Algorithm for Identifying Nursing Home Days Using Medicare Claims and Minimum Data Set Assessment Data

Abstract

Background: 

No consensus exists about methods of measuring nursing home (NH) length-of-stay for Medicare beneficiaries to identify long-stay and short-stay NH residents.

Objectives: 

To develop an algorithm measuring NH days of stay to differentiate between residents with long and short stay (≥101 and <101 consecutive days, respectively) and to compare the algorithm with Minimum Data Set (MDS) alone and Medicare claims data.

Research Design: 

We linked 2006–2009 MDS assessments to Medicare Part A skilled nursing facility (SNF) data. This algorithm determined the daily NH stay evidence by MDS and SNF dates. NH length-of-stay and characteristics were reported in the total, long-stay, and short-stay residents. Long-stay residents identified by the algorithm were compared with the NH evidence from MDS-alone and Medicare parts A and B data.

Results: 

Of 276,844 residents identified by our algorithm, 40.8% were long stay. Long-stay versus short-stay residents tended to be older, male, white, unmarried, low-income subsidy recipients, have multiple comorbidities, and have higher mortality but have fewer hospitalizations and SNF services. Higher proportions of long-stay and short-stay residents identified by the MDS/SNF algorithm were classified in the same group using MDS-only (98.9% and 100%, respectively), compared with the parts A and B data (95.0% and 67.1%, respectively). NH length-of-stay was similar between MDS/SNF and MDS-only long-stay residents (mean±SD: 717±422 vs. 720±441 d), but the lengths were longer compared with the parts A and B data (approximately 474±393 d).

Conclusions: 

Our MDS/SNF algorithm allows the differentiation of long-stay and short-stay residents, resulting in an NH group more precise than using Medicare claims data only.