16 Comments
User's avatar
Felippe Lazar's avatar

Hi, I’m a practicing onc-fellow based in Brazil with great interests in technology and health interface (specifically process optimization in health).

As far as I understand, the NCCN guideline is more a reimbursement guidelines for insurance companies than anything else. I don’t see many doctors using it often for choosing therapy. The NCCN is large, confusing, unusable because it has to include everything that can be done, not what should be done.

That being said, many oncology societies have friendly PDFs (yes, but nice) to look up. You can check ESMO guidelines (they have an app as well) and the Brazilian one are nice and free (https://sboc.org.br/diretrizes-publicas/versoes-finalizadas-2023). This other site (https://hemonc.org/wiki/Main_Page) is quite useful as well.

I don’t see this as a big problem for oncologists, many get used to checking society guidelines or just asking colleagues. I would be happy to talk about this topic if you want to get further on this.

Expand full comment
Sean Geiger's avatar

Thank you so much for taking the time to read and respond! It's interesting to hear that these guidelines are used by payers for reimbursement decisions. And thanks for sharing the Brazilian guidelines. I'd love to hear more of your thoughts on guidelines and what would actually be useful from a process perspective--I'll DM you.

Expand full comment
Jim StClair's avatar

Yes, my experience as well, speaking from the pt side of oncology. I venture to say different Oncs have preferred guidelines.

Trivia - one of the specific criticisms of Watson's effort with MD Anderson was it recommended therapies considered "not geographically appropriate" which I took to mean "we don't do that here" - so was Watson wrong?

Expand full comment
sarhaan's avatar

Very cool! Has the research on which these guidelines are based been released publicly?

If so, it would be interesting to see what guidelines models like O3 develop and use the NCCN’s guidelines as a baseline to evaluate their accuracy.

Expand full comment
Nils Hapke's avatar

Hey, I am an internal medicine resident from germany - love your approach! We use some options to make the guidelines easier to use with some nice flowcharts: https://www.onkopedia-guidelines.info/en/onkopedia/guidelines

The UX you designed is really cool, would love to get an update, if you develop this further or have a look at your code if you decide to open-source it :)

Expand full comment
Shashaank Bhaskar's avatar

Hey Sean, this was a fun read! I'm presently exploring or looking out for areas where novel interfaces could be a game changer. Guidelines are something that struck out to me. Would you be interested in discussing this further?

Expand full comment
Jim StClair's avatar

Hi Sean! Would love to chat and discuss collaboration informally. DM me or X or LI.

Expand full comment
Stephen Pardy's avatar

I'm involved in a project to scan and quantify these guidelines and have had difficulty getting the accuracy rate high enough for our use cases. Would love to compare notes.

Thank you for the post. It is such a a shame that the raw data is locked away - these charts and diagrams are already so close to machine readable format.

Expand full comment
Shoaib Rao's avatar

This is such an important topic, Sean, and absolutely necessary. At Precimmo, we’ve come across these challenges while providing personalized cancer care with our team of oncologists. I’ve also experienced this personally when my father was diagnosed with hepatocellular carcinoma. As a cancer biologist, I could navigate it, but even then, it wasn’t easy.

Your work really resonates with what we’re doing, and I’d love to collaborate. I’ve sent you a message on LinkedIn—let me know when we can connect!

Expand full comment
Haris's avatar

Cool stuff. This brings so many memories back for me personally. You might find this interesting, something me and my colleagues built almost a decade ago https://youtu.be/FPTIjDwirDI?t=2291

Expand full comment
Sean Geiger's avatar

Oh wow that's so cool! Very similar concept but yours is much more thoroughly executed. Thanks for sharing.

Expand full comment
Ricardo Vidal's avatar

I had this very thought while doing my PhD. Why are these guidelines in PDFs and not some other structured format that can be leveraged by systems/tools. Thanks for the write up, lovely read.

Expand full comment
Jamie Walsh's avatar

Yes, but... What happens when the research is not replicable, has errors, or is outright fraudulent? We like to think these are edge cases. However, some of the stats are concerning. How do you only scale the "good" research?

Expand full comment
Peter's avatar

Interesting read; the NHS needs thought like this

Expand full comment
Gary Takahashi's avatar

Retired oncologist here with a background in applied AI. I agree with your sentiment and would be interesting in working together on a novel approach, different from vendor-driven top-down decision support. Hospitals and private equity love these, but I think doctors really want something else. I have some thoughts on this, and am always interested in exchanging ideas.

Expand full comment
User's avatar
Comment deleted
Dec 24
Comment deleted
Expand full comment
Janis's avatar

There you go. Just another group extracting more money from the health care system. How about just doing it like Sean? Just to help people...

Expand full comment