Background
Classic heatstroke (CHS) is a life-threatening illness characterized by extreme hyperthermia, dysfunction of the central nervous system and multiorgan failure. Accurate predictive models are useful in the treatment decision-making process and risk stratification.This study was to develop and externally validate a prediction model of survival for hospitalized patients with CHS.
Methods
In this retrospective study, we enrolled patients with CHS who were hospitalized from June 2022 to September 2022 at 3 hospitals in Southwest Sichuan (training cohort) and 1 hospital in Central Sichuan (external validation cohort). Prognostic factors were identified utilizing least absolute shrinkage and selection operator (LASSO) regression analysis and multivariate Cox regression analysis in the training cohort. A predictive model was developed based on identified prognostic factors, and a nomogram was built for visualization. The areas under the receiver operator characteristic (ROC) curves (AUCs) and the calibration curve were utilized to assess the prognostic performance of the model in both the training and external validation cohorts. The Kaplan‒Meier method was used to calculate survival rates.
Result
A total of 189 patients (median age, 75 [68–81] years) were included. Social isolation, self-care ability, comorbidities, body temperature, heart rate, Glasgow Coma Scale (GCS), procalcitonin (PCT), aspartate aminotransferase (AST) and diarrhea were found to have a significant or near-significant association with worse prognosis among hospitalized CHS patients. The AUCs of the model in the training and validation cohorts were 0.994 (95% [CI], 0.975–0.999) and 0.815 (95% [CI], 0.596–0.956), respectively. The model's prediction and actual observation demonstrated strong concordance on the calibration curve regarding 7-day survival probability. According to K‒M survival plots, there were significant differences in survival between the low-risk and high-risk groups in the training cohort and borderline significant differences in the external validation cohort.
Conclusion
We designed and externally validated a prognostic prediction nomogram for CHS. This model has promising predictive performance and could be applied in clinical practice for managing patients with CHS.