Fig 3. Univariable comparisons of risk of combined negative outcome related to measurements of dyspnea.

Panel A–Pilot Study 1: (a) measurements of current dyspnea on admission and (b) measurements of exertional dyspnea prior to admission; Panel B–Pilot Study 2: (a) measurements of current dyspnea during first nursing shift and (b) measurements of any elevated dyspnea throughout the hospitalization. Error bars represent 95% confidence intervals calculated using Wald confidence limits for binomial proportions; tests of difference were performed using chi-squared tests with the exception of Pilot Study 2, upper right panel, which used Fisher’s exact test given small cell sizes.