Researchers at UC San Francisco trained a model on the electronic health records of 749 patients later diagnosed with Alzheimer's and 250,545 without it, searching ordinary visit histories for patterns that showed up years before diagnosis.
The model could flag elevated Alzheimer's risk up to seven years ahead of diagnosis, with accuracy improving the closer it got to diagnosis day, and it surfaced high cholesterol, and osteoporosis especially in women, as early warning signs.
The signals the model found, such as osteoporosis in women, are not causes of Alzheimer's on their own; they are patterns in routine records that happened to precede diagnosis in this dataset, and they need testing in other health systems before they mean much for an individual patient. The tool has not been used to make a clinical diagnosis.
Read the original study
Tang et al., Nature Aging, 2024 · doi.org/10.1038/s43587-024-00573-8
Years of checkups held a warning unread.
Source
- Tang, A. S., Rankin, K. P., Cerono, G., Miramontes, S., Mills, H., Roger, J., Zeng, B., Nelson, C., Soman, K., Woldemariam, S., Li, Y., Lee, A., Bove, R., Glymour, M., Aghaeepour, N., Oskotsky, T. T., Miller, Z., Allen, I. E., Sanders, S. J., . . . Sirota, M. (2024). Leveraging electronic health records and knowledge networks for Alzheimer's disease prediction and sex-specific biological insights. Nature Aging, 4(3), 379–395. https://doi.org/10.1038/s43587-024-00573-8