Perioperative medicine & outcomes

Prediction of Cardiopulmonary Resuscitation Outcomes for Arrest in Surgical Settings

JAMA Netw OpenPublished Prognostic study

Authors' conclusion

In this prospective prognostic study of outcomes following perioperative CPR, extreme gradient boosting generated individualized predictions of outcomes following perioperative CPR that may inform prevention strategies and goal-concordant surgical care.

Study context

Population
Adults ≥18 y undergoing noncardiac surgery who underwent cardiopulmonary resuscitation (CPR) on the day of surgery, ACS-NSQIP (nearly 700 US hospitals), 2012–2023
Participants
6,405 adults
Outcome
30-day mortality following CPR
Follow-up
30 days
Methodological detailsprospective, multicentre, machine learning, 10-fold cross validation, internal validation
  • Prospective
  • Multicentre
  • Machine learning
  • 10-fold cross validation
  • Internal validation

Performance

As reported in the abstract

30-day mortality prediction
Test or modelAUCAccuracySensitivitySpecificityBrier score
COMPASS0.80 (95 % CI, 0.78-0.82)0.73 (95% CI, 0.71-0.75)0.77 (95% CI, 0.74-0.79)0.68 (95% CI, 0.65-0.71)0.18

COMPASS

AUC0.80 (95 % CI, 0.78-0.82)
Accuracy0.73 (95% CI, 0.71-0.75)
Sensitivity0.77 (95% CI, 0.74-0.79)
Specificity0.68 (95% CI, 0.65-0.71)
Brier score0.18
Died within 30 days
3710 (57.9%)

Secondary outcomes

OutcomeValues
Nonhome discharge among survivors admitted from homeExtreme gradient boosting AUROC 0.78 (95 % CI, 0.74-0.82) · accuracy 0.76 · Brier score 0.17
Nonhome discharge among survivors admitted from home — Survivors admitted from home discharged to a facility
822 (33.2%) of 2478

Key predictors (SHAP analysis)

ASA status, case urgency and frailty identified as key predictors

Clinical utility (decision curve analysis (DCA))

Greater net benefit of extreme gradient boosting over default strategies (treat all or treat none) across wide threshold ranges

Original publication

Prediction of Cardiopulmonary Resuscitation Outcomes for Arrest in Surgical Settings. JAMA Netw Open. Published 1 October 2025. doi:10.1001/jamanetworkopen.2025.39767

Composed by AI from the published abstract and checked against it. How Briefs are made →

Get them every week

New studies on this topic in My Week and in your Monday e-mail.

We e-mail you a sign-in link and a code. No password, no spam.