Abstract
The optimal pre-arrest prediction model for survival with favorable neurological outcomes following in-hospital cardiac arrest (IHCA) to aid Do-Not-Attempt-Cardiopulmonary-Resuscitation-decisions remains uncertain. This study aimed to externally validate the Prediction of Outcome for In-Hospital Cardiac Arrest (PIHCA) score and the Good Outcome Following Attempted Resuscitation (GO-FAR) 2 score and compare their performance for IHCA survival prediction. We collected data from IHCAs across three Swedish hospitals covering all care levels. We identified IHCAs through the Swedish Registry of Cardiopulmonary Resuscitation during 2015-2021 and extracted prediction model data through electronic medical records. Outcome measure was survival with favorable neurological survival defined as Cerebral Performance Category 1-2 at discharge, obtained from the registry. Prediction model performance was assessed with AUROC-curves (c-statistic), calibration coefficients and nonparametric calibration plots. We identified 647 IHCAs of which 501 were eligible for inclusion. Overall, 28% had survival with favorable neurological outcome. AUROC for the PIHCA score was 0.73 (95% CI: 0.67-0.78), compared to 0.70 (95% CI: 0.65-0.75) for the GO-FAR 2 score. The 95% CI for difference of AUROC-score when PIHCA score was compared to GO-FAR 2 score was -0.01 to 0.06, i.e. not significant. The PIHCA score calibration coefficient was 0.69 (95% CI 0.51-0.88) compared to the GO FAR 2 score calibration coefficient of 1.02 (95% CI 0.73-1.34). The calibration plots were adequate for both scores, although with some degree of miscalibration at certain ranges for each model. PIHCA score and the GO-FAR 2 score performed well and comparatively for prediction of survival with favorable neurological outcome following IHCA in a Swedish setting. Both scores could be considered for clinical use, in addition to sound clinical judgment, where the PIHCA score seems more adapted for clinical use.