Report : Framework for Designing Evidence-Based Patient-Physician Matching Protocols

This document presents and evaluates a framework for designing an intelligent system for matching physicians with patients seeking elective procedures. The evaluation consists of using, as a training set, data on surgical orthopedic procedures from one time period and evaluating the system’s utility by comparing its matching recommendations with what actually unfolded in the next time period. The relevant testing and training datasets were obtained from AHRQ’s HCUP program [1] and New York state’s SPARCS inpatient database [2] for the years 2013 and 2014. The contributions of this paper are threefold: (i) we explore metrics that are useful to consider for physicians operating within health systems, especially those health systems switching to new payment models such as value-based payment, (ii) we show the prescriptive power of these metrics over two years for orthopedic surgeons in New Jersey state, and (iii) we show preliminary results supporting the benefit of an intelligent matching system that integrates machine learning techniques, statistical techniques, and stochastic process modeling to propose personalized physician-patient matches. These benefits are expressed in terms of cost and length of stay, which is considered in this report to be a sign of care quality. Technical details of this analysis and the intelligent matching system will be presented in future publications.