Statistical analysis
For the analysis, only those who tested positive on both WANTAI and Abbott assays were classified as SARS-CoV-2 antibody positive (Figure 1). Those who tested negative on only one or both assays were classified as negative.
Sampling weights were calculated based on selection probabilities at each site. Non-response weights (for questionnaire and each laboratory test) were calculated at the individual level. Weights were calibrated to regional population estimates by age and sex, and then normalized to the total sample size.14 Weighting was carried out in R using the glm, anesrake, and survey packages.
Seroprevalence estimation was done in SAS using the SURVEYFREQ procedure and the calibrated survey weights. Taylor series variance estimation was used to account for the clustered sample design, treating each survey site as a separate cluster. Unweighted counts and proportions were also computed in SAS. The χ2 test was used to assess the differences in prevalence across demographic characteristics.
We conducted unadjusted and multivariable logistic regression, incorporating survey weights, to assess the associations between demographic and behavioral characteristics with seropositivity. The analysis included participants with complete age and sex data. Final multivariable models included variables with p<0.05 and variables (sex, age, residence, region, number of household members, comorbid conditions, and history of COVID-19, travel) chosen a priori based on their likely association with the risk of SARS-CoV-2 past infection.
Regional-level seroprevalence estimates were applied to 2020 region-specific population projections from the Cameroon Statistics Agency to estimate the total number of SARS-CoV-2 infections in each region that had occurred in the survey period (October-December 2020). These numbers were compared with the total number of reported cases in the country at the end of the study (December 2020) to estimate the ratio of reported cases to total SARS-CoV-2 infections.
Agreement in test results between the WANTAI and Abbott assays was also evaluated using kappa statistics.