# Using Machine Learning to Predict Resilience Among Nurses in a South African Setting

## TL;DR

This study uses machine learning to predict resilience among nurses in South Africa, finding that a brief screening tool is effective despite limited demographic insights.

## Contribution

The study introduces a machine learning model for predicting nurse resilience using a brief screening tool in a South African context.

## Key findings

- A brief 4-item resilience screen achieved 86.41% classification accuracy.
- Demographic variables added limited predictive value in the model.
- The study focused on nursing staff in the Western Cape Province, South Africa.

## Abstract

Nursing is a stressful profession. Stress can affect the mental health of nurses. A positive response to stress, resilience, is known to be a protective factor against mental health issues. This study aimed to use machine learning with secondary data from five survey studies, conducted between 2022 and 2023, to identify factors predicting high versus low levels of resilience in South African nursing samples from the Western Cape Province, South Africa. The sample included (1134 records (male = 250, 22.0%, female = 874, 77.1%, and other = 10 (0.9%) included all data on all categories of nursing staff (student nurses (567, 50%), professional registered nurses (315, 27.8%), and non-professional nurses (246, 21.7%) who completed a survey using a response to stress scale. We used random forest analysis, demographic variables, years of experience, and a brief 4-item screen of resilience to predict resilience. The model yielded limited added value from demographic groupings in this model, but the brief screening had an overall classification accuracy of 86.41% (95% CI: 0.810; 0.908).