Minimum Data Set (MDS) Data


Dataset Summary


The Minimum Data Set (MDS) is a standardized assessment tool that measures health status in nursing home residents. MDS assessments are completed every 3 months (or more often, depending on circumstances) on nearly all residents of nursing homes in the United States. These assessments are performed and recorded by nursing home staff, and include information on a number of aging-relevant domains including functional and cognitive status, psychosocial functioning, geriatric syndromes, and life care wishes. As such, MDS is an extremely valuable resource for studying function and disability on a large scale in vulnerable older adults. MDS data are collected and made available as one of the many data products of the Centers for Medicare and Medicaid Services (CMS).

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

MDS assessments are performed on all residents of nursing homes in the United States that are eligible to receive funding from Medicare or Medicaid. This includes long-stay residents as well as people in nursing homes for a brief period of rehabilitation or skilled nursing care following acute care hospitalization. The Department of Veterans Affairs (VA) also collects MDS data for residents of its nursing homes; these data are available through VA (see separate page on VA data).

Per CMS requirements, MDS assessments occur on a set schedule. For long-term residents, the maximum interval between assessments is 3 months. Interim assessments are also required for patients who have had a recent major change in health status. MDS assessments occur more frequently during periods of care covered by the Medicare Skilled Nursing Facility (SNF) benefit, which mainly covers short-stay rehabilitative and skilled nursing care in adults following hospitalization.

MDS assessments are completed by staff at the nursing facility, with different facilities using different types of staff for the purpose. Because these are clinical, rather than billing-focused assessments, they contain richer clinical data than is typically found in claims datasets. However, because MDS assessments are completed by front-line staff with variable levels of training who may be under time pressure, in most cases the data are less accurate than one might find in a dedicated research study with trained staff.

What kind of information is in the data?

Prior to October 2010, the MDS version 2.0 assessment form was used. Starting in October 2010, a new version (3.0) has been used. The biggest innovation in the newer version was incorporation of assessments obtained from direct resident interviews, for example asking residents cognitive screening questions and about symptoms of depression, rather than relying on staff assessments of these conditions. The lack of exact comparability between the 2 versions makes it difficult to evaluate longitudinal trends crossing MDS 2.0 to MDS 3.0 for a number of measures, although activities of daily living (ADL) measures are consistent in both versions. Also, several scales designed to summarize cognitive functioning and other domains that were created for and validated in MDS 2.0 may not be transportable to MDS 3.0, although new scales are being created for MDS 3.0 data.

Both versions of MDS contain a wealth of information. These include:

  • Data on functional and cognitive status and mobility – MDS assessments contain standardized questions about several elements of functional and cognitive status and mobility.
  • Behavioral symptoms – a number of items about behavioral symptoms common in dementia
  • Psychosocial functioning – resident participation in activities and resident preferences
  • Symptoms and geriatric syndromes – for example pain, continence, falls, nutritional status, and many others
  • Diagnoses and medications – selected diagnoses and recent use of selected medications (e.g. antipsychotics, sedative-hypnotics)
  • ….and much more

MDS data may be merged with other Medicare files (for example, Part D prescription drug data, and Part A acute hospitalization data) to further enrich data available for analysis.

Information about Medicare claims data can be found at the Research Data Assistance Center (ResDAC), which contracts with CMS to support research using CMS data. They have extensive online documentation and a very helpful help desk.

For investigators interested in working with MDS data collected within the Department of Veterans Affairs (VA) Health Care System, the VA Geriatrics and Extended Care Data Analysis Center may be able to offer assistance.

Practical issues with acquiring and using MDS claims data:

Like other Medicare data, MDS data are highly complex and require a substantial learning curve to use. As one of many examples, there are many ways in which missing data are recorded in the dataset, each with a different meaning. As a result, these data are best used as a long-term resource for an investigator or research group that can invest substantial start-up time and analyst effort and are less well-suited for a one-off project by someone who has little prior experience with the data.

As noted above, there are substantial differences between MDS 2.0 and MDS 3.0, with the transition date from the former to the latter being October 2010. Given these differences, it may be difficult to evaluate longitudinal trends that cross this transition date.

When searching for information on MDS files on the RESDAC and CMS websites, note that they are often listed under the term “Resident Assessment.” The process for obtaining Medicare data depends on the level of individual detail available in the files requested. Most research questions will require patient-level Research Identifiable Files (RIF). Requests for these data require a data use agreement and are reviewed by a CMS privacy board. This process typically takes months. Each year of MDS data costs $2000 for up to 1 million subjects, with slightly higher fees for higher numbers of subjects. There are additional costs for other Medicare files, and for requesting that CMS construct specific cohorts. 

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

MDS is a gold mine of information for research on function and disability, as it contains numerous measures of physical, cognitive, and psychosocial functioning. MDS data can also be linked with Medicare claims data, for example, data on prescription drug utilization (Part D), outpatient care (Part B), and inpatient care (Part A), to construct comprehensive histories of patient health status and health care utilization across sites of care.

Measures from MDS of particular interest include:

  • Functional status and mobility
  • Cognitive status
  • Mood symptoms and behavioral symptoms
  • Patient preferences (MDS 3.0)
  • Hearing and visual acuity
  • Geriatric syndromes (e.g. falls, incontinence, pressure ulcers, weight loss)
  • Symptoms (e.g. pain, shortness of breath)
  • Selected chronic conditions and medication use

Additional Information