*2.3. Data Analysis*

We analysed compliance with screening for active TB, LTBI, and the whole protocol, also taking patient characteristics into account. We also assessed the capacity of the adopted procedures to detect LTBI and active TB. In order to do this, we computed odds ratios (ORs) along with their 95% confidence interval (CI) using crude and multivariable logistic regression models. The potential confounders included in multivariable analysis were adjusted for sex, age, presence of cough, education, knowledge of Italian, employment status (employed versus unemployed), years from arrival in Italy and homelessness. We included knowledge of Italian language because it could affect compliance, influencing patient's ability to understand the motivations of the clinical examinations and therapies proposed. When the use of multivariable analysis led to exceedingly high statistical instability, we limited our assessment to crude estimates only (unadjusted for other variables). Risk analysis was performed by calculating odds ratios estimated from conditional logistic regression with crude and multivariate models. When the odds ratios could not be calculated, we applied the chi2 test and probability according to Fisher's exact Test. For the evaluation of the differences between continuous variables, we applied the *t*-test. For PE, HS, and IPT, we performed only the crude analysis because of the smallness of the sample, causing too much instability in the statistical analysis.
