Uncovering hidden resource allocation decisions: An application in hospital bed management

Abstract

Developing proper decision-making policies can improve the performance of organizations, but it relies on the study of previous decisions and their consequential effects on performance metrics. While information systems provide large amounts of data stored in distributed databases, often it is not enough data to precisely identify past decisions. Multiple database mining is one solution to extract information about these man-made decisions. In this study, we extend the concept of rule-based entity resolution, in the context of resource allocation, to link two heterogeneous spatio-temporal databases and extract information related to non-automated decisions in a healthcare setting. This problem is motivated by a hospital bed assignment problem and the solution is applied to uncover bed managers’ decisions to assign patients to inpatient wards. We present a polynomial time algorithm that allows for uncovering resource allocation decisions regarding choices among a preferred resource and substitute resources. The algorithm is tested on a variety of simulated databases, with varying underlying parameters, and is applied in a hospital case study. The performance of the algorithm in uncovering the hidden decisions is shown to be very effective and the algorithm is robust to missing data on various system sizes with respect to accuracy, precision, and recall.

Keywords:

• Bed assignment
• decision mining
• multiple database mining
• resource allocation
• spatio-temporal database