I published a story on Color Health’s virtual cancer clinic, and the news that they just received this ASCO Certified designation. This is a virtual care model that supports cancer patients through their treatment journey, and it’s sparked a lot of feedback and discussion from readers. In particular, Color Health provides this interesting model of how systems might deploy AI to improve the health care that gets delivered to patients. If you want to learn more about Color’s AI tools, they have a bunch of blog posts describing them and the work that went into them.
More generally thinking about AI in cancer care, the quality of these models and what underlies them is so important as these tools are being built. Kyle Edmonds, a palliative care physician at UCSD, wrote to me after reading the story, and we had a good discussion on the phone after that. One of the things that he pointed out was that models like OpenEvidence don’t do that well with palliative medicine generally since “it’s not trained on our datasets.”
That makes me wonder about what kind of work can be effectively and safely deployed in larger models with or without really expert, human oversight, and whether there are fields that make a difference in cancer care and outcomes (like palliative medicine) that will always need humans for every direct health care interaction.
Well, another interesting thing in cancer. The FDA approved Revolution Medicine’s daraxonrasib for expanded access, meaning that patients with advanced pancreatic cancer actually can get access to this drug now before approval. However, it does take a fair bit of paperwork, as some physicians have pointed out. Nonetheless, the reception from the oncology community seems positive since it’s a mechanism enabling earlier access to this breakthrough drug.
Finally, I'm flying out to Hawaii tomorrow for a reporting trip focusing on cancer care, trials access, and more in rural Hawaii. If you work on this area, or on Hawaiian health, feel free to reach out!
