Abstract
BACKGROUND: Both the pattern and duration of breast feeding are important determinants of health outcomes. In vertical HIV transmission research, reliable documentation of early breast feeding practices is important in order to correctly attribute postnatal transmission to feeding pattern. AIMS: To validate methods of collecting data on the duration of exclusive breast feeding (EBF) in an area of South Africa with a high HIV prevalence rate. METHODS: A total of 130 mothers were interviewed weekly, postnatally. At every interview a 48 hour and a seven day recall breast feeding history were taken. A subset of 70 mothers also received two intermediate visits per week during which additional 48 hour, non-overlapping, recall interviews were conducted. Ninety three infants were revisited at 6-9 months of age when mothers' recall of EBF duration from birth was documented. The different methods of recalling EBF status were compared against an a priori "best comparison" in each case. RESULTS: Reported breast feeding practices over the previous 48 hours did not reflect EBF practices since birth (specificity 65-89%; positive predictive value 31-48%). Six month EBF duration recall was equally poor (sensitivity at 2 weeks 79%; specificity 40%). Seven day recall accurately reflected EBF practices compared with thrice weekly recall over the same time period (sensitivity 96%, specificity 94%). CONCLUSIONS: 48 hour EBF status does not accurately reflect feeding practices since birth. Long term recall data on EBF are even more inaccurate. We recommend that data on duration of EBF be collected prospectively at intervals of no longer than one week.