AI in Cancer Care: Separating Fact from Fiction
James Hamrick, MD, Flatiron Health, discusses his presentation at the 2023 Clinical Pathways Congress + Cancer Care Business Exchange meeting on how artificial intelligence is being used in oncology operations and care today and how it may impact the delivery of care, including clinical pathways, in the future.
Transcript:
Dr Hamrick: I'm Dr. James Hamrick. I'm a medical oncologist and vice president of Clinical Oncology at Flatiron Health.
What are you the most excited to share with attendees from your panels?
Dr Hamrick: I'm really excited about the conference this year and our panels in particular because I feel like we're at a moment where the problem that we're trying to solve, which is how do we drive high quality care for all cancer patients where we're poised to make some meaningful inroads into that. So, the solutions that we're talking about today and here at the conference are clinical pathways.
And I'm interested to hear about novel ways that people are building these, adopting them, the content that they're layering into them, how they arrive at that content, and mostly how these pathways can serve, most importantly, patients by making sure that everyone can benefit from all the great innovations that are happening in diagnostics and therapeutics and oncology. But also the physicians and their teams who are super busy and need to be able to have this stuff baked into their workflow so it doesn't slow them down.
How do you think AI will impact cancer care?
Dr Hamrick: Artificial intelligence has been in use in cancer care for a while now, but we're certainly very excited about all the new energy around it with all the popularity of language models, which have made it more available and accessible to everyone. At Flatiron, we've been thinking about how to use machine learning and using it for a long time now. We've kept that pretty far from direct patient care, point of care decisions, because we know that the models have to be really good and work really well in order to impact a patient care decision directly.
And where we're comfortable using it, where we've been comfortable to date has been in things like curating research data sets, so building models to help extract key clinical variables from large segments of deidentified data in order to make smarter data sets. What I'm really excited now is we have the confidence to get a little bit closer to the point of care. So for a while now, in some of our products, we've been doing things like using machine learning, which is a form of artificial intelligence to predict metastatic status, or the presence of NGS testing, in order to match patients to clinical trials, for instance. So as these advances happen, I'm very optimistic about this, not only making research teams and researchers more efficient, but actually making frontline caregivers more efficient as well.
What barriers would you like to see precision medicine address for cancer care?
Dr Hamrick: Well, I think something that's happening beautifully in cancer care right now is the rapid pace of innovation. So understanding more about tumor biology through things like genomic sequencing, developing new therapies that are rationally designed to target a cancer or to leverage the immune system to fight a cancer. So we're really, in the course of my career, we've seen some amazing innovations with precision medicine. What I really want to make sure in terms of barriers is that everyone can benefit from this.
So how can we make sure that all of these great advances aren't only used on the patients who are best able to advocate for themselves, but instead can be used to the benefit of all cancer patients? I think that's vital to our infrastructure and I think pathways are a good example of how we can do this. All of the pathways products that we're looking at and all the philosophies around pathways really are about reducing variation in care and driving high-quality care for everyone, both standard of care and research. So, I'm interested in some of these historic barriers where we've seen inequities in care getting broken down through precision medicine.
What are some of the main benefits and challenges for AI? And what are the benefits and challenges for precision medicine?
Dr Hamrick: Artificial intelligence is incredibly promising. We've been using it for a number of years now at Flatiron to do things like curate real-world evidence data sets, and so it's a very powerful tool. But some of the challenges are you've got to make sure that the tool is accurate enough and is protected against bias so that it's actually driving good, high-quality care and enabling providers to source through all of the volumes of new data that are coming out every month, and have access to the key data they need to take care of the patient in front of them.
So, any application that we do with that has to be built very carefully by domain experts, people who really understand oncology, built upon data sets that are largely purpose built for training models like this. And then it's got to be continuously monitored. It's not going to be an off the shelf product that you just throw out into the wild and turn it loose and let people use it. That can really introduce bias and harm people. So ongoing maintenance to mitigate against bias is really important. But I'm very excited about the potential for it to take away some of the grunt work and allow doctors to practice at the top of their license and spend more time looking patients in the eye.
As far as precision medicine, I think some of the barriers, some of them are clearly the frontiers of innovation and discovering the next great biomarker, the next great line of therapy. But there's also some slightly lower hanging fruit, which is more of an operational issue, which is, "Hey, how can we make sure that every cancer patient gets the appropriate test for their cancer?" So if they need to have multiple biomarkers checked or a panel checked, let's make sure that gets done. But it doesn't end there. We have issues where people are getting the test done, but then the results aren't readily available, or they require almost a PhD in genomics to interpret them properly to then match them with a therapy.
So there's sort of a chain with a number of areas where the journey on route to the best decision for the patient can break down. And I think in the precision medicine space, we think, at Flatiron, about technology as a tool to help make sure that it's easier for doctors and patients to make the right choices based on the best data and information. And it's harder for them to go astray.