To estimate seroprevalence (objective 1), we used a Bayesian logistic regression model, adjusted for the antibody test sensitivity and specificity performances [24 (link)]. Seroprevalence estimates were weighted by the age and sex distribution of the population of each canton. We investigated the association between potential risk factors and seropositivity (objective 2) using Poisson regression models and expressed as prevalence ratios (PR) and 95% confidence intervals. Robust variance estimators were used to relax the assumption that the outcome distribution followed a Poisson distribution. Sex, age, educational level, BMI, income, employment status, number of children in the household, comorbidity score and smoking habit were included in the models (hereafter, model 1). Results were stratified by study period. Models for period 3 were adjusted for vaccination status (hereafter, model 2; in Switzerland the vaccination campaign started at the end of December 2020, during the second period of this study). To investigate seropositivity risk factors and their changes over time (objective 2), we used multiple imputation by chained equations to impute any missing data (30 imputations). Statistical analyses were conducted using Stata 17 software (Stata Corp, TX, 2021) and R Statistical Software (version 4.1.2; R Foundation for Statistical Computing, Vienna, Austria).
We also performed several sensitivity analyses: (1) including participants who had completed the questionnaire on demographic and socioeconomic characteristics, adherence to COVID-19 preventive measures, and health status, more than 60 days before and after their blood sample; (2) including a third age category (from 20 to 34 years old; based on the hypothesis that people in this category could have had more social interactions and therefore an increased risk of being infected) and (3) using a score computed from the preventive behaviours variables (hereafter, preventive behaviours score). The score goes from 0 to 4; one point for every “occasionally/rarely” answer to the questions on preventive behaviours. The higher the score, the less frequent the adherence to preventive behaviours.
We also performed several sensitivity analyses: (1) including participants who had completed the questionnaire on demographic and socioeconomic characteristics, adherence to COVID-19 preventive measures, and health status, more than 60 days before and after their blood sample; (2) including a third age category (from 20 to 34 years old; based on the hypothesis that people in this category could have had more social interactions and therefore an increased risk of being infected) and (3) using a score computed from the preventive behaviours variables (hereafter, preventive behaviours score). The score goes from 0 to 4; one point for every “occasionally/rarely” answer to the questions on preventive behaviours. The higher the score, the less frequent the adherence to preventive behaviours.
Free full text: Click here
