David Artz, M.D., CMIO at Memorial Sloan-Kettering Cancer Center in Manhattan, has been very busy lately working with his colleagues to enable new waves of clinical—and IT—innovation. Indeed, he is helping to support exciting new realms of cancer research—the data from which is becoming bigger and bulkier by the day. Dr. Artz spoke recently with HCI Editor-in-Chief Mark Hagland about his current work. Below are excerpts from that interview.
What have you and your colleagues been up to lately?
If you look at new things that have come along in the industry in the last ten years—for instance, beginning in the early to mid-1990s through the end of the 1990s—PACS [picture archiving and communications systems] kind of came through as an innovation in the workflow; and more recently, digital pathology has emerged. But now we’re noticing a new field in diagnostics, where molecular pathology is branching off from pathology, involving genomics and personalized care.
David Artz, M.D.
And it’s really affecting cancer care. You can think of cancer as a genetic disease, not necessarily an inherited genetic disease, but one that moves forward at the cellular level, within the DNA and RNA of the cell. And there’s been a research modality called sequencing, which is now making its way into the clinical arena. We have a few specific mutations that have a known significance in terms of the care of the patient. And you can characterize an illness not only by the way it looks under a microscope, or by the different chemicals it may absorb, but also by the profile of the mutations within a particular tumor.
For actual treatment purposes, with a known actual mutation, it’s been only a handful so far. But the way that the instruments work now, and they’ve evolved significantly over the past three years—when you run a panel, you run a panel that captures, for the same price, the small number of known, potentially actionable mutations, and a much larger panel of what we’ll call mutations of interest, which differentiate the tumor from normal tissue, but we don’t yet know their significance. And the equipment and the testing modalities that we use, are evolving rapidly. The newest modalities, which we started running this year, are called next-generation sequencers.
In the last few months, we’ve switched to next-generation sequencers in the clinical lab. And now we have a new “green space” in health IT, which we’ve been working to address, because now we have a new challenge. At a research institution, first, we get a massive amount of new data, and have to figure out how to manage it. We have to create ways to incorporate this data into the medical record; we have to inform clinicians when their patients have a positive mutation of interest; and at a research institution, we have to inform the clinicians on a routine basis, kind of as a push, so that they know, if they’re doing a trial or study of a particular tumor or mutation, that they’re aware of which patients are positive for the mutation of interest that they want to study.
And what are the mechanics of how you do that informing?
We now do a push e-mail on a routine basis that says to Dr. X, these are your patients in the last week who have this tumor site with this histology and this mutation, since the last time we’ve tested. And that lets them know that they could enroll them in a particular study.
And things are complex on multiple levels, correct?
Yes; we’ve had to create the computer systems that allow people to be able to search for how many patients we have who have a particular mutation or combination of mutations, and combine that search query with all of the other data in the data warehouse. And at a cancer center, that includes some very rich, very specific data. There are two other things that we have at cancer centers that make the data very high-quality: one is that we have the pathology on all patients. Because in cancer care, particular with solid tumors, we don’t treat until we know the type of cancer, with few exceptions. So at this institution, by policy, if someone is treated elsewhere, we re-run the pathology. And in addition to that, we have something that many hospitals have, which is that we’re part of a program on tumor registries that captures about 70 percent of all tumors in the United States. That program is run by the Division on Cancer at the Chicago-based American College of Surgeons.
And because of a number of factors, including the use of a special code set called ICD-O, for oncology, we have this rich description of clinical data that is accurate; and it’s very useful for us to merge that with the molecular data, because now I have a very accurate description, for example, of everyone who has colon adenocarcinoma. We have a strict definition of that, and we can now look and see among those patients, who has particular mutations.
Are you storing actual images?
As we use these diagnostic instruments, and as we move these formerly research modalities into the clinical space, it’s a new challenge that we are now addressing, and there are several reasons why it’s a new challenge. You now have to combine the expertise of lots of different people: you have to have IT/informatics folks; the pathology and lab folks; and bioinformatics researchers, who know how to manage the massive amount of data created by these instruments. And now we have a new challenge. If you look at the data in healthcare, we have several sizes of data, if you will. And prior to this, big data might be data you’d have in PACS files; and imaged pathology slides were much larger. Now, the molecular pathology information is even bigger than that, and requires new file forms.
So in essence, the challenge is creating a whole new IT infrastructure for handling, in a clinical environment, the raw data, the processing of the data, the interpretation, the results—where do all those elements go? To whom do they go? How do they get used? It’s all green space. And in addition to that, the raw data that comes out of the next generation of sequencers is much larger than in the past.
That sounds like a very broad set of tasks and challenges to address.
Yes; the bottom line is that, in the last three years, we have learned a lot, and we’ve had to develop a new modality from scratch; and we have had to teach IT staff lots of molecular biology, when they had no biology background whatsoever. We’ve had to teach them that so that they could help us translate this information into a format that was useful to our clinical researchers and clinicians; and this was all new stuff, out of the blue. And I’ve been a CMIO since 1999; and I’ve not been here before when a fundamentally new modality has appeared out of the blue; and this has the potential to morph into its own thing, its own big field. So it’s funny, because when we come in and want to work on this stuff, it’s totally green space. When I got into this, people were enhancing PACS, but PACS already existed. And in my first job at Northwestern, I walked into a functional EHR. But this stuff is totally new. And now, in the last three months, with the introduction of these massive machines into the clinical space, it totally changed everything. Prior to this year, the tests had been much more cut-and-dried; the machines were not next-generation sequencers—they were sophisticated, but only detected one type of mutation, and the panels of tests were much smaller. But starting this year, they have become much bigger, and have the potential to branch even bigger and bigger and bigger.
Are you in touch with the people at the other big cancer centers?
Yes, we are. And I have also not seen in the industry… The lack of correlation between heat and light… That is to say, folks who are seemingly proud of what they talk about, but if you scratch the surface, actually, not a lot has occurred; perhaps they have great hopes for the future. But in the clinical arena, there’s a lot of planning going on, but not a lot of direct activity yet.
You’re the only ones doing this, then?
Well, certainly not the only ones, but in comparison to places where the IT and lab manage everything, we’re doing this for the whole institution from a central IT department so that the lab doesn’t have to do it all; and the advantage is that we as an IT organization are able to hire possibly more sophisticated programmers than a lab would be able to do, because we provide an IT career path for those folks. And that’s probably the differentiator that we have. And it has become a major focus for what we’re doing.
This is definitely on the leading edge of the intersection of cancer research and information technology development.
Well, what we’ve done is that we’ve created a centralized organization for managing a massive amount of raw data, invisible to the clinician. And we feel that this raw data is important for the integrity of the medical record. As I mentioned, there’s a gray area of interpretation. And just as when a radiologist reads an MRI, I don’t throw out the MRI. And the same thing applies here. And this is what we’ve been working on. And it’s a massive effort. And it’s very hard to do, because there’s a cultural challenge to bridge, between the basic science people and bioinformaticists, I call them the folks who wear sandals to work—and they have the skills and knowledge to deal with this modality. They don’t have the discipline that we have in clinical informatics, where tests and things are locked down, and change control is a sacred concept. You can’t just change things around easily in an accredited clinical space. So we’ve had to combine those folks, and have them work with the guys in clinical informatics and IT, our folks, and have our folks teach them this new discipline, and help them work with these massive new files. And as simple as the concept seems, it’s actually really hard to do. It’s kind of a cross-cultural exercise. And that actually is probably the hardest part of this whole thing. So that’s where we’ve put a lot of this effort. And you will not get to what I’ve shown you, unless you can do that.