Risk prediction models of compassion fatigue among nurses: A systematic review and meta-analysis

Compassion fatigue has serious consequences for nursing staff, patients and healthcare organizations. The development of predictive models specifically designed for nurses' compassion fatigue has received considerable attention in recent years, the diversity of high-risk factors makes them difficult to measure and use in clinical settings, and the quality and applicability of these models have not been systematically reviewed.

This systematic review evaluates existing risk prediction models for nurse compassion fatigue, assesses their performance and clinical utility to guide early risk identification and model optimization.

Systematic review and meta-analysis of observational studies.

The Cochrane Library, CINAHL, Embase, ScienceDirect, Web of Science, PubMed, China National Knowledge Infrastructure (CNKI), Wanfang Database, China Science and Technology Journal Database (VIP), and Chinese Biomedical Database (CBM) were searched from inception to January 21, 2026. Data from selected studies were extracted, including study design, data source, outcome definition, sample size, predictors, model development and performance. The Prediction Model Risk of Bias Assessment Tool (PROBAST) checklist was used to assess the risk of bias and applicability.

A total of 1268 studies were retrieved, and after the selection process, ten prediction models from seven studies were included in this review. Seven models used logistic regression, while the remaining respectively utilized a nomogram, decision tree, and random forest. Only one study conducted both internal and external validation. Social support, job satisfaction, job stress and working years were important predictors of compassion fatigue among nurses. The incidence of compassion fatigue among clinical nurses ranged from 49.41 % to 83.88 %. The reported area under the curve (AUC) ranged from 0.704 to 0.939. Most studies were found to have a high risk of bias, primarily due to poor reporting of the statistical analysis domain. The pooled AUC value of the six validated models was 0.819 (95 % confidence interval=0.746∼0.892), indicating a fair level of discrimination.

Although the included studies reported a certain level of predictive performance in the compassion fatigue prediction model for nurses, most of them were found to have a high risk of bias according to the PROBAST checklist. Future research should prioritize the development of new models with larger sample sizes, rigorous study designs, and multi-center external validation.

The study protocol was registered on PROSPERO (registration number: CRD42024504226).