Teasing out the numbers in obesity study

Posted: March 10, 2014

Perhaps you heard the astonishing good news that obesity among toddlers has dropped 43 percent in eight years. It made headlines, and was based on findings in a prestigious medical journal by respected researchers using gold-standard data.

Is it true?

Technically, yes. But here are some other statistics derived from the same paper:

  • Toddler obesity dropped 39.6 percent (before rounding off the raw numbers) in eight years.
  • Or 21 percent in six years - or 31 percent in just two years.
  • Or, using confidence intervals at each end - the range of statistical possibilities rather than the most likely point near the middle - the eight-year change comes out somewhere between a 7 percent increase and a 66 percent decrease.

"I'm not so convinced by these figures," said Temple University math professor John Allen Paulos.

In fact, none of these figures appeared in the Journal of the American Medical Association on Feb. 26. The authors reported that obesity among 2- to 5-year-olds declined from 13.9 percent in 2003-04 to 8.4 percent in 2011-12.

That decrease of 5.5 percentage points is neither sexy nor easily understood by the public. So the press office for the Centers for Disease Control and Prevention used a common practice among publicists and journalists: It expressed the absolute number in relative terms, in this case rounding off at both ends before calculating a drop of 43 percent.

Nearly every major news outlet, including The Inquirer, used that figure. The New York Times and the Washington Post put it in the headline. (The Bloomberg News story rounded it further, to "almost half.")

Most mentioned the other eight-year change that was statistically signficant: a 17 percent increase in obesity among women age 60 and older. Most included at least some of the other findings: no significant change for infants, or for ages 2 to 19, or for the subcategories of 6 to 11, or 12 to 19, or for adults 20 and older, or the groupings 20 to 39, 40 to 59, and men age 60 and older.

The study's conclusion: "Overall, there have been no significant changes in obesity prevalence in youth or adults between 2003-04 and 2011-12. Obesity prevalence remains high and thus it is important to continue surveillance."

News stories did refer to the conclusion, and most contained caveats and skeptical voices. Cynthia Ogden, an epidemiologist for the National Center for Health Statistics who leads this analysis every two years, described the seemingly dramatic decline among toddlers as just "a glimmer of hope."

But you are more likely to remember a "thrilled" Michelle Obama proclaiming that "healthier habits are beginning to become the new norm."

Obesity experts cited the first lady's highly publicized campaign as one reason the drop might be real. Soda consumption is down. Toddlers' diets are controlled by their mothers, who are concerned about health.

Plus, obesity rates, after tripling over three decades, had held steady for several years. And small declines had been seen in narrower populations: 6 percent in public school students in New York, 5 percent in Philadelphia.

"There are reasons to believe these results," said Aviva Must, a nutritional epidemiologist at Tufts University medical school.

Still, she found it hard to disagree with University of Pennsylvania professor Mark Liberman, who noted on his "Language Log" blog that, with enough comparisons, some statistically significant findings will appear at random.

"A pretty good rule of thumb," he said, when reading about unexpected jumps in statistics like these, "is to reserve judgment and say, 'Hmm, that is pretty interesting. I wonder if it's true.' "




comments powered by Disqus