Can AI Replace a Clinician?

Most artificial intelligence (AI) evaluations in healthcare focus on individual tasks. AI can spot an ailment. It can flag a high-risk situation. AI can quickly generate summary charts. But, can AI replace your clinician?

In new research conducted by the Mack Institute, researchers asked and challenged AI: Can a large language model manage an entire clinical decision-making workflow, rather than just individual tasks within it? To explore the question, researchers placed a multimodal large language model inside a realistic medical training simulation. These simulations are used to evaluate medical students and practicing clinicians. The virtual patient's condition evolves in real time. Vital signs change. Test results arrive with delays. Inaction has consequences to the patient . AI must decide what to do at every step. It can ask questions, turn on monitors, order lab tests or imaging, administer treatments, and escalate care. Click the video below the see a sample patient session:

https://www.youtube.com/watch?v=I4KiopTSV-A

How Did AI Do?

The researchers concluded that -

• AI consistently stabilized patients and completed cases at rates comparable to - and in some cases higher than - medical students.

• AI completed cases substantially faster.

• AI diagnostic accuracy was similar to or closely resembled expert clinical practitioners.

• AI tended to order tests that provided large amounts of new information and quickly narrowing the range of plausible diagnoses.

• Errors were more likely in AI's decision-making when it was uncertain in its diagnoses or the diagnoses were ambiguous.

• AI engaged in less patient communication than medical students or clinicians.

Researchers concluded that the better current use of AI may be to use AI as a "second set of eye" in the clinical setting. Human clinicians are better at focusing on judgment, communication, and oversight.

To read the white paper, click here.