(P17-043-23) Effect of Different Methods of Measuring Early-Infant Feeding Practices on Estimates of Prevalence and Individual-Level Classification of Early-Infant Feeding Outcomes

Objectives: To assess the effect of different methods of measuring infant feeding on estimates of population prevalence of early-infant feeding outcomes and classification of infants regarding feeding outcomes up to 6 months of age.

Methods: We compared methods of measuring infant feeding in the Synbiotics for the Early Prevention of Severe Infections in Infants Observational Cohort study conducted from November 2020 to September 2022 in Dhaka, Bangladesh. Infant feeding in the past 24 hours and 7 days were collected longitudinally during routine visits from baseline (0-4 days of life) to 6 months of age (median 187 days, interquartile range 180, 204), with an additional since-birth recall at 6 months. Early initiation of breastfeeding (EIBF) was defined as initiation of breastfeeding within 1 hour of birth, exposure to prelacteal feeds was defined as anything other than breastmilk within 2 days of birth, and breastfeeding patterns were classified as exclusive (EBF), predominant, partial, and not breastfed. Proportions of infant feeding outcomes derived from variations in recall period, timing, and number of data collection events (i.e., metrics) were compared (McNemar’s test) and the individual-level concordance was assessed (Cohen’s kappa (κ)).

Results: Overall, 25042 visits were included from 1926 infants. Baseline and since-birth metrics produced similar estimates for EIBF (18% vs 20%, p-value=0.58) and prelacteal feeds (5.1% vs 6.6%, p-value=0.082), but agreement was poor (EIBF κ = 0.067, prelacteal κ=0.038). The prevalence of breastfeeding patterns up to 2 and 3 months were significantly different by metric. For example, EBF up to 2 months was 50% using longitudinal data, 84% using one 24-hour recall at 2 months, and 87% using since-birth recall (p-values < 0.001), and there was more agreement between one 24-hour recall and longitudinal metric (κ=0.33) than between since-birth and longitudinal metric (κ=0.16). Similar differences were observed, but less pronounced, for breastfeeding patterns up to 6 months.

Conclusions: Cross-sectional and retrospective recall lead to inaccurate estimates of breastfeeding patterns. Since indicators of early-infant feeding patterns inform public health programs, consideration must be given to the methods used to obtain infant feeding data.

Funding Sources: The Bill and Melinda Gates Foundation