AI That Saves Lives

Christopher Hughey just published a wonderful article called “How WellAI is Clearing the Path from Patients to Doctors” explaining WellAI’s business philosophy and demonstrating how WellAI works.  One of the conclusions is that WellAI’s technology is not meant to replace doctors, but, rather, to facilitate an interaction between doctors and patients and to make that interaction more efficient.  It’s akin to flying commercial airplanes: autopilot was invented decades ago.  But human pilots are still in the cockpit, working hand in hand with sophisticated technology.

Full disclosure: While the WellAI Digital Health Assistant may suggest its top three diagnoses, it is still a human doctor who, just like an airplane pilot, makes the final decision on the diagnosis and how to treat a particular patient.

I would like to expand on Christopher’s story.  In particular, I would like to address the following questions:

• Why should doctors not be afraid a machine learning technology such as WellAI and in fact should love it and embrace it?
• How do we know whether the diagnosis is accurate?  How can we trust artificial intelligence with our health?
• When is it beneficial to use artificial intelligence for a medical diagnosis, and when are human doctors at their best?

However, first, I’d like to answer the following two customer questions that come up from time to time:

1. The patient may not be a medical expert and may put in information (symptoms) into the WellAI app she may not really understand.  Would that lead to a misdiagnosis? 
2. What if, to the contrary, she had some other information that she did not think was important.  Would that also lead to misdiagnosis?

Short answer to both questions is NO.  Let me elaborate.

1)  We think everyone needs to try the WellAI app for yourself to see how it works.  The experience may eliminate many questions naturally. The basic version is available for free in both App Store and Google Store.

2)  We thought Christopher gave a very illustrative example that underlines WellAI’s motto: facilitate, not aggravate.  WellAI wants patients to talk to their doctors.  In fact, we bring that moment of human interaction closer: 5 minutes for WellAI-guided onboarding vs 23 minutes for traditional onboarding (national average).  WellAI mitigates doctor burnout by eliminating the most routine and boring part of the patient visit: waiting room and the questionnaire (23 minutes national average) and the doctor/nurse note taking (15 minutes national average).

3)  We are a team of data scientists.  We look at facts and evidence. We have run backtests and model validation for about 12 months.  We have used a Bayesian-type approach to model validation: machine learning model cross-validation, as well as human doctor validation.  Based on randomly selected 2,400 vignettes (diseases), the accuracy of WellAI diagnoses has been 83%, on average. For comparison, human doctor accuracy is at around 60%.  Traditional commercial symptom checker accuracy is around 30% (!).  Unlike some other startups, we have not been marketing ourselves.  We have not been looking for investors.  We have not been making any claims, until we are finished with a thorough scientific backtest.  We have been silent about our research success for 12 months.  Now that we are done with core modeling, we can look finally at evidence.  And the evidence has impressed us…

4)  Everyone is error prone. But here is how WellAI’s algorithm mitigates misdiagnoses: we ask 12 to 18 science guided questions, as opposed to 5-10 questions asked by a doctor.  We do it for two reasons: (1) To eliminate unrelated outcomes, in order to come up with a more accurate diagnosis, (2) To get the patient back ‘on track’ with targeted questions, in case she may not fully understand what’s happening with her.

5)  Most importantly, a technology like WellAI may prevent detrimental or even deadly outcomes. When you are having possible early signs of stroke like sudden dizziness, what do you do?  Do you call 911 because you are dizzy?  Do you call your doctor?  What if you have to wait 23 minutes to talk to a doctor?  What if it’s middle of the night?  What it it’s a weekend?  Most likely you won’t call anyone if you feel sudden dizziness.  You would probably google first.  If you ever tried to google a medical outcome based on one or two symptoms, you would know it could be a frustrating exercise as there could be hundreds of conditions under a ‘sudden dizziness’ symptom.  Amazon’s Alexa, as of now, unfortunately, has the same issue as Google when it comes to diagnosing health outcomes. However, we strongly believe that this same patient may want to spend 2 minutes on an app like WellAI that’s available 24/7, with zero wait time, in order to eliminate anxiety of “what if”.  If after a few questions the WellAI app determines that you have early signs of stroke, it will stop questioning immediately and will present a big button ‘call 911’ and will in fact strongly suggest calling 911.  AI may save lives…

In fact, emergency rooms would like to partner with a technology like WellAI’s because, when wasting even one extra minute could be a matter of life and death, ERs would want to have the WellAI questionnaire transcript and the WellAI patient medical history ‘concept cloud’ instantaneously, the moment this patient calls 911.

6)  It’s important to remember that the accuracy numbers quoted above (83% for WellAI vs 60% for human doctors) were the average numbers.  For example, for the primary care situations and some newer diseases like COVID-19 the latest machine learning tools like WellAI are very accurate.

However, there are other areas where humans may prevail.  One area where machine learning algorithms are still struggling is oncological diagnosis.  The topic of one cancer, namely thyroid cancer, is near and dear to our heart. Sergei has been fighting to prevent thyroid cancer since he was 13, when he was exposed to the Chernobyl radiation fallout.  Human doctors and dogs are most likely still more accurate in detection various cancers than the machines.  That’s one big reason the WellAI team is working hard in the areas of oncological diagnostics (in fact, there is a PhD in surgical oncology on WellAI staff) and genomic diagnostics (where Sergei and Daniel are part of the international working group under the auspices of the International Federation of Clinical Chemistry and Laboratory Medicine, or IFCC).

It’s human nature to be suspicious of anything new.  However, at least in the next few decades we don’t foresee machines replacing human doctors.  Humans require social interaction, especially when they experience pain, discomfort or anxiety.  Machines will make doctors more creative and more efficient.  In the end, the patient wins.

To go back to the disclosure in the beginning of this blog: WellAI’s Digital Health Assistant is a tool to assist human doctors.  It is the human doctor who makes the final decision on patient’s condition.

As a reminder, the basic version of WellAI app is now available for free on your smartphone.  Please use the following links to schedule a 1-on-1 demo:

• If you are an employer and would like a cool benefit that makes your employees healthier and happier, and saves 34% to 45% on your healthcare costs, please schedule a virtual meeting here.
• If you are a potential investor who would like to be part of a unique once-in-a-lifetime investment opportunity that will forever revolutionize healthcare, please schedule a virtual meeting here.
• If you are a potential customer, a patient or you are just curious about any of the WellAI high tech products – voice-guided Digital Health Assistant, data-driven Telehealth 2.0, or the scientific chronic disease management program – please schedule a virtual meeting here.

Please follow WellAI  on LinkedIn,  Twitter,  Facebook,  Instagram,  Telegram and  YouTube for news and updates!