Urine output as one of the most important features in differentiating in-hospital death among patients receiving extracorporeal membrane oxygenation: a random forest approach

Table 3 Logistic regression model used to predict in-hospital death with 11 clinical variables, including model description, odds ratio, p value, and 95% confidence interval

The model used to estimate the probability of in-hospital death
Probability of in-hospital death = exp(Y)/(exp(Y) + 1), where
Y = logit (probability of in-hospital death) =
1.007
+ 0.022 × Age
+ 0.035 × Ventilation rate
+ 0.007 × Inotropic equivalent
+ 0.591 × Extracorporeal ECPR
+ 0.767 × Infection before ECMO
− 0.005 × Systolic blood pressure
− 0.033 × Urine amount
− 0.071 × Glasgow coma scale
− 0.432 × [ECMO category = respiratory]
− 1.721 × [ECMO category = others]
− 0.573 × [Bil = missing]
− 1.150 × [Bil = 0–2 mg/dL]
− 0.733 × [Bil = 2–5 mg/dL]
− 0.101 × [CKMB = missing]
− 0.516 × [CKMB = 0–23 U/L]
− 0.538 × [CKMB = 23–69 U/L]

Variables	Odds ratio	p	95% confidence interval
Age	1.022	< 0.001	1.014	1.03
Ventilation Rate	1.035	0.004	1.011	1.06
Inotropic equivalent	1.007	0.001	1.003	1.012
Extracorporeal CPR	1.806	< 0.001	1.356	2.407
Infection before ECMO	2.152	< 0.001	1.548	2.993
Systolic blood pressure	0.995	0.017	0.991	0.999
Urine amount	0.967	< 0.001	0.957	0.977
Glasgow coma scale	0.931	< 0.001	0.909	0.954
ECMO category		0.045
Respiratory	0.649	0.045	0.425	0.991
Others	0.179	0.119	0.021	1.558
Cardiovascular	Reference
Bilirubin		< 0.001
Missing	0.564	0.077	0.299	1.063
0–2	0.317	< 0.001	0.174	0.577
2–5	0.48	0.025	0.253	0.913
> 5	Reference
Creatine kinase MB		0.004
Missing	0.904	0.599	0.622	1.316
0–23	0.597	0.01	0.403	0.884
23–69	0.584	0.007	0.396	0.862
> 69	Reference

ISSN: 2047-783X