Data-Driven Approach To Patient Flow Management And Resource Utilization In Urban Medical Facilities

Abstract

Healthcare services are tightly connected with complex data analysis techniques to enable optimal resource allocation in medical institutions. This paper proposes a detailed analysis of incoming patient flow to local polyclinic by integrating clustering techniques, process mining and a concept of self-organizing systems. The study takes into account concepts based on models of managing social networks, the participants of which today can be both people and intelligent software. How could patient flow model be developed using a clinical pathways approach that combines clinical pathways tool, social media analysis, hierarchical agglomerative clustering method and probabilistic topic modeling to investigate the optimal resource utilization of medical facility? The methodology to answer this research question was demonstrated using a time- series clustering (kmedoids, Ward's method, Latent Dirichlet Allocation, Additive Regularization of Topic Models), Naive Bayes classifier based on public real data of 64668 depersonalized patient- doctor of 32 specialties conversions. In this paper, a modeling methodology for heterogeneous patient flow segmentation is proposed. The presented approaches serve as the foundation for the further development of a queuing system model of a medical institution. In addition, the shared economy principles are applied by the development of such service that would reduce the workload of appointments to therapists by matching patients to needed doctors.