Comparison of methods to identify long term care nursing home residence with administrative data

Table 3 Validation of different methods of identifying a long term care (LTC) nursing home stay using Medicare and MDS data, vs. data in Medicaid as the gold standard, restricted to patients who were hospitalized and discharged to a skilled nursing facility

Method	LTC nursing home length of stay in Medicaid data
	Cut off Length of stay >30^a	Cut off Length of stay >100^a
	N = 1,666	N = 317
1) Part A and B with E&M Charge for NH Services
Sensitivity	88.79%	92.42%
PPV^b		64.59%^b
2) Part A with MDS episode of care
Sensitivity	85.05%	78.86%
PPV	87.38%	88.07%
3) CMS Method: MDS alone
Sensitivity	69.39%	80.12%
PPV^b		72.70%^b

MDS minimum data set, E&M evaluation and management, NH nursing home, PPV positive predictive value, CMS Centers for Medicare and Medicaid Services
^aFor the cut off of > 30 or >100 day lengths of stay, the Medicaid data demonstrated > 30 and > 100 day length of stay in the validation
^bOnly one PPV could be detected for the different lengths of stay in methods one and three. Method one relies on a provider charges to identify nursing home stays, and length of stay cannot be reliably estimated. Method three uses any length of stay in a nursing facility (whether skilled nursing facility or LTC) > 100 days as the criterion

ISSN: 1472-6963