Researchers use AI to predict who will get Alzheimer’s disease - News Article

Researchers at the University of California, San Francisco used artificial intelligence (AI) and electronic medical records to predict who will get Alzheimer’s disease in the next seven years.

A new study shows AI can use data from electronic health records to predict if someone will start having Alzheimer’s disease (AD) symptoms within seven years. The researchers found AD predictions could be linked to other medical conditions. These include hypertension, high cholesterol, vitamin D deficiency, and the bone-weakening disease osteoporosis for women.

“Identification of AD onset risk can facilitate interventions before irreversible disease progression. We demonstrate that electronic health records from the University of California, San Francisco, followed by knowledge networks (for example, SPOKE) allow for prediction of AD onset,” said the study’s senior author, Marina Sirota, Ph.D., associate professor at the Bakar Computational Health Sciences Institute at UCSF.

In the study, published in February 2024 in the journal Nature Aging, the authors identified 2,996 patients with AD who had completed a dementia evaluation at the UCSF memory and aging center. These patients have an AD diagnosis from a neurologist or were prescribed their first cognitive drug. They selected 749 participants with AD diagnosis out of a group of nearly 3,000 and 250,545 UCSF patients without dementia and over one year of medical records as the control group.

The researchers used the data from the AD patients to map onto the UCSF Observational Medical Outcomes Partnership (OMOP) electronic health record database. This data allowed researchers to develop models to predict AD.  The researchers took the background medical history from patients already diagnosed with AD. Then they trained AI models to identify these factors in patients’ medical records. 

The AI program used risk factors of AD to identify patients who would develop AD seven years prior with up to 72% accuracy. Researchers demonstrated the ability to identify the medical factors and the patients likely to be diagnosed with AD. The medical factors researchers found can guide future research on diagnosing and treating AD.

 “This is a great example of how we can leverage patient data with machine learning to predict which patients are more likely to develop Alzheimer’s, and also to understand the reasons why that is so," said Marina Sirota, Ph.D., in a press release.

This study has limitations, because electronic health record data is complex and its quality can affect AI prediction abilities, providing only a snapshot of a patient's health and potentially leading to misdiagnoses of health conditions.

Additionally, misdiagnosing AD could affect the entire prediction percentage. Another limitation of this study is the study subjects were only from the University of California, San Francisco Medical Center.

This study suggests that some medical conditions can be warning signs and potentially predict AD. The medical factors that influenced the prediction of AD were hypertension, high cholesterol, allergic rhinitis and atrial fibrillation, osteoporosis, major depressive disorder, cognitive impairment, and vitamin D deficiency.

“Several factors, including hypertension, high cholesterol and vitamin D deficiency, were predictive in both men and women. Erectile dysfunction and an enlarged prostate were also predictive for men. But for women, osteoporosis was a particularly important predictor,” the researchers added.

According to the Alzheimer’s Association, seven out of 10 Americans would like to have an early warning if they have AD.  Additional research can address AD warning signs doctors can observe during appointments and help doctors decide which tests to have patients participate in.  Future research can focus on confirming the connection between the medical factors identified in this study and AD onset.