Table 2.
Association between breath-holding measurements and adverse outcomes of COVID-19
| Primary model β ± SEM (p value) |
Parsimonious model β ± SEM (p value) |
|
|---|---|---|
| Breath-holding measurements | ||
| Mean desaturation (%Hb) |
1.27 ± 0.59 (0.002) |
1.25 ± 0.54 (0.001) |
| Maximal breath-hold duration (s) |
0.10 ± 0.05 (0.037) |
0.10 ± 0.05 (0.020) |
| Ventilatory response (%baseline) |
0.00 ± 0.01 (0.9) |
— |
| Covariates | ||
| Body mass index (kg/m2) |
0.22 ± 0.22 (0.3) |
0.27 ± 0.19 (0.157) |
| Baseline SpO2 (%) |
− 0.44 ± 0.22 (0.001) |
-0.45 ± 0.22 (0.001) |
| Cardiovascular disease |
4.47 ± 2.71 (0.064) |
5.27 ± 2.38 (0.019) |
Association between breath-holding measurements at admission and the adverse composite primary outcome in COVID-19 (multivariable logistic regression). Data shown are β ± SEM (p value); β describes the increase in log-odds of the adverse outcome per change in exposure variable. Primary model: The breath-holding measurements significantly improved the model (likelihood ratio 0.0073, p = 0.02) over a reference model with covariates only (body mass index, baseline SpO2, cardiovascular disease [1 = Present, 0 = Absent], plus age and sex [not shown]). p values are based on likelihood ratio tests. The parsimonious model is a simplified and refined version of the primary model (age, sex, and ventilatory response were removed [p > 0.2]; Intercept = 25.31 ± 17.31). The potential predictive value of the model is illustrated in Fig. 2D and a tool for risk calculation is provided in Additional file 2