Patient segmentation and resource allocation for tailored healthcare delivery

Healthcare systems often face the challenge of providing quality care to a diverse patient population while effectively utilizing limited resources. While some patients with complex conditions may require specialized care at dedicated clinics, others could benefit from receiving treatments at home, rehabilitation centers, community health centers, or other tailored service programs. Evaluating resource allocation and directing patients to appropriate care settings are particularly important when planning new healthcare initiatives with limited prior data or operational experience. This work introduces a methodology that uses patient segmentation techniques to address the challenges of resource allocation in healthcare settings, specifically focusing on planning new programs that lack prior data or clinical experience by leveraging existing electric patient records. First, we use unsupervised learning, specifically ensemble clustering, to identify distinct groups of patients. Next, an algorithm for rule-based representation of clusters is proposed to generate simple data-driven recommendations that can be applied to practical settings for selecting target patient groups, and lastly, we introduce a resource delivery priority score function that can guide decision-making and patient prioritization under resource constraints. Our methodology is applied to a case study of home transfusion delivery of Red Blood Cell (RBC) products, a proposed program for patients who are required to regularly visit outpatient clinics for receiving transfusions. The results highlight the potential of our methodology in efficient resource allocation and improving patient care outcomes beyond the current heuristic-based approaches in clinical practice.