Neural-Network-Based Resource Planning for Health Referrals Creation Unit in Care Management Organizations

摘要

Care Management Organizations (CMOs) provide a wide array of healthcare services to their members. Post-discharge healthcare services (e.g., health referrals creation) are an important example of their services; assuring inpatients get their needed health services immediately after being discharged is very critical to reduce readmission rates. Therefore, successful resource planning at the CMO can result in the creation of health referrals in a timely manner. This study proposes a neural-network-based resource planning framework. The framework is composed of prediction and optimization. First, the demand quantity (number of referrals to be created by a specific date) and type (post-discharge destination) of the CMO referral creation unit are predicted. Then, a resource planning method is deployed to optimize the referral creation unit resource allocation given the predicted demand and unit limited resources. To increase the integrity of the framework, a neural-network approach is suggested (e.g., Boltzmann machine). The results of prediction models demonstrate the feasibility of utilizing real data to predict the demand of the referral creation unit of CMOs. This real data-driven demand prediction will increase the validity of the resource planning model.