After development of AI tools The company recommends antidepressants based on medical history, but researchers at George Mason University are currently considering whether adding patient demographics, such as race and ethnicity, could improve the tool’s effectiveness. According to their new research, the answer is yes.
Led by a multidisciplinary team at George Mason University Farokh AlemiThe machine learning and AI expert looked at how effective recommendations from AI-guided tools/models that recognize patient race and associations specific to African-American patients are compared to tools/models that don’t. The researchers found that recommendations based on “race-blind” AI models that don’t know a patient’s race tend to recommend drugs that are less effective for African-American patients.
“Antidepressant recommendations from race-specific models exceeded recommendations from the general model across all antidepressants studied. Our findings highlight why, as in clinical practice, clinical AI should not rely solely on general population patterns when prescribing to African Americans with depression,” said Vladimir Cárdenas, Master of Health Informatics.
why is this important
“If an AI system is not trained on the right information, including patient demographic information such as race, it may provide inaccurate or inaccurate information, resulting in patients receiving less effective drugs,” Alemi said.
Alemi and his colleagues observed this when advising patients about treatment options for depression. “AI systems may be biased against African-Americans, recommending antidepressants that are effective for general, predominantly white patients but not effective for African-Americans,” Alemi said.
detail
Researchers investigated bias in an AI system aimed at guiding treatment for major depressive disorder (MDD) and whether race-blind models missed important signals for African American patients. The AI system used medical history, such as whether a patient had taken a full dose of an antidepressant, to recommend medication. The researchers coded whether patients discontinued their antidepressants as a measure of treatment failure or success with the AI.
This study emphasizes that race is not a biological determinant of depression or treatment response, and highlights social and environmental factors that influence depression. Factors that are more common among African American patients include poverty, lower education, exposure to violence, discrimination, cultural bias and negative attitudes toward mental health, and lower access to mental health treatment resources.
“These data highlight the need to tailor antidepressants to a patient’s individual medical history. This is done by clinicians, but if done properly, AI systems can help clinicians make adjustments as well,” Cárdenas said.
“We hope our approach will help inform AI in healthcare design and governance, allowing us to truly pursue AI that improves health for all,” Cárdenas said.
reference: Alemi F, Ribalgar K, Ramezani N, Cárdenas V, Kurian M, King RC. Biases in AI-based management of patients with major depressive disorder. J Health Equity. 2026;3(1):2606724. doi: 10.1080/29944694.2025.2606724
This article has been reprinted from the following material: Note: Materials may be edited for length and content. Please contact the citation source for details. You can access our press release publishing policy here.
