UBC Theses and Dissertations
Dynamic patient scheduling for a diagnostic resource Patrick, Jonathan
We take an in-depth look at the scheduling of patients for a diagnostic resource. Our aim is to schedule patients in order to maintain reasonable waiting times for minimal cost. We assume a fixed capacity with stochastic demand coming from multiple priority classes. We further assume that it is the lower priority patients that must be booked first, therefore requiring the resource manager to implement a booking policy to assure room for later arriving higher priority patients. If too much capacity is reserved for higher priority patients then there will inevitably be unused capacity resulting in longer waiting times than might otherwise be the case. If not enough capacity is reserved then higher priority patients will have to be served through overtime or forced to wait longer than recommended. We begin, in Chapter 1, with an international overview of efforts to reduce waiting times. Chapter 2 proposes a scheduling policy assuming only two priority classes and a fixed limit on expected overtime. The higher priority class are inpatients who must be served the day the demand for a scan is placed. The lower priority class consists of outpatients who may be booked weeks in advance. We present a model that gives the optimal reservation policy and examine the benefit of introducing some flexibility into the higher priority (inpatient) class. Chapter 3 then restricts the modeling to outpatients where demand is divided into multiple priority classes. We present a Markov Decision Process that we solve through approximate dynamic programming in order to derive an approximately optimal booking policy that maintains reasonable waiting times for minimal cost. Chapter 4 presents some strong theoretical results as to the nature of the optimal policy from chapter 3 as well as providing bounds on the "deviation from optimality" associated with our approximation. Chapter 5 then adds inpatients to the model in chapter 3 and compares the results of the full model to those given in chapter 2. Finally, we conclude with possible enhancements to the model and policy insights for the resource manager.