An integrated continuity of care measure improves performance in models predicting medication adherence using population-based administrative data

Objectives Continuity of care (COC) is considered an important determinant of medication adherence based on measures such as the usual provider continuity index (UPCI) that are derived exclusively from physician visit claims. This study aimed to: a) determine if high UPCI values predict physicians who deliver different clinical services; and b) compare UPCI with an integrated COC measure capturing physician visits, prescribing, and a complete medical examination in a multivariable model of patients receiving statin medications. Methods This was a retrospective cohort study of new statin users between 2012 and 2017 in Saskatchewan, Canada. We calculated sensitivity/specificity of a high UPCI value for predicting physicians who were prescribers of statins and/or providers of complete medical examinations. Next, we used logistic regression models to test two measures of COC (high UPCI value or an integrated COC measure) on the outcome of optimal statin adherence (proportion of days covered >= 80%). The DeLong test was used to compare predictive performance of the two models. Results Among 55,144 new statin users, a high UPCI was neither a sensitive or specific marker of physicians who prescribed statins or performed a complete medical examination. The integrated COC measure had a stronger association with optimal adherence [adjusted odds ratio (OR) = 1.56, 95% confidence interval (CI) 1.50 to 1.63] than UPCI (adjusted OR = 1.23, 95% CI 1.19 to 1.28), and improved predictive performance of the adherence model. Conclusion The number of physician visits alone appears to be insufficient to represent COC. An integrated measure improves predictive performance for optimal medication adherence in patients initiating statins.