ABSTRACT
BACKGROUND: Rapid and reliable health data on SARS-CoV-2 infection among pregnant individuals are needed to understand the influence of the virus on maternal health and child development, yet the validity of self-reported COVID-19 testing and diagnosis remains unknown. OBJECTIVES: We assessed the validity of self-reported COVID-19 polymerase chain reaction (PCR) testing and diagnosis during delivery among postpartum respondents as well as how diagnostic accuracy varied by respondent characteristics. METHODS: We validated receipt of a COVID-19 PCR test and test results by comparing self-reported results obtained through an electronic survey to electronic medical record data (gold standard) among a cross-sectional sample of postpartum respondents who delivered at four New York City hospitals between March 2020 and January 2021. To assess validity, we calculated each indicator's sensitivity, specificity and the area under the receiver-operating curve (AUC). We examined respondent characteristics (age, race/ethnicity, education level, health insurance, nativity, pre-pregnancy obesity and birth characteristics) as predictors of reporting accuracy using modified Poisson regression. RESULTS: A total of 276 respondents had matched electronic record and survey data. The majority, 83.7% of respondents received a SARS-CoV-2 PCR test during their delivery stay. Of these, 12.1% had detected SARS-CoV-2. Among those tested, sensitivity (90.5%) and specificity (96.5%) were high for SARS-CoV-2 detection. The adjusted risk ratio (aRR) of accurate result reporting was somewhat lower among Hispanic women relative to white non-Hispanic women (aRR 0.90, 95% CI 0.90, 1.00) and among those who had public or no insurance vs. private (aRR 0.91, 95% CI 0.82, 1.01), controlling for recall time. CONCLUSION(S): High recall accuracy result reporting for COVID-19 PCR tests administered during labour and delivery suggest the potential for population-based surveys as a rapid mechanism to obtain accurate data on COVID-19 diagnostic history. Additional psychometric research is warranted to ensure accurate recall across respondent subgroups.