Table 3.
Multivariable logistic regression model evaluating predictors for cryptogenic stroke in Covid-19 patients
| Variable | Univariate analysis | Multivariable analysis | ||
|---|---|---|---|---|
| OR (95% CI) | p-value | OR (95% CI) | p-value | |
| Age, per decade | 0.98 (0.75–1.27) | 0.86 | ||
| Male sex | 1.32 (0.64–2.73) | 0.45 | ||
| White race | 1.06 (0.46–2.42) | > 0.9 | ||
| Hispanic ethnicity | 1.44 (0.99–2.10) | 0.05 | Dropped from model due to non-significance | |
| Medical history | ||||
| Hypertension | 0.94 (0.43–2.06) | 0.87 | ||
| Diabetes mellitus | 0.94 (0.46–1.93) | 0.87 | ||
| Dyslipidemia | 0.70 (0.33–1.45) | 0.34 | ||
| Congestive heart failure | 0.35 (0.12–1.02) | 0.05 | 0.15 (0.05–0.44) | < 0.01 |
| Active tobacco use | 0.76 (0.21–2.75) | 0.68 | ||
| Prior stroke | 1.86 (0.53–6.47) | 0.33 | ||
| Chronic renal insufficiency (stage III/IV or dialysis-dependent) | 0.61 (0.15–2.50) | 0.50 | ||
| Chronic obstructive pulmonary disease and/or asthma | 0.39 (0.08–1.97) | 0.26 | ||
| Cancer | 2.08 (0.44–9.74) | 0.35 | ||
| Baseline NIHSS a | 0.003 (−0.006–0.013) | 0.49 | ||
| Imaging | ||||
| Cortical infarction | 0.40 (0.14–1.16) | 0.09 | Dropped from model due to non-significance | |
| Large vessel occlusion | 0.95 (0.37–1.95) | 0.70 | ||
| Laboratory data a | ||||
| Admission WBC (× 1000/mL) | 0.02 (0.002–0.04) | 0.03 | Dropped from model due to non-significance | |
| Admission lymphocyte count (× 1000/mL) | 0.0001 (−0.12–0.12) | > 0.9 | ||
| Admission platelet count (× 1000/mL) | 0.0004 (−0.0003–0.001) | 0.22 | ||
| D-dimer (per mcg/mL) | 0.009 (0.002–0.017) | 0.02 | 1.73 (0.98–3.04) | 0.06 |
| C-reactive protein (per mg/dL) | 0.001 (−0.0008–0.003) | 0.23 | ||
Variables were entered into the multivariable model if they were significant to p < 0.2 in univariate regression (Hispanic ethnicity, congestive heart failure, cortical infarction, elevated d-dimer, elevated admission white blood cell count). Variables were retained if they remained significant to p < 0.2. Laboratory values indicate serologic studies collected closest to the time of stroke onset, unless otherwise noted
a Variables in which β with 95% confidence interval were used to estimate the effect on the outcome of cryptogenic stroke in univariate analysis. If included in the multivariable model, this effect is displayed as an odds ratio with 95% confidence interval. Multivariable regression model was clustered by site