Predicting nursing workload in digestive wards based on machine learning: A prospective study.

Area of Science:

• Nursing
• Health Informatics
• Machine Learning

Background:

• Nursing workload is critical for staff allocation and patient safety.
• Cross-sectional studies limit accurate workload prediction.
• Static models fail to account for longitudinal changes, impacting management decisions.

Purpose of the Study:

• To develop a machine learning model for dynamic nursing workload prediction.
• To utilize patient characteristics for predicting nursing time requirements.

Main Methods:

• Prospective cohort study (March 2019-August 2021) in two Chinese general hospitals.
• Collected data on 133 patients over 1339 hospital days, including patient characteristics and direct nursing time.
• Applied multiple linear regression and machine learning (Random Forest) for longitudinal workload analysis and prediction.

Main Results:

• Mean direct nursing workload varied significantly across hospitalizations.
• Key factors influencing workload included complications, comorbidities, BMI, income, past illness, SCS, and ADL.
• The Random Forest model achieved high predictive performance (R²: 0.74).

Conclusions:

• Patient self-care capacity, complications, and comorbidities are primary drivers of nursing workload variability.
• Machine learning, specifically the Random Forest algorithm, effectively predicts nursing workload using diverse patient data.
• The developed quantitative model aids nursing managers in proactive workload management and resource planning.