Skip to main content
Fig. 3 | Critical Care

Fig. 3

From: Predicting neurological outcome after out-of-hospital cardiac arrest with cumulative information; development and internal validation of an artificial neural network algorithm

Fig. 3

Illustration of the impact of features for a patient-specific prediction. An example of patient-specific prediction using the Shapley additive explanations algorithm (SHAP). This patient is predicted to have a 23% risk of a poor outcome on day 1, and 18% and 13% risk of a poor outcome on days 2 and 3, respectively (using the level C model). The patient’s age is a is the most important risk-increasing variable while the modest levels of biomarkers like NFL, NSE decrease the risk a poor outcome. TNT, Troponin-T (ng/L). NFL, Neurofilament light (ng/L). UCHL1, Ubiquitin carboxy-terminal hydrolase L1 (ng/L). NSE, Neuron-specific enolase (ng/ml.) GFAP, Glial fibrillary acidic protein (ng/L)

Back to article page

Critical Care

ISSN: 1364-8535