Statin adherence among patients with hypertension in Swedish primary health care: a QregPV cohort study using explainable machine learning

AIMS: Evidence regarding statin adherence among patients with hypertension in primary care is limited. We assessed statin adherence in Swedish primary care and examined associations between patient characteristics and adherence. METHODS AND RESULTS: In this retrospective register-based cohort study from a large Swedish region, data from all primary health care centres were linked to data on comorbidities, dispensed drugs, and socioeconomic factors. Patients with hypertension who initiated statin therapy in 2012-2015 were included. Adherence during 2 years after initiation was measured using the proportion of days covered (PDC). Explainable machine learning with XGBoost and SHAP values identified baseline characteristics associated with adherence. Associations were estimated using adjusted multilevel regression and presented as odds ratios (ORs) and median odds ratios (MORs). Among 30 497 patients (mean age 66 years; 48% women), the mean 2-year PDC was 73.2%, while 60.2% achieved adequate adherence (PDC > 80%). Adequate adherence differed by indication (P < 0.001): 54.0% in primary prevention, 61.3% in diabetes without cardiovascular disease, and 65.5% in secondary prevention. The five most important characteristics associated with adequate adherence according to SHAP values were dispensed antithrombotic therapy (OR 1.38, 95% CI 1.26-1.50), primary health care centre (MOR 1.16, 95% CI 1.12-1.20), 10-year increase in age (OR 1.14, 95% CI 1.12-1.17), initiation with atorvastatin (OR 1.33, 95% CI 1.24-1.44), and countries of birth outside Sweden (OR range 0.73-0.78). CONCLUSION: Only 60% of patients achieved adequate statin adherence, which varied substantially by indication. Patient- and provider-level factors identified could guide targeted interventions to optimize preventive cardiovascular care.