According to a Harvard University study published in the journal Science, artificial intelligence (AI) demonstrated higher diagnostic accuracy than human physicians during initial patient assessments in emergency departments—situations where both time and information for decision-making are scarce.
An experiment conducted at a Boston hospital involving data from 76 patients revealed that, under conditions of limited information, AI provided an accurate diagnosis in 67% of cases, while the accuracy of doctors ranged between 50-55%. Furthermore, in formulating long-term therapeutic plans (such as antibiotic therapy), the machine algorithm significantly outperformed a group of 46 physicians, achieving 89% accuracy compared to the doctors’ 34%. In one real-world case from the study, AI correctly identified that a patient’s pulmonary issues were caused by underlying lupus, a detail that the physicians had overlooked.
Despite these impressive results, researchers emphasize that technology cannot replace doctors. During the experiment, the algorithm relied solely on documentary data and was unable to assess a patient’s visual condition or stress levels. Consequently, AI is viewed not as a replacement for physicians, but as a generator of a professional “second opinion.”
Specialists also highlight the risks associated with implementing this new technology. A primary challenge remains the issue of legal liability in the event of a medical error made by the software. There is also the risk that doctors might reflexively agree with the machine’s decisions instead of thinking independently. Moreover, when making life-critical decisions, the trust-based relationship with a human doctor remains the primary and decisive factor for patients.
