Two major Chinese clinical trials of AI in diagnosis and care this year are offering some of the most rigorous evidence yet on how AI performs in real hospital settings — a randomized trial of an AI-assisted rare disease diagnosis tool reviewed by Tsinghua University’s ethics committee, and an 8,240-patient orthopedic study at Shanghai’s Ruijin Hospital.
The Rare Disease Trial
The rare disease diagnosis trial, registered in the Chinese Clinical Trial Registry and approved by Tsinghua University’s Ethics Committee, tested a specialized reasoning large language model designed to assist physicians with diagnosing rare conditions — a category where human physicians, who may encounter a given rare disease only once or twice in a career, often struggle with pattern recognition that AI systems trained on much larger datasets can potentially surface faster. Physician recruitment for the trial ran in early 2026, with the randomized design meant to isolate the AI tool’s genuine diagnostic contribution from the natural variation in physician skill and experience.
The Ruijin Hospital Orthopedic Study
Separately, Ruijin Hospital ran a large-scale randomized study from September 2025 through April 2026, enrolling 8,240 consecutive musculoskeletal inpatients and randomizing them 1:1 between standard care and an AI-assisted treatment pathway — one of the largest single-site randomized AI healthcare trials conducted anywhere to date, by sheer patient volume. The scale of the study, run entirely within one hospital system, reflects China’s increasingly aggressive approach to generating large real-world evidence bases for clinical AI, often faster than equivalent multi-site trials could be organized in the U.S. or Europe.
Why Scale Doesn’t Settle the Debate
Despite the impressive patient numbers, AI health researchers caution that large single-institution trials can still face generalizability limits — results from Ruijin Hospital’s specific patient population, clinical protocols, and AI tool configuration may not translate directly to hospitals with different patient demographics or existing care pathways. A separate, more modest but methodologically telling study found that giving 50 physicians access to a large language model did not significantly improve their diagnostic-reasoning scores compared with conventional resources, underscoring that AI access alone doesn’t guarantee better outcomes without careful integration into clinical workflow.
The Broader Chinese AI Health Research Push
These trials are part of a much larger wave: one analysis counted more than 8,500 AI-related clinical trials registered globally as of this year, with over 1,800 actively enrolling patients, and a substantial share of that volume originating from Chinese research institutions moving quickly to generate evidence for domestic AI health tools — partly driven by the same push toward technological self-sufficiency visible in China’s semiconductor and AI chip strategy.
What International Researchers Are Watching
Western health AI researchers have expressed interest in seeing full peer-reviewed results from both trials, particularly methodological details on how “AI-assisted” care was defined and what specific decision points the AI tools influenced versus purely human-driven ones. Transparency around these details will determine how much international weight the results carry beyond China’s own regulatory and clinical context.
What’s Next
Full results from both trials had not been published in peer-reviewed journals as of this reporting, with the rare disease trial’s findings still pending and Ruijin’s orthopedic study results still being analyzed following its April 2026 completion. Their eventual publication will be closely watched as some of the largest and most rigorously designed evidence yet on AI’s real clinical impact rather than its benchmark performance.
Photo: DΛVΞ GΛRCIΛ / PEXELS via Pexels
