

The autonomous medical AI agent MIRA has shown higher diagnostic accuracy in simulated clinical cases, working with Electronic Health Record (EHR) data, compared to practicing physicians. In 87.8% of cases, MIRA outperformed specialists, opening new perspectives for automating routine but critically important medical processes.
If your clinic is losing time and resources on routine data processing and preliminary diagnoses, message our manager โ he will run a free analysis of your business and niche and show exactly how to get a real business result from an AI agent in your case, not a nice-looking picture. Message the manager
Large Language Models (LLMs) have already proven their ability to successfully pass standardized medical exams and answer complex clinical questions. However, translating this knowledge into the actual operational workflow of a hospital has remained a major challenge. This is because most traditional medical AI tools function as narrow, task-specific search engines or text generators rather than active partners in the treatment process.
True clinical decision-making is a complex multi-step process involving history taking, ordering tests, synthesizing conflicting results, and updating hypotheses. Virtually all of this work occurs within Electronic Health Record (EHR) systems, which require complex, standardized coding protocols.
Researchers developed MIRA, an autonomous AI agent capable of operating within sandboxed EHR environments. Unlike previous implementations, MIRA independently analyzes patient histories, orders relevant diagnostic tests, and uses this data to formulate diagnoses and treatment plans within a controlled simulation.
MIRA employs 11 specialized digital tools and has over 85,000 operational choices. The agent can request physical examinations, order targeted laboratory values, look up medical histories, and generate medication orders within the simulated EHR environment.
The study, published in Nature, showed that MIRA achieved 88.9% diagnostic accuracy across 574 MIMIC-IV cases. In a comparison with a group of physicians (311 cases), MIRA's accuracy was 87.8%, significantly higher than that of experienced human specialists under identical simulated conditions.
| Participant | Average Diagnostic Accuracy |
|---|---|
| MIRA (all 574 cases) | 88.9% |
| MIRA (compared to physicians, 311 cases) | 87.8% |
| Board-certified physicians | 78.1% |
| Mixed-seniority team (residents + board-certified physicians) | 71.1% |
MIRA particularly excelled in diagnosing conditions like appendicitis and pancreatitis, achieving 100% recall for laparoscopic appendectomies. Notably, the AI agent did not resort to excessive test ordering; its selection remained below historical baselines.
Safety evaluations were also encouraging: an independent medical review of 56 patient-level outputs and 468 prescriptions written by MIRA showed zero high-severity drug-drug interactions, zero renal dosing incompatibilities, and zero medication-allergy mismatches. The agent also achieved a perfect recall score (1.00) for critical hospitalization decisions.
Source: news-medical.net
The implementation of AI agents like MIRA can significantly enhance the efficiency and accuracy of diagnostics in medical institutions:
Despite the impressive results, the study authors emphasize that MIRA and similar AI agents are not replacements for expert human staff. They require continuous human oversight and patient-level safeguards. However, their potential to transform healthcare is immense.
To learn how AI agents can boost your clinic's efficiency, message the manager. We will analyze your processes for free and show you areas for growth.