« Statbusters: To divide or not to divide by 365 | Main | Statbusters: Were you drunk on Labor Day? »


Feed You can follow this conversation by subscribing to the comment feed for this post.


BMI is a case of using the numbers you have simply because you have them, not because they are the right numbers. It is a prime example of what is wrong with a lot of the "Data speaks for itself" by prognostications around Big Data. Yes, BMI is easy to measure, but then again so are student test scores. Doesn't make either of them the best tool for evaluating a human's health risk or a teachers skill/performance.

As to the correlation, yes it exists, but their is no inherent reason why height/weight ratio should be predictive of health (which is what we are actually interested in). DXA, for its practical flaws, is more precise and accurate and prone to fewer "False Positives".

What the costs of misclassification you ask?
1. Well there is the tendency of many who are overweight to avoid going to the doctor so that they won't get nagged for not doing anything about it. False positives could keep those people from getting checked out as often for OTHER health problems.
2. Dr. OZ exemplifies what is wrong with the American approach to dieting. They are always looking for the Miracle cure that won't involve getting up and walking. How much money is wasted on treatments that don't work. If 18% of those identified as obese via BMI are not in-fact unhealthy, then it is probable that there is roughly 18% more money being wasted on these miracle cures than would otherwise be.
3. There are all sorts of programs out there to help people lose weight who need it, but the funding for these programs comes from somewhere. Having 18% fewer people attempting to take advantage of these programs would enable them to more aggressively target those who actually need them, or it could result in that money being more wisely spent somewhere else for another at-risk group.
4. Finally, we all carry insurance (or should) as result of the new health care laws. Premiums are tied to tables, and BMI is one of those metrics used to calculate your risk. If I'm one of those 18% for whom the BMI says I belong in a higher risk level, but DXA says I don't (which is the more direct measure of that which BMI is attempting to estimate) I'd rather use DXA and pocket the difference.


There is a simple correction to BMI used by the US Navy which addresses one of the concerns that 'big-boned' or very muscular people have higher BMI scores without fat.

For the Navy, only if you have a high BMI, do they do an additional measurement of neck and waist circumference. For women, its neck, waist and hips. Neck circumferences are subtracted from the others to create a score used with weight.

This simple extra measurement corrects one side of the errors.

However, the extra measurements are not universally collected, so cant be used for population health measures today.


Joshua: For 80 to 90 percent of the population, BMI and DXA come to the same conclusion. If we do not have effective remedies for the BMI-obese, we do not have remedies for the DXA-obese either. That is a much bigger problem facing us.

Chris: yes, BMI is a simple measure that works for almost everyone. Unusual cases can be teased out with secondary measures.


@Kaiser, I'm not advocating tossing out BMI, but I don't think your repeated characterization of DXA as a solution in search of a problem is fair or accurate.

For those 10 to 20% of the population for where they disagree, there are real world consequences (financial, psychological, etc.) to that disagreement. As long as we are aware of the short comings, and using BMI primarily for science and indications when an intervention might be appropriate I have no objections. However, the wider the acceptance of BMI without acknowledging its limitations, the more likely we are to see it used inappropriately like as the primary driver of insurance premiums.

There is a lot of data showing that the poor are disproportionately likely to be obese. They are the segment of the population least capable of absorbing higher than necessary premiums due to false positive diagnosis as obese.

@Chris, My brother had to have those extra measurements done before the Naval Academy would accept him because he was so muscular for his height that he came up as obese. However, in his case they did not satisfy, so he ended up having to have a DXA (or it's predecessor) scan to determine his percent body fat before they would accept him. Those additional measurements may help, but they are not foolproof.

The comments to this entry are closed.

Get new posts by email:
Kaiser Fung. Business analytics and data visualization expert. Author and Speaker.
Visit my website. Follow my Twitter. See my articles at Daily Beast, 538, HBR, Wired.

See my Youtube and Flickr.


  • only in Big Data
Numbers Rule Your World:
Amazon - Barnes&Noble

Amazon - Barnes&Noble

Junk Charts Blog

Link to junkcharts

Graphics design by Amanda Lee

Next Events

Jan: 10 NYPL Data Science Careers Talk, New York, NY

Past Events

Aug: 15 NYPL Analytics Resume Review Workshop, New York, NY

Apr: 2 Data Visualization Seminar, Pasadena, CA

Mar: 30 ASA DataFest, New York, NY

See more here

Principal Analytics Prep

Link to Principal Analytics Prep