2019 NHIS Redesign
The NHIS questionnaire was substantially redesigned in 2019. According to the 2019 NHIS Survey Description, "In 2019, the structure and content of the NHIS were redesigned to better meet the needs of data users, the Centers for Disease Control and Prevention, and the Department of Health and Human Services (DHHS). The goals of the redesign were to reduce respondent burden by shortening the length of the questionnaire, harmonize overlapping content with other federal health surveys, establish a long-term structure of ongoing and periodic topics, and incorporate advances in survey methodology and measurement."
Data Collection Methods
For 1997-2018, NHIS was a sample of households, with each household potentially containing multiple families. One person, the family respondent, was selected to give information about all family members. Information was collected on demographics, health status, chronic conditions, health insurance, and injuries. Interviewers then randomly selected one adult and one child per family to complete the additional "sample adult" and "sample child" interviews. The sample adult and sample child questionnaires collected more detailed information on the individual, including Body Mass Index, mental health, access to health care, health behaviors, and (for adults) sexual orientation and details about paid employment. For these years, NCHS released separate data files for households, families, family members, sample adults, and sample children. Please see the blog post for more information on NHIS data collection in 1997-2018.
The survey design dramatically changed in 2019. The NHIS is still a sample of households, but there is no longer a family respondent who answers questions about all family members. Instead, the interviewer completes a brief household rostering interview to collect basic demographic information of all people who usually live or stay in the household. The household roster also identifies the members of the sample adult's and sample child's families in the household. Then, one sample adult and one sample child are selected per household to complete the sample adult and sample child questionnaires. In contrast to earlier years, family-level content is collected within the sample adult and/or sample child questionnaire. If the sample adult and sample child are from the same family, the family-level content is collected in whichever interview occurs first and is copied to the other data file in a post-processing step (i.e., if the sample child interview occurs first, the family-level information is later replicated on the sample adult file). Because there is no longer an entire family questionnaire, there is loss of content on family members beginning in 2019, including information on "health insurance, access to health care and utilization of select health care services, country of birth (if not US), some employment and earnings information, active duty military time periods, disability-associated conditions, and receipt of medical advice by phone" (2019 Survey Description).
In addition, the Sample Adult and Sample Child data files now include four types of content: annual core content, rotating core content, sponsored content, and emerging content. Annual core content will be included in the NHIS every year, whereas rotating core content appears either every other year, one out of every three years, or two out of every three years. Sponsored content is funded by external agencies. If funded by a sustaining sponsor, this content appears every year, otherwise it is included in the questionnaire when funding is provided. The NCHS sponsors emerging content to ask questions about emerging topics. For example, in 2019, the emerging content includes questions about prescription opioid use and pain management.
Differences in File Structure and Content
In contrast to earlier years, beginning in 2019, the NCHS will release data files for the Sample Adult, Sample Child, Imputed Income (Sample Adult), Imputed Income (Sample Child), and Paradata. There will no longer be household, family, or person data files, as household and family information is now contained in the Sample Adult and Sample Child files. Family-level content can be asked in either the Sample Adult or Sample Child questionnaires if both the sample adult and sample child are in the same family. If they are in different families, the family-level content is asked in both questionnaires.
Changes to Procedures to Create Sampling Weights
Sampling weights are constructed so that each person can be inflated or expanded to represent the total population of the United States. The approach for creating NHIS sampling weights changed significantly in 2019; large declines in response rates, from approximately 90% to 60% or less, required the introduction of more sophisticated adjustment techniques to better correct for nonresponse. Major differences from past approaches to nonresponse adjustment and creation of the final sampling weights include the use of multilevel models to adjust for nonresponse and raking (rather than post-stratification) to adjust weights to match population totals. For more information, please refer to the user note on sampling weights.
Implications for Analysis
As a result of the new data collection structure, most variables that were previously collected for all family members are now only collected for sample adults and/or sample children. Users will note that universes previously may have been "all persons," whereas, beginning in 2019, they may be "sample adults age 18+", "sample children age 0-17", or "sample children age 0-17 and sample adults age 18+." Therefore, users must be careful to use the proper weights when using data from before 2019 and from 2019 forward. For example, variables collected for "all persons" prior to 2019 and for "sample adults age 18+", "sample children age 0-17", or "sample children age 0-17" for 2019 forward should use the weight PERWEIGHT for years prior to 2019 and the weight SAMPWEIGHT for 2019 forward. For more guidance on selecting the proper weights for analysis, please see the user note on sampling weights.
At the time of this writing, NCHS advises against analyses that pool 2019 with earlier years, as it will be difficult to detect whether differences between 2019 and earlier years are due to true underlying change or due to changes in measurement and data collection.