Medicare’s home health benefit has remained a stable and essential component of community-based care in the U.S. over the past decade. More than 12,000 home health agencies serve millions of beneficiaries annually. These agencies provide skilled nursing, therapy, and social care services that substitute for institutional care and align with patient preferences. Spending has remained consistent at about 4-5% of total fee-for-service (FFS) Medicare expenditures. A key data source to evaluate home health care quality and outcomes is the Outcome and Assessment Information Set (OASIS), which captures detailed functional, clinical, and social information not available in the claims data. OASIS data must be linked with Medicare enrollment and claims files using encrypted beneficiary identifiers
A study published in JAMA Network aimed to evaluate the completeness and performance of linkage between OASIS assessments, Medicare enrollment data, and FFS home health claims from 2017 to 2023. The objectives were to quantify annual match rates, examine trends over time, and assess variation by payer type and geography to better understand the implications for research and policy analyses using these datasets.
This cohort study used national Medicare administrative and assessment data from January 1, 2017, to December 31, 2023, with analyses conducted from June 2025 to January 2026. The study included two primary populations: all OASIS assessments to evaluate linkage with the Master Beneficiary Summary File (MBSF) and all beneficiaries with at least one FFS home health claim to evaluate matching between claims and OASIS records. Key variables included assessment dates, claim episode periods, and payer source (FFS Medicare, Medicare Advantage, and Medicaid).
Statistical analyses included calculating annual linkage rates between OASIS and the MBSF, which examines payer distribution in unmatched records, and comparing trends in the OASIS-linked and claims-based FFS beneficiaries. Two methods were used to assess claims to OASIS matching: beneficiary year concordance, defined as having at least one OASIS assessment in the same year as a claim, and overlap of episode windows between claims and OASIS records. State-level variation was assessed using overlap-based matching rates.
Results showed that the overall volume of OASIS assessments (approximately 18 million annually) and unique beneficiaries (approximately 6 million) remained stable. However, the proportion of assessments successfully linked to a Medicare beneficiary declined significantly from 89.8% in 2017 to 76.4% in 2023. Among unmatched assessments, the proportion listing Medicaid as the payer decreased substantially, while those listing Medicare increased by 2023, 77% of unmatched assessments had a Medicare designation.
The number of individuals with OASIS-linked assessments declined by 38.9% as compared to a smaller 23.5% reduction in beneficiaries with claims in FFS beneficiaries. It indicates increasing incompleteness in OASIS linkage instead of solely reduced utilization. Matching rates between claims and OASIS also deteriorated markedly: beneficiary-year concordance reduced from 95.6% to 76.9%, and episode-overlap matching declined from 96.8% to 73.9%. Geographic variation enlarged, with all states exceeding 90% match rates in 2017, but some dropping as low as 62% by 2023. These findings indicate a substantial and worsening breakdown in data linkage across datasets.
The study has several limitations. The use of deidentified data prevented direct validation of linkage accuracy at the individual level. The exact cause of declining linkage could not be determined, whether due to issues in identifier generation, mapping, or upstream data processing. Analyses were largely limited to FFS beneficiaries due to incomplete Medicare Advantage data. Additionally, potential changes in data submission or processing practices could not be fully ruled out.
Overall, this study showed a significant and ongoing reduction in the ability to link OASIS assessments with Medicare enrollment and claims data from 2017 to 2023. This deterioration has serious risks to the validity of research and policy analyses that rely on these linked datasets, specifically during a period marked by major healthcare changes like the COVID-19 pandemic and the implementation of the Patient-Driven Groupings Model. These findings underscore the need for the Centers for Medicare & Medicaid Services to address issues in identifier mapping and restore high linkage quality. Researchers should interpret findings based on OASIS-linked data with caution, and greater transparency about the data limitations is important for informed decision-making.
Reference: Rahman M, Wang X, Smith JM, et al. Match Rates Between Home Health Assessment and Medicare Claims Data. JAMA Netw Open. 2026;9(4):e264788. doi:10.1001/jamanetworkopen.2026.4788
