Photo Credit: Gpointstudio
The following is a summary of “Machine learning identifies higher survival profile in extracorporeal cardiopulmonary resuscitation,” published in the July 2024 issue of Critical Care by Crespo-Diaz et al.
Extracorporeal cardiopulmonary resuscitation (ECPR) has demonstrated improved neurological outcomes in out-of-hospital cardiac arrest (OHCA) patients with shockable rhythms, highlighting the need for better patient selection to optimize the resource-intensive treatment.
Researchers conducted a retrospective study developing a patient selection model for ECPR in cases with refractory cardiac arrest using machine learning (ML) techniques.
They studied adults aged 18–75 with refractory OHCA due to shockable rhythms in a cardiac ICU at a quaternary care center. Among 376 patients, 301 received ECPR and venoarterial extracorporeal membrane oxygenation. Clinical variables present at cannulation were assessed and ranked according to the ability to predict neurologically favorable survival. Supervised ML models were trained to detect favorable neurological outcomes for patients with ECPR. The most effective models were internally validated with a separate test set.
The result showed that 40% (119 out of 301) patients receiving ECPR had favorable neurological outcomes. Various factors were analyzed, including rhythm at the time of cannulation, intermittent or sustained return of spontaneous circulation, arrest to extracorporeal membrane oxygenation perfusion time, and lactic acid levels. A training model incorporating all the analyzed factors achieved an in-sample AUC of 0.89 and a misclassification rate of 0.19. The model’s generalizability was confirmed with an out-of-sample AUC of 0.80 and a misclassification rate of 0.23, indicating good prediction accuracy.
Investigators concluded that ML offers a promising tool for selecting patients with ECPR based on risk stratification and specific arrest characteristics.
Source: journals.lww.com/ccmjournal/fulltext/2024/07000/machine_learning_identifies_higher_survival.8.aspx
