Thread by @EricTopol

Some features that differentiate Shallow Medicine, as currently practiced, and the potential of #DeepMedicine, powered by #AI
Scratches the surface of the book, some supportive data in this thread

There are at least 12 million diagnostic errors/ year and the leading cause is failure to consider the diagnosis (which relates to System 1 thinking)

Humans have 185 different types of biases which leads to misdiagnosis. Example: failing to diagnose heart attacks because doctors think they don't occur in people age < 40

Presently, about half of physicians suffer from burnout.
This leads to a doubling of medical errors, which perpetuates a vicious cycle of further disillusionment.
jamanetwork.com/journals/jamai… @JAMAInternalMed

Clinicians have minimal time w/ patients (avg return visit 7 mins, new patient 12 mins). If a doctor considers a diagnosis in the 1st 5 minutes, the accuracy is 98%, but if doesn't come to mind in that time the final accuracy is only 25%.

Would keyboard liberation, which > 20 companies are working on using #NLP and #ML, provide the gift of time to clinicians?
Our #NHS review showed the value of just 1 minute of not being a data clerk
topol.hee.nhs.uk @NHS_HealthEdEng

OK, I could go on with this thread all day ;-)
I hope I piqued your interest in the topic of what we can get out of #AI in #healthcare
#DeepMedicine is out Tuesday.
amzn.to/2XO4TPV or deepmedicinebook.com