What physicians want from precision medicine is to understand what works and what doesn't in treating conditions and diseases, according to Dr. Dan Riskin, a physician and CEO of Verantos, a Silicon Valley startup that runs advanced, real world evidence studies.
Physicians currently rely on evidence-based medicine, using data from randomized clinical trials. This approach gets information from clinical studies to determine what works best for most people, on average.
"Right now people are being treated in general ways even though they're unique individuals," Riskin said.
Also, because people are living longer, they are often living with co-morbidities, which increases their uniqueness from a medical perspective and makes general care for a single disease less likely to be optimal.
Precision medicine goes beyond evidence-based medicine and tailors care, but it requires technology and interpreting unique patient characteristics, complexities that present a significant roadblock for physicians, Riskin said.
The challenge is getting information that exists through clinical care and other documentation to understand what has been tried for specific subgroups of patients and what has worked best. Some of the information lives in narrative text in the EHR.
"Now we're talking about using AI to identify patterns," Riskin said. "The next step is to use the information in the system to run tailored studies for unique groups of people. That is shockingly hard to achieve. Getting between here and there is hard."
WHY THIS MATTERS
"If I've got someone coming in to me who has two injuries, such as a hole in the intestines and a head injury, I want to know what works and doesn't work," Riskin said.
Combinations of conditions such as these have been seen by many doctors in last 20 years, but different choices have been made that result in varying outcomes.
"I know what the protocol says for an intestinal injury, but there's no published literature on combinations of problems. It's simply too hard to study. But, a data-driven approach can answer what worked best across natural variation in practice."
In another example, an 80-year old woman who has both heart failure and lupus may be treated in the same way as any other lupus patient. But, heart failure may change the equation. No one has studied or published on this, but natural practice has led to many doctors seeing the combination and trying different approaches.
What's happening is that healthcare is undergoing a revolution. With more available data, increased compute power, and smarter AI technologies, it's finally possible to perform subgroup analytics and comparative effectiveness. This is leading to more precise care tailored to individuals, Riskin said in an article for Forbes.
But, the entire system for physicians to learn what other doctors are doing and what is working is set up through peer-review literature. This requires a cycle of more than a year and often a decade to distribute what is learned. Even specialists can't keep up with all of the available literature in the field.
In an ideal world, the AI reads and understands what's going on with a patient and other similar patients, and does a custom study and match that creates a plan for the patient, Riskin said. This may be a decade or more into the future.
The ideal is for all doctors to practice as man plus machine, the AI collaborating with the physician, using information in real time.
The Holy Grail of healthcare would be subgroup analytics performed to the patient level to optimize for their unique characteristics, Riskin said.
"This would allow physicians to know that certain subgroups would do far better with another treatment," Riskin said. "I think this will lead to tailored therapy and transform healthcare."
Email the writer: email@example.com