Evaluating SARS-CoV-2 Pediatric Seroprevalence in Montréal: Marginal Standardization vs Raking

Overview

This study assessed the representativeness and generalizability of pediatric SARS-CoV-2 seroprevalence estimates in Montréal by comparing two weighting methods: marginal standardization and raking. Both methods yielded similar results, but raking was preferred due to its ability to adjust for multiple underrepresented population characteristics simultaneously.

Background

Seroprevalence studies are critical for estimating SARS-CoV-2 infection prevalence, especially in children who often have asymptomatic infections and are underrepresented in testing data. Many early seroprevalence studies used nonrandom convenience samples, limiting their generalizability. Weighting methods such as marginal standardization and raking can adjust for sample nonrepresentativeness by aligning study samples with target population characteristics. This study used baseline data from the Enfants et COVID-19 (EnCORE) cohort to compare these two weighting approaches in children aged 2-17 years in Montréal.

Data Highlights

Baseline data were collected from October 16, 2020, to April 18, 2021, from children aged 2-17 years across four Montréal neighborhoods. Missing data were addressed using multiple imputation, with the highest missingness in household income (31.6%) and parent’s birthplace (31.9%). Serological analyses used enzyme-linked immunosorbent assays with 95% sensitivity and 100% specificity. Weighting adjustments were applied to improve representativeness based on 2016 Canadian census data.

Key Findings

• Both marginal standardization and raking methods produced similar SARS-CoV-2 seroprevalence estimates in the pediatric population.
• Raking allowed simultaneous weighting across multiple demographic characteristics, improving adjustment for underrepresented groups.
• Nonrandom convenience sampling in seroprevalence studies can bias prevalence estimates if unadjusted.
• Weighting based on external census data is essential to enhance generalizability of seroprevalence findings.
• Multiple imputation effectively addressed missing data, particularly for socioeconomic variables collected at follow-up.

Clinical Implications

Clinicians and public health professionals should interpret pediatric SARS-CoV-2 seroprevalence data with consideration of sampling biases and the weighting methods used. Applying comprehensive weighting techniques like raking can yield more accurate prevalence estimates, aiding in better understanding of infection spread among children and informing targeted interventions.

Conclusion

Weighting methods are crucial to improve the representativeness of pediatric SARS-CoV-2 seroprevalence estimates from nonrandom samples. Raking is preferred for its flexibility in adjusting multiple population characteristics simultaneously, enhancing the applicability of study findings to the general pediatric population.

References

1. Enfants et COVID-19 (EnCORE) Study -- Baseline Seroprevalence Data Montréal 2020-2021