Healthcare Expenditure Prediction for Crowd with Co-existing Medical Conditions

Healthcare has been one of the most important issues in modern society. However, expenditure on healthcare is always a high burden for both individuals and countries. Aiming to take full advantage of the limited budget of healthcare expenditure, many medical institutions resort to estimate the future healthcare cost and carefully decide their spending plans. This enlightens us that the accurate prediction of healthcare cost in the future contributes to the effective utilization of the medical resources. In this paper, we propose a healthcare expenditure prediction system for the crowd with co-existing medical conditions. We analyze the diagnoses and payment records, and leverage a gated recurrent unit (GRU) network model to learn the contextual features which would contribute to the prediction. Evaluations over diagnoses from 37,734 patients and the financial report from a real-world medical institution indicate that our system can provide dependable prediction of healthcare expenditure and it is practical to use our system to assist the medical budget decision.