Development and validation of portion size food photographs to determine maize intake of young children in rural Eastern Cape Province

Abstract

Objective: The objective of the study was to develop and validate a portion size food photograph series to more accurately determine maize intake of infants aged 6 - 24 months living in deep rural areas in the Eastern Cape (EC) province of South Africa. Design: This was a community based, cross sectional, observational study. Participants: Mothers/caregivers aged 18 years and older taking care of infants and young children (6 - 24 months). Outcome measure: This study developed a food photograph series to improve portion size estimation of maize dishes consumed by infants and young children. The food photograph series were developed to be used alongside the validated quantitative food frequency questionnaire (QFFQ). The food photograph series consisted of photos representing portions ranging from teaspoons, tablespoons to large serving ladles. Participants were first shown the food photograph series and asked to identify the portion size most frequently given to the infant / child. Then the fieldworker recorded the portion. As part of validation participants were asked to dish up the amount of food usually consumed by the infant or child and the fieldworker recorded the dished up portion size. Results: The data distribution was tested with the Shapiro-Wilk‟s test and found normal. Thus all statistical tests were conducted on parametric data. The paired t-test showed a significant difference (p < 0.05) between the photograph portion sizes and the dished up portion sizes for two dishes (soft porridge and crumbly pap). Agreement at group level was good for all dishes accept soft porridge and crumbly pap when the t-test was conducted. The percentage difference was acceptable for only three dishes (soft porridge, maize meal and pumpkin and crumbly pap). However, agreement at group level in terms of percentage difference was acceptable for soft porridge, maize meal and pumpkin and crumbly pap and not for the other dishes. Thus, even though the t-test indicated poor agreement for soft porridge and crumbly pap, it is acceptable when using the percentage difference. Strength of association was measured with Pearson correlation coefficients. Results indicated that the association was acceptable, if not strong for only stiff pap and samp and beans. For all other dishes this was weak. Lastly, the Bland-Altman analyses indicated good agreement at individual level for all dishes, although this was mostly due to the wide limits of agreements. Bias was present for all dishes with the exception of maize and pumpkin. When looking at the overall validity of the dishes it is clear that the food photograph series is valid for three dishes, maize meal and pumpkin, stiff pap and samp and beans. Conclusion: The accuracy of portion size estimates is critical in the assessment of food consumption patterns. The food photograph series is valid for three dishes (maize meal and pumpkin, stiff pap and samp and beans).