Machine Learning Improves Nurse Workload Prediction
Accurate nurse staffing is essential to maintaining high-quality care in hospitals. A new methodological study by the Krauthammer Lab at DQBM, in collaboration with the University Hospital Zurich, shows how data-driven models can enhance workload forecasting. Using five years of nursing activity data across eight hospital wards, the team developed machine learning models to predict shift-level nursing workload 72 hours in advance.
Among the tested models, lasso regression performed best, reducing prediction error by 25% over baseline approaches. The model accurately flagged significant workload increases or decreases while minimizing critical misclassifications. The study highlights how historical shift data and ward-level trends are key predictors of future demand.
The findings demonstrate the potential of machine learning to support nurse scheduling and hospital operations. Future research will focus on incorporating richer clinical data and testing models in real-world settings.
Read the full publication: https://www.jmir.org/2025/1/e66667