Skip to main content
. Author manuscript; available in PMC: 2022 Aug 1.
Published in final edited form as: Crit Care Med. 2021 Aug 1;49(8):1312–1321. doi: 10.1097/CCM.0000000000004966

Table 2:

Performance metrics for DRNN and other early earning scores

Performance metric DRNN Logistic Model NEWS MEWS qSOFA
Area under precision-recall curve 0.042 (0.040, 0.043) 0.043 (0.041, 0.045) 0.034 (0.032, 0.035) 0.028 (0.027, 0.030) 0.021 (0.021, 0.022)
Positive predictive value 0.034 (0.034, 0.035) 0.031 (0.031, 0.032) 0.020 (0.020, 0.020) 0.015 (0.015, 0.015) 0.015 (0.015, 0.015)
Number needed to evaluate 29 (29, 30) 32 (32, 33) 50 (49, 51) 67 (66, 67) 67 (66, 67)
Area under receiver operating characteristics curve 0.723 (0.720, 0.726) 0.711 (0.708, 0.714) 0.600 (0.596, 0.604) 0.574 (0.570, 0.577) 0.551 (0.548, 0.554)

Performance metrics reported with 95% confidence intervals. Positive predictive value and number needed to evaluate were calculated for a fixed sensitivity of 50%. Abbreviations: DRNN, deep recurrent neural network; NEWS, National Early Warning Score; MEWS, Modified Early Warning Score; qSOFA, quick Sepsis-related Organ Failure Assessment.