A Comparison of Measures for Identifying Possible Dementia in Veterans Affairs Nursing Home Residents

Identifying people with possible dementia in health care systems is important to study outcomes and target improvements in care. This study sought to compare the performance of diagnostic codes and Minimum Data Set (MDS)-based measures for identifying dementia and cognitive impairment in older veteran nursing home residents.

Retrospective, cross-sectional analysis.

We used real-world health care data from the Veterans Affairs (VA) Residential History File, VA Corporate Data Warehouse (CDW), Medicare claims, and the MDS to assemble a cohort of VA Community Living Center (CLC) admissions over 2015 to 2021 for veterans aged ≥ 65 with dual VA and Medicare enrollment (n = 54,234).

We defined 3 measures of possible dementia: (1) claims/CDW diagnoses using Chronic Conditions Warehouse (CCW) algorithms for Alzheimer's disease or non-Alzheimer's dementia; (2) MDS active diagnosis items for Alzheimer's disease and non-Alzheimer's dementia; and (3) MDS Cognitive Function Scale (CFS) assessment indicating at least mild cognitive impairment. We calculated proportions identified with each definition, and sensitivity, specificity, and positive predictive value of claims/CDW diagnoses and MDS indicators for dementia for identifying CFS impairment.

Among VA CLC residents, 61.4% met at least 1 criterion for possible dementia (38.6% claims/CDW, 23.3% MDS active diagnosis, 50.8% CFS). Diagnoses from claims/CDW had 56.5% sensitivity and 80.0% specificity for identifying veterans with CFS cognitive impairment. Active diagnoses from the MDS exhibited poorer sensitivity (38.1%), but higher specificity (92.0%) identifying veterans with cognitive impairment on the CFS.

Consistent with what has been reported in Medicare nursing home residents, we observed only partial overlap between indicators of possible dementia across diagnosis codes and other indicators vs cognitive assessments in MDS. Our findings support the utility of these measures for identifying individuals with possible dementia across different systems, but further work is needed to understand implications when using diagnosis codes or cognitive assessments.