OCM Data Analysis, Predictive Analytics, and Future Value-Based Payment Modeling
At the Association of Community Cancer Centers (ACCC) Annual Meeting and Cancer Center Business Summit (March 20-22, 2019; Washington, DC), Journal of Clinical Pathways sat down with Charles Saunders, MD, chief executive officer, Integra Connect, and Lucio N Gordan, MD, managing physician and president, Florida Cancer Specialists, to preview their session on current Oncology Care Model (OCM) performance data and the future of value-based payment modeling post-OCM.
To read more of the session coverage, click here.
So, this first question is for Dr. Saunders. Now that OCM is halfway complete, what have been, in your estimation, some of the larger key learnings or trends thus far?
Dr Saunders: Well, to start I will say that in the beginning it was a learning year because practice transformation is what CMS was buying and that required a lot of infrastructure and new capabilities to put in place by practices. So the performance period 1 was not surprisingly, there were not a lot of wins there for practices. But over the last, let us say three performance periods, the practices have, in aggregate, gotten progressively better in terms of being able to close the gap between CMS's target price and their own episode cost and that has been driven a variety of things. Some of them have been efficiencies in their practice but others have been, for example, cost efficiencies and reduction in the hospital admissions and ER visits and an increase in use of end-of-life services, hospice, and palliative care. But a lot of them have just been better documentation, better comorbidity coding and things of that nature and smarter use of high cost novel therapies. So those things are starting to occur.
Some of the challenges that maybe were not as deeply appreciated in the beginning but have turned out to be reality is how difficult it is to get the data necessary to do the work because all of the programs are data-driven. Quality programs are data-drive, the reporting requirements require the access to data. Even if you want to most efficiently use your care management and care coordinated resources you want you to do those for people that are at highest risk so doing things like predictive analytics to risk stratify your population, that is data-driven. So all of this relies on data and very and difficult to get structured data out of the EHR. So data liquidity is something that has to be a part of the solution in the future and that has probably got to be mandated and enforced by CMS.
Anything to add, Dr. Gordan?
Dr Gordan: Yes, I echo Dr. Saunders’ comments. In the beginning it was an initiative cultural shift. The idea of fee-for-service to value-based care, as expected, our staff and our physicians took quite a long time to get there and a lot of training to go on extended hours of operation in the clinics so we would avoid expensive sites of service like emergency departments and admissions.
So, as Dr. Saunders said, in PP1-PP2, we did not do that great, but as we get more mature in the OCM we have been able to be successful for instance, as an example, mentioned today that we have been able to diminish emergency department visits by 29% as compared other peers in OCM, and 41% as compared to non-OMC practices in the country. But in other endpoints, we will give more details through the presentation today. But to get there, there was a tremendous burden on us in terms of getting documentation out of the unstructured EMR. We are trying to perfect that. We are trying to get the physicians and assistants to get the data to the right places for us to harvest in analytics. We invested millions of dollars in a practice of our size, we hired 100 oncology-certified nurses to run our care management. We have very intense analytic teams and companies helping us out with that. So it all culminates into a success story, but it has been an arduous pathway.
And then what is the future? Can we continue to tweak and improve emergency room avoidance? Hospitalization avoidance? Duration of hospice stay? I think we can. But the fruits are getting a little bit higher on the tree and so that will require other thought processes and you may be getting to that in the next questions as far drug management and pathways or others.
I do want to just jump back to something you alluded to, Dr Saunders, and that is just predictive models and data analytics. So, can you just speak more to that? How can predictive models and such be utilized to improve cost efficiency?
Dr Saunders: Well, you know, about 20% of the patients drive about 50% of the costs. So you want to identify those 20% up front so that you can aggressively case manage those patients. So to manage the adverse events that put them into the emergency department and ultimately the hospital.
Also, you also want to identify people who are likely to die and that are really out of life-saving alternatives or curative therapeutic options so that you can start the conversations about hospice or introducing palliative care. That means not waiting until they are already dying in the ICU but getting out ahead of it. And so there are a lot of predicative analytics tools that you can use, including AI, machine learning, and so forth in trying to identify who those patients are and then focus your precious resources on managing those people with the greatest risk.
At Florida Cancer Specialists, has AI sort of entered the folds of what you are doing now? I am curious, AI is such a hot, everyone is talking about AI and I am curious to know to what extent community practices are beginning to use it.
Dr Gordan: I should say we are still in the infancy of AI utilization in practice, in general, in the United States. I think our practice is progressive, but we are not quite there yet as far as identifying a patient who has a dropping performance status for instance, with high comorbidities and we are not matching real change in laboratory studies, for instance, to detect the declining patient. And then, once we get there then we can alert the physician or healthcare provider team that this patient is to be seen in the clinic maybe once a week instead of every 3 weeks so we can limit ED visits and hospitalization.
So we are not leveraging as much as we can or should, but I think obviously, companies like Integra and others may be able to help us and others in the country to get there.
Dr Saunders: I will give an example of one large practice that we support with predictive analytics. So, once a week we will give them a roster of all the new patients they have seen, risk-stratified, broken down by risks for four or five different things and risk scores and then the ones that are in red or the ones that have the highest risk of ED visits or highest risk of hospitalization or the greatest risk of dying in that same period of time. There are models out there that will allow you to predict the risk of death in 1 month, 2 months, 3 months, 4 months, 5 months, and 6 months so that you can actually introduce them into hospice. In fact, in the presentation that we are giving, I am going to show the results of that where they actually implemented a predictive analytics program and they significantly reduced both the cost per episode and increased the number of hospice days, decreased the number of acute care days.
Dr Gordan: Yeah, we do perform quite a bit of that manually via care management team. I think I mentioned we have 100 care managers, certified oncology nurses, and at any given time they are managing 26,000 patients at FCS. So imagine the burden. But they do a phenomenal job, we have, like, a 96% satisfaction rate from the patients being taken care of by the care management team in addition to us. But we do need to get predictive analytics to help because there is leakage there, right? There is no way we can capture everything and I think we can make the success story even better with help of models like yours.
Dr Saunders: There are other ways that predictive analytics are helpful and that is to predict where you are going to come out on costs. So you are in a performance period, you would kind of like to know whether or not you are likely to get a performance-based payment, you are likely to get a clawback on the MEOS payments from CMS after the fact so that you can reserve for it in your financials. Otherwise, it comes as a shock and a surprise and it is kind of late.
In both of your opinions, how has OCM revolutionized value-based payment modeling for the future? OCM will end, we are already halfway through it as you have said. What aspects of OCM do you expect future value-based payment models to keep or to use? What aspects of OCM do you think will be done away with?
Dr Saunders: Well, OCM did not invent the medical model. But what it did, it took the medical model which is based on the patients in a medical home and has been around for a while. If fully implemented and realized, it has benefits, and it has been well-documented to do that. When I was at Aetna, we had 350 medical homes, about 35 oncology medical homes, and they were this model before OCM actually implemented it. So practices that just simply collect the MEOS payments and do not do anything, they are not going to show any results. But if they actually implement team-based navigation and case management of high-risk patients and managing end of life and they use care pathways that are more thoughtfully considered based on evidence and they use technology and EHRs, they actually can have an impact.
So I do not suspect that they are going to abandon that. My guess will be is they will go down the same path that the ACOs went down where they listened to the feedback and they see what worked and what does not work and then tweaked the formula.
I also suspect that there is going to have to be greater emphasis on managing high costs drugs because in the performance period 1 the cost of drugs was about 50% of the total episode cost. Now, it is about 60%. It is looming to be far more important as a cost driver, and the new novel therapies are coming out at a faster and faster pace so I think they have to get control of that. I am not sure how they are going to do it. A lot of things have been talked about, but I think that has got to be a big part of the equation.
Dr Gordan: Yeah, excellent points. From your last comment, Dr Saunders, as you said, the price of drugs have kind of wiped away our shared savings in the last PP3 and something is going to have to give at some point in terms of us being able to control that. How can we control? We can control with some stiff pathway, or formulary navigator, or options to tell the physicians option one, two, and three for the practice for different OCM or value-based care programs. It will be quite complicated because we have to model not only one or two drugs, but all the others and the number of visits, etc. But I think we are going to have to do something with that.
But I think in the meantime, there is room yet to continue to improve upon what we have done so far, emergency room hospitalization, readmissions. But as it was said, 60% of the cost has to be addressed eventually.
In terms of the OCM longevity, as it was said, we are 50% into the OCM half-life. We do participate with another four or five private companies doing value-based care, all of them with good success. I do not think this is going to be something that will go away. I think we finally got our physicians and other health care providers to switch mentality and culture from fee-for-service to VBC. I think it makes sense for everybody that patients are better cared for. But it will continue to be a very investment-driven, commitment practice-driven approach to make this successful. Back in 2015 or 2014, we had zero value-based care contracts. Now, we have 65% of our business is VBC and I do not see that going down.
Dr Saunders: That pretty much aligns with the goals of CMS and the major carriers. They want to see 70% of their business in value-based contracts by 2021-2022.
Dr Gordan: In our practice, I could see that happening if the cost of operation outpaced the cost of the shared savings or the net, because it is not cheap to maintain 100 care managing people. It is literally thousands of a team member hours to deliver what we deliver and help of analytic companies to assist us in the understanding of our data.
But I think it is much more exciting to couple treatment to value to outcomes to decreasing ER visit, hospitalization. I do not think it should change. I just do think that we should have a better answer from a national standpoint as far as the drug costs because at some point there would be no way we can do better from a clinical management.
Dr Saunders: Yeah I think that pharma is going to have to step up too to be part of this equation with value-based pricing on their drugs. Those conservations are going on.
Do either of you have anything else you just want to add before we stop recording?
Dr Saunders: I think it will be interesting to see the evolution of the use of clinical pathways and formularies, and you alluded to it, but I think they have evolved from something the doctors hated because it was cookbook medicine and it was just forced down the throat by payors as to align with their payment reimbursement policy. But I think increasingly, if you have the various regimen choices, annotated in terms of their impact on total cost of care and the adverse event rates and things like that, then physicians can make choices based on real medicine. The fact that this is a better pathway because it produces a better result, longer disease-free progression periods, fewer adverse events, and a lower total cost of care. Physicians will do the right thing, and they want to, so they need to be informed better and the pathways need to be better informed than just expert consensus based on literature that occurred in a carefully controlled trial and in a cancer center or someplace where this does not represent the real world.
Dr Gordan: I agree. I think one thing companies like Dr. Saunders' and ours, we are going to have to understand better also is how to sequence drugs. For instance, I think most of us can agree with first line therapy for a cancer, let us say renal cell carcinoma, or multiple myeloma for instance, but there are several different ways of doing the sequencing. So maybe one day we will be able to have total cost care and outcomes by sequence A versus B versus C, which one is more cost-effective and at the same time maintain progression-free survival, quality, and overall survival. As the data accumulates, maybe we can go there.
But it gets very difficult for us to be gatekeepers of drug utilization if you have an indication like, for instance, small cell cancer with atezolizumab, carboplatin etoposide, expensive combination because of the addition of immunotherapy, but it is what it is. The data is out there, the patient, the consumer is demanding you have to use it. So, it is difficult to control.
