Statistical analysis

We used either the t-test (for normally distributed data) or the Mann-Whitney U test (for non-normally distributed data) for continuous variables and the χ2 test for categorical variables for comparisons of the results among all samples or for comparisons of the results for each of the three groups (ACPE, COPD, and OHS) between the patients with and without severe acidosis. The following parameters were analyzed: 1) baseline values; 2) changes in arterial blood gas parameters (at baseline and discharge); 3) percentage of patients with a pH ≥7.35 after one hour, after 24 h, and at the time of discharge (to assess the RICU courses of the patients); 4) hospitalization and RICU durations (days); and 5) incidence of NIV treatment failure. Comparisons 3, 4 and 5 were repeated for each of the DNI sub-groups (without and with a DNI order). Missing values due to NIV treatment failure were replaced with the last available value. Finally, we performed logistic regression analysis using the variables with a p < 0.2 in univariate analysis to identify predictive factors for NIV treatment failure in each disease group.