Skip to main content
Fig. 2 | Critical Care

Fig. 2

From: A deep learning model for real-time mortality prediction in critically ill children

Fig. 2

Depiction of time and feature contributions for mortality using PROMPT. Measured contribution (%) for mortality at the critical time point and serial trend of vital signs over 24 h are plotted on each panel. The last sub-figure presents the time contribution. The height of the graph represents the level of importance, and the positive/negative conversion distinguishes the time point contributed to make positive or negative predictions for mortality. In the presented case, the critical time point (i.e., a peak of time contribution) was about 10 h, of which fluctuations in SpO2, blood pressure, and HR are shown to contribute to instability which can be associated with mortality. SBP, systolic blood pressure; DBP, diastolic blood pressure; MBP, mean blood pressure; HR, heart rate; RR, respiratory rate; SpO2, peripheral capillary oxygen saturation; BT, body temperature

Back to article page