AI in Clinics: Study Shows Mixed Results for Patient Care

Researchers tested whether large language model (LLM) decision support could help health workers improve patient care in two outpatient clinics in Nigeria. Health workers first created treatment plans for patients and then had the option to revise them after receiving feedback from the LLM. These plans were later evaluated by on-site physicians who independently examined the same patients, as well as through laboratory tests and retrospective reviews by academic physicians.

After receiving AI feedback, health workers changed their prescriptions for more than half of the patients and reported high satisfaction with the recommendations. Academic reviewers who later examined the medical notes also tended to rate the AI-assisted plans more positively.

AI in Clinics: Study Shows Mixed Results for Patient Care

However, real-world results were less impressive. On-site physicians found little to no improvement in diagnosis accuracy or treatment decisions. Laboratory tests also showed mixed outcomes. The AI support reduced unnecessary malaria tests but increased testing for urinary tract infections and anemia, without significantly improving detection rates. Overall, the findings suggest that while AI tools may influence decision-making, their impact on clinical outcomes remains uncertain.

National Bureau of Economic Research. (2026). Does LLM assistance improve healthcare delivery? An evaluation using on‑site physicians and laboratory tests (Working Paper No. 34660). https://doi.org/10.3386/w34660