OBJECTIVES: To identify potential predictors of Alzheimer’s Disease (AD) using longitudinal data captured in Medicare claims data. METHODS: AD and non-AD patients were identified from a random 5% sample of Medicare beneficiaries reporting Fee-for-Service claims across the inpatient, outpatient and office settings from 2010-2013.  AD patients were defined as those with an ICD-9 code of 331.0 first appearing in 2013, while non-AD patients were those who reported no AD or AD-related dementias from 2010-2013.  AD and non-AD patients were matched on sex and age group.  Using claims from 2010-2012, the proportion of patients reporting each diagnosis code was calculated and the frequencies were compared between AD and non-AD patients.  The top 10 codes that appeared more frequently among AD patients were isolated and risk differences (RDs) were computed. RESULTS: CONCLUSIONS: This study isolated those conditions that appear more frequently in AD patient claim histories, suggesting that these conditions may have the most value in predicting whether a patient progresses to AD.  With the proper leading indicators in hand, we may be able to develop detection algorithms based on readily-available claims data that prove useful in identifying the disease in its earlier stages.