High Sparrow

Implausibly High BMI

Methods matter. A new study published today in Obesity shows that the prevalence of severe early childhood obesity may be twice as high as previously thought. This is because of limits for what are considered to be impossibly high BMI values in standard methods for estimating obesity prevalence. Those limits are no longer working right because severe obesity in children has become so much more common and more severe. As a result, values that were once simply errors are now valid observations of children with severe obesity. The authors conclude:

Our results indicate that among 2- to 4-year-olds in PedNSS, many of the high values of weight, height, and BMI considered to be biologically implausible are unlikely to be errors.

As compared with the estimated prevalences based on the current WHO cut points, increasing the BIV [biologically implausible value] cut points, either through the use of higher z-scores or by using the age-specific NHANES maximums for weight, height, and BMI, increased the prevalence of both obesity and severe obesity by about 1%.

That one percent difference in severe obesity for two to four-year-olds takes the prevalence estimate from 1.18% to anywhere from 2.09% to 2.34%, depending upon a range of different methods for determining which values are actually errors.

These methods matter not only for measuring the prevalence of childhood obesity, but for discerning where the trends are headed. Early childhood obesity prevalence estimates are increasingly affected by what were once thought to be impossibly high BMI values. So an increasing rate of severe obesity results in increasingly inaccurate estimates because more kids who have the most severe obesity simply don’t get counted.

This is one more reason that cherry picking statistics for preschool childhood obesity is a particularly bad idea. Two years ago false claims about plummeting early childhood obesity rates made headlines. Likewise, researchers from the NYC Department of Health published an analysis of declining rates of severe childhood obesity in the city. But those estimates were questionable because the decline resulted from discarding a growing number of implausibly high BMI values.

Without careful attention to methods, bias creeps into research findings. We start seeing the results that we hope to see, instead of reality.

Click here to read this fascinating new study.

High Sparrow, photograph © Gemma Louise / flickr

Subscribe by email to follow the accumulating evidence and observations that shape our view of health, obesity, and policy.


March 17, 2016

9 Responses to “Implausibly High BMI”

  1. March 19, 2016 at 8:06 am, Allen Browne said:

    I like that: “Without careful attention to methods, bias creeps into research findings. We start seeing the results that we hope to see, instead of reality.”

    • March 19, 2016 at 8:16 am, Ted said:

      Thanks Allen.

  2. March 20, 2016 at 12:33 pm, Miroslav Toms said:

    I think it is another reason for using new approach for obesity definition. I personally use it and it works perfectly. I recommend it for general use. (As AHA makes)

  3. March 20, 2016 at 1:38 pm, Miroslav Toms said:

    Sorry, my comment was not exactly to the topic of the text. Every measured value collected to database has its technical cut off (BIV) of course, but in term of BMI there is problem with SDS (Z score) for extreme values over 99.perc. I believe that statistically is more precise to use method of Flegal – in percent of obesity limit.

    • March 20, 2016 at 7:20 pm, Ted said:

      Thanks, Miroslav.

  4. March 23, 2016 at 2:36 pm, Carolyn said:

    Note that the NHANES data on obesity in children does not use these cutoffs for implausible values and is not affected by this problem.

    • March 23, 2016 at 5:27 pm, Ted said:

      Thanks, Carolyn. NHANES is just a dataset. Individual researchers decide how they will screen the data for BIVs when they are analyzing it.

  5. March 25, 2016 at 7:54 am, Carolyn said:

    The published estimates of childhood obesity from NHANES do not apply the implausible value cutoffs and are not affected by this problem.

    • March 25, 2016 at 8:40 am, Ted said:

      Thanks, Carolyn.