Agentic AI in Medicine
Agentic AI has come to medicine, from Eric Topol (emphasis mine):
The overall diagnostic accuracy for MIRA was 87.8% compared with BC physicians at 78.1%. That increase was especially notable for specific diagnoses like pancreatitis (95.2% vs 78.6%), and appendicitis (100% for MIRA, 88% for BC physicians). While MIRA ordered more blood tests (51% vs 28%), resource consumption was countered this by ordering substantially less scans. For therapy, MIRA surpassed the BC physicians 53.5% vs 38.3% for correctly ordering procedures such as laparoscopic appendectomy or cholecystectomy (Figure below).
Those results are really interesting: the AI seems to have picked the cheaper test (blood tests vs scans) to achieve equal diagnostic quality as doctors. This resource preservation is valuable on its own right, the question is who’s benefitting and where is the incentive? Did doctors order more scans because they are economically driven to do so or because scans providers them with more certainty? Does time-to-diagnosis play any role in the decision making process, as blood test results take longer than scan results? My hunch is here is more to the story.
Both agents excelled at adhering to complex and lengthy clinical guidelines, and did so at a fraction of the time real clinicians needed. From the same article (emphasis mine):
Other advantages for MIRA therapy included better IV fluid management and analgesic adherence to guidelines, and an overall 35% increased alignment of clinical guidelines compared with the BC physicians. Of 468 medications ordered by MIRA, 99.8% were correct for indication and safety (such as allergy, interactions, and kidney dosing). MIRA triaged more cases than physicians for hospital admission, which reflects lack of being economically driven.
And again:
The massive >600 clinical guidelines were fully tokenized (not just parts of them) to provide the grounding and citations for management. This ensemble only took about 80 seconds to produce.
As impressive as all of this is, of course, it remains to be seen how such models perform in a real setting with all the messiness that this implies. And yet, what does all of this mean for the future of healthcare? Are we going to see a decline in administrative work or an increase in regulation? How will the incentives for doctors change, as the system shifts towards flat rates, subscriptions and care? Is this shift going to disrupt existing players and open the doors to young and innovative startups? Only time will tell.