An Epidemic of Misdiagnosis: Using AI To Solve A Quiet Crisis In HealthcareAn Epidemic of Misdiagnosis: Using AI To Solve A Quiet Crisis In Healthcare
An Epidemic of Misdiagnosis: Using AI To Solve A Quiet Crisis In Healthcare
May 29, 2018
By Jonny Ainslie
AUSTIN, TX - Receiving results from Clinical Pathology Laboratories Inc. in Austin, Texas, Vicky Phelan was quietly relieved to learn that her cervical smear sample had returned negative results in 2011. The same could probably said of 208 other women from Ireland using US companies for screening programmes without being told of testing abnormalities. Today, seventeen of these women have since died.
A single medical misdiagnosis is now sadly not uncommon. Johns Hopkins researchers in Baltimore reviewed tissue samples from 6,000 cancer patients across the US, and found that one in every seventy-one cases was misdiagnosed, and up to one in five cases were misclassified. After independent re-diagnosis Vicky was finally able to begin treatment in 2014, although even her settlement of €2.5m awarded this year will not be able to change her prognosis: there is now no cure for her cancer, she has been given a year to live.
There is no traditional solution to this problem
When reviewing 25 years of US malpractice claim payouts, the researchers revealed that it was diagnostic errors, not surgical mistakes or medication overdoses, that accounted for the most severe patient harm, the largest fraction of claims, and the highest total of penalty payouts. In fact, the most recent figures from the National Academies’ Institute of Medicine reports that diagnostic errors contribute to approximately 10% of patient deaths, and up to 17% of hospital complications.
It is extremely important to note that the performance of physicians is typically not the direct cause of diagnostic error; instead, the researchers found that inefficient collaboration and integration of health information, gaps in communication, and the healthcare work system itself inadequately supports the diagnostic process. In short: it’s not the human face of treatment that’s even the issue, but how hospitals are able to support their doctors. Perhaps one problem, as noted by Dr. Leonard Zwelling of the MD Anderson Cancer Center, is that “we still make the diagnosis pretty much the way we did for the last 50 years”, and there is no traditional solution to this problem.
As people live longer they need exponentially greater combinations of care and medicine; the number of data points to keep track of when properly formulating prescriptions and diagnosing an illness with many symptoms keeps getting harder. So in recent years, countries and companies have turned to the latest breakthroughs in Artificial Intelligence. This week, UK Prime Minister Theresa May has become the latest global leader to pledge her support for a revolution in healthcare by deploying AI in the NHS, insisting that it’ll prevent cancer from taking 22,000 more lives every single year.
AI is the missing link that connects our vast yet disparate supply of healthcare data to its true potential
Optum, part of the Fortune giant UHG, is one of the companies at the forefront of providing this new sort of solution. They’ve developed a Care Coordination Platform to aggregate vast sums of data into a comprehensive overview of every individual patient’s entire medical history. This allows any doctor or nurse access to a big picture and the smallest details of a case history in a glance.
It suggests immediate treatment options automatically, and the OptumIQ system underneath it means that with every newly inputted data point, the algorithms adapt. One of Optum’s other ventures is a collaboration with NYUPN, which uses these individual data profiles to identify high-risk patients before they even begin to feel sick. Their performance analytics combines clinical, claims, and socio-economic data to suggest the most appropriate and cost-effective treatments. This is going a long way to bolster one of any doctor’s most important assets: time. Perhaps in Vicky Phelan’s case, this technology could have been lifesaving.
[caption id="attachment_11540" align="aligncenter" width="5616"] Kerrie Holley, Technology Fellow @ Optum[/caption]
To better understand the possibilities for AI in diagnostic healthcare, AI Business sat down with Kerrie Holley, Optum’s first Technology Fellow, hired by CIO Jonathan Telley to drive their emerging tech departments. Holley’s self-assurance is quickly apparent, his easy manner belying a vast technical and professional experience.
Named an IBM Fellow in 2006, the most prestigious position in the company’s tech community, Holley made his name as the CTO of SOA’s Centre of Excellence, and later moved on to design high-performing applications in the financial service market. At Optum, his mandate has been diverse, combining machine learning and deep learning with applications in IoT, Genomics, and Cybersecurity.
Beyond the human eye: AI applications for healthcare go far beyond misdiagnosis
When asked for his opinion on the value of AI for avoiding misdiagnosis, his answer is succinct: “these patterns just can’t be detected by the human eye.” For Holley, AI is the missing link that connects our vast yet disparate supply of healthcare data to its true potential. He estimates that machine learning algorithms can make a diagnosis accurate 87% of the time – but combined with his fostered work in deep learning, they can reach a world-beating 97-98% success rate.
But he doesn’t see this as a slow-burn takeover effort in making doctors defunct. Navigating an industry that has person-person interaction at its heart, Holley is extremely quick to insist that he is not in the business of replacing human specialists. “We just want to make those people better at their jobs. We don’t want to replace a call centre, we want to make their agents more helpful.” What this looks like in practice: “Let’s say someone calls in asking about ‘ABA’, and their agent isn’t familiar with autism. The AI can still look at our call logs and give them all the relevant information, just like that.”
Beyond this, Holley is also keen to emphasise the sheer breadth of uses Ai has for healthcare providers. Edge computing can strengthen any localised centre’s performance power, their systems can help with customer service, but mainly – their applications aren’t limited to cancer diagnosis. They can do much the same with heart disease, diabetes, even identifying strains of flu.
The Optum Ventures fund is endowed with $250m of capital investment for start-ups and early-stage companies who are advancing the healthcare system; and the Optum team provide decades of experience and strategic guidance to their nascent wards. Of course, Holley has a hand in that too. He feels a palpable excitement for the entire healthcare industry as they uncover the many applications of AI in providing life-changing solutions.
Jonny Ainslie is an Editorial & Content Executive for AI Business. He is the editor of Journalists On Truth.
You May Also Like