During FY2001 the VA began collecting health and functional status assessments on nursing home care unit patients using the Minimum Data Set (MDS) Version 2.0 as developed by the Centers for Medicare & Medicaid Services (CMS). Prior to that time, VA used the Patient Assessment Instrument (PAI) to collect a smaller set of data, stored in the Patient Assessment File (PAF) at the Austin Automation Center. To date, VA MDS data have been made available to only a few research groups (including CHQOER), pending validation of the data and finalization of a data use agreement. Since PAF data are no longer collected, there is an urgent need to validate the MDS data and make the results widely available to VA researchers.
This project will validate a new source of health and functional status data for VA nursing home care unit patients, MDS Version 2.0. At the same time, VA and CMS are jointly sponsoring the development of a new version of the MDS (Version 3.0), scheduled to be ready for adoption by FY2005. Timely validation of MDS 2.0 in the VA setting will support refinement of the draft MDS 3.0 instrument as well as ensure that discontinuities in the data are well understood and documented. It will also permit MDS-based research to proceed now, rather than waiting several years for MDS 3.0 data to be collected and validated.
The project is an observational study in which we will tabulate and compare secondary data from multiple sources. We analyze data from all VA MDS assessments from October 2000 to May 2004, from PAF files from January 1998 to December 2000 and from NPCD files from October 2000 to May 2004. We obtain five chronic care quality indicators constructed from MDS 2.0 assessments from all Medicare or Medicaid certified nursing facilities and made available through CMS's Nursing Home Compare website. We evaluate the consistency of VA MDS data relative to itself, checking different data elements on the same assessment and checking longitudinally across assessments for the same patients. We evaluate the comparability of VA MDS data relative to PAF data from prior years and other VA administrative data. We compare basic MDS-based quality indicators constructed from VA MDS data to the same measures constructed from MDS data collected from community nursing facilities.
Analysis of the internal consistency of VA MDS data finds the quality of data to be very good. The overall percentage of records flagged as potentially inconsistent is very low for all comparisons. Our comparison of VA MDS data to PAF data and to diagnoses recorded in the NPCD leads us to draw two strong conclusions: (1) the coding of quality-related elements on the MDS differs substantially from the coding on the PAF and (2) the diagnostic data found on the MDS is different in its volume and specificity that is found in the NPCD. Lastly, our comparison of VA MDS-based quality indicators to community nursing facility quality indicators finds that the quality measures derived from VA MDS data are, in most cases, similar to those derived from community MDS data.
VA MDS data appears to be consistent and reliable enough to support a variety of uses by administrators and researchers. Administratively, MDS data could be used for casemix adjustment, care planning, and quality monitoring. Among other uses, researchers may find MDS data invaluable in developing new risk-adjusted outcome-based measures of quality of care.
External Links for this Project
None at this time.