VU mathematicians predict occupancy levels of Covid-19 patients in hospitals

Mathematicians Ger Koole and René Bekker have developed a model for the National Coordination Center for Patient Distribution (LCPS, Landelijk Coördinatiecentrum Patiëntenspreiding) to predict the occupancy levels of Covid-19 patients in Dutch hospitals.

12/02/2020 | 4:53 PM

Bekker and Koole are experts in the field of healthcare logistics. The two VU scientists, together with Michiel uit het Broek, a postdoc in Groningen, have developed a model to predict occupancy levels of Covid-19 patients in Dutch hospitals. The model statistically predicts arrivals, which subsequently are translated into occupancy levels of hospitals using queuing theory. They started developing the model on 1 November and it is already being used to redistribute patients across hospitals in the Netherlands.

Professor Ger Koole is working on this project three days a week: “It is difficult to predict what will happen in the corona crisis. We predict based on historical data from the past weeks. Every day, all hospitals in the Netherlands provide information about their patient numbers and inflow. We calculate with that data and with the historical data. We always predict one week in advance how many new patients will join, both in the Intensive Care Unit and in the clinic. On this basis, the hospitals transfer patients every day.”

Deferred care
Koole, Bekker and Uit het Broek also look at the impact of Covid-19 on deferred care. Think of all kinds of postponed knee and hip operations, but also cardiovascular operations and oncological care. Many hospitals do not or less perform these due to corona.

Koole: “If a hospital can handle the inflow of Covid-19 patients again, it should not immediately stop transferring patients, because you first want to get that deferred care back up to standard. With our model we prevent the hospital in one region from performing knee operations again, while the hospital in another region cannot even help its heart patients.”

Political decision-making
The long-term predictions from the model are intended for political decision-making. Koole: “For example, we also calculate all kinds of scenarios: what will be the effect on the hospital beds if we celebrate Christmas together? And what will happen if a flu wave comes? Because we have that every year in the Netherlands, but will it be a large or a small flu wave and what are the different consequences?” The three scientists are now building a dashboard to process these long-term predictions and to calculate different scenarios.