Table 3.

Determination of intensive care unit (ICU)-mortality-related parameters through logistic regression analysis

Parameters	Univariate analyses			Multiple analyses
	OR	95% CI	p	OR	95% CI	p
Neurological causes (non-CVA)	0.271	0.087–0.844	0.02*
CVA	0.334	0.104–1.074	0.07
Mechanical ventilation support1	9.111	2.499–33.212	0.001*	15.155	1.457–157.660	0.023
Steroids	9.519	2.657–34.103	0.001*
Septic shock	15.600	4.252–57.240	<0.001*	8.558	1.460–50.171	0.017
Age, years	1.021	0.986–1.059	0.24
GCS score	0.776	0.643–0.936	0.008
APACHE-II score	1.219	1.086–1.368	0.001*	1.146	1.003–1.310	0.045
SOFA score	1.814	1.308–2.516	<0.001
CRP levels (pre-CZA treatment), mg/L	1.009	1.000–1.017	0.04*
PCT levels (pre-CZA treatment), µg-L	1.398	1.032–1.893	0.03
PCT levels (post-CZA treatment), µg-L	2.266	1.171–4.384	0.02
Colistin therapy2	4.950	1.569–15.618	0.006*

APACHE-II: Acute Physiology and Chronic Health Evaluation; CI: Confidence Interval; CRP: C-Reactive Protein; CVA: Cerebrovascular Accident; GCS: Glasgow Coma Scale; HR: Hazard Ratio; PCT: Procalcitonin; pre: Before CZA treatment; post: After CZA treatment; SOFA: Sequential Organ Failure Assessment. “Mechanical ventilation support” was categorized into two groups: -Absent: Patients not needing mechanical ventilation support, given only oxygen (reference group). - Present: Patients given mechanical ventilation support using either non-invasive mechanical ventilation (NIMV) or invasive mechanical ventilation (IMV) devices. “Colistin therapy” was categorized into two groups: - Absent: Patients who did not use colistin therapy (reference group). - Present: Patients who used colistin therapy either before or concomitantly with CZA treatment. Since all patients with solid tumors passed away and there were no survivors, they were not included in the regression analysis model. Parameters marked with an asterisk (*) were included in the multiple binary logistic regression analysis. The SOFA score and GCS were not included in the multiple analyses because the APACHE-II score evaluates similar clinical parameters, such as GCS, creatinine levels, mean arterial blood pressure, and PaO₂ and FiO₂ levels, more comprehensively than the SOFA score. Since the missing values in PCT parameters were high (12 patients), we added only the CRP levels to the multiple analyses. Binary logistic regression was analyzed using the backward method. In the final step (step 3), the statistically significant findings are presented in Table 3. For this step, the p-values of the Omnibus test and the Nagelkerke R square were found to be <0.001 and 0.671, respectively.