Table 3.
Univariate and multivariate logistic regression analyses for ICU admission.
| Variable | Univariate |
Multivariate |
||
|---|---|---|---|---|
| (1) |
(2) |
(3) |
||
| OR (95% CI) p-value [R2] | OR (95% CI) p-value [R2] | OR (95% CI) p-value R2 = 0.368 | OR (95% CI) p-value R2 = 0.533 | |
| Demographic and clinical variables | ||||
| Gender (male) | 1.062 (0.48–2.35) p = 0.882 [0.000] | |||
| Age (years) | 1.7 (1.21–2.49) p = 0.003 [0.094] | 1.733 (1.206–1.491) p = 0.003 [0.094] | ||
| Lymphocytes (∗103/μL) | 0.49 (0.22–1.06) p = 0.034 [0.01] | |||
| CRP (mg/dL) | 12.1 (0.22–1.06) p = 0.0001 [0.274] | 1.107 (1.053–1.163) p = 0.0001 [0.256] | 1.091 (1.019–1.169) p = 0.012 | |
| LDH (U/L) | 2.53 (1.08–5.93) p = 0.033 [0.055] | 1.005 (1.002–1.009) p = 0.003 [0.183] | ||
| D-dimer (ng/mL) | 4.62 (1.91–11.17) p = 0.001 [0.134] | |||
| Temperature (°C) | 1.84 (0.84–4.05) p = 0.13 [0.125] | |||
| Hypertension | 1.56 (0.63–3.87) p = 0.338 [0.013] | |||
| Cardiovascular disease | 2.32 (0.49–10.95) p = 0.288 [0.019] | |||
| ACEi/ARB | 0.77 (0.25–2.43) p = 0.658 [0.005] | |||
| Diabetes | 2.15 (0.58–7.98) p = 0.251 [0.021] | |||
| Malignancy | 2.49 (0.30–20.75) p = 0.400 [0.013] | |||
| Kidney failure | 2.8 (0.34–23.14) p = 0.339 [0.017] | |||
| Radiological variables | ||||
| LSS | 5.79 (2.34–14.35) p = 0.0001 [0.165] | 1.173 (1.101–1.249) p = 0.026 [0.312] | 1.121 (1.042–1.206) p = 0.002 | 1.262 (1.071–1.488) p = 0.005 |
| TAT (mm2) | 2.22 (0.99–4.93) p = 0.049 [0.041] | 1.59 (1.057–2.392) p = 0.026 [0.167] | ||
| VAT (mm2) | 3.13 (1.36–7.19) p = 0.007 [0.081] | 1.577 (1.051–2.365) p = 0.028 [0.168] | 2.474 (1.017–6.019) p = 0.046 | |
| EFT (mm) | 1.29 (0.60–2.76) p = 0.504 [0.005] | |||
| SAT (mm2) | 1.60 (0.73–3.50) p = 0.239 [0.015] | |||
Notes: Univariate analysis was performed by converting continuous variables into dummy dichotomic variables based on median values, while continuous variables were used for multivariate analyses. To build a multivariate logistic regression model with ICU admission as the dependent variable, we used a forward stepwise approach and investigated the following variables/models: (1) multivariate analysis including age (quartiles) and gender, together with variables with significant univariate association (p value ≤.05) analyzed one by one as regressors (age + gender + lymphocytes; age + gender + CRP; age + gender + LDH; age + gender + D-dimer; age + gender + LSS; age + gender + TAT; age + gender + VAT; the variable age was adjusted for gender only). (2) Multivariate analysis including all statistically significant variables of the univariate analysis as regressors in one single model. (3) Multivariate analysis including all the variables of model 2 with the addition of hypertension, ACEi or ARB use prior to admission and diabetes as clinically important variables. The forward stepwise selection method does not provide ORs and 95% CI for the variables not retained in the model because they do not significantly improve prediction. Therefore, only the variables with statistically significant results were added in the table, reporting their OR and 95% CI, [R2]. For the forward stepwise analysis, a P-IN = 0.05 and a P-OUT = 0.10 were used. The effect estimate is reported as Nagelkerke's R2, which informs on how much the model explains the variance of the dependent variable. The ORs represent the mean change in the dependent variable per one unit of change in the independent variable while holding other predictors in the model constant. The results were considered statistically significant when p < 0.05. Age in years. Pearson coefficient values are highlighted in bold when correlation was statistically significant at the p < 0.05 level and below. OR, Odds ratio; CI, confidence interval; CRP, C reactive protein; LDH, Lactate dehydrogenase; ACEi angiotensin-converting-enzyme inhibitor; ARB, angiotensin II receptor blocker; TAT, Total Fat Area; VAT, Visceral Fat Area; SAT, Subcutaneous Fat Area; EFT, Epicardial Fat Thickness; LSS, Lung Severity Score.