Social Desirability Bias in a Randomized Controlled Trial That Included Breastfeeding Promotion in Western Kenya

Background: Breastfeeding promotion is associated with improved measures of breastfeeding practice; however, most studies rely on participant-reported outcomes. Objectives: This study aimed to evaluate the likelihood of bias in self-reported breastfeeding outcomes. Methods: We used data from the WASH Benefits randomized controlled trial in Kenya (clinicaltrials.gov NCT01704105), which included intervention arms that received infant and young child feeding messages (nutrition group) and arms that did not (non-nutrition group) to examine recall bias in the assessment of early initiation of breastfeeding (EIBF) and exclusive breastfeeding (EBF) duration. Pregnant women were eligible for inclusion, but 40% of infants were born before the first visit by the health promoter. We used the random variability in the timing of the infant birth relative to the first contact with the health promoter to examine recall bias in EIBF. We used ≤5 rounds of repeated surveys to examine bias in recall of EBF duration. Results: There was a significant effect of the intervention group on EIBF [prevalence ratio: 1.50, 95% confidence interval (CI): 1.43, 1.57] and a significant difference in the duration of EBF [nutrition group: 6.0 mo; interquartile range (IQR): 3.0–6.0; nonnutrition group: 3.0 mo; IQR: 0.5–6.0; P < 0.001]. There was not a significant interaction with birth timing relative to the first contact with the health promoters (P = 0.915), suggesting that the observed main effect on EIBF was due to recall bias. We found that 75.9% of mothers who initially reported having ceased EBF before 6 mo in the nutrition group changed their response to 6 mo or later in subsequent surveys. Only 32.5% of mothers in the non-nutrition group followed this pattern. Conclusions: These data provide strong evidence of bias in reporting of breastfeeding practices in this nonblinded intervention trial and should serve as a caution to researchers who rely on self-reported outcomes of behavior change interventions.This trial was registered at clinicaltrials.gov as NCT01704105.