Bayes network analysis for data-driven decision support in healthcare
Constantly rising costs in the healthcare sector require economic action without compromising the quality of care. Hospital catering is cost-intensive, but also very significant for patient satisfaction. In addition to purely economic optimization, numerous qualitative factors such as sustainability and employee satisfaction are also of key importance. Of particular interest here are their interdependencies, which are often not directly apparent.
The increasing collection of data in the health sector (including hospital catering) enables a systematic analysis of such questions. In particular, the use of Bayesian networks enables the causal and probabilistic modeling of numerous factors that are in a complex, hierarchical interdependency. Bayesian networks are a particular kind of graph, which allow the representation of variables and their interdependencies in an easily accessible manner – for both computation and interpretation. In the present project, extensive data collected from various hospitals is examined. In addition, simulated data is analyzed to improve the model quality.
The resulting models in the form of Bayesian networks will allow conclusions to be drawn about how various factors influence one another directly or indirectly. In this way, systematic foundations for management decisions can be found and a contribution can be made towards patient-centered and resource-optimized services in health institutions.