Table 4.

Adjusted Effect of IET on Mortality and LOS in Multivariable Models^a

	Mortality (n = 39 203)		Hospital LOS, d (n = 39 203)		ICU LOS, d (n = 11 221)
Characteristic	OR (95% CI)	P value	Estimated Mean (95% CI)	P value	Estimated Mean (95% CI)	P value
Empiric therapy		<.001		<.001		.397
AET	Ref		14.5 (13.9–15.1)		8.0 (7.4–8.7)
IET	1.21 (1.10–1.33)		16.1 (15.5–16.7)		8.2 (7.6–8.9)
SARS-CoV-2 test result		<.001		<.001		<.001
Negative	Ref		13.0 (12.6–13.5)		6.0 (5.5–6.5)
Positive	4.04 (3.67–4.45)		18.0 (17.2–18.7)		11.0 (10.1–12.0)

Abbreviations: AET, adequate empiric therapy; ICU, intensive care unit; IET, inadequate empiric therapy; LOS, length of stay; OR, odds ratio; Ref, reference value; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

^aOther covariates or adjusting variables included discharge month, culture source, age group, sex, Candida albicans test status (positive or negative), polymicrobial (yes/no), baseline comorbidities, ICU or ventilator criteria met, hospital characteristics (bed size, facility type, teaching status), and geographic region based on US Census regions. Respiratory source (yes/no) was included in the LOS models but was not included in the mortality model because it did not improve the model fit, despite the fact that this variable significantly affected mortality in univariate assessment. One possible reason is that a highly correlated factor, such as ICU/ventilator status, accounted for part of its effect on mortality. Other reasons, such as confounding from unknown or unobserved factors, may also have played a role in the multivariable modeling analysis.