Sign up FAST! Login

Babylon Health App AI scored 92% accuracy for “check a symptom” compared to the doctor’s 82% and the nurse’s 77%.

Stashed in: Awesome, Medicine, Apps, Healthcare!, Training

To save this post, select a stash from drop-down menu or type in a new one:

The title is an excerpt from the middle of the article. 

In June, an in-house “live challenge” pitted Babylon’s AI against a senior A&E nurse and an Oxford-educated junior doctor. UCL professor Irwin Nazareth compared the results. According to Nazareth, “check a symptom” was consistently faster and more accurate in triaging patients than its human rivals: it scored 92% accuracy compared to the doctor’s 82% and the nurse’s 77%.

The article also makes the following point:

“The machine doesn’t get worked up, it doesn’t get angry, it doesn’t get pushed because there are 20 people outside in the queue,” says Parsa. “A machine learns every time a diagnosis happens. A doctor might do 7,000 consultations. The machine will do thousands of times more. Therefore the speed at which it learns and everything it sees increases. So we believe in time we will do this more accurately than a human.

The future of healthcare may be AI indeed?

Reddit comment:

The headline is quite overstated and deceptive. The article admits that this was not a scientific study. The health app discussed in the article has a symptom checker that triages patients and tells the patient if they are fine, if they need to see their primary care physician, or if they need to go to the ER. This app does not actually diagnose any diseases. It is unclear how they measured whether or not the patients were accurately triaged. This is not something that can easily or quickly be determined, especially by a non scientific and non peer reviewed study such as this. There are many triage apps which already exist which are based on telemedicine triage protocols which are evidence-based and which have low rates of over-referral and under-referral to the ER. One such app is KidsDoc by the American Academy of Pediatrics. In my opinion this article is not a very impressive example of artificial intelligence and is not very futuristic. Fair disclosure I am a physician.

You May Also Like: