Validity of food photographs for portion size estimation among Nairobian women of reproductive age

Introduction The coexistence of obesity and undernutrition referred to as the double burden of malnutrition (DBM), has been reported in low- and middle-income countries, including Kenya. Women are especially vulnerable to malnutrition due to their nutritional needs during pregnancy and lactation, highlighting the importance of assessing their dietary intake to address their food and nutrition security. However, accurate portion size estimation has been found to be challenging during dietary surveys. Food photographs are useful for estimating portion sizes and are easy to use in dietary surveys. Objectives To assess the validity of food photographs in portion size estimation of commonly consumed foods among women of reproductive age (13-45 years) in Nairobi. The second aim was to examine the association of sociodemographic characteristics such as age and educational level on the accuracy of portion size estimations. Methodology A validation study was carried out among 206 women of reproductive age (13-45 years) in Kahawa West, Nairobi County. Eleven commonly consumed Kenyan foods (ugali, chapati, rice, beans, beef, sukuma wiki, tilapia, sweet banana, orange, pawpaw and watermelon) were chosen to be tested from the Photographic Food Atlas for Kenyan Adolescents (9-14 years). The participants were randomly served pre-weighed food portions based on those in the Food Atlas (in most cases there were three different portion sizes). In most cases, the food portions were similar in weight to those depicted in the food photographs. After the meal, participants were asked to estimate the amount of food consumed using food photographs from the Food Atlas. Any leftover food was recorded and the amount of food consumed was calculated as the difference between the amount weighed and the amount left. Validity was assessed by calculating the mean percent difference between estimated and consumed portions, percentage of estimates within -10 to 10% of consumed portion size, Spearman’s correlation coefficients and Bland–Altman limits of agreement. Pearson’s Chi-square test was used to examine the associations between the accuracy of estimations and participants’ age and educational level. Results Correlations between consumed and estimated portions were significant for all food items except for watermelon (p=0.380). The proportion of participants with estimates within ± 10% of the consumed portion size was above 50% for four of the food items including chapati, sukuma wiki, rice and beef; four other food items, including ugali, pawpaw, orange and tilapia, correct estimates ranged from 43-47%; and for the remaining three food items, including beans, sweet banana and watermelon, correct estimates ranged from 15-34%. Extreme mean differences between the consumed and estimated portion sizes by photograph were between -45% for beans to 563% for pawpaw. In most of the food items, small portions were overestimated while large portions were underestimated. Bland–Altman plots showed wide limits of agreement. The accuracy of estimations was not associated with participants’ age or educational level. Conclusion The food photographs for chapati, sukuma wiki, rice and beef from the Food Atlas for Kenyan adolescents seem to be a valid tool for quantification of portion sizes for women of reproductive age in Nairobi. However, the findings of this study suggest that further improvements are needed to the Food Atlas for wider use in Kenya