Jornal de Ciência da Computação e Biologia de Sistemas

Data Mining Risk Score Models for Big Biomedical and Healthcare Data

Abstract

Emad Elsebakhi, Ognian Asparouhov, Anton Berisha, Kris Latenski, Eric Schendel, Anwar Haque and Rashid Al- Ali

Recently, big data is becoming the key to improve the future healthcare. The era of big biomedical data comes with significant challenges in querying, storage, visualize, and analyze the available petabytes of biomedical data, which makes healthcare industry a data-driven field. Currently, the available Concurrent Risk Model (CRM) is limited to the availability of patient episodes that are sensitivity to its cost. Herein, we propose a novel hierarchical data mining based on functional networks to develop a new CRM. This new risk score evaluates the last twelve-month period of patients’ expected risk/cost/severity/illness burden/disease intervention using both medical and drugs claim-based predictors: diagnoses, medications (yes/no), and demographics. Our novel CRM predicts $50,000 permember- per month (PMPM) tracks risk trends over time for any particular group, especially severe chronic diseases. Our CRM model has R2=0.57 in comparison with the best results of Society of Actuaries.