Whatever the issue, state rankings make for popular news headlines. This state is the most/least fit. That state is the most/least obese.
Such was the case for a recent report by Gallop – Healthways, which has tracked adult obesity prevalence in the US for several years now. As a polling organization, Gallup provides some very helpful data on the state of public opinion on particular issues. In the case of their State of American Well-Being, they also collected height and weight information of respondents to calculate BMI.
The state-by-state breakdown of obesity in the US is somewhat helpful, showing where it clusters geographically and how individual states have fared over time (see figure).
They mistakenly use “incidence” when they really mean “prevalence.” Now, these terms sound similar, but they mean very different things when it comes to epidemiological studies.
- Incidence – a measure of change from non-disease to disease (which is the numerator) in a “population-at-risk” (which is the denominator) over a specific time period. In basic terms, incidence is often expressed as [Newly diagnosed cases] divided by [population-at-risk (or person-time-at-risk)].
- Prevalence – a static measure of the proportion of a population that is diseased, whether the disease cases occurred recently or at some time in the past. Prevalence is expressed as [Existing cases] divided by [total population].
Unfortunately, the Gallup-Healthways data on obesity trends use these terms interchangeably. For example, this is the chart in the most recent report launched earlier this week. Notice it uses incidence in the tile.
The data are, however, exactly the same as those previously reported by Gallup. The data below are from a poll released in January of this year. These data are portrayed as prevalence of adults within each of the three different weight categories. Look at the obesity trend line. Look familiar?
The Gallup-Healthways project is a helpful data source that provides some interesting insights into well-being in the US. I hope in future reports they will be a bit more cautious when it comes to characterizing data. Though the average reader may not pick up on the nuances of epidemiological research and data, it still matters. Using the appropriate language to accurately describe data is an easy way to help ensure research integrity in the future.