Two-Year Test-Retest Reliability of the Breastfeeding Duration Question Used By the Pregnancy Risk Assessment Monitoring System (PRAMS): Implications for Research

Introduction: A large literature exists on positive sequelae of breastfeeding, relying heavily on maternal self-reported infant feeding behaviors. Many such studies use PRAMS data, though estimates of reliability for the breastfeeding duration question on PRAMS have not been published. Methods: We used data from Oregon PRAMS (respondents are a median 3.5 months postpartum) and PRAMS-2 (median 25 months) to assess test-retest reliability of maternal self-reported breastfeeding duration, among women who had weaned prior to completing the PRAMS survey. Results: The sample-wide kappa for the paired, self-reported breastfeeding duration was 0.014, and the intraclass correlation coefficient was 0.17, both of which indicate poor agreement. More than 80% of women reported a longer duration on PRAMS-2; the median (interquartile range) difference was +1.0 (0.31 – 2.1) months. Discussion: Recent literature on this topic from high-income countries falls into two categories: entirely retrospective versus “prospective” reliability assessments. Entirely retrospective assessments (both inquiries occur well after weaning) universally report exceedingly high reliability, whereas “prospective” assessments (women report infant feeding behavior during infancy, immediately after weaning, and some years later are asked to replicate their original answer) universally report poorer reliability. Interestingly, all “prospective” reliability studies, including ours, found that women over-report past breastfeeding durations by about 1 month upon the second inquiry. Researchers need not refrain from using maternal self-reported breastfeeding durations, because participants are largely still ranked correctly, relative to each other. However, such research efforts must avoid attempting to determine any optimal threshold duration.