πŸ—žοΈ News - April 28, 2025

AI Outperforms Physicians in Diagnostic Accuracy and Treatment Recommendations

AI shows higher diagnostic accuracy and treatment recommendations than physicians, with optimal ratings at 77% vs. 67%. πŸ€–πŸ“Š

🌟 Stay Updated!
Join AI Health Hub to receive the latest insights in health and AI.

AI Outperforms Physicians in Diagnostic Accuracy and Treatment Recommendations

Study Overview

A recent study conducted by Prof. Dan Zeltzer from the Berglas School of Economics at Tel Aviv University has revealed that artificial intelligence (AI) provides more accurate diagnoses and treatment recommendations than physicians. This research was carried out at Cedars-Sinai Connect, a virtual urgent care clinic in Los Angeles, in collaboration with the Israeli startup K Health. The findings were published in the Annals of Internal Medicine and presented at the annual conference of the American College of Physicians (ACP).

Research Methodology

The study analyzed a sample of 461 online clinic visits over a month in the summer of 2024, focusing on adult patients with common symptoms such as:

  • Respiratory issues
  • Urinary problems
  • Eye conditions
  • Vaginal concerns
  • Dental issues

Patients were first assessed by the AI algorithm, which provided recommendations before they consulted with a physician via video. A panel of four experienced doctors evaluated all recommendations on a four-point scale: optimal, reasonable, inadequate, or potentially harmful.

Key Findings

The results indicated significant differences in the quality of recommendations:

  1. AI recommendations were rated as optimal in 77% of cases, compared to 67% for physicians.
  2. Only 2.8% of AI recommendations were deemed potentially harmful, versus 4.6% for physicians.
  3. In 68% of cases, both AI and physicians received the same score.
  4. In 21% of cases, AI outperformed physicians, while in 11% of cases, physicians were rated higher.
Advantages of AI

The evaluators noted several advantages of the AI system:

  • Adherence to medical guidelines, such as avoiding unnecessary antibiotic prescriptions for viral infections.
  • Comprehensive identification of relevant medical history, including recurrent infections.
  • Accurate detection of symptoms that may indicate serious conditions.

While AI demonstrated strengths in adherence to guidelines and data analysis, physicians excelled in assessing the patient’s overall condition and context.

Conclusion

Prof. Zeltzer emphasized that the study highlights AI’s potential as a decision-support tool in medical practice. He noted that while AI can enhance diagnostic accuracy, further research is needed to understand how best to integrate AI recommendations into clinical practice and the optimal collaboration between human expertise and AI.

Reference

Zeltzer D, Kugler Z, Hayat L, et al. Comparison of Initial Artificial Intelligence (AI) and Final Physician Recommendations in AI-Assisted Virtual Urgent Care Visits. Ann Intern Med. 2025 Apr;178(4):498-506. doi: 10.7326/ANNALS-24-03283

Share on facebook
Facebook
Share on twitter
Twitter
Share on linkedin
LinkedIn
Share on whatsapp
WhatsApp

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.