Three Practical Problems in Healthcare Analytics

Abstract

This thesis investigates three critical problems faced by healthcare service providers and proposes analytical solutions. In the first manuscript, we aim to set wait time targets in a multi-priority patient setting using simulation, statistical regression and supervised machine learning. Using illustrative examples with two and three patient classes, and a clinical study with four patient priority classes, we demonstrate potential managerial and societal savings based on the proposed approach. Numerical analyses show that typically wait time targets are quite low, and thus the additional wait imposed by long wait time targets cannot be justified. In the second manuscript, we investigate scheduling policies to book patients for follow-up appointments with their service providers dynamically. We develop a Markov decision process model to efficiently allocate available capacity to consults and follow-up visits in a dynamic fashion. We solve this model using the linear programming approach to Approximate Dynamic Programming (ADP) and discuss the characteristics of the approximate optimal booking (AOP) policy for multi-class patients with repeat visits. Finally, we compare the AOP policy's performance to that of existing policies through simulation and show the superior performance of the AOP policy over a First Available Slot Policy with booking limits (i.e., a Myopic policy). In the third manuscript, we propose a framework for the drug formulary decision with the help of Multi-Criteria Decision Analysis. We use a recent extension of the UTilities Additives DIScriminantes approach, UTADISᴳᴹˢ and demonstrate the method using the oncology drugs reviewed through pan-Canadian Oncology Drug Review (pCODR) in Canada between 2011 and 2017. Finally, we show the method's prescriptive and predictive ability using an open-source decision support tool. Each paper makes a worthwhile contribution to the healthcare operations literature through the modelling and case studies using real-life data. Also, each essay conclusion has the potential to inform healthcare policy decision making. Approaches presented in the thesis can help managers allocate resources efficiently with proven, analytics-based methods.