The Impact of Positive Fluid Balance on Sepsis Subtypes: A Causal Inference Study

Table 2

Average treatment effect (ATE) of sepsis subtypes using inverse probability weighting (IPTW) with logistic regression and T-learners machine-learning models.

	ATE	Confidence interval
Sepsis all
Binarized	0.042	(0.034, 0.047)
Continuous	0.034	(0.028, 0.040)
Pulmonary sepsis
Binarized	0.047	(0.037, 0.055)
Continuous	0.28	(0.22, 0.34)
Urinary sepsis
Binarized	−0.013	(−0.024, −0.0035)
Continuous	−0.28	(−0.34, −0.22)

IPTW with a logistic regression model corresponds to binarized values. T-learners model corresponds to continuous values.