Health, United States banner

National Health and Nutrition Examination Survey (NHANES)

National Center for Health Statistics

Print
Related Pages
On This Page

Overview

NHANES is designed to assess the health and nutritional status of adults and children in the United States. The survey is unique because it combines interviews and health examinations. NHANES collects data on the prevalence of chronic and infectious diseases and conditions (including undiagnosed conditions) and on risk factors such as obesity, elevated serum cholesterol levels, hypertension, diet and nutritional status, and a number of other measures.

Coverage

NHANES III, conducted during 1988–1994, and the continuous NHANES, which began in 1999, target the U.S. civilian noninstitutionalized population.

Top of Page

Methodology

NHANES includes clinical examinations, selected medical and laboratory tests, and self-reported data. NHANES interviews people in their homes and conducts health examinations in a mobile examination center (MEC), including laboratory analysis of blood, urine, and other tissue samples. During 1988–1994, as a substitute for the MEC examinations, a small number of survey participants received an abbreviated health examination in their homes if they could not come to the MEC.

The survey for NHANES III was conducted from 1988 to 1994 using a stratified, multistage probability design to sample the U.S. civilian noninstitutionalized population. About 40,000 people age 2 months and older were selected and asked to complete an extensive interview and a health examination. Children ages 2 months through 5 years, adults age 60 and older, Black people, and people of Mexican origin were oversampled to provide precise descriptive information on the health status of selected population groups.

Beginning in 1999, NHANES became a continuous annual survey, collecting data every year from a representative sample of the U.S. civilian noninstitutionalized population, newborns and older, through in-home personal interviews and health examinations in the MEC. The sample design is a complex, multistage, clustered design using unequal probabilities of selection. The first-stage sample frame for continuous NHANES is primary sampling units (PSUs). Typically, an NHANES PSU is a county. For 1999, 12 distinct PSUs were in the annual sample. For each year during 2000–March 2020, 15 PSUs were selected. (Due to the COVID-19 pandemic, only 3 were visited in 2020 before field operations were suspended.) The within-PSU design involves: a) forming secondary sampling units within counties, b) selecting dwelling units within secondary sampling units, and then c) selecting sample people within dwelling units. Selection of the final sample person involves differential probabilities of selection according to the demographic variables of sex (male or female), race and Hispanic origin, and age. Because of the differential probabilities of selection, dwelling units are screened for potential sample people.

Beginning in 1999, NHANES oversampled low-income people, adolescents ages 12–19 years, adults age 60 and older, Black or African American people, and people of Mexican origin. The sample for data years 1999–2006 was not designed to give a nationally representative sample for the total Hispanic population living in the United States. Starting with the 2007–2010 data collection, all Hispanic people were oversampled, not just people of Mexican origin, and adolescents were no longer oversampled. For 2011 through 2014, the sampling design was changed and the following groups were oversampled: Hispanic people; Black non-Hispanic people; non-Black Asian non-Hispanic people; White non-Hispanic and other people at or below 130% of the poverty level; and adults age 80 and older. For 2015 through 2018, the sampling design was revised again, changing the cut point for low-income oversampling from 130% of poverty level or below to 185% of poverty level or below, and oversampling children ages 0–11 years. Oversampling in the 2019–2022 sample design is consistent with the 2015–2018 design. For more information on the sample design for 1999–2006, see: The National Health and Nutrition Examination Survey: Sample Design, 1999–2006; for 2007–2010, see: National Health and Nutrition Examination Survey: Sample Design, 2007–2010 ; for 2011–2014, see: National Health and Nutrition Examination Survey: Sample Design, 2011–2014; for 2015–2018, see: National Health and Nutrition Examination Survey, 2015−2018: Sample Design and Estimation Procedures; and for 2017–March 2020, see: National Health and Nutrition Examination Survey, 2017–March 2020 Prepandemic File: Sample Design, Estimation, and Analytic Guidelines.

The estimation procedure used to produce national statistics for NHANES involves inflation by the reciprocal of the probability of selection, adjustment for nonresponse, and poststratified ratio adjustment to population totals. Sampling errors also are estimated, to measure the reliability of the statistics.

During the NHANES 2019–2020 cycle, field operations were suspended in March 2020 due to the COVID-19 pandemic. The partial 2019–2020 sample was not nationally representative. To create a data set that could be used to produce nationally representative estimates, the partial 2019–2020 data were combined with the full 2017–2018 cycle to create the 2017–March 2020 prepandemic file using methodology to combine a probability sample with a partial probability sample. Thorough evaluations showed that the final data set produced reliable national estimates. Trend comparisons for subgroups (for example, by age, sex, race and Hispanic origin) between the 2017–March 2020 prepandemic file and previous NHANES cycles should be interpreted with caution. The magnitude and direction of the differences between two cycles may vary by subgroup. This could cause changes in the magnitude or the direction of the trend within a certain subgroup, relative to overall changes. When conducting analyses and interpreting results, analysts should consider the historical context of the trends in addition to the methodological approach to create the 2017–March 2020 prepandemic file. For more detailed information, see National Health and Nutrition Examination Survey, 2017–March 2020 Prepandemic File: Sample Design, Estimation, and Analytic Guidelines.

Top of Page

Sample Size and Response Rate

  • Over the 6-year survey period of NHANES III, 39,695 people were eligible, the unweighted household interview response rate was 86% (33,994), and the unweighted medical examination response rate was 78% (30,818).
  • For NHANES 1999–2000 through NHANES 2013–2014, the number of eligible people selected ranged from 12,160 to 14,332. The unweighted response rates ranged from 71% to 84% for the interviewed sample and from 68% to 80% for the examined sample.
  • For NHANES 2015–2016, a total of 15,327 people were eligible, of which 9,971 were interviewed and 9,544 completed the health examination component. The unweighted response rates were 61% for the interviewed sample and 59% for the examined sample.
  • For NHANES 2017–2018, a total of 16,211 people were eligible, of which 9,254 were interviewed and 8,704 completed the health examination component. The unweighted response rates were 52% for the interviewed sample and 49% for the examined sample.
  • For NHANES 2017–March 2020, a total of 27,066 people were eligible, of which 15,560 were interviewed and 14,300 completed the health examination component. The unweighted response rates were 51% for the interviewed sample and 47% for the examined sample.

In addition to accounting for sample person nonresponse, weights are also poststratified to match the population control totals for each sampling subdomain. This makes the weighted counts the same as an independent estimate of the U.S. civilian noninstitutionalized population. For NHANES 2011–2020, the sample weights were poststratified (2011–2016) or calibrated (2017–March 2020) to population totals from the American Community Survey (ACS). The weights for earlier NHANES cycles were poststratified to population totals from the Current Population Survey (CPS). This change from CPS to ACS was made, in part, because the larger sample size of ACS provides more reliable population estimates for Asian people within age and sex categories, which are required due to the addition of the Asian oversample starting in the 2011 survey. For more detailed information on unweighted NHANES response rates and response weights using sample size weighted to ACS and CPS population totals, see: NHANES Response Rates and Population Totals.

For the 2017–2018 NHANES cycle, enhanced weighting methods were required to minimize errors of representation resulting from sample location characteristics and nonresponse. To further reduce any error, combining multiple comparable cycles is recommended. For more information on nonresponse bias in the 2017–2018 NHANES, see: An Investigation of Nonresponse Bias and Survey Location Variability in the 2017−2018 National Health and Nutrition Examination Survey.

Top of Page

Issues Affecting Interpretation

Data elements, laboratory tests performed, and the technological sophistication of health examination and laboratory equipment have changed over time. Consequently, trend analyses should carefully examine how specific data elements were collected across the various survey years. Data files are revised periodically. If the file changes are minor and the impact on estimates is small, then the data are not revised in Health, United States. Major data changes are incorporated.

Periodically, NHANES changes its sampling design to oversample different groups. Because the total sample size in any year is fixed due to operational constraints, sample sizes for the other oversampled groups (including Hispanic people and non-low-income White and other people) are decreased. Consequently, trend analyses on demographic subpopulations should be carefully evaluated to determine if the sample sizes meet the analytic guidelines. (For more information, see National Center for Health Statistics Data Presentation Standards for Proportions. In general, any 2-year data cycle in NHANES can be combined with adjacent 2-year data cycles to create analytic data files based on 4 years of data or more, which improves precision. If provided, NHANES 4-year weights should be used. Otherwise, the user should apply adjusted sampling weights. However, because of the sample design change in 2011–2012, the data user should be aware of the implications of combining these data with data from earlier survey cycles. Users are advised to examine their estimates carefully to see if the 4-year estimates (and sampling errors) are consistent with each set of 2-year estimates.

Top of Page

References

 

For more information, see the NHANES website.

Top of Page