Skip to main content

Currently Skimming:

5 Analytic Approaches and Considerations
Pages 123-160

The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.


From page 123...
... • Published reports based on surveys use response rates as a gauge of the potential for nonresponse bias. A high response rate, however, does not always mean the data are unbiased.
From page 124...
... Some analytic approaches, such as the reference population selected to classify body mass index (BMI) , specifically pertain to the assessment of obesity among children, adolescents, and young adults.
From page 125...
... . This list of themes is not exhaustive, but it illustrates that a range of purposes can fall under the purview of a "report on obesity prevalence or trends." Although some analytic approaches are common across all published reports (e.g., classification of obesity status)
From page 126...
... selected affects results, interpretation of findings. Compare obesity Helps elucidate how prevalence May not account for all prevalence estimates estimates relate to each differences between locations between multiple other geographically.
From page 127...
... Growth References for Classifying Obesity Status 2000 CDC BMI-for-Age Growth Charts  The 2000 CDC sex-specific BMIfor-age growth charts are designed for individuals ages 2 to 20 years (see Figure 2-1 for an example) (Kuczmarski et al., 2000)
From page 128...
... reference growth reference, merged with the MGRS data NOTE: CDC, Centers for Disease Control and Prevention; IOTF, International Obesity Task Force; MGRS, Multicentre Growth Reference Study; N/A, not applicable; WHO, World Health Organization. a Only pertains to the BMI-for-age growth charts.
From page 129...
... The age in which an individual transitions from the growth charts to the adult cut point for obesity is inconsistent across reports. Some investigators use the growth charts for individuals through age 19 years (Gee et al., 2013; Ogden et al., 2014)
From page 130...
... Although rarely reported in the published literature, the extent to which the estimate of obesity prevalence changes with the inclusion and exclusion of the 30 kg/m2 criterion for an analytic sample provides evidence of its utility and need for a given population. IOTF BMI Cut Points  Although the IOTF BMI cut points are not as common as the 2000 CDC BMI-for-age growth charts in the published literature, they have been used in reports on U.S.
From page 131...
... Thus, obesity classification using the IOTF is a comparison of a BMI to the distribution of children's BMIs that existed at the time points the data were collected from various international locations. WHO Growth Standards and Growth Charts  The WHO developed growth standards for children 0 to 5 years of age and growth references for children 5 to 19 years of age.
From page 132...
... The WHO growth references are aligned with the age 0 to 5 years growth standards, providing continuity as children younger than age 5 years transition to the growth reference for those ages 5 to 19 years. For the WHO growth charts, a child with a BMI exceeding +2 standard deviations (approximately the 98th percentile)
From page 133...
... . Because IOTF cut points correspond to a higher threshold, obesity prevalence calculated using the IOTF will be lower than prevalence based on CDC growth charts (Freedman et al., 2011; Lang et al., 2011)
From page 134...
... Obesity prevalence is typically lower when using the IOTF cut points, compared to CDC growth charts. Biologically Implausible Values Extreme values for height, weight, and BMI occur in datasets.
From page 135...
... Both inclusion of BIV data that are errors and exclusion of BIV that are legitimate data points can affect an obesity prevalence or trend estimate. Because the purpose of identifying BIVs is to locate extreme values in the data, BIV criteria are often based on evaluation of growth reference z-scores rather than percentiles (Flegal and Ogden, 2011)
From page 136...
... When the 2000 CDC BMIfor-age growth charts were developed, skewness in the distribution was handled in such a way that extreme values converge to high but still plausible z-scores (CDC, 2002, 2016b; Flegal and Cole, 2013)
From page 137...
... . This data collection approach minimizes data entry error, and presumably all captured height and weight data represent legitimate data points.
From page 138...
... The committee identified three key elements that investigators assess to establish the representativeness of an analytic sample used in a published report: response rate, missing data, and weighting. These elements are not exclusive to reports on obesity prevalence or trends, but rather are general principles of epidemiologic research and data analysis.
From page 139...
... Factors that can contribute to the degree to which missing data might bias results include the amount of data that are missing and the mechanism that generated missing values. Analytic procedures have been developed to handle missing data.
From page 140...
... , the interpretation of resulting statistics changes. Another approach that has been used is to fill in the missing data using the average from the sample or group.
From page 141...
... The weights also can be adjusted to account for response rate, with those from groups that had lower response rate being assigned larger weights, to make up for the data that are missing from those who did not respond. Furthermore, the sample can be weighted to match the distribution of demographic characteristics within the target population for which the estimate is designed to represent.
From page 142...
... Instead of evaluating statistical procedures individually, the committee identified considerations that would broadly apply to a wide range of published reports. The topics in this section cover considerations related to the sample size, determining the prevalence, assessing the prevalence over time, and performing comparisons.
From page 143...
... Similarly, the desire to obtain a more precise estimate with less error would necessitate a larger sample size. Although these concepts are not specific to obesity prevalence or trends, they are reflected in statistical approaches present in published reports.
From page 144...
... This concept can be illustrated by discussing NHANES, although it should be noted that it occurs across data sources and published reports. As described in
From page 145...
... However, published reports on obesity prevalence and trends among children and adolescents using NHANES data present estimates for only a limited number of race and ethnicity groups: non-Hispanic white, non-Hispanic black, Hispanics (previously just Mexican American) ,2 and more recently Asians (Freedman et al., 2006; Ogden et al., 2012, 2014; Ver Ploeg et al., 2008)
From page 146...
... Determining Prevalence Prevalence of obesity has been presented in published reports in several formats. The simplest prevalence estimates are point estimates calculated as the raw percentage of those in the sample who have obesity.
From page 147...
... Assessing Prevalence Over Time Published reports evaluating prevalence of obesity over time in a population have approached the assessment in different ways. The number of time points, the span of time represented by each time point, and the span of time the trend represents vary across reports.
From page 148...
... In isolation, these reports describe obesity trends among children from different perspectives and at first pass appear incongruous. However, in considering the time frames included in the analyses, the findings across the reports appears to present different aspects of the same overall trend.
From page 149...
... The Presentation of Change Published reports have used both absolute and relative change to describe how prevalence differs between two points in time. Absolute change is independent of the baseline prevalence, and is defined as the simple difference between the two estimates of prevalence (i.e., Prevalence2 – Prevalence1)
From page 150...
... 150 ASSESSING PREVALENCE AND TRENDS IN OBESITY Percent of Population That Has Obesity a a FIGURE 5-2  Two hypothetical scenarios illustrating the difference between absolute and relative change with varying population sizes. a The two numbers represent the population size for Scenario 1 and Scenario 2, respectively.
From page 151...
... Consistent with its task, the committee also provides guidance for comparing obesity trends among diverse populations, both within and between reports (see Box 5-7)
From page 152...
... Recognize that use of different reference populations (e.g., Centers for Disease Control and Prevention, International Obesity Task Force, World Health Organization) can lead to different estimates of obesity prevalence, and are therefore not interchangeable.
From page 153...
... Both datasets based BMI on directly measured height and weight and defined obesity as having a BMI at or above the 95th percentile on the 2000 CDC sex-specific BMI-for-age growth charts. Shustak et al.
From page 154...
... In preparing the data, investigators must classify obesity status, and can elect to identify BIVs and evaluate the representativeness of the data source, as appropriate. For children, adolescents, and young adults, the 2000 CDC BMI-for-age growth charts are most typically used, but others exist, such as the IOTF cut points and the WHO growth charts.
From page 155...
... 2002. 2000 CDC growth charts for the United States: Methods and development.
From page 156...
... 2009. Characterizing extreme values of body mass index-for-age by using the 2000 Centers for Disease Control and Prevention growth charts.
From page 157...
... 2000. CDC growth charts: United States.
From page 158...
... 2011. Obesity in preschool children is more prevalent and identified at a younger age when WHO growth charts are used compared with CDC charts.
From page 159...
... Geneva: World Health Organization. WHO MGRS (World Health Organization Multicentre Growth Reference Study)


This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More information on Chapter Skim is available.