Validation of the Minimum Data Set in identifying hospitalization events and payment source

Abstract

Objectives: To evaluate the accuracy of the Minimum Data Set (MDS) in identifying hospitalization events and payment source among nursing home residents.

Research design: The 2003 MDS, Medicare Provider Analysis and Review File (MedPAR), Medicare denominator file, Medicaid Analytical Extract (MAX) long-term care file, and MAX personal summary file for 4 states (California, Ohio, New York, and Texas) were obtained and merged.

Setting: All Medicare/Medicaid-certified nursing ho-mes in these 4 states during 2003.

Participants: All nursing home residents who were eligible for Medicare. Medicare or Medicaid managed care enrollees were excluded.

Measurements: Using the identification by linking the MDS and claims data as the "gold standard," we calculated false negative and false positive error rates of the MDS in identifying hospitalization events and payment source.

Results: As for the accuracy of the MDS in identifying hospitalization events, the false negative error rates ranged from 6.8% to 19.5% and the false positive error rates were between 12.0% and 15.7%, depending on the state. With regard to the identification of Medicare payment source, the MDS had a low false negative rate (varying from 0.4% to 1.1%), and a relatively high false positive rate (ranging from 6.1% to 14.9%). The MDS alone did not seem to be a sufficient source for identification of Medicaid payment source (false negative rate ranging from 11.0% to 55.3%).

Conclusions: The accuracy of the MDS in identifying hospitalizations and payment sources varies across the study states, and should be evaluated carefully with regard to the intended uses of the data.