Mexican-American mothers' reports of the weights and heights of children 6 months through 11 years old.

OBJECTIVE: To determine the accuracy of mothers' reports of their children's weights and heights. DESIGN AND
SETTING: Cross-sectional survey of Mexican Americans in five southwestern states.
SUBJECTS: Interviews were held with mothers of 2,578 children aged 6 months to 11 years old. MAIN OUTCOME MEASURES: Sensitivity and specificity of categories formed from reported values, and correlation of reported and measured values.
RESULTS: Probability of mothers answering "don't know" was 24% for children's weights and 51% for heights. On the average, mothers overestimated weights at the 15th percentile or lower for age and sex and underestimated weights at the 85th percentile or higher. On the average, they underestimated heights. Categories of low and high weight, height, and body mass index were created by applying absolute-value cutoffs to reported values. All the categories had low sensitivity or specificity. Age-group-specific correlation coefficients between reported and measured values ranged from .79 to .89 for weight and from .32 (for 6- through 23-month-olds) to .70 (for 9- through 11-year-olds) for height. APPLICATIONS: The use of categories formed by applying absolute-value cutoffs to mother-reported values results in frequent misclassification of individuals. Therefore, such categories should not be used to estimate relative risks associated with weight, height, and body mass index. The good correlation of mother-reported and measured weights indicates that despite their inaccuracies, reported weights well reflect the relative ranking of measured weights. Thus, the use of reported weights as a continuous variable in multivariate analyses might cause only small errors in the coefficient for weight.