The association between the severity of out-of-hospital cardiac arrest and the effectiveness of target temperature management: a retrospective study based on prediction models.

Aim: This study aimed to develop prediction models and conduct risk stratifications for patients with out-of-hospital cardiac arrest (OHCA) to identify patients who could benefit from targeted temperature management (TTM) at 33°C.

Methods: A retrospective analysis was carried out on 368 patients and the primary outcome was the neurological outcome at discharge evaluated by the Cerebral Performance Categories (CPC) scale. Six variables were utilized to construct prediction models via six methodologies, and the Chi-square test or Fisher's exact test was used to analyze the efficacy of TTM at 33℃ under diverse risk stratifications.

Results: A total of 264 eligible patients were divided into the development cohort and test set. The identified predictors comprised bystander cardiopulmonary resuscitation (CPR), pupillary light reflex, Acute Physiology and Chronic Health Evaluation II (APACHE II) score, lactate, serum calcium (Ca²⁺), and base excess (BE). The AUC of different prediction models in the test set ranged from 0.7592 to 0.9304. Patients with a predicted probability of 80-100%, 75-100%, and 67-100% in the Random Forest model, and 40-60% in the K-Nearest Neighbors model, can benefit from 33℃ TTM (OR [95% CI]: 3.21[1.44-7.19], 2.73[1.25-5.97], 2.18[1.09-4.36], 6.42[1.09-37.73], respectively). Among patients who had successfully undergone TTM at 33 °C, there was a higher prevalence of patients classified as CPC 3 and CPC 4 and a lower incidence of those classified as CPC 5 (OR [95% CI]: 3.90[1.12-12.58], 2.29[1.24-4.26], 0.31[0.19-0.51], respectively).

Conclusion: Prediction models developed from early variables can predict the neurological prognosis of OHCA, and the efficacy of 33℃ TTM may be related to severity.