USER GUIDE

It’s usually not feasible to measure every person in a population. Obesity prevalence and trends estimates, therefore, are typically based on data from select individuals who agreed to take part in the assessment.

There are different ways investigators can select participants. Some approaches are designed to arrive at a sample that is representative of a specific target population. Other approaches and techniques can help ensure adequate representation of select subpopulation groups.

Some data sources, like electronic health records and registries, collect data on almost every person who makes use of the system. Although sampling isn’t used in these instances, it’s still important to consider who is included—and who is not included—in the data.

An estimate describing a broad population, like a state or the nation, provides insight into the general status of obesity. Subgroup comparisons shed light on how obesity prevalence and trends differ within and between populations.

While any variable can be used to create subgroups, the most common ones used in obesity prevalence and trends reports are demographic characteristics. Some key factors to consider are age, sex, race, ethnicity, socioeconomic status, rurality, and geographic location.

Sample size is a primary determinant of the number of subgroups that can be evaluated in an analysis and the way they are defined. A sample size that is too small may not lead to a reliable estimate. This often leads to the omission or combination of subgroups.

Over the past several decades, the demographic landscape of the nation, states, and communities has changed. The median age has risen. Net immigration has increased. The population has become more racially and ethnically diverse.

Because different population groups can have different risks and rates of obesity, demographic shifts can affect an obesity trend. An influx or efflux of groups with higher or lower obesity prevalence can affect the prevalence estimate for the broader population. These changes over time can affect obesity trends. An assessment of the stability of the underlying or target population provides context for interpretation.

Body mass index, or BMI, is typically used to classify a person’s obesity status.

Data can be collected by directly measuring heights and weights or by asking people to report such information about themselves (self-report) or about someone else (proxy-report).

Reported heights and weights are not always accurate. This can make the obesity prevalence estimate higher or lower than it would be if the people were actually measured.

Inaccurate height or weight data can lead to misclassification of obesity status. Extreme values in height, weight, and BMI within a dataset may be due to measurement or data entry error OR due to actual extreme values. Both including these values when they’re a mistake, or excluding them when they’re legitimate, can affect obesity prevalence estimates.

Investigators often look for these values when they’re preparing data for analysis. The criteria they use to find the extreme values and what they do with them once identified differs across reports.

Data from collection systems that automatically detect extreme values at the point of data collection with a way to distinguish true extreme values from measurement error may need different criteria than those with limited ability to check data until analysis.