Harnessing Clinical Data Streams for Nursing Workload Prediction Using Artificial Intelligence.

Area of Science:

• Healthcare Management
• Applied Machine Learning
• Nursing Informatics

Background:

• Increasing workload and staff shortages in inpatient care necessitate innovative personnel planning strategies.
• Accurate prediction of nursing workload is crucial for efficient resource allocation and demand-oriented staffing.

Purpose of the Study:

• To develop and evaluate a machine learning approach for predicting daily nursing workload in inpatient settings.
• To identify key data predictors for accurate workload forecasting.
• To assess the impact of incorporating clinical data on prediction accuracy.

Main Methods:

• Utilized routine nursing and medical data to train tree-based machine learning models.
• Examined various administrative, nursing, diagnostic, and procedural data sets for predictor selection.
• Evaluated model performance using R2adj and Normalized Root Mean Square Error (NRMSEstd).

Main Results:

• Machine learning models achieved R2adj values from 0.58 to 0.78, with NRMSEstd between 47% and 64% of the target variable's standard deviation.
• A selection of administrative, nursing, diagnostic, and procedural data proved effective as predictors.
• Inclusion of clinical data sources improved prediction accuracy for both independent nursing scope and specific task areas.

Conclusions:

• Machine learning offers a viable solution for demand-oriented personnel planning in healthcare by predicting nursing workload.
• The developed models provide valuable insights into factors influencing nursing workload, enabling more precise staffing adjustments.
• Integrating diverse data sources, including clinical data, enhances the predictive power and applicability of workload forecasting models.