An Artificial Intelligence-Based Probabilistic Forecasting Model for Predicting Nursing Workforce Demand in Hospital...

Area of Science:

• Healthcare Management
• Data Science
• Nursing Workforce Analytics

Background:

• Appropriate nurse staffing is vital for healthcare quality but remains a challenge.
• Advanced machine learning for nursing demand forecasting is underutilized.
• Optimizing resource utilization and hiring requires reliable workforce predictions.

Purpose of the Study:

• To propose and evaluate a probabilistic forecasting approach for nursing workforce demand.
• To utilize the DeepAR algorithm for predicting demand across multiple hospital units.
• To provide a reliable tool for healthcare administrators to improve staffing decisions.

Main Methods:

• Applied DeepAR, an autoregressive recurrent neural network, to forecast nursing demand.
• Analyzed 5-year historical workforce data from a large US health system.
• Developed a single global model for heterogeneous time-series, including COVID-19 impact via feature encoding.

Main Results:

• The DeepAR model accurately captured temporal patterns, including trends and sudden changes.
• Most ground truth values fell within the 90% prediction intervals.
• The model demonstrated consistent performance and strong generalizability across various business unit-specialty combinations.

Conclusions:

• The proposed probabilistic forecasting approach using DeepAR is a reliable tool for nursing workforce demand prediction.
• This method enables informed staffing decisions in diverse hospital settings.
• It optimizes resource utilization and supports proactive hiring strategies.