Development of a Nurse Turnover Prediction Model in Korea Using Machine Learning

Area of Science:

• Healthcare Management
• Data Science in Healthcare
• Nursing Workforce Studies

Background:

• Nurse turnover poses a significant challenge in Korea, impacting patient care quality and healthcare costs.
• High turnover rates necessitate proactive strategies for retention and workforce stability.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) model for predicting nurse turnover in Korea.
• To identify key factors influencing nurse turnover decisions.
• To provide a cost-effective tool for healthcare institutions to manage nursing staff retention.

Main Methods:

• Comparative analysis of three ML models: decision tree, logistic regression, and random forest.
• Development and optimization of a random forest model for nurse turnover prediction.
• Analysis of feature importance to determine key drivers of nurse turnover.

Main Results:

• The random forest model demonstrated high predictive accuracy, achieving 0.97 initially.
• Optimized random forest model improved one-year nurse turnover prediction accuracy to 98.9%.
• Analysis identified salary as the most significant factor influencing nurse turnover.

Conclusions:

• Machine learning, specifically the random forest model, offers an efficient and cost-effective method for predicting nurse turnover in Korea.
• The developed prediction model can aid hospitals and nursing units in proactively managing nurse turnover.
• Understanding key factors like salary is crucial for implementing targeted retention strategies.