ABSTRACT
Minimum dietary diversity for women of reproductive age (MDD-W) was validated as a population-level proxy of micronutrient adequacy, with indicator data collection proposed as either list-based or open recall. No study has assessed the validity of these two non-quantitative proxy methods against weighed food records (WFR). We assessed the measurement agreement of list-based and open recall methods as compared to WFR (i.e., reference method of individual quantitative dietary assessment) for achieving MDD-W and an ordinal food group diversity score. Applying a non-inferiority design, data were collected from non-pregnant women of reproductive age in Cambodia (n = 430), Ethiopia (n = 431), and Zambia (n = 476). For the pooled sample (n = 1337), proportions achieving MDD-W from both proxy methods were compared to WFR proportion by McNemar's chi-square tests, Cohen's kappa, and receiver operating characteristic (ROC) analysis. Ordinal food group diversity (0-10) was compared by Wilcoxon matched-pairs signed-rank tests, intraclass correlation coefficients (ICC), and weighted kappa. MDD-W food groups that were most frequently misreported (i.e., type I and II errors) by the proxy methods were determined. Our findings indicate statistically significant differences in proportions achieving MDD-W, ordinal food group diversity scores, and ROC curves between both proxy methods and WFR (p < 0.001). List-based and open recall methods overreported women achieving MDD-W by 16 and 10 percentage points, respectively, as compared to WFR (proportion achieving MDD-W: 30%). ICC values between list-based or open recall and WFR were 0.50 and 0.55, respectively. Simple and weighted kappa values both indicated moderate agreement between list-based or open recall against WFR. Food groups most likely to be misreported using proxy methods were beans and peas, dark green leafy vegetables, vitamin A-rich fruit and vegetables, and other fruits. Our study provides statistical evidence for overreporting of both list-based and open recall methods for assessing prevalence of MDD-W or ordinal food group diversity score in women of reproductive age in low- and middle-income countries. Operationalizing MDD-W through qualitative recall methods should consider potential trade-offs between accuracy and simplicity.
