Could AI Replace My Mother?

I have a caring mother who always looks out for me. Sometimes I feel she worries too much — she often assumes the worst-case scenario. It’s not that she’s wrong though, I could fall off my bike on my daily commute to work and yes, my lack of diligently sanitizing my hands during grocery shopping could put me at risk of the coronavirus. Still, I’ve often thought to myself — “she’s overreacting”. So far, my daily minor risk-taking ventures have luckily not resulted in injury. My mother’s overprotectiveness has, on the other hand, been less of a nuisance and more of a blessing on a number of more serious occasions.

Like most people, I’m occasionally in need of the healthcare system. It generally starts with a visit to my general practitioner, most often related to something minor and for the sake of my peace of mind. And then there are the more serious occasions. Those days that we fear, but luckily experience rarely. This is where my mother’s overprotectiveness comes into play: for my genuine benefit and less for my trouble. Besides being “mum”, she is also a 30+ year veteran in the healthcare sector and has seen many things. She knows the pitfalls, the ins and outs, and she has a second-to-none gut-feeling of the healthcare system. Those serious occasions could have ended up having catastrophic consequences, but having someone by my side who knows me well, the inner workings of the healthcare sector, and a built-in capacity to challenge the doctors’ authority was exactly what was needed. While I’ll never know if things would have turned out okay without my insisting sidekick, I am happy I didn’t have to find out.

In the United States, approximately 5% of all outpatients experience a delayed or wrong diagnosis. Autopsies show that 10% of all fatalities are directly related to diagnostic errors and finally, diagnostic errors are responsible for 17% of all unintended events at hospitals — only surpassed by surgical mistakes. As a result of complaints from patients in cases of missed or faulty diagnoses, 5.7 billion US dollars was paid out in compensation in a single year. These numbers appear outlandish and otherworldly for a Dane like me. I live in a country that touts one of the best healthcare systems in the world, surely we are doing better than that? Well, a quick glimpse at similar stats for Denmark shows that a “missed — or misdiagnosis” is the most common cause for patient complaints and of all the patient complaints that ended with compensation since 2008, misdiagnosis accounted for 26% of them. 500 million DKK was paid out in compensation over a five-year period. https://patientsikkerhed.dk/nyt-projekt-kortlaegger-diagnosefejl-danske-sundhedsvaesen/

Now, if you’re anything like me — oftentimes (read: mostly), blindly trusting the authority and the ability of the doctor to always provide me with a correct diagnosis and forward me to the correct specialist for the right treatment, as well as trust that specialist’s ability — these are worrying and anxiety-inducing numbers.

To me, these facts, more than anything, tell the tale of the medical interview; the conversation between patient and doctor — the single most-used diagnostics tool in healthcare. This tale is not about doctors not being skilled or caring enough, as it is a tale about the sheer difficulty and complexity of performing a medical interview, even for a highly trained doctor, under, oftentimes, difficult and stressful circumstances. Circumstances so complex, that a mother’s gut-feeling oftentimes is the best available second opinion.

“It is in the initial phase of listening to the patient’s history, examining the patient and perhaps ordering the first examinations, where it usually goes wrong. Based on the information we have had access to, we cannot pinpoint exactly why the errors are happening. However, factors such as time constraints, inadequate communication between patient and doctor, lack of access to collegial sparring or lack of involvement of the patient’s relatives are all good suggestions” Doctor, ph.d. Eva Beneke

Now, since I can’t count on my mother to come to my rescue and everyone else in the world at all times, what could be the next best option? To me, no answer is more apparent and potent than Artificial Intelligence.

According to the World Medical Association, we will soon lack 18 million doctors globally, which will only lead to an increase in the disparity of healthcare services and overall medical errors. Furthermore, around 3.5 billion people currently lack access to proper healthcare services and more than 100 million people are pushed to extreme poverty every year due to healthcare expenses. So, if the current statistics mentioned above caught your eye as worrying, there is all the reason to assume it will only get worse if we don’t do something drastically different.

With COVID-19, we have witnessed a rapid uptake of digital solutions in healthcare, where no initiative has been more apparent than the wide use of video consultations. While this has been specifically excellent in terms of consulting your doctor from a distance and helping to contain the movement of pathologies around society, it has little impact in solving the overall issue of quality in diagnosis.

These types of bottom-up solutions won’t be impactful in meeting the demand of medical expertise and they certainly won’t improve quality. We need to look elsewhere and we need to collectively be more ambitious. My ambition is to put the medical expertise in the hands of everyone through artificial intelligence and use every single data point available to completely eliminate the guesswork that leads to issues of misdiagnosis and medical error.

The prospect of having ultra-personalized healthcare, which draws on the experience and knowledge of every single diagnosis made in the past, available, may sound like a pipe dream of Blade Runner-like proportions. But the truth is, the technology needed to turn this dream into reality is available today. And not in the “someone created a quantum computer that theoretically works — all it takes is a sterile, -273C degree cooled, lab-grown diamonds-filled server room” — but in actual live-production, helping medical professionals make correct diagnoses on actual, real-life patients. And it is working.

Corti’s AI has as an example helped emergency medical services achieve a drastic increase in the quality of diagnosis of patients with very critical conditions, such as cardiac arrest. They are now on average 20% faster at arriving at the correct diagnosis (Think: timely care), 92% accurate in diagnosing the correct condition (Think: misdiagnosis) and they identify 40% more of previously undetected cardiac arrests (Think: missed diagnosis). And the AI only gets better the more data it is exposed to.

This is all made possible by applying proper machine learning frameworks, access to relevant data, and delivery of real-time decision support in an intuitive interface. Now, if this can be done for one of the most critical and difficult-to-treat illnesses we know of, it can be done for every single diagnosis. The principle is and will be the same. It just requires access to the right data.

So, while I hope AI won’t replace my mother in any sense, I hope it replaces my mother’s need to be the vigilant watchdog towards the quality of healthcare I receive.

After all, we could all use someone who takes care of us as mothers do.