Burnout and resilience among healthcare nurses in Saudi Arabia: a cross-sectional study with exploratory multiclass classification approach

Abstract

Burnout is a significant occupational and psychological concern among nurses. However, there is an evidence gap in research considering the dimensions of burnout and resilience in healthcare settings where the nursing workforce is predominantly expatriate. Therefore, this study aims to assess burnout and resilience levels among nurses in a Saudi tertiary hospital with a predominantly expatriate workforce, and to identify key predictors using a multiclass classification approach. A cross-sectional study was conducted among 501 nurses at Prince Sultan Military Medical City (PSMMC), Saudi Arabia. The study data were collected using the Maslach Burnout Inventory–Human Services Survey for Medical Personnel (MBI-HSS(MP)) and the Nicholson McBride Resilience Questionnaire (NMRQ). An MLP-based multiclass classification approach was applied to classify burnout dimensions and resilience levels. Model performance was evaluated using classification accuracy, precision, recall, F1-score, area under the receiver operating characteristic curve (AUC), and calibration metrics (Brier score and expected calibration error), while predictor importance was summarized using normalized permutation importance. The study was analyzed using SPSS version 28, and five-fold stratified cross-validation was implemented using scikit-learn (Python), and data visualization was performed using RStudio (R-4.6.1). Among 501 nurses, the distribution of burnout dimensions showed that 37.7% of participants had a moderate level of EE, 40.3% reported low DP, and 41.9% reported moderate PA. Emotional exhaustion showed a significant positive correlation with depersonalization ( r = 0.626, p < 0.001), showing that nurses with higher exhaustion levels also tended to report greater depersonalization. Perceived stress about going to work was positively correlated with emotional exhaustion ( r = 0.507, p < 0.001). The model showed acceptable to good discrimination for high/low EE and DP, while PA and resilience showed weak discrimination (AUC ≈ 0.6). Within the exploratory models, age and marital status showed the highest normalized importance for DP, while age and night shifts showed the highest normalized importance for PA; these patterns are hypothesis-generating and should not be interpreted as inferential associations. This study identified exploratory predictor patterns of burnout and resilience in a tertiary military nursing workforce that may inform future hypothesis-driven research. External validation and longitudinal designs are needed before these exploratory associations can guide targeted interventions.