Stata v 14.1 se

All variables were tabulated as frequencies and valid percentages and each was described by time period: pre-pandemic (before February 21, 2020) and during the pandemic (after February 21, 2020), taking into account that 21 February was the date in which the first native Italian COVID-19 case was detected. We used Multiple Correspondence Analysis (MCA) to identify variables associated with vaccine stance. The following variables were used to build the graph (“active” variables): tone of voice, stance, kind of author, information sources, kind of vaccine, and target population.
Data analysis was carried out using STATA V.14.1 SE (StataCorp, College Station, Texas, USA).

Data were summarised using means and standard deviations (±SD) for continuous parametric data, and median (IQR) for non-parametric data. A Student’s t-test was used to test for differences between men and women on parametric continuous variables within each site, while Kruskal-Wallis test was used for non-parametric data. Chi-square test was used for categorical data. Due to significant sex differences within sites (online supplemental tables 1–4), all further analyses were stratified by sex. Since the outcome had three categories, that is, zero conditions, one condition and at least two conditions (multimorbidity) of the seven conditions under study, multinomial logistic regression was used to explore the factors associated with either one condition or at least two conditions, with none as the reference group, in men and women separately. A p value<0.05 was considered statistically significant in all tests carried out using STATA V.14.1 SE.
To understand the different multimorbidity patterns between men and women, within and between countries, and between regions, that is, South (South African sites), East (Kenyan site) and West (Ghana and Burkina Faso) Africa, different diseases were plotted using UpSetR, an R package.22 (link) Separate analyses were done for men and women, separated by site and then geographic region.