Today, hospitals are experiencing great day-to-day variation in demand and capacity at emergency departments, wards and other service functions in general. That leads to overflow situations, which are handled locally at the hospital with patients in hallways and patients in borrowed beds. Consequently, patients experience less attention. Prolonged waiting time is caused by understaffing issues and patients placed in hallways and borrowed beds are due to bed capacity issues resulting in more adverse events. The challenge is to diminish these issues, which is possible by statistically learning about periods of high workload on the hospitals.
The solution is Real Time Demand/Capacity Management, where the primary goals are to diminish unnecessary wait time and to place the right patient in the right bed. The solution rely on a managerial change with capacity conferences and white board meetings. The chosen approach is minimal analysis and maximum hands-on experimentation. In addition, the solution is IT supported with training data generating forecasts in order to deliver trustworthy predictions of workload and insights in “surplus” and “shortage”. The predictions of workload give the hospital staff the necessary overview that enables them to diminish wait time and to place the right patient in the right bed.
The solution is based on machine learning, and the applied method is a model that is able to make forecasts, classifications, predictions, etc. about the workload on various hospital departments. Training data/historical data and algorithms, as well as new unknown data, provide the statistical foundation that enables the model to find patterns automatically from the historical data. These patterns then turn into trustworthy predictions of workload.
The process and used methods can easily be adopted by all hospitals. All management units and other staff members can gain insights into how patient flow can be optimized for the benefit of both hospital personnel and patients. The new knowledge centres around the ability to make predictable forecasts about workload on the basis of patterns from training data and algorithms.
A clear potential of this work is that hospital management and other staff members are able to diminish understaffing issues and bed capacity issues by getting predictable forecasts about the workload on different departments. This will also lead to a reduction in adverse events. Moreover, this benefits the patient to the same extent, because the patient will experience more attention from staff members.
Aarhus University - Dept. of Computer Science
Central Denmark Region
North Denmark Region