Table 4. Multivariable logistic regression analysis of possible predictors of severe or critical COVID-19 disease among hospitalized patients with a pre-infection measurement of 25(OH)D level.
The multivariable analyses before and after cosinor correction models for the annual mean as well as each patient’s date of hospitalization are shown. This is the final step in the logistic regression (backward elimination).
| Pre-infection 25(OH)D level (ng/mL) | Predicted 25(OH)D (ng/mL) annual mean | Predicted 25(OH)D level (ng/mL) corrected to the hospitalization date | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Values obtained from patient’s chart | Values corrected according to cosinor model (Eq 1) | Values corrected according to cosinor model (Eq 3) | |||||||||||
| Mild/moderate cases | Severe/critical cases | OR | 95%CI | p value | OR | 95%CI | p value | OR | 95%CI | p value | |||
| variable | N | % | N | % | |||||||||
| 25(OH)D levels | p < 0.001 | p < 0.001 | p < 0.001 | ||||||||||
| ≥40 | 37 | 92.5% | 3 | 7.5% | baseline | baseline | baseline | ||||||
| 30–39.9 | 40 | 90.9% | 4 | 9.1% | 1.45 | 0.29–7.25 | p = 0.65 | 1.23 | 0.25–6.20 | p = 0.80 | 1.11 | 0.22–5.48 | p = 0.90 |
| 20–29.9 | 32 | 88.9% | 4 | 11.1% | 1.32 | 0.26–6.78 | p = 0.74 | 1.36 | 0.29–6.42 | p = 0.70 | 2.48 | 0.55–11.26 | p = 0.24 |
| <20 | 57 | 42.9% | 76 | 57.1% | 14 | 4–51 | p < 0.001 | 14.49 | 4.01–51.73 | p < 0.001 | 12.43 | 3.52–43.85 | p < 0.001 |
| Mean age, years (std) | 58.9 (18.9) | 72.7 (14.3) | 1.04 | 1.02–1.06 | p = 0.001 | 1.04 | 1.02–1.06 | p < 0.001 | 1.03 | 1.007–1.05 | p = 0.01 | ||
| Diabetic | |||||||||||||
| yes | 45 | 50% | 45 | 50% | 2.03 | 1.04–3.97 | p = 0.04 | 2.00 | 1.02–3.90 | p = 0.04 | (/) | ||
| no | 121 | 74.2% | 42 | 25.8% | baseline | baseline | |||||||
| COPD | (/) | (/) | |||||||||||
| yes | 71 | 52.6% | 64 | 47.4% | 2.01 | 0.97–4.14 | p = 0.06 | ||||||
| no | 95 | 80.5% | 23 | 19.5% | baseline | ||||||||
| Constant | 0.004 | p < 0.001 | 0.005 | p < 0.001 | 0.009 | p < 0.001 | |||||||
Multivariable logistic regression analysis, backward elimination method. Dependent variable: COVID-19 disease severity (severe or critical vs. mild or moderate). The variables included in step 1 of this model were age, sex (male), BMI, and presence of chronic obstructive pulmonary disease (COPD), chronic heart failure (CHF), chronic ischemic heart disease (CIHD), chronic renal failure (CRF), or diabetes. Included are the 25(OH)D levels [Pre-infection 25(OH)D level obtained from patient’s chart; predicted 25(OH)D annual mean corrected according to cosinor model (Eq 1); and predicted 25(OH)D level (ng/mL) corrected to the hospitalization date corrected according to cosinor model (Eq 3)]. OR = odds ratio; CI = confidence interval; (/) Removed in the last step of backward elimination.