There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
| Principal Investigator | |
| Principal Investigator's Name: | Sang Kyu Cho |
| Institution: | University of Houston - Houston, TX |
| Department: | Pharmaceutical Health Outcomes and Policy |
| Country: | |
| Proposed Analysis: | For this proposed research, I will be working with Dr. Soeren Mattke at University of Southern California. The potential approval of a disease-modifying Alzheimer’s treatment will create an unprecedented challenge for healthcare system because of the combination of a large prevalent patient pool and a complex diagnostic process. In previous work, we have predicted substantial wait times, and ensuing disease progression, even in high-income countries. As building up the required infrastructure for specialist evaluation and confirmatory biomarker testing is neither practically not economically feasible, improved triage at the primary care level would be a viable solution. The triage approach would have to determine (1) whether a patient with early-stage cognitive decline, i.e., MCI or mild dementia, has the Alzheimer’s pathology and (2) how likely the patient is to progress to more advanced stages in a short timeframe, such as three years. This approach would allow to prioritize patients with likely treatment benefit and more urgent need for immediate attention. Conversely, patients with other pathologies and low risk of progression could be held back for a timeframe, in which most healthcare systems will have been able to handle the backlog of prevalent cases. Our objective is to create a risk model based on ADNI data for such predictions. For the model to be practically useful, it needs to be based on variables that are likely available to primary care physicians, such as brief cognitive instruments (e.g., MMSE), blood tests (e.g., ApoE, p-tau, metabolic indicators) and information on family risk, demographics and comorbidities. We will also test a model that takes CSF-based parameters into account, as primary care physicians could conceivably conduct or order a lumbar puncture, and capacity for this procedure is not constrained in most settings. We will split the data into a learning and a validation sample and use deep learning techniques to understand which combination of variables and cutoff points allow for a sufficiently accurate prediction (defined as an area under the ROC curve of at least 0.75) of Alzheimer’s pathology and progression risk. We will train the model on the learning sample and test it in the validation sample to avoid overfitting. We might seek to apply the model to different datasets at a later stage. The output of the product will be a set of risk equations to prioritize patients for further evaluation and treatment, which would help to manage waitlists for specialist evaluation and, if access is limited by budget constraints, treatment delivery. |
| Additional Investigators |
