Today both Jasmijn Manders and Nicky Schuermans succefully defended their PDEng theses. Jasmijn's project focuses on the invitation strategy for colon cancer screening:
Each year 2.2 million clients are invited for the Dutch colon cancer screening program. An invitation consists of a self-test which results in either a positive or a negative result. In case of a positive (undesirable) result the client should get an intake-appointment in a nearby hospital within 15 days after submission of the result. Invitations are send based on available capacity and hospital service areas such that a future intake-appointment can guaranteed. Until now these hospital service areas were constructed manually based on gut feeling and trial and error. As a result, linking clients in municipalities to capacity of hospitals was inefficient. In this research we develop a prototype algorithm that determines the optimal hospital service areas for the entire Netherlands, taking into account possible future intake-appointments in case of a positive test result. Via a Mixed Integer Linear Program we maximize the number of clients linked to the nearest hospital and minimize the total travel time for clients, while satisfying the limited capacity constraints of the hospitals. With these optimal hospital service areas we can invite all clients of which 83% to the nearest hospital and 95% within a travel time of 30 minutes. The optimal hospital service areas for 2020 are currently used in practice where we anticipate to see a decrease of rescheduling intake-appointments. We are developing a software architecture that can be added to the current IT-software of Bevolkingsonderzoek.
Nicky's project focuses on the forecasting of the hospital's bed census:
Forecasting the bed census is of utmost importance for hospitals to optimally assign nurses to wards, especially in light of the considerable shortage of nurses in Dutch hospitals. We develop a tactical model for both surgical and internal departments that allows us to forecast the bed census three months ahead, which enables rostering of nurses to follow the predicted workload at the nursing wards. The characteristics of the flow of patients in the surgical and internal chain differ considerably. The majority of patients in the surgical chain have an elective surgery and are scheduled in operating room sessions that are scheduled three months ahead and may therefore be used as input for the bed census predictions. In contrast, the majority of patients in the internal chain do not undergo surgery, but originate from the emergency department or are admitted due to an appointment at the outpatient clinic. Our model considers the combined flow of surgical and internal patients. This research is done in collaboration with Rijnstate, a large hospital in the Netherlands, where the results are incorporated in the planning environment.