclusion: Adherence (or non-adherence) to medications could be predicted with considerable accuracy based on information regarding adherence (or non-adherence) to another medication in type 2 diabetic patients treated with polypharmacy.