Estimating Nurse Workload Using a Predictive Model From Routine Hospital Data: Algorithm Development and Validation.

Area of Science:

• Healthcare Management
• Nursing Informatics
• Predictive Analytics

Background:

• Nurse staffing management is complex, influenced by fluctuating patient demand, acuity, and dependency.
• Real-time monitoring of nurse staffing adequacy is crucial for safe and efficient staff deployment.
• Current patient classification systems (PCSs) for workload measurement require frequent administrative input from nursing staff.

Purpose of the Study:

• To explore the potential of an algorithm to estimate ward workload.
• To determine if routinely recorded data can be used for workload estimation.

Main Methods:

• Utilized anonymized admission records and PCS assessments from a UK hospital (February 2017-February 2020).
• Developed a predictive model using routinely recorded administrative data and National Early Warning Scores.
• Outcome variable: ward workload measured as whole-time equivalent (WTE) nursing staff per patient.

Main Results:

• The predictive model achieved a mean absolute error of 0.078 and a mean percentage error of 4.9% on a test set of 11,592 ward assessments.
• 95% of predictions fell within 0.21 WTE per patient, as shown by a Bland-Altman plot.
• Moderate accuracy was observed for general wards using a limited set of routinely collected variables.

Conclusions:

• Automating nurse staffing requirement assessments from routine data is feasible.
• This approach can reduce non-clinical administrative overhead for nursing staff.
• Improved real-time monitoring of nursing staffing pressures can be achieved.