Fig. 2

Association between VR and mortality in the total ICU population. a Scatter plot showing the relationship between VR and ICU days. The blue line with the gray area represents the 95% confidence intervals for the fitted nonlinear trend (p < 0.001). b VR in survivor and nonsurvivor group. The VR levels were significantly higher in the nonsurvivor group than in the survivor group by the Mann–Whitney nonparametric test (p < 0.0001). c The curve of HR versus VR using univariable Cox regression with restricted cubic splines. A Cox model with restricted cubic splines identified the lowest mortality risk when VR was 1.3. There were significant positive correlations between VR and HR for VR larger than 1.3 (increase of 0.1 per VR; HR 1.05, 95% CI 1.04–1.07). d The curve of HR versus VR using multivariable Cox regression with restricted cubic splines. The curve indicated a nadir of 1.3 even after adjusting for PEEP and the P/F ratio, and a highly positive correlation between VR and HR was also observed when VR was greater than 1.3 (increase of 0.1 per VR; HR 1.11, 95% CI 1.08–1.14). Cubic spline curves are shown as a solid line, with the shaded area representing the 95% confidence intervals