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

Background: The process of assessing and allocating nursing staff, as well as evaluating performance, relies heavily on nursing workload, which is strongly associated with patient safety outcomes. Nevertheless, most previous studies have utilized cross-sectional data collection methods, which limit the precision of workload prediction. Static workload models do not incorporate longitudinal changes in influential factors, potentially resulting in delayed or erroneous nursing management decisions and ultimately causing imbalances in nurses' workload.

Aim: To employ machine learning algorithms to facilitate the dynamic prediction of nursing workload on the basis of patient characteristics.

Methods: This prospective cohort quantitative study was conducted between March 2019 and August 2021 in two general hospitals located in China. Data on the characteristics of 133 patients over the course of 1339 hospital days, as well as direct nursing time, were collected. A longitudinal investigation of nursing workload was carried out, applying multiple linear regression to identify measurable factors that significantly impact nursing workload. Additionally, machine learning methods were applied to dynamically predict the nursing time needed for each patient.

Results: The mean direct nursing workload varied greatly across hospitalizations. Significant factors contributing to increased care needs included complications, comorbidities, body mass index (BMI), income, history of past illness, simple clinical score (SCS), and activities of daily living (ADL). The predictive performance improved through machine learning, with the random forest model demonstrated the best performance (root mean square error (RMSE): 1148.38; coefficient of determination (R²): 0.74; mean square error (MSE): 1318744.64).

Conclusions: The variability in nursing workload during hospitalization is influenced primarily by patient self-care capacity, complications, and comorbidities. The random forest algorithm, a machine learning algorithm, effectively handles a wide range of features, such as patient characteristics, complications, comorbidities, and other factors. This algorithm has demonstrated good performance in predicting workload.

Implications for nursing management: This study introduces a quantitative model designed to evaluate nursing workload throughout the duration of hospitalization. By employing the model, nursing managers can consider multiple factors that impact workload comprehensively, resulting in enhanced comprehension and interpretation of workload variations. Through the application of a random forest algorithm for workload prediction, nursing managers can anticipate and estimate workload in a proactive and precise manner, thereby facilitating more efficient human resource planning.