Systematic review of prediction models for post-traumatic hypothermia risk

Post-traumatic hypothermia is a prevalent complication in trauma care, affecting up to 66 % of multiple trauma patients upon emergency admission and doubling trauma-related mortality. While risk prediction models for post-traumatic hypothermia have been developed, guidelines and evidence specific to their clinical utility-especially in pre-hospital and low- to middle-income country (LMIC) settings-remain scarce. This systematic review evaluates existing post-traumatic hypothermia risk prediction models, their performance characteristics, and applicability in diverse clinical contexts.

Databases including China National Knowledge Infrastructure (CNKI), Wanfang Database, VIP Database, China Biomedical Literature Database, PubMed, Web of Science, and Cochrane Library were searched from the time of database establishment to July 2025. Two researchers independently screened the literature, extracted relevant data, and conducted quality assessments. The included studies were analyzed to comprehensively evaluate the predictive models for post-traumatic hypothermia risk.

A total of 9 studies were included, comprising 9 predictive models for the risk of post-traumatic hypothermia, with a total sample size of 91 to 732 cases and 24 to 117 outcome events. The area under the receiver operating characteristic curve (AUC) for the predictive models ranged from 0.704 to 0.990, with specificity ranging from 50.6 % to 95.2 %. and sensitivity ranging from 70.9 % to 92.8 %. The most frequently identified predictive factors in the models were trauma severity, wet clothing, lack of warming measures, fluid resuscitation, and environmental temperature at the time of injury. The included risk prediction models demonstrated overall good applicability, but they had a high risk of bias, which was associated with limitations in sample selection, indicator selection and measurement, study design flaws, and inadequate model validation.

Although the predictive model for post-traumatic hypothermia risk demonstrates certain advantages in overall applicability and can provide reference for clinical prediction of post-traumatic hypothermia risk, it has a high risk of bias. Future studies should be conducted in a multicenter, large-sample setting to strengthen external validation of the model and test its performance in different clinical environments to ensure its stability and accuracy.