Fig. 2
From: Breath-holding as a novel approach to risk stratification in COVID-19
Individual breath-holding measurements in COVID-19 patients who experienced the adverse primary composite outcome (and required ventilatory support “VS + ”, N = 11), in COVID-19 patients without adverse outcomes (“VS − “, N = 39), and controls (N = 23). Horizontal black bars overlying individual data indicate group mean values. A. Mean desaturation after 20-s, unadjusted. Inset: desaturation profile for each group (mean ± SEM desaturation at any time, delay corrected). B. Baseline oxygen saturation, a potential confounder for mean desaturation, was different between groups and independently associated with adverse COVID-19 outcomes. C. Maximal breath-hold duration (*shown adjusted for baseline SpO2 [+ 2.6 s per %Hb below 97.6] and mean desaturation [+ 0.6 s per %Hb above 1.57]); shorter adjusted durations were interpreted as increased chemosensitivity (or sensitivity to dyspnea). Data on ventilatory response to breath-holds (not shown) were similar between groups. D. Left: Multivariable logistic regression model output for each individual patient (parsimonious model, Table 2) shows that mean desaturation and maximal breath-hold duration (+ covariates baseline SpO2, body mass index, cardiovascular disease) provides good outcome discrimination (threshold ~ 0); a score of 0 represents 50% probability of the adverse primary outcome. Each 1-point increase represents a log (2.7-fold) increase in likelihood of the primary outcome. Right: Reference model without breath-holding measures (baseline SpO2, body mass index, cardiovascular disease) showed significantly reduced outcome discrimination (p = 0.007, random permutation analysis)