Behavioral pattern analysis for adherence in clinical trials using sequence mining

• We transform raw clinical visit logs into behavioral sequences for adherence analysis. • cSPADE sequence mining identifies patterns linked to both dropout and completion. • Event encoding merges visit type, rescheduling, and timing into interpretable tokens. • Sunburst visualizations reveal subgroup trajectories and distinct risk behaviors. • The framework supports decision-making in resource-limited clinical trial settings. • Repeated rescheduling and visit modality predict dropout, highlighting high-risk groups. Clinical trials play a crucial role in evaluating the safety and efficacy of drugs and therapies before market approval. To ensure their intended outcomes, these interventions undergo multiple phases aimed at validating or rejecting them efficiently. A critical challenge in this process is maintaining patient adherence throughout all trial phases, as consistent participation is essential to generate reliable results. This study analyzed a dataset capturing patient behavior across multiple clinical visits, with particular emphasis on scheduling dynamics such as appointment rescheduling and time-of-day variations. Using the cSPADE algorithm, the behavioral subsequences were extracted and visualized through the Sunburst method, enabling the identification of patterns that are often obscured in traditional tabular data. The findings highlight the importance of flexible scheduling and minimizing rescheduling, as frequent appointment changes were associated with higher patient dropout rates. This sequence mining approach offers valuable information on adherence-related behaviors, paving the way for improved strategies to enhance patient retention in clinical trials. This methodological approach can inform future protocol design and operational decision-making in real-world clinical research settings.