Demand and Capacity Management for Medical Practices

This thesis on tactical demand and capacity management for medical practices consists of four main parts. In the first part, we analyze the general planning and control decisions that need to be taken by a practice manager when opening and then running a medical practice. We further present a best-case data set containing all relevant information on interactions between patient and practice. We compare several real-world appointment data sets to this best-case data set, commenting on the consequences of not collecting specific data. We discuss the fundamental problem of defining model parameters from data and give recommendations for modelers and practitioners to bridge the gap between theory and practice.

In the second part, we present a flexible analytical queueing model to investigate the relationship between the physician's daily capacity, the panel size, and the distribution of indirect waiting times of patients. Essential features of the basic model are the consideration of queue length-dependent parameters such as the appointment request rate, the no-show probability, and the rescheduling probability. We present several extensions to the basic model, including the consideration of queue length-dependent service times. ... mehrFinally, we investigate the model behavior by conducting extensive numerical experiments.

In the third part, we propose deterministic integer linear programs that decide on the intake of new patients into panels over time, considering the future panel development. Here, we minimize the deviation between the expected panel workload and the physician's capacity over time. We classify panel patients and define transition probabilities from one class to another from one period to the next. Experiments are conducted with parameters based on real-world data. We use the programs to define upper bounds on the number of patients in a patient class to be accepted in a period through solving the programs several times with different demand inputs. When we use those upper bounds in a stochastic discrete-event environment, the expected differences between workload and capacity can be significantly reduced over time, considering several future periods instead of one in the optimization. Using a detailed classification of new patients decreases the expected differences further.

In the last part, we present further integer linear programs to decide on the intake of new patients. For example, we consider several physicians with overlapping panels and capacities as decision variables. Last but not least, we investigate how the queueing model and the panel management programs could be combined.