(905) Prognostic Predictions After Pediatric Cardiac Arrest: A Single-Center Prospective Study

Introduction: Half of children who have a cardiac arrest die after withdrawal of life sustaining therapy (WLST), which is influenced by provider outcome prediction. We describe the concordance of provider outcome prediction in the first 7 days post-arrest with patient survival and functional outcomes. We also report the difference in concordance based on patient baseline neurologic function (PCPC>1 vs =1). We hypothesized that concordance would increase over time and be higher for survivors than non-survivors.

Methods: Single-center prospective cohort study from 7/2023 to 4/2024. We included children with an index in- or out-of-hospital (OHCA) cardiac arrest who received post-arrest care in the PICU. Multidisciplinary team members completed surveys on post-arrest days 1, 2, 3, and 7, eliciting outcome prediction at hospital discharge (Pediatric Cerebral Performance Category [PCPC] score). Prognostic concordance was defined at each time point as percent agreement between provider prediction and patient outcome. We evaluated concordance for survival and death using sensitivity (percent agreement for survivors) and specificity (percent agreement for non-survivors). We evaluated concordance of functional outcome as 1) categorization as favorable (PCPC 1, 2, 3 or no change from baseline) or unfavorable (PCPC 4, 5, 6) and 2) absolute difference between predicted and discharge PCPC < 1. Differences in concordance were explored using Chi-square test.

Results: Fifty patients (median age: 4.2 years [IQR: 1.2, 9.6], 60% OHCA, and 52% pre-arrest PCPC=1) were analyzed; 73% survived to hospital to discharge. 461 surveys were completed by 217 providers. Sensitivity was >80% at all time points, with highest on day 7 (88%). Specificity decreased over time, 70% day 1 and 36% day 7. Sensitivity was higher for patients with baseline PCPC>1 than with baseline PCPC=1 (92% vs 72%, p< 0.001). Prognostic concordance of functional outcome did not differ between patients with baseline PCPC>1 vs =1 for favorable or unfavorable outcome [78% vs 82%, p=0.351] or PCPC prediction +/-1 [81% vs 86%, p=0.187].

Conclusions: Despite generally good sensitivity for predicting survival, provider prediction of post-arrest neurologic outcome is frequently inaccurate and varies temporally and in relation to baseline neurologic function.