Fresh NPJ Health Systems study confirms the reliability of Daniel Nadler’s OpenEvidence

Opinions expressed by Digital Journal contributors are their own.

Winning the trust of users will be prove essential for finding practical uses for AI in the medical sector.

Daniel Nadler’s OpenEvidence are showing that a years-long debate about the role AI can play in medicine may now have some practical answers. As with any technological leap forward, the question of what increasingly powerful AI capabilities can do to help heal the sick and prevent disease has been high on the agenda. The difficulty to date has been two-fold. Turning lofty ambition into practical reality and winning the trust of both patients and medical professionals. 

Nature audit

On the latter, OpenEvidence has received a major boost. Earlier this month NPJ Health Systems, part of the world leading Nature science journal group, confirmed that of the nearly 5,000 references returned on 150 standard OpenEvidence prompts, none were fabricated. 

For a tool that depends on the quality of its training data, this finding will go a long way to inspiring confidence in those using it. The fear any doctor has in using such a tool to help diagnose patients is following a course of action based on incorrect information. This latest study helps address such concerns and shows that OpenEvidence can make a positive contribution to the diagnostic process. 

OpenEvidence backstory

For Daniel Nadler, a long-established Canadian tech entrepreneur, developing tech tools that can help improve medical outcomes is not just a commercial matter. It is personal. In 2021, Nadler says he lost his grandfather to medical error, something he told Forbes was a major motivation for creating OpenEvidence. The exam question, so to speak, was ‘how can doctors more efficiently and effectively process the mounds of information related to diagnosing illnesses? 

The answer is OpenEvidence, founded by Nadler in 2022. The tool acts as a medical search engine for healthcare professionals, sometimes dubbed the ‘ChatGPT of medicine’. This term is misleading as OpenEvidence deliberately differentiates itself from general AI tools by being highly selective with the information it trains its model on. 

This represented an entirely different path from his previous venture, Kensho Technologies, which provided cutting edge data analytics and machine learning services to world leading investment banks and financial institutions. It was the sale of Kensho to S&P in 2018 that allowed Nadler to ruthlessly pursue his vision for building OpenEvidence. 

An expansive medical library

In a nutshell, Nadler says OpenEvidence draws upon the world’s leading medical texts, studies and peer-reviewed research to train its model. This helps improve the accuracy of responses, based on high quality medical data. The ambition is that when a doctor types a patient’s symptoms into the tool, it can get a quick, reliable and scientifically backed response in seconds.

The efficiency this creates could prove a boost for the medical profession. Traditionally, when presented with a complex set of symptoms, a doctor would have to consult a range of medical texts, conduct several database searches and generally take a considerable amount of time to draw a conclusion.  The problem of human error is considerable in such an approach, risking misdiagnosis and delays in the appropriate treatment. It would be unfair to place the blame on doctors, they are simply human after all. 

The efficiency problem

The other issue is the amount of time the ‘traditional’ diagnosis approach takes. Attempting to get to the root of a complicated set of symptoms can often prove time and labour consuming. OpenEvidence provides a possible solution to that problem. This is demonstrated in the company’s claim that 45% of America’s doctors are already using it. “We’ve become the default operating system for Doctors” Nadler told Forbes earlier this year. 

This is what AI in medicine could look like in reality. AI cannot, will not and should not seek to replace humans in the dispensing of healthcare. Ultimate judgement must remain in the hands of people, for the purposes of care as well as accountability. However, technology must be harnessed to ensure people can make those critical judgement calls using the best available information and in the most timely and efficient manner possible. This is where OpenEvidence’s potential lies. 

For many, the future role AI is anticipated to play in our lives is a cause for concern and even fear. Despite the excitement about the obvious and massive potential for change in sectors such as medicine, that cause for concern is real, legitimate and must be addressed before any progress can be made. Real life case studies of practical, beneficial uses of the technology to improve patient experience and outcomes can go a long way to alleviating those fears.

The ongoing debate about how AI will embed into our lives will rumble on for years, if not decades to come. However, in tools like Daniel Nadler’s OpenEvidence, we are getting some early glimpses as to what the practical reality of this technological leap might look like.