Analyzing longitudinal antidiabetic medication patterns: a data-driven clustering framework

BACKGROUND: Antidiabetic medications play a critical role in supporting glycemic control, which is essential for preventing or delaying diabetes-related complications. While prior studies have often focused on isolated treatment transitions, such as first- or second-line therapy initiation or single events like intensification, few have examined long-term medication trajectories. This study applied a novel time-series clustering approach to identify common antidiabetic medication use patterns among Medicare beneficiaries over a three-year period and to compare demographic and clinical characteristics across patient clusters with similar treatment pathways. METHODS: A retrospective cohort study was conducted among Medicare beneficiaries who initiated metformin monotherapy and received antidiabetic medications during a three-year follow-up period, with a one-year washout period to ensure they were antidiabetic medication-naïve prior to initiation. Five types of transitions (switch, intensification, de-intensification, discontinuation, and re-initiation) were defined based on changes in the number of concurrently used medications to describe longitudinal medication patterns. Dynamic time warping was applied to measure pairwise similarities in medication use sequences, accommodating variations in timing and duration. Partitioning Around Medoids clustering was used to group patients with similar medication use patterns. RESULTS: Among 4,616 eligible patients, 222 distinct medication patterns were identified. The most common patterns included continuous metformin monotherapy (36.3%), recurrent cycles of discontinuation and re-initiation (23.1%), one-time discontinuation (14.2%), switching (3.0%), and intensification (2.8%). These patterns were grouped into five clinically meaningful clusters. Cluster-level comparisons indicated that Non-Hispanic Black patients were more likely to experience discontinuation, whereas female patients were more likely to undergo intensification. Patients who switched or re-initiated therapy tended to be older, had more comorbidities, and higher rates of emergency and inpatient visits. CONCLUSIONS: This study demonstrates the value of combining sequence analysis and unsupervised clustering to reveal complex medication use patterns over time. Understanding these trajectories and associated disparities can support data-driven decision-making and guide personal interventions to improve medication adherence and equitable diabetes care.