This is the second article in a blog series about the technology trends that will fundamentally redefine healthcare in the not so distant future. In this article we focus on A.I in Healthcare.
Almost two centuries ago, when the first passenger railway, the Stockton-Darlington Railway, opened, the public reacted with fear. Common belief held such that the human body would melt – that we were simply not designed to travel at such incredibly high speeds…30mph. When we are faced with the unknown and unfamiliar, we err on the side of fear and caution. With hindsight, we are able to put technological progression into an evolutionary context, but it’s much harder to do looking forward.
Artificial Intelligence is our Stockton-Darlington Railway, and soon we will look back and wonder how we survived without it. The capital-T Truth is that we’re not. Medical errors are now the third leading cause of death in America, while there is currently a global deficit to the cacophonic tune of 7.2 million health care workers. Artificial Intelligence will not replace clinicians – at least not yet – but it stands in their symbolic shadow, waiting, watching and learning.
Here are a select few ways that A.I will improve healthcare in the not too distant future:
Lucy in the Sky with Data Driven Drug Design
The trial-and-error method of researching, developing and bringing new drugs to market is outdated, tedious, and primitive. A 21st century solution is required to predict which medicines will work, and which ones will not. By employing A.I, it is possible to rapidly analyse millions of molecular structures to understand how a hypothetical molecule will react, without ever needing to go through the arduous process of manufacturing a drug in order to test its efficacy.
Lucy in the Sky with Data Driven Drug Design
The trial-and-error method of researching, developing and bringing new drugs to market is outdated, tedious, and primitive. A 21st century solution is required to predict which medicines will work, and which ones will not. By employing A.I, it is possible to rapidly analyse millions of molecular structures to understand how a hypothetical molecule will react, without ever needing to go through the arduous process of manufacturing a drug in order to test its efficacy.
Using A.I, 2015 Y Combinator alumni, Atomwise, discovered two existing drugs that could be modified to reduce Ebola infectivity. Accomplished in less than a day, this was a striking contrast to the months or years typically required to complete such a task. “If we can fight back deadly viruses months or years faster, that represents tens of thousands of lives,” said Alexander Levy, COO of Atomwise. A.I driven drug design will become more accurate and powerful as genetic and social data points are incorporated over time, allowing for the creation of truly personalised drugs. N-of-1 individualised and targeted patient-specific drugs will be available at a fraction of the cost, and in a fraction of the time.
Shortsighted Blind Spots
#airbnbwhileblack, the hashtag that launch’d a 1000 headlines. Minority and LGBT guest requests for Airbnb rentals were being denied at a disproportional rate; Airbnb had a serious discrimination problem. To combat this bias, Airbnb applied machine learning and pattern recognition techniques to identify conscious and unconscious predispositions in hosts’ language, tone and attitudes when screening potential guests.
That same bias exists in healthcare. In a recent survey by Medscape, 40% of clinicians admitted that they are bias against certain groups of patients. Psychological temperament, weight, language disparities, and level of insurance coverage are some of the factors that trigger biases. While clinicians may be prepared to admit biases, the deeper underlining problem is unconscious bias, the “subconscious associations gathered over a lifetime that can override conscious beliefs and cause people to unknowingly act in ways that are inconsistent with their true values.”
Conscious and unconscious biases profoundly affect the way clinicians interact with their patient, their line of questioning, attitude, and subsequent diagnosis. These subtle, verbal and physical cues from the clinician can, in turn, affect the level of trust that a patient places in their doctor. When trust is lost, the patient can withhold information or fail to follow medical advice. Unlike a clinician, A.I can scan the records of millions of patients who have presented with similar symptoms, and provide an unbiased diagnosis, while using pattern recognition to identify and address clinician bias.
The First Law of Robotics
The word ‘triage’ originates from the French verb ‘triar’ – its meaning: to separate out. We separate patients based on their medical urgency, especially in a crisis where the allocation of scarce medical resources is necessary. In a recent study of 1,342 patients, the triage nurse correctly predicted the disposition in just 75.7% of patients.
Applying Artificial Intelligence to triage will enable a more accurate assessment of a patient’s clinical needs, and provide more specific recommendations for the clinician or nurse. Contending clinicians against its A.I triage system, Babylon Health accurately triaged 90.2% of patients compared to the clinicians 77.5%. A.I can already more precisely predict the disposition of a patient, and will eventually supersede clinicians to make healthcare more affordable and accessible to all.
But the real question is the process of triage. While A.I can already more accurately triage a patient than a clinician, we need to reexamine the archaic notion of triage itself, which has more-or-less remained the same since Napoleon’s surgeon in chief, Dominique-Jean Larrey conceived it 10 score years ago. The process itself often upsets less critical patients when other patients are given priority, and sometimes results in those patients leaving without being seen. A.I can, and will eventually be used to remove the need for triage itself, pioneering the move from reactive, to a precise, proactive and preventative approach to healthcare.
For all its promise and potential, Artificial Intelligence will make mistakes. In moments like these, we need to be slow to blame and quick to recognise that A.I will be swift to learn. Like those averse to the thunderous roar of the approaching train in the 18th century, it is irrational to let fear of the unknown delay or halt technological progression. If we are to move the human race forward, here lies our moral imperative: to experiment, to challenge the status-quo, and to acknowledge that maybe, just maybe, A.I can create a better world for us all.
