Abstract
BACKGROUND: Medication adherence, or how patients take their medication as prescribed, is suboptimal worldwide. Improving medication-taking habit might be an effective way to improve medication adherence. However, habit is difficult to quantify, and conventional habit metrics are self-reported, with recognized limitations. Recently, several objective habit metrics have been proposed, based on objective medication-taking data. OBJECTIVE: We aim to explore the correlation between objective habit metrics and objective medication adherence on a large dataset. METHODS: The Medication Event Monitoring System Adherence Knowledge Center, a database of anonymized electronic medication intake data from ambulant participants enrolled in past clinical studies, was used as the data source. Electronic medication intake data from participants following a once-daily regimen and monitored for 14 days or more were used. Further, two objective habit metrics were computed from each participant's medication intake history: (1) SD of the hour of intake, representing daily variability in the timing of medication intakes, and (2) weekly cross-correlation, representing weekly consistency in the timing of medication intakes. The implementation component of medication adherence was quantified using (1) the proportion of doses taken and (2) the proportion of correct days. RESULTS: A total of 15,818 participants met the criteria. These participants took part in 108 clinical studies mainly focused on treatments for hypertension (n=4737, 30%) and osteoporosis (n=3353, 21%). The SD of the hour of intake was significantly negatively correlated with the 2 objective adherence metrics: proportion of correct days (Spearman correlation coefficient, ρ(S)=-0.62, P<.001) and proportion of doses taken (ρ(S)=-0.09, P<.001). The weekly cross-correlation was significantly positively correlated with the 2 objective adherence metrics: proportion of correct days (ρ(S)=0.55, P<.001) and proportion of doses taken (ρ(S)=0.32, P<.001). A lower daily or weekly variability in the timing of medication intakes is thus associated with better medication adherence. However, no variability is not the norm, as only 3.6% of participants have 95% of their intakes in a 1-hour window. Among the numerous factors influencing medication adherence, habit strength is an important one as it explains over 30% of the variance in medication adherence. CONCLUSIONS: Objective habit metrics are correlated to objective medication adherence. Such objective habit metrics can be used to monitor patients and identify those who may benefit from habit-building support.