AI Boosts Medical Records, Not Patient Outcomes in Large-Scale Clinical Trial

A recent large-scale clinical trial conducted in Kenya has revealed that while artificial intelligence (AI) can substantially enhance the quality of medical records and treatment plans, its direct impact on improving patient treatment outcomes remains unconfirmed. The study, involving 9,691 patients across 16 primary care institutions, utilized a generative AI-based clinical support tool called "AI Consult," powered by GPT-4o, to assist healthcare workers. Published in the international journal *Nature Medicine*, the research was led by an international team from the University of Birmingham, UK. The trial divided 103 clinical healthcare workers into two groups: one using AI Consult alongside their existing electronic medical record system, and the other using only the traditional system. The final analysis included 4,693 patients receiving AI-assisted care and 4,654 receiving standard care. The AI system analyzed patient symptoms and test results, providing diagnoses and treatment recommendations aligned with Kenyan clinical guidelines, complete with risk indicators. While medical staff retained final decision-making authority, the study found a marked improvement in the thoroughness of medical records and the appropriateness of treatment plans in the AI-assisted group, as evaluated by independent Kenyan family medicine specialists. However, the primary outcome measure, "treatment failure" (defined as patients returning to the hospital within 14 days due to unresolved symptoms or referral to a higher-level institution), showed no statistically significant difference between the AI and control groups. The treatment failure rate was 2.2% in the AI group and 2.0% in the control group. Similarly, rates of hospitalization, death, and other serious adverse events were comparable across both groups. Interestingly, the study noted a slight reduction in the average cost of antibiotics per patient in the AI group, suggesting AI might have guided providers toward equally effective but more affordable options. Despite the lack of direct patient outcome improvement, the findings confirm AI's potential as a valuable supportive tool for medical professionals in real-world clinical settings, particularly in enhancing documentation and decision support. The researchers emphasized the need for continued large-scale verification to fully ascertain AI's role in directly improving patient health.