We compare the proportion of patients presenting with comorbid conditions among fee-for-service Medicare beneficiaries with heart failure obtained by using (a) home health care claims data, (b) data from the Medicare Provider Analysis and Review (MedPAR), and (c) the Outcome and Assessment Information Set (OASIS), using a unique database developed by linking the records of Medicare administrative data and the OASIS. Data from the MedPAR yielded a higher proportion of patients with comorbid conditions compared to the OASIS or home health care claims data (77%, 51%, and 51%, respectively), leading us to conclude that non-home health care claims data must be used for risk adjustment in studies involving home health care patients.