Health and Retirement Study

Health and Retirement Study

Dataset Summary


HRS is a large, nationally-representative longitudinal panel survey of approximately 20,000 persons age 50 and older who are followed every 2 years until death. It started in 1992 and has continued through the present.

Who is in the data, and how are they followed?

HRS comprises a nationally representative sample of adults age 50 and older who were non-institutionalized at the time of enrollment. Subjects are enrolled between age 50 and 56, and followed every 2 years until death. In the mid-1990s, an additional cohort of adults age 70 and older were enrolled in a substudy entitled Assets and Health Dynamics among the Oldest Old (AHEAD). African-Americans and Hispanics are oversampled in HRS, allowing greater statistical power for studies focused on these populations. Patient weights and other survey design factors are provided in the data so that results can be adjusted to be nationally representative.

Following their enrollment interview, all subjects receive followup interviews conducted in person or by telephone every 2 years. In addition, selected subgroups of subjects are given "leave-behind" questionnaires to complete and send in via mail, as well as off-year questionnaires. Medicare claims data can be obtained and linked to the survey data to provide a more complete picture of health status and health services utilization (see further information below).

What kind of information is in the data?

HRS contains extensive self-report information about functional abilities, cognitive status, and social factors, and as a result has been a major source of research on function and disability in older adults. See below for additional information about items of particular interest for studies of function and disability. The longitudinal nature of HRS makes is particularly useful for studies that evaluate outcomes over time.

While most core survey modules have remained consistent over successive interviews and waves of the study, a number of substudies and experimental modules have been implemented over the years on a wide variety of topics pertinent to function and disability. Earlier iterations of the study relied mostly on subject interviews, but starting in 2004 selected physical measures such as grip strength, walking speed, and weight have been collected, as well as selected biomarkers based on saliva and blood spot samples.

Technical guidance about the origin and validity of measures in HRS can be obtained through their excellent help desk, as well as user guides and a searchable bibliography

Practical issues with acquiring and using HRS data:

The core dataset is de-identified and is available free of charge with a minimum of regulatory barriers. As a result, it can be obtained with minimal delay. Because the data are de-identified, many IRBs will provide an exemption from review. Additional data, including linked Medicare data, can be acquired with special permissions and are subject to stringent data security procedures. This adds a major additional layer of complexity and time to obtain the data.

HRS data are organized in modules which are well-documented, but require close attention due to skip patterns in the data. HRS provides survey weights and design features that can be incorporated using standard procedures into analyses in SAS, Stata, and other common statistical packages to generate nationally representative adults. Not incorporating these features into analyses may lead to misleading results. HRS data are typically available about 2 years after the data were collected.

Overall, HRS data are voluminous given the number of questions, modules, and substudies available across more than 20 years. However, the data are relatively easy to obtain, easy to use, and well-documented, making them amenable for new users to use with a moderate amount of effort.

Areas of Particular Interest for Research on Function and Disability in Vulnerable Populations

The wide range of measures, long length of follow-up, and large sample size make HRS particularly valuable for research on the predictors and outcomes of disability. Also, the survey’s inclusion of numerous sociodemographic measures provides excellent opportunities to evaluate socioeconomically vulnerable populations. Within the HRS, it is possible to generate cohorts for a wide range of research question that assess risk factors over time or cross-sectionally.

In addition to the measures included in the primary study, HRS frequently conducts additional studies including one-time modules administered to a random subset of the sample, off year self-administered questionnaires, and various substudies. Data from all of these substudies are publicly available, and often present excellent research opportunities.

Measures from the main study of particular interest include:

  • Health status, including self-reported health conditions and treatments, symptoms (e.g., fatigue and depressive symptoms), and health behaviors.
  • ADL/IADL difficulty and dependence, extent of caregiver help with ADL and IADL, mobility, other functional limitations, falls and incontinence.
  • Cognition, including self- and proxy-reported cognition, behavioral problems, and cognitive tests.
  • Work related disability, including extensive measures of limitations working and available accommodations,
  • Demographics, and various modules on assets, wealth, work history, and socioeconomic status. This includes useful measures of wealth in older adults that are not dependent on current income (since many older adults are retired, and their current income does not reflect their wealth).
  • Attitudes to health, worth, and quality of life, and self-perceived social status. However, there are no direct measures of health literacy.
  • Physical measures and biomarkers, including direct physical measures such as grip strength, gait speed, and weight, and selected biomarkers derived from saliva and blood spot samples.

Supplemental studies of particular interest include:

  • Aging, Demographics, and Memory Study (ADAMS) – Focused on patients with dementia, cognitive impairment, not dementia, and a "control" group of subjects with normal cognition. Contains extensive cognitive tests and neuropsychiatric measures, as well as extended information on health status, functional limitations, caregiver burden, and more. In assessing these factors, the study provides information that can be used to study the natural history of cognitive decline as well as the impact of cognitive impairment on a variety of outcomes including health status, caregiver utilization and burden, and societal costs of care.

Experimental modules:

  • Each wave, HRS hosts a variety of experimental modules on a subset of participants. These can be valuable resources for specific research topics. For a full list, visit the HRS site here.

Additional Information - HRS Links

Home page


Survey Design