Accuracy of self-reported birth outcomes relative to birth certificate data in an Internet-based prospective cohort study

Background: The accuracy of birth outcome data provided by Internet-based cohort study participants has not been well studied.

Methods: We compared self-reported data on birth characteristics in Pregnancy Study Online (PRESTO), an Internet-based prospective cohort study of North American pregnancy planners, with birth certificate data. At enrolment, participants were aged 21-45 years, attempting conception, and not using fertility treatment. Women completed online questionnaires during preconception, early and late pregnancy, and postpartum. We requested birth certificate data during 2014-2019 from seven health departments in states with the most participants. After restricting to singleton births, we assessed specificity, sensitivity, and agreement comparing self-reported data from postpartum questionnaires with birth certificate data for gestational age at delivery (GA) and birthweight (grams). Our primary measure of self-reported GA (weeks) was calculated as [280-(due date-birth date)]/7. We used log-binomial regression to assess predictors of agreement.

Results: We linked 85% (771/909) of women in selected states. Median age of women was 30 years (range: 21-42), 84% had ≥ 16 years of education, nearly 96% were married, 12% had household incomes <$50 000, 32% were parous, and 85% identified as non-Hispanic White. Median recall interval was 6 months. Among those with self-reported data, 89% reported the same GA as the birth certificate and 98% reported GA within 1 week of the birth certificate. Self-report of preterm birth (GA < 37 weeks) agreed with information from birth certificates for 100% of women; sensitivity was 100%, and specificity was 99%. Self-reported low birthweight (<2500 grams) agreed with birth certificates for 93% of women; sensitivity and specificity were 93% and ≥99%, respectively. Predictors of poorer agreement included higher parity and longer pregnancy attempt time for GA, and lower education and longer recall interval for birthweight.

Conclusion: Self-reported data on GA and birthweight from an Internet-based cohort showed high accuracy compared with birth certificates.