The data collection methodologies discussed in Chapter 3 do not exist or operate in isolation, but rather coalesce to form a data system. The study design, sample selection, and data collection approaches interact with each other and ultimately inform the interpretation of the associated analyses. The purpose of this chapter is to review and compare data sources used to assess obesity prevalence and trends. Throughout this chapter, key components of data collection methodologies will be highlighted.
Through its review of the evidence, the committee identified four broad categories of data sources used to assess obesity prevalence and trends among children, adolescents, and young adults. These include population surveillance surveys, direct measurement in the school setting, clinical and public health setting administrative data, and cohort studies. Although they are presented as distinct, these categories can overlap, depending on the design of the data source. These intersections, and the inability to discretely classify data collection efforts, emphasize the nuances, inconsistencies, and complexities that currently exist.
Detailed information about design and methods often reside in reports, protocols, and other documentation specific to a particular data source. It is the intent of this chapter to bring together condensed overviews of various data sources used to estimate obesity prevalence and trends, particularly among individuals ages 18 years and younger, to demonstrate how they are designed and how their differences would affect estimates presented in published reports.1 The total sampled population2 and methodologies, for example, will be described for the purposes of comparison. The committee
___________________
1 For additional information about a range of surveillance systems that collect weight-related data, the reader is referred to the National Collaborative on Childhood Obesity Research’s Catalogue of Surveillance Systems (http://tools.nccor.org/css [accessed June 7, 2016]).
2 This chapter provides a broad overview of select data sources and their total sample populations. Published reports using these specific datasets often do not use the entire sample. When evaluating a report on obesity prevalence or trends, end users are advised to assess the sample used in the analysis in addition to considering the total sampled population.
acknowledges that the list of data sources in this chapter is not exhaustive and others exist beyond those included here.
POPULATION SURVEILLANCE SURVEYS
Population surveillance surveys are a key source of data used in reports on obesity prevalence and trends in a range of U.S. populations. Some population surveys that capture height and weight data are designed to describe the country as a whole (“nationally representative”). Those in which children and adolescents are included in the surveyed population include the National Health and Nutrition Examination Survey (NHANES), the National Health Interview Survey (NHIS), and the Medical Expenditure Panel Survey-Household Component (MEPS-HC). Other surveys and surveillance systems are designed to be both nationally representative and representative of multiple states and localities. This has been accomplished in two ways. The first way is the approach taken by the Youth Risk Behavior Surveillance System (YRBSS), which conducts a nationally representative survey in addition to separate surveys for participating states and large urban school districts. The second way is the approach taken by the National Survey of Children’s Health (NSCH), which used samples collected in each state to generate statistics of prevalence for each state and at the national level. The following sections describes each of these surveys to highlight the similarities, differences, and gaps that exist across nationally representative data sources and data sources designed to represent the nation, multiple states, and various localities (summarized at the end of this section in Table 4-1). The barriers to comprehensively evaluating and comparing individual state and local population surveys also are described. The committee’s synthesis of this information is summarized at the end of this section.
Population Surveillance Surveys Designed to be Nationally Representative
Population surveillance surveys with the primary intent of being nationally representative describe the country as a whole. These surveys are generally not designed to generate estimates of obesity prevalence or trends for specific regions, states, or localities within a given year or cycle of data collection.3 The sample size and sampling procedures in these surveys,
___________________
3 This statement comes with three caveats. First, some researchers have been able to use the nationally representative data to calculate estimates for populous states and counties that are well represented within the data source (Johnson et al., 2013b; Porter et al., 2011). This, however, cannot be done for most locations. Second, the committee acknowledges that model-based estimation approaches are being used to generate estimates for smaller areas than those for which a survey or study was designed to represent. For that reason, the statement in
however, may provide enough data to allow for comparisons of select population groups. At the present, data sources with this general sampling approach that capture height and weight data and have samples that include individuals ages 18 years and younger are NHANES, NHIS, and MEPS-HC.
National Health and Nutrition Examination Survey
NHANES is a nationally representative, repeated cross-sectional survey that explores the relationship between nutrition and health, identifies emerging public health issues, and provides baseline information on the health and nutritional status of the nation (CDC, 2016a). The first iteration of the NHANES program, called the National Health Examination Survey I, was conducted in 1960-1962, with various cycles following in the years thereafter. In 1999, NHANES became a continuous survey. To produce reliable estimates and to decrease the likelihood of identifying individual participants, data are released in 2-year cycles (Curtin et al., 2012).
Approximately 5,000 individuals of all ages are surveyed as part of NHANES each year (Johnson et al., 2014). Participants are selected through a complex, four-stage sampling design. Each year, individuals are sampled from 15 locations from across the country, with data being collected throughout the year (NHANES, 2013a). Both the locations4 and the individual participants change annually. NHANES oversamples various population groups in an effort to provide more precise and stable estimates of health parameters. In the 2011-2012 and 2013-2014 cycles of NHANES, the oversampled groups included: Hispanic persons, non-Hispanic black persons, non-Hispanic non-black Asian persons, low-income non-Hispanic non-black non-Asian white and other persons (≤130 percent of the federal poverty level), and adults ages 80 years and older (Johnson et al., 2014). The groups that are oversampled have changed over time, which affects the amount of data available for trend analyses for population groups that have comprised or currently represent a small portion of the total U.S. population.
Information is gathered from participants through an interview and a physical examination (Zipf et al., 2013). Heights and weights are directly measured from trained data collectors. The digital scale currently used to obtain the body weight is linked to the study database, as is the stadiometer used to measure height (CDC, 2013b). Demographic variables collected include, among others, sex, date of birth, income, occupation, and highest
___________________
the text pertains to direct survey estimates rather than synthetic estimates. Third, some data sources are both nationally representative and representative of multiple states and localities. These data sources will be discussed later in the chapter.
4 While the selected locations for NHANES data collection do change annually, a location can be used in multiple cycles.
level of educational attainment. NHANES asks participants to choose from a large number of race and ethnicity categories, with recent demographic questions offering 29 Hispanic, Latino, or Spanish origin and ancestry groups, 35 Asian origin and ancestry groups, and 4 Native Hawaiian or Pacific Islander origin and ancestry groups (NHANES, 2013b).
NHANES is widely considered the gold standard for nationally representative estimates of obesity among the U.S. population at large. Results from NHANES analyses have been used as benchmarks to which state and local estimates are compared. A key strength of NHANES for assessing obesity prevalence is that heights and weights are measured by staff trained “to follow standardized examination protocols, to calibrate equipment according to a prescribed schedule and method, and to measure and record the survey data with precision” (CDC, 2013b).
Despite its methodologic strengths, NHANES may not meet the information needs of stakeholders at the state and local level. NHANES is not designed to assess variations in prevalence estimates that exist across and within regions, states, and localities. The purpose of NHANES is to provide nationally representative estimates overall and in select subgroups of interest while maintaining high-quality data (including clinical examinations and biomarker measurements). This limits the number of participants who can be included in the sample in a given year. Assessing obesity prevalence in select population groups, therefore, may require the use of multiple cycles’ worth of data. Although this can improve the stability of estimates, it also expands the time frame a prevalence estimate describes. This, in turn, affects the interpretation and comparability of the estimate.
National Health Interview Survey
NHIS, which has been continuously operating since 1957, is a cross-sectional household survey that assesses a variety of health topics, including the prevalence of, outcome of, and services received related to illnesses and disabilities. Although data are collected from participants in each of the 50 states and the District of Columbia, the annual prevalence estimates are representative at the national- and U.S. Census region-levels only (NHIS, 2014). Select data can be combined across years to produce a stable state-level estimate (NHIS, 2015a). The questionnaire used in NHIS is periodically updated, with the most recent revision implemented in 1997, and another redesign scheduled for 2018 (NHIS, 2015a).
The NHIS uses a multistage, stratified sampling approach (Parsons et al., 2014). One adult and one child (if applicable) are randomly selected from each household (CDC, 2016b). When the sampling design is fully executed, approximately 35,000 households (approximately 87,500 individuals) provide complete interviews each year (NHIS, 2015a). The NHIS currently oversamples black, Hispanic, and (more recently) Asian persons
(NHIS, 2015a). Additionally, adults ages 65 years and older who are of one of these three race or ethnicity groups are at an increased odds of being selected to be the sample adult (NHIS, 2015a). In 2014, the NHIS evaluated approximately 3,000 additional Native Hawaiian and Pacific Islander (NHPI) households throughout the country to better characterize the health status, health needs, and well-being of this population (CDC, 2015f).
NHIS is conducted as an in-person interview. Heights and weights, however, are not directly measured. Instead, data for children ages 12 years and older are obtained through proxy-report (i.e., an adult in the household knowledgeable about the child) (NHIS, 2015c). In 2008, NHIS discontinued collecting proxy-reported height and weight data on children younger than age 12 years due to concerns about the accuracy of these values (NCHS, 2015a). In addition to height and weight, the NHIS interview questions cover a range of health topics, including health insurance coverage, health care use, health conditions, health behaviors, and general health status. Demographic characteristics also are captured. For example, participants are asked to identify with one or more of 16 different options for race and, if applicable, 8 different options for Hispanic origin or ancestry (NHIS, 2015b).
NHIS has several strengths in relation to the assessment of obesity prevalence and trends at the national level. For instance, it is a rich source of data not only on adolescents’ and adults’ obesity status, but also on health behavior, health status, and other sociodemographic indicators. The oversampling procedures employed allow for evaluation of select subpopulations. Furthermore, NHIS data are available annually, generally within 6 months of the end of data collection (NCHS, 2015b). In spite of these strengths, the NHIS data also have limitations with respect to assessing obesity prevalence and trends in children. First, height and weight data are not collected on children younger than age 12 years. Height and weight that are collected for children ages 12 to 17 years are based on proxy-report, which is subject to bias (see Chapter 3). Another consideration for the NHIS data is that data files with state-level and other geographic identifiers can be accessed only through 1 of 20 Federal Statistical Research Data Centers across the country. The limited access to state identifiers, the national sampling frame, and relatively small state samples restricts the regular use of these data for state or local area analysis. Researchers have developed model-based estimates of state obesity prevalence, but these are difficult to replicate, which limits their use for surveillance purposes.
The Medical Expenditure Panel Survey-Household Component
MEPS-HC is a nationally representative survey sponsored by the Agency for Healthcare Research and Quality (AHRQ). The MEPS-HC has collected data on health conditions, health status, and the use and cost of health care services since 1996. Participants in the MEPS-HC are drawn from house-
holds that participated in the previous year’s NHIS. This sampling approach allows for oversampling of demographic characteristics and health-related conditions identified through the NHIS responses (Mirel and Machlin, 2013).
Data are collected from participants through five interviews over the course of 30 months (Ezzati-Rice et al., 2008). Paper-based questionnaires, provided in English and Spanish, are occasionally sent out to participants to gather supplemental information (MEPS, 2011). Height and weight are proxy-reported for children and self-reported for adults.
A primary advantage to the MEPS-HC is that participants are drawn from NHIS. As such, data from the surveys can be linked, providing complementary information. The opportunity for linking data across the NHIS and the MEPS-HC, however, can be incomplete. New members of the household (e.g., through marriage, birth) may not be represented in NHIS, and the interval between NHIS participation and beginning of MEPS-HC participation can vary (Mirel and Machlin, 2013). As with any longitudinal study, attrition grows as time passes. Furthermore, estimates of obesity prevalence in children derived from this data source are not pervasive in the literature and not currently included in MEPS-HC summary tables or query system (MEPS, 2009). AHRQ has, however, published prevalence of obesity for adults ages 20 and older using data from the 2009 MEPS-HC (Carroll and Rhoades, 2012).
Population Surveillance Surveys Designed to Represent the Nation and Individual States and Localities
Nationally representative statistics provide invaluable insight into the overall health status of the country. However, nationally representative estimates encompass a considerable amount of variability that exists at the state and local levels. To have a better sense of who is affected and where, surveillance systems and surveys have been developed that generate estimates for individual states and localities, in addition to generating national estimates. This section describes two such data sources that include children and adolescents: YRBSS and NSCH.
The committee also identified two additional data sources that can be used to assess obesity prevalence across states. First, described in Box 4-1, is the Behavioral Risk Factor Surveillance System (BRFSS). The BRFSS is designed to produce state-specific prevalence estimates of behaviors and practices associated with disease and injury in U.S. adults ages 18 years and older. Although it does not currently collect information about the weight status of children or adolescents, it represents a critical state-level data source related to obesity surveillance, and has therefore been included in this chapter. Second, Box 4-2 describes how administrative data from the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) have been used to produce estimates of obesity prevalence
among participants in WIC. The discussion of WIC data is included here, rather than in the section on clinical and public health administrative data, because it is a prominent data source used to estimate prevalence for individual states and multiple localities, albeit for a select population.
Youth Risk Behavior Surveillance System
The YRBSS monitors health risk behaviors contributing to the leading causes of death and disability among America’s youth. The YRBSS is not a single evaluation, but rather a collection of assessments.5 Within the system, the school-based Youth Risk Behavior Survey (YRBS) is administered every other year to high school students in a nationally representative sample, in participating states, and in select large urban school districts.6
On average, approximately 14,500 students are included in the national sample each cycle (CDC, 2013a). The sampling frame includes both public
___________________
5 In addition to the Youth Risk Behavior Survey (YRBS), the YRBSS includes one-time, specialty population, and methods surveys.
6 The CDC provides both the YRBS questionnaire and an implementation guide through its website, allowing schools, districts, and communities that are not part of the YRBSS to conduct their own assessment (CDC, 2014, 2015j). The committee acknowledges that reports using these data may exist at the state and local level. The discussion in this chapter, however, pertains only to data collected and analyzed through the YRBSS.
and private high schools and includes strategies for oversampling black and Hispanic students (CDC, 2013a). The large sample size allows for estimates of prevalence to be calculated by a single demographic characteristic (i.e., sex, grade, race and ethnicity categories [white, black, Hispanic]), and by interactions between demographic characteristics (i.e., grade*sex; race/ethnicity*sex; race/ethnicity*grade) with a high degree of confidence (CDC, 2013a).
The sampling frame for the state and localities typically consists of public schools, although some samples are derived from both public and private schools and others include alternative schools (CDC, 2013a). Some jurisdictions forego sampling and collect data from all schools (CDC, 2013a). Accordingly, sample sizes vary considerably across the state and local YRBS, ranging from 1,102 to 53,785 participants in 2013 (Kann et al., 2014). Sample sizes are often adequate to stratify the analyses by sex, but estimates by race and ethnicity groups can be unstable. The state and local YRBS assessments are designed to be representative of students in grades 9 to 12 within the jurisdiction (CDC, 2013a). The representativeness however, is contingent on the overall response rate. If the overall response rate is at least 60 percent, the sample is weighted during analysis to be representative of the jurisdiction; otherwise, the results are unweighted and represent only those who participated (CDC, 2013a).
The type of parental consent obtained for any of the YRBS evaluations is determined by the state, the school district, or the individual schools. Some states require active consent (i.e., having to take an action to opt in) for all participants (CDC, 2015a), though the majority of assessments use passive consent (i.e., participating unless an action is taken to opt out) (CDC, 2013a). As discussed in Chapter 3, the type of consent used affects the sample size.
The YRBS is an anonymous, voluntary, self-report, paper-based survey (CDC, 2013a). Questions about height and weight are required on all administrations at the national, state, and local levels, which eliminate issues related to comparability of data collection instruments across sites. For demographic characteristics, students are asked about their age, sex, and grade. Age is not based on date of birth, but rather students select from one of seven age categories. Students are asked whether they are Hispanic or Latino, and are also instructed to select from five race categories (American Indian or Alaska Native; Asian; black or African American; Native Hawaiian or Other Pacific Islander; white) (CDC, 2015j). An indicator of socioeconomic status is not collected from the student.
Because the YRBS is based on adolescent self-report, it is limited in its ability to estimate obesity prevalence (Brener et al., 2003). A validation study demonstrated that students tend to underestimate their weight by approximately 3.5 pounds and overestimate their height by approximately
2.7 inches (Brener et al., 2003). Underestimation of weight was more common among females. Overestimation of height was positively associated with grade level and is more common among white students. Collectively, the underreporting of weight and overreporting of height leads to estimates of obesity prevalence lower than would be obtained using direct measured height and weight data. The errors in reporting height were the primary drivers of the errors seen in body mass index (BMI) estimates (Brener et al., 2003).
Another consideration for YRBS is that the target population does not necessarily represent all U.S. adolescents. Rather, the national sample describes students enrolled in grades 9 to 12 in private and public schools. The sample does not typically represent students who attend alternative or charter schools, who are home-schooled, or who have dropped out. Furthermore, state and local YRBS evaluations can be limited in their comparability across jurisdictions, as sites vary in terms of sampling approach, parental consent process, and response rate. Changes to these factors over time also can affect the trends analysis for a particular jurisdiction, as the data may not be comparable. Finally, results are not generated by zip code, census tract, or individual schools due to student confidentiality and instability of estimates due to sample size (CDC, 2015k). This can limit the application of state YRBS data at the local level.
National Survey of Children’s Health—2003, 2007, and 2011-2012
The NSCH, which is currently being redesigned (see next section), was a cross-sectional survey designed to produce both state-specific and national prevalence estimates for a variety of health-related topics for children younger than age 18 years. NSCH data are available through the Data Resource Center for Child and Adolescent Health (Data Resource Center for Child and Adolescent Health, 2016c) and can be queried through Web-based interactive tools to generate national, Health Resources and Services Administration region, and state estimates on select parameters (Data Resource Center for Child and Adolescent Health, 2016b). Because data are still accessible, a discussion highlighting the strengths, limitations, and methodologies of the previous cycles of the NSCH is included here.
The three cycles of the NSCH (2003, 2007, and 2011-2012) were conducted through the State and Local Area Integrated Telephone Survey (SLAITS) program (CDC, 2015i). Developed by the CDC’s National Center for Health Statistics, SLAITS is a mechanism used to supplement data collected through ongoing surveillance programs (CDC, 2015i). SLAITS is not a single survey but a vehicle that government agencies, nonprofits, and other survey sponsors can use to collect customized data from select or defined populations. The NSCH sampling frame is based on the cross-
sectional telephone sample frame (land and cell phone lines) used for the National Immunization Survey. Households owning a wireless phone were included in the sample beginning in 2011 (Data Resource Center for Child and Adolescent Health, 2016a).
The target population for the NSCH was children birth to age 17 years in all 50 states, Washington, DC, and the U.S. Virgin Islands (added in 2011-2012). Each cycle of the NSCH had minimum enrollment goals: 2,000 participants per state in NSCH 2003, 1,700 per state in NSCH 2007, and 1,800 per state in NSCH 2011-2012 (NSCH, 2003, 2007, 2012b). This relatively small sample size limits the ability to perform subgroup analyses, especially in racially and ethnically diverse states. For example, national NSCH estimates for Asian, American Indian, Alaska Native, and Native Hawaiian/Pacific Islander children were categorized in a single “Other” category. Data for these groups individually are available only for states where the group represented at least 5 percent of the total population (Data Resource Center for Child and Adolescent Health, 2016a).
Because it was a telephone survey, the data on height and weight were collected through proxy-report. The person in the household with the most knowledge about the child’s health and health care needs was asked to report the child’s weight and height (NSCH, 2012a). Although the child’s age was not listed as a criterion for asking about height or weight (NSCH, 2012a), Web-based tools presenting prevalence estimates from NSCH data restrict the results to children ages 10 to 17 years (Data Resource Center for Child and Adolescent Health, 2016b). Similar to the NHIS, the rationale for such a restriction was based on findings that proxies generally underreported heights and over-reported weights of young children.
The rapidly changing technological climate also limited the ability to accurately sample NSCH’s target population. Although the 2011-2012 cycle included sampling of wireless phones, their inclusion presented some methodologic barriers. Sampling wireless phones is an expensive endeavor and, because of their portability, an area code no longer represents a current residence (MCHB, 2015). This consideration, among others, led to the redesign of the NSCH, which will be combined with the National Survey of Children with Special Health Care Needs (NS-CSHCN).
National Survey of Children with Special Health Care Needs
Similar in design to the NSCH, the NS-CSHCN was a telephone-based survey conducted through SLAITS. NS-CSHCN was designed to produce nationally representative and state-level estimates of children ages birth to 17 years “who have or are at increased risk for a chronic physical, developmental, behavioral, or emotional condition and who also require health and related services of a type or amount beyond that required by children
generally” (CDC, 2015e; McPherson et al., 1998). The NS-CSHCN did not collect information about the child’s height or weight, so it is not a data source that can be used to assess obesity prevalence or trends. It is, however, a survey that is currently being integrated with the redesigned NSCH.
Redesigned National Survey of Children’s Health
Although the NSCH and the NS-CSHCN served unique purposes, they shared many elements. Because of this, it was decided that the two separate surveys should be combined and redesigned as a single, continuous survey moving forward, providing both state-level and national estimates on a range of health indicators. The redesigned survey will no longer rely on telephone-based sampling or interviewing, but instead will use a household address-based approach, with data collected through mail or Web-based surveys. Content from the two previous surveys will be merged and streamlined to address current and emerging priorities. The redesigned NSCH/NS-CSHCN is expected to be a data source of national and state-level estimates. The redesigned survey will continue to provide accessible data and summaries to users of all levels through the Data Resource Center for Children and Adolescent Health. Efforts are being made to be able to generate local-level estimates through approaches such as model-based estimation (see Chapter 5 for additional details). At the time of this report, the redesigned survey’s content has been reviewed and edited, cognitive interviews and associated survey revisions have taken place, and the mode of delivery has been tested. The survey is currently being pretested, with the full fielding of the survey anticipated to take place in mid- to late-2016. The first public release of data is anticipated in 2017 (MCHB, 2015). Questions about the participant child’s height and weight have been included. As with the prior cycles of NSCH, height and weight data are collected through proxy-report.
Considerations for Assessing Population Surveys Used to Estimate Obesity Prevalence and Trends
The nationally representative population surveillance surveys used to assess obesity prevalence and trends among children, adolescents, and young adults have different goals and objectives. As a result, the target population and data collection methodologies differ across these data sources.
At present, NHANES is the only nationally representative, continuous survey that directly measures heights and weights of participants. The high quality and quantity of data collected limits the number of individuals who can be measured in any given year. The restricted sample size, in turn, limits which population groups have sufficient data to estimate obesity prevalence
and trends. Although the other nationally representative population surveillance surveys have total sample sizes significantly greater than NHANES, height and weight data are currently collected through self- or proxy-report rather than direct measure. Concerns over the accuracy of proxy-reported height and weight for young and school-aged children (younger than ages 10 to 12 years) have led to the discontinuation of capturing or reporting on such data for some of the nationally representative data sources (i.e., NHIS, previous cycles of NSCH).
Some data sources are designed to be both nationally representative and representative of individual states and locations. For example, WIC administrative data are used to characterize low-income participants enrolled in the program, at the national, state, and agency level. Directly measured height and weight data are collected as part of program delivery. Estimates of obesity prevalence, however, cannot be generalized to those who do not participate in the program. The YRBSS, a population survey-based surveillance system, is specifically designed to produce obesity prevalence and trends estimates from a nationally representative sample and across multiple states and localities. YRBS obesity prevalence estimates, however, describe only high school students and are based on self-reported heights and weights. The upcoming redesigned NSCH/NS-CSHCN is intended to be a source of state-level estimates. Although this survey plans to capture data in a large range of ages (birth to 17 years), it will be doing so through proxy-report, which has inherent limitations especially with younger children, as noted above and described in Chapter 3. No population surveillance survey or system currently produces state-level estimates of obesity prevalence in school-aged children, adolescents, or young adults from directly measured height and weight data collected in a consistent manner across multiple states.
The committee was unable to comprehensively assess or compare population surveys being conducted in individual states and localities. No centralized resource comprehensively catalogues the wide array of population survey efforts that have been and are being conducted in these jurisdictions. The availability of information about data collection procedures for such population surveys varies, which further limited the committee’s ability to comprehensively assess methodologies. For these reasons, it was not possible to describe the current practices of, or draw comparisons between, population surveys conducted in individual states and localities. In spite of these evidentiary barriers, the committee recognizes that population surveys are used to assess prevalence and trends in obesity at the state and local levels. Box 4-3 provides an illustrative example of one such survey. The committee acknowledges, however, that other state and local surveys of different designs and objectives exist.
TABLE 4-1 Comparison of Data Sources Used to Estimate Obesity Prevalence and Trends Among Children, Adolescents, and Young Adults Nationally and Across Multiple States or Localities
| Data Source | Approximate Sample Size | Representativeness of Sample | |
|---|---|---|---|
| Nationally | States, Localities | ||
| NHANES |
5,000 per year |
U.S. population |
N/A |
| NHIS |
87,500 individuals (35,000 households per year) |
U.S. populationb |
N/Ac |
| MEPS-HC |
33,000 persons per year (13,000 households per year)d |
U.S. population |
N/A |
| YRBS, nationale |
14,500 per survey year |
U.S. high school students |
N/A |
| YRBS, state and locale |
Varies by locationf |
N/A |
U.S. high school students; locations vary by yearg |
| WIC PC Data |
9.3 million nationally; varies by locationh |
WIC participants, as of April of the assessment yearh |
WIC participants, as of April of the assessment yearh |
| NSCH, (2003, 2007, 2011-2012) |
96,000 per cycle (1,800 per state in each cycle) |
U.S. children ages 0 to 17 years |
Children ages 0 to 17 years, in:
|
| Redesigned NSCH/NS-CSHCNj | N/Ak |
U.S. children ages 0 to 17 yearsj |
Children ages 0 to 17 years, in: All 50 states Washington, DCj |
| Height and Weight Data | Potential Advantagesa | Potential Disadvantagesa |
|---|---|---|
|
Directly measured |
Height and weight directly measured. Continuous survey. |
Confined sample size, limiting subgroup comparisons. |
|
Interview (proxy-, self-report) |
Large sample size. Generates Census region-level estimates. Data rapidly available. |
Height and weight not directly measured. Height and weight data not captured for children younger than age 12 years. |
|
Interview (proxy-, self-report) |
Panel design allows evaluation over time. |
Height and weight not directly measured. Longitudinal data subject to attrition. Reports based on this data source are not common in the literature. |
|
Paper-based survey (self-report) |
Large sample size. |
Height and weight not directly measured. Only captures students attending public and private schools. |
|
Paper-based survey (self-report) |
Large sample size. Most states and several localities participate. |
Height and weight not directly measured. Differences in sampling frame, consent process, and response rate can limit comparability across sites. |
|
Directly measured |
Height and weight directly measured. Large sample size. All agencies provide data. |
Only describes participants enrolled in WIC. Does not describe all participants in a given year, only those enrolled in the month of April. |
|
Telephone survey (proxy-report) |
Estimates were generated for each state using the same protocol. User-friendly data querying tool available through the Web. |
No longer being conducted. Height and weight not directly measured. Cellular phone sampling limited ability to select based on geography. Limited ability to perform subgroup analyses. Obesity status not typically reported for children younger than age 10 years. |
|
Web- and mail-based questionnaire (proxy-report)j |
Sampling will be address-based rather than telephone-based.j |
Height and weight data will be collected through proxy-report, limiting utility of obesity prevalence estimates, especially for children school-age and younger.j |
NOTE: MEP, Medical Expenditure Panel Survey; N/A, not applicable; NHANES, National Health and Nutrition Examination Survey; NHIS, National Health Interview Survey; NS-CSHCN, National Survey of Children with Special Health Care Needs; NSCH, National Survey of Children’s Health; PC Data, Participant and Program Characteristics; WIC, Special Supplemental Nutrition Program for Women, Infants, and Children; YRBS, Youth Risk Behavior Survey.
a The potential advantages and disadvantages are contingent on the population assessed, the methodology employed, the analytic approach, and the end user seeking to apply such information. Population and methodologic considerations are discussed in Chapter 3. The analytic considerations are more fully explored in Chapter 5, while considerations related to end users are discussed in Chapter 6.
b Sampling is nationally representative. However, height and weight data are not collected on children younger than age 12 years.
c Participants are drawn from all 50 states and Washington, DC. However, sample size for each state is generally too small to generate precise state-level estimates from a single year (NHIS, 2015a).
d Based on average participation from 2004-2013; includes all participants surveyed in the year (i.e., both panels being followed in a given year) (AHRQ, 2015).
e The national YRBS and the state and local YRBS are separate surveys with different sampling procedures, and therefore presented separately.
f In the 2013 cycle, sample sizes ranged from 1,107 to 53,785 participants across states and from 1,102 to 10,778 across large urban school districts (Kann et al., 2014).
g State and large urban school districts that participate vary by YRBS cycle; not all states participate (CDC, 2015b).
h Participants in the WIC program must meet eligibility criteria, including income guidelines. Participants include pregnant women, postpartum women, breastfeeding mothers, infants, and children younger than age 5 years.
i Added in the 2011-2012 cycle (CDC, 2015i).
j Anticipated, based on information about the pretesting phase (MCHB, 2015).
kAt the time of this report, this survey has yet to be conducted.
DATA COLLECTED IN THE SCHOOL SETTING
As described in Chapter 3, schools are a common setting for capturing data used to assess obesity prevalence and trends in children and adolescents. In gathering evidence related to data collection methodologies of school-based efforts, the committee encountered barriers similar to those it faced when gathering evidence related to state and local population surveys. Centralized information on the range of state and local initiatives capturing height and weight in the school setting does not currently exist. Information regarding these efforts must be sought out from each individual state or locality. For illustrative purposes, the following sections describe three school-based efforts (Arkansas, California, and Texas) that directly measure students’ heights and weights. Their similarities and differences are presented in Table 4-2 at the end of this section, followed by a summary of the committee’s assessment of current school-based efforts.
The committee elected to differentiate the efforts included in this section from population surveys conducted in schools that collect self-reported height and weight data (i.e., YRBS), in order to compare different ways that states are collecting directly measured height and weight data from students. It is for this reason that the Texas School Physical Activity and Nutrition (SPAN) Survey appears here rather than the preceding discussion on population surveys. The committee acknowledges that school-based assessments employing sampling strategies, like SPAN, have different considerations than those that seek to capture data on all students.
Arkansas
In 2003, Arkansas passed legislation that required an assessment of each public school child’s BMI and a confidential report sent to the student’s parents on an annual basis (Act 1220 of 2003, HB 1583, 84th General Assembly, Regular Session [AR 2003]). Initially, height and weight measurements were collected on all students. In 2007, however, legislation was passed to assess public school students only in kindergarten, and grades 2, 4, 6, 8, and 10 and to further clarify that parents had the right to opt out of the BMI assessment of their child (Act 201 of 2007, HB 1173, 86th General Assembly, Regular Session [AR 2007]).
The Arkansas Center for Health Improvement (ACHI), a nonpartisan health policy center, coordinates the assessment of BMI within the schools. Schools are provided with scales and involved staff members, including nurses and teachers, are trained in the measurements procedures. BMI assessments results are reported back to ACHI through a “secure, web-based computer system that is used to generate individual, confidential Child Health Reports for parents” (ACHI, 2014).
Participation, both at the school and individual level at participating schools, has been high since its inception, exceeding 90 percent for each year7 (ACHI, 2015). Reports are produced with summary data on the percent of students within each BMI classification for a given school, school district, and county. Data also are summarized for the entire state, and shown by sex, grade, and race and ethnicity groups. Schools or districts that provide too little data do not receive summary profiles. Data are collected annually, which provides a substantial number of data points for trends analyses. Although the Arkansas BMI assessments broadly cover public school students, considerations should be made for the students not represented in the data. The results, for example, may not be readily applicable to the more than 25,000 students enrolled in private schools in Arkansas (Broughman, 2013).
California
All California schools are required to administer an annual physical fitness test to students in grades 5, 7, and 9 (California Education Code section 60800). In 1996, the State Board of Education designated the FitnessGram®8 to be the test used to evaluate students’ physical fitness.
The FitnessGram® contains six components, including an assessment of body composition. Body composition can be measured in one of three ways: skinfold measurements, BMI, or bioelectric impedance analysis (see Chapter 2). The vast majority of schools opts to measure height and weight and performs the BMI assessment (98 percent in 2013-2014) (California Department of Education, 2015). Results for the FitnessGram® are submitted electronically (California Department of Education, 2016). In recent years, approximately 92 to 94 percent of enrolled students across the three grades have been assessed through the FitnessGram®. The annual dataset has more than a million records per year of children throughout the state. As such, estimates have been generated for select groups that are not adequately represented in nationally representative surveys on a consistent basis (e.g., Filipinos).
The California Department of Education maintains a comprehensive website and provides a tool for evaluating FitnessGram® results. The tool can query data since the 1998-1999 school year, by school, district, county, and state levels (California Department of Education, 2013). Reports can provide results at the state, county, district, and school level and can be pre-
___________________
7 This number represents only the participation, not the percent of students with valid measurements used for analyses.
8 The FitnessGram® was developed and is a registered trademark of The Cooper Institute®, Dallas, Texas.
sented by sex, economic disadvantage (yes or no), or race and ethnicity category (black, American Indian/Alaskan Native, Asian, Filipino, Hispanic, Native Hawaiian/Pacific Islander, white, or two or more races) (California Department of Education, 2013). The BMI classification approach used by the California FitnessGram® has changed over time and differs from approaches typically taken in the literature. As such, caution should be taken when assessing BMI-related results generated from these data (see Chapter 5, Box 5-2 for more details).
Texas
Another approach being taken to monitor obesity prevalence in school-aged children is the Texas SPAN Survey.9 SPAN is conducted by university researchers supported by Texas Department of State Health and Services funding. In addition to establishing a surveillance system for monitoring obesity in school-aged children in the state, SPAN collects contextual data on dietary practices, nutrition knowledge, and physical activity (SPAN, 2016). SPAN is currently in its fourth cycle (2015-2016), having collected data in 2000-2002, 2004-2005, and 2009-2011. Students in grades 4, 8, and 11 were evaluated in each of the cycles. The 2009-2011 cycle added a parental survey for students in grade 4 and the current cycle has added students in grade 2 and a parental survey to their target population (SPAN, 2016).
The sampling approach seeks to be representative of the entire state and provide subgroup estimates by grade level, sex, race and ethnicity categories, and state health service regions (HSRs)10 (Hoelscher et al., 2004). The sample size has changed over time, as additional evaluations have been added (Hoelscher et al., 2010). Across the survey cycles, SPAN participants’ heights and weights are measured by study staff or by state or county personnel (Hoelscher et al., 2010). As a quality assurance measure, repeated measurements are performed on 5 percent of the students—more than 98 percent of these measurements were within 0.2 kilograms and 1.2 centimeters of the original values (Hoelscher et al., 2010).
___________________
9 Each school district throughout Texas is required to conduct an annual fitness assessment of students in grades 3 through 12 (Texas Education Code, § 38.101 and § 38.103). Like other states and localities, Texas uses the FitnessGram® (Texas Education Agency, 2016), which has many similar considerations, advantages, and disadvantages as the California administration. The Texas School Physical Activity and Nutrition (SPAN) Survey is conducted in addition to the FitnessGram® assessments.
10 Health service regions (HSRs) are regional divisions that allowed for administrative management and program implementation. Texas was originally divided into 11 HSRs, but was later reorganized into 8 HSRs (Texas Department of State Health Services, 2016).
SPAN produces both state and regional estimates of obesity using measured height and weights. However, where many school-based assessments are collected annually, SPAN data have been collected approximately every 5 years. This expanse of time between assessments can be a challenge for those seeking current estimates.
Considerations in Assessing School-Based Efforts That Collect Directly Measured Height and Weight Data
School-based assessments that directly measure students’ heights and weights address some of the limitations of population surveys (e.g., large samples of children, directly measured height and weight data). As illustrated by California, school-based data collection initiatives in locations that are racially and ethnically diverse can allow for assessments of population groups that are not sampled in sufficient quantities to produce reliable estimates from nationally representative population surveys. In states where direct measurement of height and weight in schools is mandated, obesity prevalence and trends have been assessed at the regional, county, and, in some cases, local levels, to the degree to which sufficient data are available and the presentation of the results cannot lead to the identification of individual students.
School-based assessments, however, can be difficult to compare across states because different grades are represented in the data, different approaches exist for determining which students are measured, and different protocols are used to measure height and weight. Furthermore, the prevalence or trends that are generated from these data sources are not necessarily representative of all children or adolescents living within a given jurisdiction.
TABLE 4-2 Comparison of Three Examples of School-Based Efforts That Directly Measure Students’ Heights and Weights
| Assessment | Students Assessed | Sample Size | Who Performs Measurement | Collaborators | Potential Advantagesa | Potential Disadvantagesa |
|---|---|---|---|---|---|---|
| Arkansas BMI Assessment |
All public school students, Pre-K/K, and grades 2, 4, 6, 8, 10 |
181,000 per yearb |
School staff member |
Arkansas Dept. of Education Arkansas Dept. of Health ACHI Schools, districts |
Data collected annually since 2003. Estimates available for schools, districts, counties, and state. |
Only represents public school students. |
| California Annual FitnessGram® |
All public school students, grades 5, 7, 9 |
1.3 million per year |
Local education agency or county education office employee |
California Dept. of Education Local education agencies |
Data collected annually. Large sample size. Diverse populations represented in the data. |
Reports related to BMI not readily comparable over time or other non-FitnessGram® reports. |
| Texas SPAN Surveyc |
Public school students, grades 4, 8, 11 in sampled schoolsd |
17,000 per cycle |
Project staff and Dept. of State Health and Services employees |
University researchers Texas Dept. of State Health and Services Selected schools, districts |
Produces both state and regional estimate. |
Collected approximately every 5 years. |
NOTE: ACHI, Arkansas Center for Health Improvement; BMI, body mass index; Dept., Department; Pre-K/K, pre-kindergarten/kindergarten; SPAN, School Physical Activity and Nutrition (Survey).
a The potential advantages and disadvantages are contingent on the population assessed, the methodology employed, the analytic approach, and the end user seeking to apply such information. Population and methodologic considerations are discussed in Chapter 3. The analytic considerations are more fully explored in Chapter 5, while considerations related to end users are discussed in Chapter 6.
b Based on average number of students who had valid measurements from 2010-2011 through 2013-2014 school years (ACHI, 2014).
c Texas also conducts a FitnessGram® assessment of public school students in grades 3-12, similar to California’s administration.
d The 2015-2016 SPAN is also including students in grade 2 (SPAN, 2016).
CLINICAL AND PUBLIC HEALTH ADMINISTRATIVE DATA
Height and weight measurements collected in clinical and public health settings have also been used as a source of obesity prevalence and trend data. Because these measurements are part of routine care or procedures, they are considered administrative data. A range of administrative data exists, including those collected in public health programs, such as WIC (see Box 4-2). This section, however, discusses the potential for and current challenges in using electronic health records (EHRs) and immunization registries for the purposes of monitoring obesity prevalence and trends.
Electronic Health Records
The concept of integrating EHRs with public health surveillance efforts is embedded in the meaningful use criteria outlined in the Health Information Technology for Economic and Clinical Health Act of 2009 and promoted by Medicare and Medicaid through bonus payments. The Federal Health IT Strategic Plan for 2015-2020 includes a specific strategy to improve community health, well-being, and resilience by increasing “public health entities’ ability to use, benefit from, and manage advances in real-time electronic health information for public health surveillance, situational awareness, and targeted alerting” (ONC, 2016).
One basic question that has been and is being explored through EHR data is the extent to which a measure is actually documented in a patient’s record. The results from such a data query would not, for example, provide insight into how many patients had obesity, but do provide information about the consistency to which height, weight, and BMI are recorded within a given medical practice, plan, or system. The Healthcare Effectiveness Data and Information Set (HEDIS®)11 is a performance measure tool that currently captures such information (see Box 4-4).
Using EHRs for the purposes of assessing obesity prevalence and trends has many appealing features. EHRs house data over the course of a patient’s participation within a given health care system. Accordingly, data can be used to assess cross-sectional prevalence within the patient population as well as over time. Investigators are using EHRs within and between health care systems to assess obesity status on a large number of patients (often millions) and can track individual patients longitudinally (see Box 4-5).
Although EHR datasets can be a particularly rich source of measured height and weight data on a large number of patients, the interpretation of assessments of obesity prevalence and trends comes with several important considerations. One caveat is that the data represent only patients mak-
___________________
11 HEDIS® is a registered trademark of the National Committee for Quality Assurance.
ing use of the medical system. As such, use of EHR data may underrepresent low-income, the uninsured, and other populations. Furthermore, the population represented in an EHR does not necessarily equate to a defined geographic area, but rather to patients who elect to be seen at a particular medical facility or system. Accordingly, the results of an EHR-based assessment are not necessarily generalizable beyond the medical facilities themselves. A final consideration is EHR interoperability. Multiple EHR software systems are currently being used throughout the country. Not all
of these systems are able to interact with each other, meaning that it may not be possible to collect longitudinal data on patients who switch between practices that use different EHR systems. This limitation has implications for trends analyses.
Immunization Registries
States are increasingly interested in using their immunization registries for BMI surveillance. One such example is highlighted in Box 4-6. The goal is to have clinics collect weight and height data and report these data to a state registry for surveillance purposes. This could include reports submitted by clinics at the time of immunization, or even to a separate database from reports required for physical exams for enrollment or school-based BMI screening. The registry concept aligns with the public health objectives of “meaningful use” criteria for clinics, in that it anticipates the ability to send information from EHRs to state public health departments. Immunization registries can be used to calculate incidence rates and prevalence
rates, if data are collected over time and tracked by individual. Additional information can be collected and transmitted to allow for risk profiles and for monitoring trends over time.
COHORT STUDIES
Data from cohort studies also have been used to assess obesity prevalence and trends. Cohort studies are observational studies of a select study population (“cohort”) that is followed over time to determine risk factors associated with changes in levels of disease incidence. Cohort studies can be used to calculate incidence, remission, and prevalence of obesity, as well as trends. Cohort studies can be a way to obtain obesity prevalence estimates on subpopulations of interest that may not be represented in the larger national surveys because of sample restrictions on size and representativeness. They also are useful to provide information on trends over time and to identify risk factors and other characteristics that might be associated with obesity prevalence and trends. However, one limitation of cohort studies is that participation may vary over long periods of time due to loss of followup. Three illustrative examples of cohort studies that collected height and weight data are presented below.
The National Longitudinal Study of Adolescent to Adult Health
The National Longitudinal Study of Adolescent to Adult Health (Add Health) is a large, comprehensive study that follows adolescents as they progress to adulthood. It originated in 1994 and is supported by grants from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), as well as other federal agencies and foundations (Add Health, 2016; Chantala and Tabor, 2010). The original questionnaire was administered in schools to a nationally representative sample of students in grades 7 to 12, and the study followed up with a series of in-home interviews in 1995, 1996, 2001-2002, and 2007-2008. Height and weight data were obtained by interviewers in each wave. A fifth wave is currently under way (2016-2018).
Add Health provides valuable longitudinal data on obesity trends in a fixed, nationally representative population starting in middle and high school through early and mid-adulthood. Data are released immediately after cleaning, with public use data available from the Odum Institute at the University of North Carolina at Chapel Hill, the Inter-University Consortium for Political and Social Research, and Sociometrics (Add Health, 2016). A larger restricted-use sample is available by contractual agreement to certified researchers who commit themselves to maintaining limited data access.
Growing Up Today Study
The Growing Up Today Study (GUTS) began in 1996 to follow the children of Nurses’ Health Study participants. The initial cohort included 16,882 children ages 9 to 14 years. In 2004, GUTS enrolled a new group of 10,923 children between the ages of 10 and 17 years (GUTS, 2013). These participants are now young adults and many continue to complete annual online health questionnaires, which collect information on self-reported weight and height along with information on health and social behaviors. Nearly 100 peer-reviewed research articles have been published from these data looking at obesity-related risk factors and trends in risk factors over time.
Minneapolis Childhood Cohort Studies
The Minneapolis Childhood Cohort Study is one of seven cohort studies included in the International Childhood Cardiovascular Cohort (i3C) Consortium.12 The Consortium is focused on identifying risk factors asso-
___________________
12 The seven cohorts include the (1) Muscatine Study, (2) Bogalusa Heart Study, (3) Cardiovascular Risk in Young Finns Study, (4) Childhood Determinants of Adult Health Study, (5) Minneapolis Childhood Cohort Study, (6) Princeton Lipid Research Clinics Study, and (7) National Heart, Lung, and Blood Institute Growth and Health Study.
ciated with childhood origin of cardiovascular disease by leveraging data collected on more than 40,000 children in the collaborating cohorts (i3C, 2011). The Minneapolis study included three cohorts: the first study recruited 1,200 children ages 7 to 9 years in 1978 with measured height and weight yearly until age 20, at age 24, and in 2007 through 2011 for follow-up; the second cohort started in 1985 and recruited students in grades 5 through 8 who were seen a second time recently at ages 25 to 30 years for additional weight and height measures; the third study recruited 400 children starting in 1995 who were seen roughly at ages 13, 15, 19, and 24 years with measured height and weight data collected. The studies also collected data on a wide range of factors that may influence the development of cardiovascular disease (i3C, 2011).
SUMMARY
Data sources being used to assess obesity prevalence and trends include population surveillance surveys, direct measurement in the school setting, clinical and public health administrative data, and cohort studies. Each has a unique approach and captures different types of data in different ways. Data sources that include directly measured height and weight data for children, adolescents, and young adults include NHANES, school-based assessments, EHRs, and select cohort studies.
REFERENCES
ACHI (Arkansas Center for Health Improvement). 2014. Assessment of childhood and adolescent obesity in Arkansas: Year eleven (Fall 2013-Spring 2014). Little Rock, AR: ACHI.
ACHI. 2015. Statewide body mass index data collection system and reporting. http://achi.net/Pages/OurWork/Project.aspx?ID=42 (accessed March 14, 2016).
Add Health (National Longitudinal Study of Adolescent to Adult Health). 2016. The National Longitudinal Study of Adolescent to Adult Health. http://www.cpc.unc.edu/projects/addhealth (accessed March 18, 2016).
AHRQ (Agency for Healthcare Research and Quality). 2015. MEPS-HC sample sizes. http://meps.ahrq.gov/mepsweb/survey_comp/hc_sample_size.jsp (accessed March 24, 2016).
Bailey, L. C., D. E. Milov, K. Kelleher, M. G. Kahn, M. Del Beccaro, F. Yu, T. Richards, and C. B. Forrest. 2013. Multi-institutional sharing of electronic health record data to assess childhood obesity. PLoS ONE 8(6):e66192.
BlueCross BlueShield. 2015. HEDIS quality measures requirements for meeting clinical goals. Durham, NC: BlueCross BlueShield of North Carolina. https://www.bcbsnc.com/assets/providers/public/pdfs/hedis_provider_quality_reference_guide.pdf (accessed June 7, 2016).
Brener, N. D., T. McManus, D. A. Galuska, R. Lowry, and H. Wechsler. 2003. Reliability and validity of self-reported height and weight among high school students. Journal of Adolescent Health 32(4):281-287.
Broughman, S. P., and N. L. Swaim. 2013. Characteristics of private schools in the United States: Results from the 2011-12 Private School Universe Survey. Washington, DC: U.S. Department of Education. (NCES-316).
California Department of Education. 2013. Dataquest. http://data1.cde.ca.gov/dataquest (accessed March 14, 2016).
California Department of Education. 2015. Report to the governor and the legislature: 2013-14 California physical fitness test report.
California Department of Education. 2016. PFT frequently asked questions and answers. http://www.cde.ca.gov/ta/tg/pf/pft11fasqa.asp#q18 (accessed March 14, 2016).
Carroll, W., and J. Rhoades. 2012. Obesity in America: Estimates for the U.S. civilian non-institutionalized population age 20 and older, 2009. Rockville, MD: Agency for Healthcare Research and Quality. MEPS Statistical Brief (364).
CDC (Centers for Disease Control and Prevention). 2011. Discontinuation of the Pediatric Nutrition Surveillance System and the Pregnancy Nutrition Surveillance System. Atlanta, GA: Centers for Disease Control and Prevention. http://www.nj.gov/health/fhs/shapingnj/news/newsletter/documents/Letter%20re%20PedNSS%20and%20PNSS.pdf (accessed June 10, 2016).
CDC. 2012. 2012 SMART BRFSS MMSA methodology. http://www.cdc.gov/brfss/smart/2012/2012_smart_brfss_mmsa_methodology.pdf (accessed March 14, 2016).
CDC. 2013a. Methodology of the Youth Risk Behavior Surveillance System—2013. Morbidity and Mortality Weekly Report 62(1).
CDC. 2013b. National Health and Nutrition Examination Survey (NHANES) anthropometry procedures manual. http://www.cdc.gov/nchs/data/nhanes/nhanes_13_14/2013_Anthropometry.pdf (accessed June 7, 2016).
CDC. 2013c. SMART: BRFSS frequently asked questions (FAQs). http://www.cdc.gov/brfss/smart/smart_faq.htm (accessed March 14, 2016).
CDC. 2014. A guide to conducting your own Youth Risk Behavior Survey. http://www.cdc.gov/healthyyouth/data/yrbs/pdf/yrbs_conducting_your_own.pdf (accessed March 28, 2016).
CDC. 2015a. Alaska Youth Risk Behavior Survey (YRBS) FAQs. http://dhss.alaska.gov/dph/Chronic/Documents/School/pubs/YRBS_FAQ.pdf (accessed March 28, 2016).
CDC. 2015b. Behavioral Risk Factor Surveillance System 2014 summary data quality report. http://www.cdc.gov/brfss/annual_data/2014/pdf/2014_dqr.pdf (accessed March 24, 2016).
CDC. 2015c. Behavioral Risk Factor Surveillance System: Comparability of data BRFSS 2014. http://www.cdc.gov/brfss/annual_data/2014/pdf/compare_2014.pdf (accessed March 24, 2016).
CDC. 2015d. BRFSS questionnaires. http://www.cdc.gov/brfss/questionnaires/index.htm (accessed March 14, 2016).
CDC. 2015e. National Survey of Children with Special Health Care Needs. http://www.cdc.gov/nchs/slaits/cshcn.htm (accessed March 14, 2016).
CDC. 2015f. Native Hawaiian and Pacific Islander (NHPI)-National Health Interview Survey (NHIS). http://www.cdc.gov/nchs/nhis/nhpi.html (accessed March 9, 2016).
CDC. 2015g. Overview: BRFSS 2014. http://www.cdc.gov/brfss/annual_data/2014/pdf/overview_2014.pdf (accessed March 11, 2016).
CDC. 2015h. Smart BRFSS factsheet. http://www.cdc.gov/brfss/factsheets/index.htm (accessed March 14, 2016).
CDC. 2015i. State and Local Area Integrated Telephone Survey. http://www.cdc.gov/nchs/slaits.htm (accessed March 24, 2016).
CDC. 2015j. YRBS questionnaires. http://www.cdc.gov/healthyyouth/data/yrbs/questionnaires.htm (accessed March 24, 2016).
CDC. 2015k. YRBSS frequently asked questions. National Center for Health Statistics. http://www.cdc.gov/healthyyouth/data/yrbs/faq.htm (accessed March 24, 2016).
CDC. 2016a. Key concepts about survey history. http://www.cdc.gov/nchs/tutorials/nhanes/SurveyOrientation/SurveyOverview/Info1.htm (accessed March 7, 2016).
CDC. 2016b. National Health Interview Survey: The principal source of information on the health of the U.S. population. Hyattsville, MD: National Center for Health Statistics. http://www.cdc.gov/nchs/data/nhis/brochure2010january.pdf (accessed March 9, 2016).
CDC. 2016c. Nutrition, physical activity and obesity: Data, trends and maps. https://nccd.cdc.gov/NPAO_DTM/IndicatorSummary.aspx?category=28&indicator=89&year=2008&yearId=10 (accessed May 11, 2016).
Chantala, K., and J. Tabor. 2010. Strategies to perform a design-based analysis using the Add Health data. Chapel Hill, NC: Carolina Population Center, University of North Carolina at Chapel Hill.
CHIS (California Health Interview Survey). 2014. CHIS 2014 county sample sizes by age. Los Angeles, CA: UCLA Center for Health Policy Research.
Colorado WIC Program. 2013. Overweight and obesity prevalence among children ages 2 through 4 years enrolled in the Colorado WIC program during 2012. https://www.colorado.gov/pacific/sites/default/files/PF_WIC_overweight-obese-Description-2012.pdf (accessed March 24, 2016).
Curtin, L. R., L. Mohadjer, and S. Dohrmann, et al. 2012. The National Health and Nutrition Examination Survey: Sample design, 1999-2006. Vital and Health Statistics (2)155.
Dalenius, K., E. Borland, B. Smith, B. Polhamus, and L. Grummer-Strawn. 2012. Pediatric Nutrition Surveillance 2010 report. Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention.
Data Resource Center for Child and Adolescent Health. 2016a. Frequently asked questions about the NSCH: Why aren’t there any data on the DRC for Asian, American Indian, Alaska Native, or Native Hawaiian/Pacific Islander children in the NSCH? http://childhealthdata.org/learn/NSCH/FAQ#Meas2 (accessed March 14, 2016).
Data Resource Center for Child and Adolescent Health. 2016b. How many children age 10-17 years are currently overweight or obese, based on body mass index (BMI) for age? http://childhealthdata.org/browse/survey/allstates?q=2416 (accessed March 24, 2016).
Data Resource Center for Child and Adolescent Health. 2016c. Request a data set. https://childhealthdata.org/help/dataset (accessed March 14, 2016).
Ezzati-Rice, T. M., F. Rohde, and J. Greenblatt. 2008. Sample design of the Medical Expenditure Panel Survey Household Component, 1998-2007. Rockville, MD: Agency for Healthcare Research and Quality. Methodology Report (22).
Forrest, C. B., P. A. Margolis, L. C. Bailey, K. Marsolo, M. A. Del Beccaro, J. A. Finkelstein, D. E. Milov, V. J. Vieland, B. A. Wolf, F. B. Yu, and M. G. Kahn. 2014. PEDSnet: A national pediatric learning health system. Journal of the American Medical Informatics Association 21(4):602-606.
GUTS (Growing Up Today Study). 2013. GUTS history. http://www.gutsweb.org/index.php/about-us/guts-then-and-now (accessed March 18, 2016).
Hoelscher, D. M., R. S. Day, E. S. Lee, R. F. Frankowski, S. H. Kelder, J. L. Ward, and M. E. Scheurer. 2004. Measuring the prevalence of overweight in Texas schoolchildren. American Journal of Public Health 94(6):1002-1008.
Hoelscher, D. M., S. H. Kelder, A. Perez, R. S. Day, J. S. Benoit, R. F. Frankowski, J. L. Walker, and E. S. Lee. 2010. Changes in the regional prevalence of child obesity in 4th, 8th, and 11th grade students in Texas from 2000-2002 to 2004-2005. Obesity (Silver Spring) 18(7):1360-1368.
i3C (International Childhood Cardiovascular Cohort). 2011. International Childhood Cardiovascular Cohort (i3C) Consortium. http://i3cconsortium.org/cohorts.html (accessed March 18, 2016).
JHHC (Johns Hopkins HealthCare). 2016. 2016 provider tips for optimizing HEDIS results. http://www.hopkinsmedicine.org/johns_hopkins_healthcare/downloads/updated_files/HEDISTipSheetJan2016.pdf (accessed March 24, 2016).
Johnson, B., B. Thorn, B. McGill, A. Suchman, M. Mendelson, K. L. Patlan, B. Freeman, R. Gotlieb, and P. Connor. 2013a. WIC participant and program characteristics 2012: Final report. Alexandria, VA: U.S. Department of Agriculture, Food and Nutrition Service.
Johnson, C. L., R. Paulose-Ram, and C. L. Ogden. 2013b. National Health and Nutrition Examination Survey: Analytic guidelines, 1999-2010. Vital and Health Statistics 2(161).
Johnson, C., S. Dohrmann, V. Burt, and L. Mohadjer. 2014. National Health and Nutrition Examination Survey: Sample design, 2011-2014. Vital and Health Statistics (2)162.
Kann, L., S. Kinchen, S. L. Shanklin, K. H. Flint, J. Kawkins, W. A. Harris, R. Lowry, E. O. Olsen, T. McManus, D. Chyen, L. Whittle, E. Taylor, Z. Demissie, N. Brener, J. Thornton, J. Moore, and S. Zaza. 2014. Youth Risk Behavior Surveillance—United States, 2013. MMWR Surveillance Summary 63(Suppl 4):1-168.
Le, A., S. E. Judd, D. B. Allison, R. Oza-Frank, O. Affuso, M. M. Safford, V. J. Howard, and G. Howard. 2014. The geographic distribution of obesity in the US and the potential regional differences in misreporting of obesity. Obesity (Silver Spring) 22(1):300-306.
MCHB (Maternal and Child Health Bureau). 2015. Dataspeak: Progress on the redesign of the NSCH and the data resource center. Paper read at DataSpeak Webinars.
MCIR (Michigan Care Improvement Registry). 2014. Body mass index growth module in the Michigan Care Improvement Registry: Frequently asked questions. https://www.mcir.org/wp-content/uploads/2014/09/BMI_Growth_Module_in_MCIR_FAQ.pdf (accessed March 14, 2016).
McPherson, M., P. Arango, H. Fox, C. Lauver, M. McManus, P. W. Newacheck, J. M. Perrin, J. P. Shonkoff, and B. Strickland. 1998. A new definition of children with special health care needs. Pediatrics 102(1 Pt 1):137-140.
MEPS (Medical Expenditure Panel Survey). 2009. MEPS topics: Obesity. Agency for Healthcare Research and Quality. http://meps.ahrq.gov/mepsweb/data_stats/MEPS_topics.jsp?topicid=13Z-1 (accessed March 9, 2016).
MEPS. 2011. MEPS survey questionnaires. Agency for Healthcare Research and Quality. http://meps.ahrq.gov/mepsweb/survey_comp/survey.jsp#supplemental (accessed March 9, 2016).
Mirel, L. B., and S. R. Machlin. 2013. Enhancing the Medical Expenditure Panel Survey through data linkages. Paper read at Federal Committee on Statistical Methodology (FCSM) Research Conference, Washington, DC.
NCHS (National Center for Health Statistics). 2015a. 2014 National Health Interview Survey (NHIS) survey description. ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/NHIS/2014/srvydesc.pdf (accessed March 7, 2016).
NCHS. 2015b. National Health Interview Survey. NCHS Fact Sheet. http://www.cdc.gov/nchs/data/factsheets/factsheet_nhis.pdf (accessed March 9, 2016).
NCQA (National Committee for Quality Assurance). 2014. NCQA’s health insurance plan rankings frequently asked questions. https://www.ncqa.org/Portals/0/Report%20Cards/Rankings/HPR2014_FAQ_Final.pdf (accessed March 24, 2016).
NCQA. 2015. The state of health care quality 2015. National Committee for Quality Assurance. http://blog.ncqa.org/wp-content/uploads/SOHC-2015-Slides.pdf (accessed March 24, 2016).
NCQA. 2016a. HEDIS measures. http://www.ncqa.org/HEDISQualityMeasurement/HEDISMeasures.aspx (accessed March 24. 2016).
NCQA. 2016b. Summary table of measures, product lines and changes. http://www.ncqa.org/Portals/0/HEDISQM/HEDIS2016/HEDIS%202016%20List%20of%20Measures.pdf (accessed April 1, 2016).
NHANES (National Health and Nutrition Examination Survey). 2013a. Interviewers procedure manual. https://www.cdc.gov/nchs/data/nhanes/nhanes_13_14/Intrvwr_Proc_Manual.pdf (accessed March 7, 2016).
NHANES. 2013b. Questionnaire: Demographics information. http://www.cdc.gov/nchs/data/nhanes/nhanes_13_14/DMQ_H.pdf (accessed March 24, 2016).
NHIS (National Health Interview Survey). 2014. National Health Interview Survey (NHIS) geocodes. http://www.cdc.gov/rdc/geocodes/geowt_nhis.htm (accessed March 7, 2016).
NHIS. 2015a. About the National Health Interview Survey. http://www.cdc.gov/nchs/nhis/about_nhis.htm#sample_design (accessed March 7, 2016).
NHIS. 2015b. Draft 2015 NHIS questionnaire-household composition. National Center for Health Statistics. ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Survey_Questionnaires/NHIS/2015/english/qhhld.pdf (accessed April 15, 2016).
NHIS. 2015c. Draft 2015 NHIS questionnaire-sample child. National Center for Health Statistics. ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Survey_Questionnaires/NHIS/2015/english/qchild.pdf (accessed March 28, 2016).
NSCH (National Survey of Children’s Health). 2003. Fast facts about the survey. Data Resource Center for Child and Adolescent Health. https://childhealthdata.org/docs/nsch-docs/2003_fast_facts_508-pdf.pdf?sfvrsn=2 (accessed March 14, 2016).
NSCH. 2007. Fast facts about the survey. https://childhealthdata.org/docs/nsch-docs/2007_fast_facts_508-pdf.pdf?sfvrsn=2 (accessed March 14, 2016).
NSCH. 2012a. 2011 NSCH questionnaire. Quarter 1 2012. http://www.cdc.gov/nchs/data/slaits/2011nschquestionnaire.pdf (accessed March 14, 2016).
NSCH. 2012b. Fast facts about the survey. https://childhealthdata.org/docs/drc/2011-12-fastfacts.pdf?sfvrsn=1 (accessed March 14, 2016).
ONC (Office of the National Coordinator for Health Information Technology). 2016. Federal health IT strategic plan goals. http://dashboard.healthit.gov/strategic-plan/federal-healthit-strategic-plan-goals.php (accessed March 14, 2016).
Parsons, V., Moriarity, C, Jonas, K, et al. 2014. Design and estimation for the National Health Interview Survey, 2006-2015. National Center for Health Statistics. Vital and Health Statistics 2(165).
Porter, K. S., L. R. Curtin, M. D. Carroll, X. Li, L. Mohadjer, M. Shih, P. A. Simon, and J. E. Fielding. 2011. Health of adults in Los Angeles County: Findings from the National Health and Nutrition Examination Survey, 1999-2004. National Health Statistics Report (42):1-14.
SPAN (School Physical Activity and Nutrition Survey). 2016. School Physical Activity and Nutrition Survey (SPAN). University of Texas.
Texas Department of State Health Services. 2016. Health service regions. http://www.dshs.state.tx.us/regions/default.shtm (accessed March 14, 2016).
Thorn, B., C. Tadler, N. Huret, E. Ayo, C. Trippe, M. Mendelson, K. L. Patlan, G. Schwartz, and V. Tran. 2015. WIC participant and program characteristics 2014: Final report. Alexandria, VA: U.S. Department of Agriculture, Food and Nutrition Service.
UCLA (University of California, Los Angeles) Center for Health Policy Research. 2015. California Health Interview Survey: Survey topics. http://healthpolicy.ucla.edu/chis/design/Pages/survey-topics.aspx (accessed March 14, 2016).
UCLA Center for Health Policy Research. 2016a. What’s new in the 2011-2012 California Health Interview Survey. http://healthpolicy.ucla.edu/chis/design/Documents/whats-newchis-2011-2012.pdf (accessed March 14, 2016).
UCLA Center for Health Policy Research. 2016b. What’s new in the 2013-2014 California Health Interview Survey. http://healthpolicy.ucla.edu/chis/design/Documents/Whats-NewCHIS-2013-2014.pdf (accessed March 14, 2016).
USDA (U.S. Department of Agriculture). 1994. WIC program and participants characteristics. http://www.fns.usda.gov/wic-program-and-participants-characteristics-1994 (accessed May 11, 2016).
Zipf, G., M. Chiappa, and K. Porter, et al. 2013. National Health and Nutrition Examination Survey: Plan and operations, 1999-2010. Vital and Health Statistics 1(56).
