2.3. Statistical Analysis

Variables were expressed as median and interquartile range (IQR) or as percentages, as appropriate. The normality of distribution of continuous variables was assessed with the Kolmogorov-Smirnov test. PCT levels on admission were considered as the independent variable. The population was divided into four groups, according to PCT levels: <0.05 ng/ml (group 1, normal), ≥0.05 ng/ml and <0.5 ng/ml (group 2, mild elevation), ≥0.5 ng/ml and <2 ng/ml (group 3, severe elevation), and ≥2 ng/ml (group 4, extreme elevation). These cut-offs were compatible with those used for prescription of antibiotic therapy in suspected bacterial infections [27]. All demographic, clinical, laboratory, and outcome data were compared across these four groups using Kruskal-Wallis test, linear regression, univariate general linear model, one-way analysis of variance (ANOVA) for continuous variables, and Pearson chi-square test and logistic regression for dichotomous variables. All demographic, radiological, and virological characteristics showing a significantly different trend across PCT classes were considered as potential confounders in regression models. In particular, age, sex, positive RT-PCR test for SARS-CoV-2, and presence of consolidations on chest HRCT were considered as possible confounders, for their established influence on COVID-19 clinical presentation and prognosis [28, 29].

Then, the association of different PCT classes upon admission with in-hospital mortality was evaluated separately for patients older than 75 and for patients aged 75 or younger with Cox regression multivariate analysis. Included as possible confounders were all demographic, radiological, and virological characteristics with significantly different trend across PCT classes (namely, age, sex, presence of consolidations on chest HRCT, and positive nasopharyngeal swab for SARS-CoV-2) and treatment with ACE-I or ARBs, for its possible association with COVID-19 outcome.

P values were considered significant when <0.05. Statistical analyses were performed with the SPSS package (v.26, IBM, Armonk, United States).