Creating Patient Segments by examining indicator's associated with Prodromal Alzheimer's Disease Lead Investigator: Dominic DeBiaso Institution : McKinsey Company E-Mail : dominic_debiaso@mckinsey.com Proposal ID : 1031 Proposal Description: The goal of the proposed research is to assist clinicians in aiding patients during the Prodromal phase of Alzheimer?s Disease. The proposed methodology is to use a rich data set, such as NACC UDS, to understand drivers that are associated with patients that are eventually diagnosed with Alzheimer?s Disease. Biomarkers and genetic information are useful in determining the state of Alzheimer?s that the patient is currently at. Furthermore, the NACC-UDS data set contains a significant number of cognitive tests and survey data that can be used to create proxy variables for the next phase of modeling. Trends that would be discovered from analyzing the rich NACC-UDS data set would be used to create patient segments and profiles. These patient segments would then be appended to medical claims and EMR data to enrich these larger patient populations. At this point, we would begin another phase of analysis and deep predictive modeling to analyze the larger EMR/medical claims patient population to construct a model to classify patients that may be susceptible to Alzheimer?s Disease. This model could then be applied to patients that may be at risk for developing Alzheimer?s Disease so that clinicians could take early intervention. The data processing, cleansing, analysis, and predictive modeling would be conducted using R/Python. We would perform regression modeling such as logistic regression as well as machine learning methods such as tree-based models including random forests and gradient boosted machines to engage the latest techniques to target patients at risk of developing Alzheimer?s Disease, specifically in the earlier phases. Diagnosing Alzheimer?s Disease often requires biomarkers such as CSF testing and a panel of cognitive tests. At this stage, the disease is already underway and the patient is already past the early stage of the disease. If a predictive model could be used to recognize early indicators of Alzheimer?s then significant progress could be made to