Using an explainable machine learning approach to prioritize factors contributing to healthcare professionals' burnout

Malvika Pillai,Chao Chin Liu,Elizabeth Kwong,Ian Kratzke,Nadia Charguia,Lukasz Mazur,Karthik Adapa
DOI: https://doi.org/10.1007/s10844-024-00862-z
2024-06-14
Journal of Intelligent Information Systems
Abstract:Burnout in healthcare professionals (HCPs) is a global concern. Few studies use theoretical and conceptual models to assess work system stressors contributing to HCP burnout. A survey consisting of demographic characteristics and work system stressors based on the National Academy of Medicine's theoretical model was administered to 589 HCPs (participation rate: 64%). We compare four classifiers with four feature selection methods to predict HCP burnout. We use SHapley Additive exPlanations (SHAP) to prioritize key work system stressors contributing to HCP burnout and explain the predictions from the best model, random forest using recursive feature elimination (AUC: 0.75). Time pressure, work-life integration, inadequate technology, administrative burden, values and expectations, professional relationships, lack of recognition for QI activities, moral distress, and patient factors are the work system stressors that have the highest impact on predicting HCP burnout. Identifying and prioritizing key factors to mitigate HCP burnout is important for the healthcare system to allocate resources and improve patient safety and quality of care.
computer science, information systems, artificial intelligence
What problem does this paper attempt to address?