National Health Interview Survey (NHIS)

Overview

Data from NHIS are used to monitor the health of the U.S. population on a broad range of health topics by many demographic and socioeconomic characteristics. During household interviews, NHIS collects information on the demographic and socioeconomic characteristics of respondents, in addition to information on activity limitation, illnesses, chronic conditions, health insurance coverage (or lack of), utilization of health care, and other health topics.

Coverage

The survey covers the civilian noninstitutionalized population of the United States. Among those excluded are patients in long-term care facilities, people on active duty with the armed forces (although their civilian family members are included), people who are incarcerated, and U.S. nationals living in foreign countries.

Methodology

NHIS is a cross-sectional household interview survey. Sampling and interviewing are continuous throughout each year. The sample design is a probability design that permits the representative sampling of households and noninstitutional group quarters. The sample design is redesigned after every decennial census to better measure the changing U.S. population and to meet new survey objectives. A new sample design was implemented in 2016, with additional changes in 2018.

Before 2016, the sample design was a multistage sample design, with primary sampling units selected at the first stage. The current sample design is not a multistage sample design. It consists instead of a sample of more than 300 clusters of addresses that are located in well-defined geographic areas. These geographic areas consist of a county, a small group of contiguous counties, or a metropolitan area, always located within the boundaries of a single state. The geographic areas are classified into one to two groups per state, and conceptual clusters of addresses are identified using 2010 census housing-unit count information. A sample of address clusters is selected independently for each group for NHIS field interviewing efforts.

In 2019, for the first time in more than 20 years, the NHIS questionnaire was redesigned to increase relevance, enhance data quality, and minimize respondent burden. The wording of some questions and answer choices was revised to improve quality of measurement. Some questions were dropped, and others were added, changing the context within which some questions were asked. As part of the 2019 redesign, the number of family members for whom data are collected has been reduced to one “Sample Adult” aged 18 and over and one “Sample Child” aged 17 years and under (if any children live in the household). Information about the sample adult is collected from the sample adult unless they are physically or mentally unable to do so, in which case a knowledgeable proxy can answer for the sample adult. Information about the sample child is collected from a parent or adult who is knowledgeable about and responsible for the health care of the sample child. From 1997 through 2018, NHIS also included a family questionnaire. Questions from the family questionnaire were answered by a member of the family on behalf of all members of the family.

As of 2019, the redesigned NHIS questionnaire consists of three modules: Household Roster, Sample Adult Interview, and Sample Child Interview. Before 2019, NHIS consisted of four components: Household Composition, Family Core, Sample Adult Core, and Sample Child Core.

Sample Size and Response Rate

The NHIS sample size varies from year to year. It may be reduced for budgetary reasons or augmented if supplementary funding is available. The normal annual sample size (in other words, the number of households or people for whom data are collected and publicly released) for both the previous 2006–2015 sample design and for the new 2016 sample design is about 35,000 households containing about 87,500 people (survey years up to 2018) or 27,000 Sample Adults and 9,000 Sample Children (starting in 2019). With the 2019 redesign, an estimated 28,000 Sample Adult and 8,400 Sample Child interviews are expected to be available annually for analysis in NHIS.

In 2018, the NHIS sample was 29,839 households containing 72,831 people in 30,309 families, with 25,417 Sample Adults and 8,269 Sample Children. In 2018, the total household response rate was 64.2%. The response rate in 2018 for the Sample Adult component was 53.1% and for the Sample Child component, 59.2%. In 2019, the NHIS sample was 33,138 households containing 9,193 Sample Children and 31,997 Sample Adults. In 2019, the total household response rate was 61.1%. The final Sample Adult and final Sample Child response rates in 2019 were both 59.1%.

Issues Affecting Interpretation

As part of the 1997 questionnaire redesign, some basic concepts were changed, and other concepts were measured in different ways. For some questions, a change was made in the reference period. Also in 1997, the data collection methodology changed from paper-and-pencil questionnaires to computer-assisted personal interviewing. Some indicators presented in Health, United States begin with 1997 data because the redesign caused a break in the trend. Moreover, starting with Health, United States, 2005, estimates for 2000–2002 were revised to use 2000-based weights, and they differ from previous editions of Health, United States that used 1990-based weights for those data years. The weights available in the public-use NHIS files for 2000–2002 are 1990-based. Data for 2003–2011 use weights derived from U.S. Census 2000. Data for 2012 and beyond use weights derived from U.S. Census 2010. In 2006–2010, the sample size was reduced, which is associated with slightly larger variance estimates than in other years when a larger sample was fielded. Starting in 2010, a geographic nonresponse adjustment was made to both the Sample Adult weight and the Sample Child weight; see Moriarity.

A new approach to weighting was implemented with the 2019 redesign to address nonresponse bias. The updated approach for nonresponse adjustment used multilevel regression models that include paradata variables predictive of survey response and selected key health outcomes. In addition, the nonresponse adjusted weights were calibrated to U.S. Census Bureau population projections and an expanded set of variables from the American Community Survey (age, sex, race and ethnicity, educational attainment, Census division, and metropolitan statistical area [MSA] status). These changes to the nonresponse adjustment approach and the calibration methods have the potential to impact comparisons of the weighted survey estimates over time.

Due to the 2019 redesign of the NHIS questionnaire and other changes in weighting and design methodology, data for 2019 are not comparable with data for earlier years.

References

• Moriarity C. 2009 National Health Interview Survey sample adult and sample child nonresponse bias analysis. 2010. Available from: https://www.cdc.gov/nchs/data/nhis/nr_bias_analysis_report_2009_NHIS.pdf.
• National Center for Health Statistics. Preliminary evaluation of the impact of the 2019 National Health Interview Survey questionnaire redesign and weighting adjustments on Early Release Program estimates. Hyattsville, MD. 2020. Available from: https://www.cdc.gov/nchs/data/nhis/earlyrelease/EReval202009-508.pdf.
• National Center for Health Statistics. National Health Interview Survey, 2019. Public-use data file and documentation. 2020. Available from: https://www.cdc.gov/nchs/nhis/data-questionnaires-documentation.htm.

For more information, see the NHIS website at: https://www.cdc.gov/nchs/nhis.htm.