Physicians in pioneering organizations have been teaming up to apply evidence and clinical consensus to medical practice in new, innovative ways. One of the organizations in which very promising activity in this area has been taking place is Hackensack Meridian Health, an integrated health system based in Edison, New Jersey that encompasses 13 hospitals, including two academic medical centers, two children's hospitals and nine community hospitals, physician practices, more than 120 ambulatory care centers, surgery centers, home health services, long-term care and assisted living communities, and other services, and 28,000 team members, and more than 6,000 physicians.
At Hackensack Meridian Health, physician leaders have been developing a broad strategy to leverage data analytics to support the ordering process at the point of diagnosis and care, beginning with the hematology-oncology service line. The physician who initiated this initiative, Andrew Pecora, M.D., is a hematologist-oncologist who has been practicing at the Hackensack organization since 1989. Since July 1, 2016, Dr. Pecora has been president of the Physician Enterprise at the Hackensack Meridian organization, and chief innovation officer for oncology for the entire system. He has received numerous awards and recognitions, and has published many articles (a summary of which can be viewed here).
At the core of the initiative that Dr. Pecora initiated with fellow specialist physicians at Hackensack Meridian Health, and whose methodology is being replicated now across numerous specialties, is a precision analytics tool called a gene expression profiling (GEP) tool, which is helping hematologists and oncologists as a clinical decision support (CDS) tool that streamlines the use of large data sets and enhances decision-making and ordering around cancer care—and which is beginning to be applied to other areas of medical care as well. Pecora and some of his colleagues, after developing and expanding the use of this tool at Hackensack Meridian Health, have commercialized this methodology, through the New York City-based Cota, Inc., where he serves as co-founder and chairman. Pecora and his colleagues continue to expand and refine the methodology at Hackensack Meridian Health.
Dr. Pecora spoke recently with Healthcare Informatics Editor-in-Chief Mark Hagland about the leveraging of analytics for clinical decision support in diagnosis. Below are excerpts from that interview.
Tell me a bit about the core initiative that you’ve helped to create and lead.
We’ve created a precision analytics tool that enables physicians, nurses, and patients to identify adverse variance, to prevent it, for all healthcare, not just oncology, though we’re starting in oncology. Adverse variance means when too much or too little care is provided, and results in adverse outcomes. About a third of our healthcare expenditure is wasted, according to current statistics. We’ve created a methodology that allows all stakeholders to see where adverse variance is occurring, and you can reduce variance and reduce adverse outcomes and total cost of care at the enterprise level.
Andrew Pecora, M.D.
What was the origin of this program?
About eight or nine years, ago, I took part in a think tank with the McKinsey Corporation, looking at how we control the rising costs of cancer care. And at the end of a three-day work session with a lot of well-known people in the country, the conclusion was that we had to ration care, though we couldn’t call it that. I left that meeting quite discouraged, because I’ve taken part in so many clinical innovation efforts. I believe our job is to cure cancer, and here we are on the precipice of curing more and more cases of cancer, and that didn’t sit well with me. And I’m an inventor, with over 50 patents in stem cell science, and I’m an administrator as well as an entrepreneur. So I sat down and spent the better part of six months trying to solve this problem, and realized it wasn’t a biology problem, but a mass statistics problem and a variance problem.
So I started to consult with biostatisticians and others, and came to the conclusion that we weren’t going to solve the problem of wasting one-third of our resources using traditional methods. And clearly, using ICD-9-based and claims data, just wouldn’t get us there. So I came to the conclusion that only the expression of numbers would allow us to identify, codify, stratify, and analyze data sets in a timeframe appropriate for the codification of care. That will allow you to intervene at the point of care, because you want to know that there’s the risk of adverse variance so that you can prevent it, because once it happens, it’s too late. We had to create a way to compress data. So I looked at how the television industry digitizes pictures and large data sets are compressed into a numeric format. And combining that with the concept of GPS, which is numbers-based, and a node, and we came up with the idea of a Coda Nodal Address. It takes all your demographic information as well as behavioral information like smoking, drinking, your family history, and all the attributes of your disease, histology, stage, level of risk, etc., and the intent of the treatment of the doctor, and where you are in the progression of your disease—first, second, third, fourth—in other words, its presentation; and we took all of that data and put it into a CNA.
And we spent the better of three years working with experts from around the country to map all of the variables that are relevant for each type and subtype of cancer, and assign numbers to them, so that we have a digital code for all stages, forms, and presentations of cancer. So you have the CNAs. And again, the way to think about this is the following: if you were asked what is the shelf life of the produce in the produce section of a supermarket, you would come up with an average number but it wouldn’t help you at all in terms of how I should order things. You’d have to take each vegetable and measure each one separately; and then you’d realize, with apples, there are different kinds of apples. And at some point, there are no more subdivisions, so you maybe end up with 50 buckets of data, but now you have much greater precision in seeing whether or not an intervention is good or bad.
So we’ve been able to do that with all cancer, and can risk-stratify at a level that’s never been done before with cancer, and determine which care process might give you the best outcome and lowest total cost of care. And we know we can probably save 25-30 percent on the cost of care using this methodology. And it’s currently in the lexicon of the insurance industry. In fact, we’ve been able to obtain two federal patents for CNA-guided care. The first patent was granted a year and a half ago, and the most recent patent was granted a few months ago.
Can you share a few highlights of the results at Hackensack Meridian?
Yes, please quote the abstract from our article, “Early economic benefits of gene expression profiling using a 21-gene panel among breast cancer patients” [published in conjunction with the 2016 annual meeting of ASCO, the American Society of Clinical Oncology].
Thank you. Here is a portion of the abstract from the article:
“A retrospective review of 227 consecutive patients, aged < 70, with early stage breast cancer at a single institution. Cost data extracted from the chart reflected actual costs not charges. Results: 68 percent underwent GEP with 52 percent, 43 percent, and 5 percent having low, intermediate and high recurrence scores. By multivariate analysis GEP was applied more frequently in patients without lymph node micro-metastasis. Adjuvant chemotherapy was utilized less in genomic profiled cohorts (19 percent versus 29 percent) and followed GEP recurrence scores. The mean actual 6-month outpatient costs were $24,955 with adjuvant chemotherapy and $2,654 with hormonal therapy. Stage II patients undergoing GEP received adjuvant chemotherapy at a lower frequency (28.6 percent vs 86.7 percent); but stage I patients with testing received slightly more chemotherapy (15.8 percent vs 14 percent). Universal GEP testing of stage II patients would result in net savings $11,494 per patient inclusive of the cost of the test; stage I testing increased costs by $4,505. Similar trends of decreased costs for universal GEP testing of grade II/III tumors ($2394) but increased for grade I tumors ($6047) were noted. Conclusions: Universal GEP testing of stage II or grade II/III lymph node negative breast cancers resulted in lower outpatient costs inclusive of the diagnostic test within the first 6 month episode of care. Among patients with stage I disease GEP identifies a small group of “higher risk” patients who may derive clinical benefit from adjuvant chemotherapy who otherwise might not have been treated, with acceptable QALY ratios. Inclusion of hospital/relapse costs would further support GEP.”
Meanwhile, with regard to the methodology, we’ve been using it at Hackensack Meridian Health, and Hackesack and Horizon Blue Cross are going into a bundle program.
When did you and your fellow specialist physicians begin to integrate the use of this tool into patient care?
We’ve been doing this for about three years now with Horizon Blue Cross, in its first iteration, and now we’re going into the second iteration. So we’ve been working with it for about three years.
Overall, what have you learned from applying this tool/methodology to patient care?
We’ve learned that you can improve clinical outcomes at the individual patient level, and reduce total cost of care at the population level. In some people, you’ll spend more money than you did before and in others, less, but you’ll wind up spending less money overall, and yet you’ll end up right.
What has the process of convincing oncologists to use this, been like?
You’ve just defined why you can’t do this without CNA. The CNA is a number. That’s why we’ve had the uptake we’ve had. I can look at all my 113s and 115s and 138s and on and on, and all yours, and doa fair comparison. And when you do it at that level, oncologists are incredibly precise. Of all the diseases doctors treat, cancer is one of the most complicated. Doctors say, don’t share claims data with me, that’s why you have to use the CNAs. Those are subgroups—not only different types of cancer, but all the variables of patients that add up to you the individual. Did your mother have colon cancer? At what age did she have it? Did your father have it? Are you diabetic? Do you exercise? And that allows you to compare group to group, like to like.
Do you see what you and your colleagues have been doing here as being part of the broader framework of moving towards value in healthcare?
Obviously, I’m conflicted, because I have personal interests here. But this is the only way it’s ever going to happen; because if you don’t account for the true differences upfront, you’ll never get the buy-in of the docs, nor should you. And we’ve created a simple yet elegant way to account for differences in patients and stratify numerically, so that large data sets can be compared, and put at the point of service for the doctor. We’ve created an end-to-end system for the doctor at a level of clarity that has never been introduced before.
What have your fellow physicians at Hackensack Meridian told you about the value of using a tool/solution/methodology like this?
They’re all huge supporters of it. It gives them insights into value and into quality of care that they otherwise couldn’t get before.
How does this integrate with data and IT?
It’s an overlay. It’s self-contained, but it can interface with all EMRs and any other applications. It’s basically a real-world evidentiary tool.