Some other ways to look at the CDC’s preschool obesity numbers

On Tuesday, the Centers for Disease Control released a study of obesity rates among low-income preschool-aged (2–4 years old) children in 43 U.S. states and territories, all participants in the Pediatric Nutrition Surveillance System. Obesity is defined as a body mass index in the 95th percentile or above. MinnPost’s Susan Perry wrote about the findings, which were mostly positive: 19 states and territories showed small but statistically significant declines in childhood obesity rates over the period studied, from 2008–2011. Three states — Colorado, Pennsylvania and Tennessee — saw increased odds of childhood obesity over the four years.

The CDC researchers visualized their findings in a pair of maps. The first showed which states had made gains, which states had setbacks, and states with no significant change in the odds of childhood obesity. The second showed the rate of childhood obesity for states in 2011, when the national average was 12.1%. 

two maps
Source: CDC
The map at left shows states that had statistically significant increases and decreases in the probability of low-income preschooler obesity. The map at right visualizes obesity prevalence among low-income preschool-aged children in 2011.

Looking at the first map, I was struck by the inclusion of Colorado — with its reputation for high levels of physical activity — among the states where childhood obesity had gotten worse. To see how the states stacked up in absolute terms over the past four years, I took data on obesity rates from the report and charted it. I highlighted the three states with increasing obesity odds in red, the three states with the biggest improvements in adjusted odds ratios in blue, and Minnesota in green.

Presented this way, it's easy to see that while Colorado did indeed see an increase in its rate of obese children, it still has, along with Hawaii, one of the lowest rates of low-income childhood obesity in the nation. And one of the biggest droppers, New Jersey, started at one of the highest levels, tied with Puerto Rico at 17.9%. Keep in mind though, these are raw percentages of obese children that do not control for factors like race and income (the report noted that childhood obesity is more prevalent among black and Hispanic youth, so states with higher proportions of those children in their samples might show higher overall rates of childhood obesity). Researchers used linear regression to control for factors like race in developing their Adjusted Odds Ratio (AOR), which was the basis for the first map above. Controlling for race and other factors allowed researchers to isolate the effects of policies or specific conditions within a state/territory on the obesity rate.

You can examine the data yourself in the table below. The table includes absolute numbers and percentages of obese children in the sample for each state and territory in each of the four years studied, the AOR, the 95% confidence interval for the AOR, and whether the change represented by the AOR is statistically significant (this is why some states with AORs that are less than or more than 1 are listed as "No Change" in the first map). The table can be sorted by clicking on column headings, and searched using the form at right. You can also download a spreadsheet with the study's findings.

For more information about the study, including important caveats about the limitations of the findings, read the full report from the CDC [PDF].

2008 2009 2010 2011
State Number Percent Number Percent Number Percent Number Percent AOR (95% CI) Sig?
Alabama 56,813 13.8 60,572 14.4 65,914 14.1 67,246 14.1 0.99 (0.98–1.00)
Arkansas 38,591 13.9 42,270 14.2 32,615 14.1 42,455 14.2 1.01 (0.99–1.02)
Arizona 75,338 14.6 85,020 14.3 87,897 14.2 86,516 14.5 1 (0.99–1.01)
California 301,643 17.3 332,663 17 282,216 17.3 260,314 16.8 0.99 (0.99–1.00) x
Colorado 43,476 9.4 51,659 9 52,292 9.1 27,467 10 1.02 (1.00–1.04) x
Connecticut 25,623 15.5 28,432 16 28,401 15.8 27,561 15.8 1 (0.98–1.01)
Washington DC 6,195 13.3 6,742 13.6 6,954 13.7 6,940 13.1 0.97 (0.94–1.00)
Florida 209,671 14.1 238,542 13.7 242,399 13.4 240,022 13.1 0.97 (0.96–0.97) x
Georgia 124,533 14.8 134,173 14.2 136,379 13.5 138,622 13.2 0.95 (0.95–0.96) x
Hawaii 16,106 9.3 17,252 9.3 17,827 9.1 17,819 9.2 0.99 (0.97–1.01)
Iowa 33,548 15.1 36,225 15 35,783 14.7 34,327 14.4 0.98 (0.97–0.99) x
Idaho 20,081 12.3 22,620 11.9 22,973 11.4 22,238 11.5 0.97 (0.95–0.99) x
Illinois 121,608 14.7 133,023 14.6 135,408 14.6 132,671 14.7 1 (0.99–1.01)
Indiana 66,499 14.5 75,671 14.3 78,634 14.2 73,247 14.3 0.99 (0.98–1.00)
Kansas 34,352 13.3 36,956 13.2 37,838 13 37,419 12.7 0.98 (0.96–0.99) x
Kentucky 62,832 15.7 68,450 15.8 75,189 15.6 33,008 15.5 0.99 (0.98–1.00)
Massachusetts 59,297 16.7 63,567 16.8 60,433 16.1 61,094 16.4 0.98 (0.97–0.99) x
Maryland 54,866 15.7 62,194 15.8 63,951 15.7 64,773 15.3 0.98 (0.97–0.99) x
Michigan 103,523 13.9 114,489 13.7 106,019 13.3 115,608 13.2 0.97 (0.97–0.98) x
Minnesota 65,607 13.4 68,997 13.1 68,594 12.7 70,353 12.6 0.97 (0.96–0.98) x
Missouri 60,908 13.9 60,150 13.9 67,547 13.6 67,650 12.9 0.97 (0.96–0.98) x
Mississippi 44,807 14.6 51,741 13.9 52,112 13.7 47,494 13.9 0.98 (0.96–0.99) x
Montana 10,428 12.4 10,105 12.5 8,958 12.2 10,681 11.7 0.98 (0.95–1.00)
North Carolina 96,381 15.7 104,323 15.2 105,392 15.5 103,565 15.4 0.99 (0.99–1.00)
North Dakota 6,551 13.8 6,968 14.1 6,836 14.1 6,665 13.1 0.99 (0.95–1.02)
Nebraska 20,658 13.9 20,811 14.2 22,194 13.8 22,136 14.3 1 (0.99–1.02)
New Hampshire 8,082 15.5 8,963 14.4 8,621 14.2 8,219 14.6 0.97 (0.94–0.99) x
New Jersey 68,163 17.9 75,191 18.4 78,181 17.3 77,476 16.6 0.96 (0.95–0.97) x
New Mexico 22,295 12 31,433 12 31,043 11.7 30,269 11.3 0.97 (0.96–0.99) x
Nevada 23,348 12.9 28,159 13.9 32,080 13.6 33,427 12.7 0.99 (0.98–1.01)
New York 209,713 14.6 224,130 14.4 224,243 14.5 229,291 14.3 0.99 (0.99–1.00) x
Ohio 125,011 12.2 130,792 12.3 128,754 12.4 121,624 12.4 1 (0.99–1.01)
Oregon 49,193 14.7 52,713 15 54,150 15.1 54,212 14.9 1.01 (1.00–1.02)
Pennsylvania 111,767 11.5 119,134 12 117,337 12 119,812 12.2 1.02 (1.01–1.03) x
Puerto Rico 99,828 17.9 100,313 18.1 89,438 18.3 89,463 17.9 1 (0.99–1.01)
Rhode Island 11,466 16.2 12,456 16.2 12,922 15.5 12,644 16.6 1 (0.98–1.02)
South Dakota 9,125 16.2 9,705 16.4 10,106 16.1 10,312 15.2 0.97 (0.95–1.00) x
Tennessee 69,015 13.5 71,914 14 71,349 14.5 69,276 14.2 1.02 (1.01–1.03) x
U.S. Virgin Islands 2,339 13.6 2,587 11.9 2,587 11.2 2,552 11 0.92 (0.87–0.97) x
Vermont 7,009 13.3 7,051 13.2 6,921 12.2 6,168 12.9 0.98 (0.95–1.01)
Washington 92,980 14.4 104,389 14.4 105,886 14.4 106,346 14 0.99 (0.98–1.00) x
Wisconsin 55,875 13.6 60,280 13.7 59,975 14.1 58,745 14 1.01 (1.00–1.02)
West Virginia 22,689 13.5 23,739 13.4 23,205 13.7 22,581 14 1.02 (1.00–1.03)

Comments (1)

  1. Submitted by Paul Brandon on 08/08/2013 - 03:35 pm.

    One problem is

    that this is a post hoc analysis.
    That is, the authors simply separated the states into those above and those below average, without any previous rationale for which states would be above and below average.
    Even if all of the states simply varied randomly in obesity rates, half would be below and half above average.
    The authors did do some sort of statistical test that showed significance; I wonder if they used an adjustment for repeated comparisons to control for the post hoc selection.
    And note the small effect sizes.
    For Minnesota, a drop in obesity rate of 0.8% over three years may be statistically significant, but still is not very impressive.

Leave a Reply