Researchers develop AI tool to identify undiagnosed Alzheimer's cases while reducing disparities
Researchers at UCLA have developed an artificial intelligence tool that can use electronic health records to identify patients with undiagnosed Alzheimer’s disease, addressing a critical gap in Alzheimer’s care: significant under-diagnosis, particularly among underrepresented communities.
Disparities in Alzheimer’s and dementia diagnosis among certain populations have been a longstanding issue. African Americans are nearly twice as likely to have the neurodegenerative disease compared to non-Hispanic whites but only 1.34 times as likely to receive a diagnosis. Similarly, Hispanic and Latino people are 1.5 times more likely to have the disease but only 1.18 times as likely to be diagnosed.
“Alzheimer's disease is the sixth leading cause of death in the United States and affects 1 in 9 Americans aged 65 and older," said the study's corresponding author from UCLA Health Department of Neurology. “The gap between who actually has the disease and who gets diagnosed is substantial, and it's more significant in underrepresented communities.”
Previous research has leveraged machine learning models to attempt to predict Alzheimer’s disease using electronic health records, but were designed using traditional frameworks that may not account for certain diagnostic biases.
The new model developed by the UCLA team took a different approach, known as semi-supervised positive unlabeled learning, that was specifically designed to promote fairness while maintaining high accuracy.
Researchers tested the model using electronic health records from more than 97,000 patients at UCLA Health, including those with confirmed diagnoses of Alzheimer’s disease and unconfirmed cases.
The model achieved sensitivity rates of 77 to 81% across non-Hispanic white, non-Hispanic African American, Hispanic/Latino, and East Asian groups compared to the 39 to 53% sensitivity of conventional supervised models.
UCLA researchers built on previous AI models used to predict various diseases including Alzheimer’s disease but had gaps as far as reducing biases and disparities. The UCLA tools analyzed patterns in the health records such as diagnosis, age, and other clinical factors. Key predictive features for Alzheimer’s were also identified, including both neurological indicators such as memory loss and unexpected patterns such as decubitus ulcers and heart palpitations that may signal undiagnosed cases.
Unlike traditional approaches that require confirmed diagnoses for all training data, the UCLA model learns from both confirmed cases and patients with unknown Alzheimer's status. The researchers incorporated fairness measures throughout the model's development, using population-specific criteria to reduce diagnostic disparities.
The tool was validated using multiple approaches, including genetic data. Patients predicted to have undiagnosed Alzheimer's showed significantly higher polygenic risk scores and genetic markers for the disease, known as APOE ε4 allele counts, compared to those predicted not to have it. Chang said the tool could help clinicians identify high-risk patients who may benefit from further evaluation or screening. Early identification is crucial as new Alzheimer's treatments become available and lifestyle interventions can slow disease progression.
The research team plans to validate the model prospectively in partnering health systems to assess its generalizability and clinical utility before potential implementation in routine care.
“By ensuring equitable predictions across populations, our model can help remedy significant underdiagnosis in underrepresented populations,” Chang said. “It has the potential to address disparities in Alzheimer's diagnosis.”