Active Voice: Flawed Method May Underestimate Childhood Obesity
By Daniel P. O'Connor, Ph.D.

Active Voice is a column by experts in science, medicine and allied health. The viewpoints expressed do not necessarily reflect positions or policies of ACSM.

Daniel P. O’Connor, Ph.D., is an ACSM member and Assistant Professor of Kinesiology in the Department of Health and Human Performance at the University of Houston. His primary focus is applied biostatistics and much of his research is interdisciplinary, involving collaborations in such areas as obesity, fitness, orthopedics and physiology. Dr. O’Connor presented research related to this commentary at ACSM’s Annual Meeting and World Congress on Exercise is Medicine®, held in Baltimore in June 2010.

Many of us have seen headlines and media reports about the obesity epidemic plaguing U.S. children, with prevalence exceeding 35 percent by some estimates. While these numbers are alarming, many of these reports use national surveys or databases containing parent-reported values for their child’s height and weight, and these values are often used to compute body mass index (BMI) and identify obesity rates. The inaccuracy of proxy reporting is well known, but we were surprised to find few reports investigating the accuracy of parent-reported height and weight across the entire span of childhood.

Using archival data from an orthopedic clinic, we compared measures of parent-reported height and weight for 1,430 children (55 percent boys) who were between 2 and 17 years old. None of these children were seeking treatment for weight-related issues. Using objectively measured values, we determined that 17 percent of the children were overweight and 19 percent were obese.

We found that parents tended to overestimate boys’ height and underestimate girls’ height, and this error was larger when the reporting parent was the opposite sex of the child. In addition, almost half of the parents underestimated their child’s weight, and errors in reporting weight tended to be larger for girls and to increase with the child’s age. Weight errors were also larger in children who were overweight or obese. Ethnicity also played a role, with larger errors reported by African-American and Hispanic parents than by Caucasian parents.

The most striking finding to us, however, was that using the parent-reported values to compute BMI and obesity status, using CDC guidelines, resulted in about one in five obese children—21 percent—being missed in the count and not identified as obese! This suggests that obesity prevalence for children may be underestimated in studies where parent-reported values, such as we studied, are utilized.

Several useful points are illustrated by these findings. First, parents’ recall of their children’s height and weight may not be reliable, so children’s stature should be measured whenever possible. Second, the errors of parent-reported height and weight vary by the child’s sex and age, and by the parent’s ethnicity, which means that there is no simple way to identify parents who may be misreporting this information. Third, we should remember these errors when reading reports that depend on parent-reported information, and note that conclusions from such reports related to childhood obesity may actually be underestimating the extent of the problem.

Finally, the errors in parent-reported height and weight are substantial enough to affect calculation of BMI, which interferes with identification of obesity using the CDC guidelines. Considering mounting evidence that the relationship between BMI and measures of adiposity (presumably a causal factor in obesity-related disease) also varies by gender, age and ethnicity, perhaps relying on BMI as an indicator of obesity should be re-examined. These points should be of interest not only to epidemiologists and other public health researchers, but also to clinicians and practitioners who are involved in screening or intervening in childhood obesity.