Stata 15.1 software package for windows

Continuous variables were reported as either mean and standard deviation (SD) or median and interquartile range (IQR) on the basis of their distribution (assessed using the Shapiro–Wilk test). Comparison of variables between groups were performed by an unpaired Student t-test or Mann–Whitney U test, as appropriate. Categorical variables were expressed as the absolute number and percentage and statistical significance was assessed by Pearson’s Chi-square test. In a second step, the analysis of follow-up data was carried out to evaluate the effectiveness of the intervention. Frequencies of eHEALS items were compared between baseline and follow-up via Pearson’s Chi-square test. The overall mean values of the eHEALS scale were compared between the two time points and statistical significance was assessed with a paired t-test. Willingness to pay was analyzed reporting the mean and SD of the program satisfaction for each cost item: statistical significance was assessed by one-way analysis of variance (ANOVA). The relationship between SOTU subscales, overall SOTU, satisfaction with the training program, and eHEALS was estimated by calculating the pairwise correlation coefficients. The statistical significance was set at p < 0.05. Statistical analysis was conducted using the Stata 15.1 Software Package for Windows (StataCorp, College Station, TX, USA).

Demographic, clinical, radiological, and laboratory characteristics of patients, both survivors and non-survivors, were compared by Student’s t-test or Mann–Whitney U test when appropriate for continuous variables and chi-square test for categorical ones. The association between each variable and mortality was explored by unadjusted Cox proportional hazard models. The CXR score, frailty, and variables significantly associated with the outcome in preliminary models were included in multivariable analysis. Five multivariable Cox proportional hazard models were built to obtain adjusted estimates of the association between exposure variables and the study outcome. The accuracy of exposure variables in predicting mortality was estimated by the Area Under the Receiver Operating Characteristic Curve (ROC). Finally, we investigated the additive effect of the CFS and other significant predictors on the predictive ability of the CXR score. Changes in Area Under the Curve (AUC) and categorical net reclassification index (NRI) with 1000 bootstrap samples to estimate 95% CIs were calculated. Statistical analysis was conducted using the Stata 15.1 Software Package for Windows (StataCorp, College Station, TX, USA).