Purpose: Predicting the prognosis of Pulseless Electrical Activity (PEA) would allow optimally directing resuscitation efforts. Contradictory results have been reported regarding the association between PEA characteristics such as Heart Rate (HR) and Return of Spontaneous Circulation (ROSC). The aim of this study was to analyse the ECG characteristics of PEA that predict ROSC.
Materials and methods: Data from 173 out-of-hospital cardiac arrest patients were extracted, all of them recorded by the DFW Center for Resuscitation Research (UTSW, Dallas), and divided into ROSC/noROSC patients. The former showed sustained QRS complexes from the onset of ROSC (tROSC, annotated by clinicians) without chest compressions and the latter were annotated as died in field at the end of the episode (tend). A total of 1439 artifact-free PEA ECG segments of 5 s were extracted during the last 10 minutes prior to tROSC (326, ROSC) or tend (1113, noROSC). The HR, Mean Slope (MS) from the first difference of the ECG and Amplitude Spectrum Area (AMSA) were computed automatically for each segment, and combined in a Logistic Regression (LR) model. Patient-wise 10-fold cross validation was adopted to train and test the model, and its performance was evaluated in terms of Sensitivity (Se), Specificity (Sp) and Area Under the Curve (AUC).
Results: Three features showed different distributions for ROSC/noROSC groups (p<0.001), mean (SD) were: 66.6(40.0)/54.8(33.72), 5.5(3.5)/2.4(1.8) and 43.1(22.6)/18.0(11.3) for HR, MS and AMSA respectively. AUC values were 0.56, 0.80 and 0.84. The LR model had Se/Sp/AUC of 80.3%/78.6%/0.85, and the posterior probability of ROSC measured using the model increased as time approached tROSC (see figure).
Conclusions: PEA characteristics are good prognostic markers of ROSC. The best ECG feature was AMSA, but combining all features provides a better prediction.