Table 3:

Estimate of the conditional treatment effect of changing antibiotics during any timepoint on to mortality, using the causal survival forest method.

CATE Model (1) - Using only DAG variables
Average Treatment Effect: −0.07619 (SE = 0.02953)
Individual Treatment Effect: Min: −0.1584, Mean: −0.0755, Median: −0.0739, Max: −0.0217
Variables	Estimate	Std. Error	t value Pr(> |t|)
Sex	0.0407	0.0604	0.5004
Race	−0.0565	0.0580	0.3305
Age	−0.0008	0.0018	0.6429
Charlson’s Comorbidity Index	−0.0031	0.0073	0.6725
ICU stay	−0.0656	0.0569	0.2492
Multi-drug resistance	0.0627	0.0575	0.2760
CATE Model (2) - Using DAG + individual comorbidities
Average Treatment Effect: −0.0810 (SE = 0.0281)
Individual Treatment Effect: Min: −0.1083, Mean: −0.0782, Median: −0.0783, Max: −0.0462
Variables	Estimate	Std. Error	t value Pr(> |t|)
Sex	0.0404	0.2723	0.8824
Race	−0.1513	0.3784	0.6903
Age	−0.0117	0.0110	0.2928
Myocardial infarction	−0.1721	0.4967	0.7299
Congestive heart failure	0.2024	0.3747	0.5906
Cerebrovascular disease	0.2644	0.4104	0.5212
Dementia	0.4225	0.6213	0.4985
Chronic pulmonary disease	0.3245	0.3663	0.3783
Rheumatoid disease	−0.2275	0.6859	0.7410
Peptic ulcer disease	0.372	0.4869	0.4471
Mild liver disease	−0.1906	0.2974	0.5234
Diabetes without complications	0.2187	0.6174	0.7241
Diabetes with complications	0.8421	0.7525	0.2664
Hemiplegia or paraplegia	0.1629	0.4264	0.7034
Renal disease	−0.0713	0.3952	0.8572
Cancer (any malignancy)	0.2896	0.4059	0.4776
Moderate or severe liver disease	−0.2539	0.6188	0.6827
Metastatic solid tumor	0.2895	0.5006	0.5647