Dr Galit Yom-Tov, Industrial Engineering and Management Technion-Israel Institute of Technology (IE&M)
Previous research has shown that placing patients in non-dedicated medical units increases patient length of stay (LOS) and mortality. However, in many cases, the “optimal” LOS and the best location for treatment of the patient is not obvious. A case in point are haematology patients, for whom this is a critical decision. These patients are hospitalised on a regular basis for chemotherapy treatments and it is debated whether these treatments should be offered during hospitalisation followed by an observation period, or whether these patients are better treated in outpatient clinics and should stay at home unless infection is suspected.
Patients with haematological malignancies are susceptible to life-threatening infections after chemotherapy. Hence, LOS optimisation of haematology patients must balance the risk of patient infection with mortality. The former is minimised by minimising hospital stay, while the latter is reduced through hospital care, where infections are identified and treated earlier.
We analyse longitudinal patient data of a large tertiary hospital and evaluate whether management of such patients in dedicated inpatient and emergency wards could provide superior infection prevention and better outcomes. Electronic medical records were analysed to retrieve infection related information and on this basis, patient-tailored risk prediction models were developed. These models help us to empirically compare different patient management policies.
We then developed a Markov Decision Process formulation to explore the connection between these non-monotonic risk functions and the optimal LOS from a single patient perspective. We further consider the social problem in which capacity constraints limit the ability of hospitals to keep patients for their optimal LOS. We find that the optimal policy under this constraint takes the form of a two threshold policy. This policy either blocks some of the patients and immediately routes them to home care observation (under the outpatient clinic guidance), or speeds up some of the patients and routes them to home care after an observation period in the hospital.
Our optimisation model assumes general hazard rate functions, and therefore, can capture the changing dynamics inpatient health through time. It can help guide physician decisions in real-time, as patient state evolves.
Galit Yom-Tov is an Assistant Professor at IE&M faculty of the Technion. Her research focuses on operations of service systems, in particular, healthcare and contact centres. Dr Yom-Tov is the Co-Director of the Service Engineering Enterprise Lab (SEE-Lab) a worldwide hub for research and teaching in Service Engineering. Her research aims to build models for understanding the impact of customer and agent behaviour on service systems and to incorporate these behaviours into operational models of such systems. Her multidisciplinary research approach applies a combination of data science and stochastic modelling to archives of digital traces from service systems. Her recent work used such data to study the dynamics of customer emotions in contact centres, the reaction of customers to waiting announcements in emergency departments, as well as other aspects of service delivery. Dr Yom-Tov has published her work in leading operations research journals, including Management Science, Manufacturing & Service Operations Management, Operations Research, and Stochastic Systems.