In the United States, Alzheimer’s disease does not strike all communities equally, and neither does diagnosis. African Americans are nearly twice as likely to have Alzheimer’s as non-Hispanic white adults but only 1.34 times as likely to receive a diagnosis, while Hispanic and Latino people are 1.5 times more likely to have the disease but just 1.18 times as likely to be diagnosed. These gaps leave many patients in underrepresented communities without answers, treatment, or support as the sixth leading cause of death in the country advances unchecked.
Researchers at UCLA have built an artificial intelligence tool that mines electronic health records to find patients with undiagnosed Alzheimer’s, explicitly targeting this unequal burden. Using records from more than 97,000 patients, the model was designed with “fair positive unlabeled learning,” a strategy intended to maintain accuracy while reducing diagnostic bias across racial and ethnic groups.
The tool’s performance is striking: its sensitivity ranged from 77% to 81% across non-Hispanic white, non-Hispanic African American, Hispanic/Latino and East Asian patients, compared with 39% to 53% for conventional supervised models. Researchers also built in population-specific fairness criteria to avoid simply replicating existing disparities in the data. Genetic validation showed that patients flagged as likely having undiagnosed Alzheimer’s carried higher polygenic risk scores and more APOE ε4 alleles, bolstering confidence that the model is surfacing true, previously missed cases.
Study leaders say such a tool could help clinicians prioritize high-risk patients for further evaluation and screening, a critical step as new treatments and lifestyle interventions can slow progression only when the disease is detected early. “By ensuring equitable predictions across populations, our model can help remedy significant underdiagnosis in underrepresented populations,” said Dr. Timothy Chang of UCLA Health, noting its potential to directly address disparities in Alzheimer’s diagnosis.
See: “AI tool can detect missed Alzheimer’s diagnoses while reducing disparities” (December 11, 2025)
