ADVERTISEMENT
Perspective on Lessons Learned from Academic Practices Participating in OCM
In an interview with the Journal of Clinical Pathways, Theresa Dreyer, MPH, the Association of American Medical Colleges (AAMC), provided an overview of her presentation from the 2023 ASCO Annual Meeting titled “Perspective on lessons learned from academic practices participating in OCM.”
Transcript:
Theresa Dreyer, MPH: I'm Theresa Dryer and I lead the value-based care portfolio at the Association of American Medical Colleges.
Please give a brief summary of your ASCO 2023 presentation, “Perspective on lessons learned from academic practices participating in OCM.”
Theresa Dreyer: I worked with 20 academic OCM practices. These are OCM practices that were affiliated with academic health systems and were implementing OCM in the context of being part of a larger teaching hospital, often alongside other value-based care programs so that they were implementing concurrently. One of the things that we did for these practices was take in their data and help benchmark them against other OCM practices that were affiliated with teaching hospitals so they could understand their performance and could talk to their peer organizations around the country to find ways to improve their quality of care.
One of the things that we found is that although our practices were all in one-sided risk arrangements, so they could share in savings but they were not experiencing losses, had they been in two-sided risks, they would've lost their shirt, millions of dollars for most practices. What we found was when we looked at their data and we looked at the top four cancers, those top four cancers by volume were breast, prostate, lung, and multiple myeloma. What we found was that the losses on lung and multiple myeloma were extraordinary. By 2021, there had been an aggregate negative 17 million in losses across those practices that we worked with, compared that to three million in savings for all other cancers put together, and you're looking at negative 14 million in losses.
When you look just at those top four cancers to try to understand what's driving these trends, because by volume these were the four cancers driving the trends. Lung and multiple myeloma saw their chemo costs grow enormously over the period of OCM, and the model didn't provide risk adjustment for chemo that was specific to each individual cancer. Instead, it blended the chemo cost growth across all cancer types, but that meant if you were a practice that was disproportionately receiving lung or multiple myeloma patients, especially those needing the highest cost drugs, the risk adjustment wasn't fully accounting for that. In contrast, we saw prostate cancer had savings for our practices. Breast cancer essentially broke even for our practices. These two cancers, unlike the others, included both high- and low-risk patients, and there was additional risk adjustment in which CMMI used the treatment regimen that these high versus low-risk patients were in, they used the treatment regimen as a proxy for clinical risk and applied a risk adjustment accordingly. These were the only two cancer types in which the chemo prescribed to the patient was adjusted for in the model itself.
We wanted to get a sense of why was there this difference. We don't actually know whether some of the trends that we saw were specific to academic medicine or whether they were experienced by every OCM practice. We suspect that every OCM practice was experiencing similar pressures around chemo growth, but one of the hypotheses that our practices had is that because they were working with teaching hospitals that were centers of excellence for things like multiple myeloma, that they might be disproportionately getting late stage patients or patients who had failed other treatment regimens previously. Luckily, we had a treasure trove of clinical data as well as claims that let us investigate this hypothesis.
We investigated a question of whether the risk adjustment could be improved by including that clinical data. The reason the clinical data was present was because CMMI had a requirement that all practices collection report on clinical data elements for the patients that were attributed to them under OCM. We were using data that the practices had collected and delivered to CMMI, that CMMI themselves had access to, and we were pairing that with the claims data, merging it together, in order to provide a more robust clinical lens on what the risk adjustment could look like if you merged these two data sets. We did that for about 15,000 episodes that we had complete clinical data for, and that was specifically for breast, lung, and multiple myeloma. We were not able to assess that for prostate cancer because the CMMI requirements didn't include the necessary elements to stage the prostate cancer.
Our core question was, is it possible to refine the risk adjustment methodology by using cancer stage or a field-culled current clinical status in which organizations reported on whether the patient was having their initial diagnosis, whether they were stable, whether it was metastatic, whether it was recurrent or progressive disease, and that was a field-culled current clinical status.
What we found in our analysis is that when you looked at cancer stage, that was a significant, unmeasured variable for breast and lung cancer. For multiple myeloma, it was not significant, which makes sense given that it's a liquid tumor. For the current clinical status, patients with recurrent or progressive disease, that was also a significant unmeasured variable in the risk adjustment methodology, and that was true for all cancer types, breast, lung, and multiple myeloma. This has face validity when you think about the fact that patients with more advanced stages or recurrent and progressive disease were more likely to basically experience very large losses because of the cost of the chemo associated with treating those patients.
Which brings us to EOM. How, if at all, was this finding applied to the EOM model? OCM did not initially have any risk adjustment based on clinical variables. Beginning in 2020, they did have a partial risk adjustment for cancers that were metastatic, and that was for any cancer type. The cancer was metastatic under OCM, there was a risk adjustment later in the model. EOM, in contrast, will only include a metastatic risk adjustment for three cancer types: breast, lung, and colorectal cancer. It will also risk adjustment for breast cancers that are HER2 positive. There will be no other clinical risk adjustment. Even so, CMS will continue the status quo that they had under OCM, which is that practices will be required to report extensive clinical data across the board for every patient, for every cancer type. But CMS will not be using that for the purposes of risk adjustment for the financial methodology or for quality measurement.
From the perspective of the AAMC, this is a key missed opportunity given what we were able to see in our data, which is a much smaller data set than what CMS has access to. We think that there would be a lot of opportunities to find other clinical variables that would be highly relevant for every cancer type and that the clinical variables used for each cancer type could be specific to that cancer type, depending on the disease in question. We would encourage CMS and other payers who are developing oncology value-based care models to explore using that clinical data to improve the risk adjustment methodology because we want practices to be rewarded for the care they're providing to patients not have wins or losses based on who walks in their door.
Other things that are different about EOM, this is on the good side, they're doing more sophisticated claims-based risk adjustment that is unique to each cancer type. That's different from OCM, and we hope will improve the methodology even though they don't have the sophisticated clinical data incorporated in a way that they could. However, on the con side, EOM is a narrower model, it included fewer cancers, and many of the cancers that are excluded were those that were winners for OCM practices; however, the losses that were primarily concentrated in among multiple myeloma, that's being kept in the model, and yet many other cancers that practices won on are being excluded. I think it might be a more challenging mix of cancers in the model for many practices. There's also the requirement that every EOM practice that is in the model has to take risk from day one.
How were the lessons from OCM incorporated into the framework of the enhancing oncology model?
Theresa Dreyer: There are some lessons from OCM that are directly incorporated into EOM, specifically in the claims-based risk adjustment. In the claims-based risk adjustment, CMS created a unique model for every cancer type, which was not the case in OCM, it was blended across all cancers. As a result, the risk adjustment itself will account for things like the growth in chemo costs for lung cancer specifically, and it will be matched to the proportion of lung cancer patients you're caring for. You don't have to worry that the mix of your patients itself is harming your target prices because you're getting a disproportionate share of lung cancer patients, for example. That's a key improvement from my perspective in EOM. I also appreciate that CMS did incorporate some additional clinical risk adjustments, such as for the HER2 positive breast cancers. I think that's a great improvement.
I don't think CMS went far enough in terms of the use of the clinical data, given that they are still requiring reporting on all cancer types, and they are now requiring reporting on all patients, not just half of patients, which was true under OCM. I believe that CMS should only have that degree of reporting requirement if they plan on using that data. They had many years under OCM to collect the data, analyze it, and see what would be valuable. And I think that CMS should very quickly update the reporting requirements to match it to the data elements that they're using in the model for either financial or quality-based risk adjustments.
In what ways do academic practices differ from community practices in general and specifically around OCM?
Theresa Dreyer: Every provider will tell you that their patients are sicker and most academic practices will tell you the same thing. I don't have the data on the non-academic practices so I can't prove or disprove that, but I can say that there are many academic practices that are centers of excellence for the treatment of a given cancer type. Maybe they have specialists in that condition, maybe they have a research arm and they do clinical trials at a higher rate than other practices. To the extent that an organization is bringing in a disproportionate share of late stage cancers, of recurrent or progressive disease patients, that is not going to be captured in the risk adjustment under the model and so they'll get the same target price as everyone else, even though they're caring for these higher risk patients. That is something that has certainly informed the academic practices in terms of their participation in EOM. Many are really scared of taking on risk in the model that they didn't do great in OCM and that they don't see large changes to under EOM.
What considerations are specific to academic practices participating in OCM and EOM?
Theresa Dreyer: Academic practices do have an obligation not only to be caring for patients and delivering high quality clinical care, but also to be doing things like research and training of residents. Those elements can be a reason to join things like EOM because you can see it as applied research since it is a test to under CMI to see how cancer care could be improved in the US, but they also have to meet the core elements of their mission in ways that are broader than your typical community practice. They do need to get those residents in and train them, they do need to be doing research, and so there can be a sense that adding one extra thing on top, such as participation in EOM, might just be a bridge too far, especially given the financial stress that all providers are under these days as a result of the COVID pandemic.
What challenges do you foresee for academic institutions around implementation of EOM?
Theresa Dreyer: It really remains the risk adjustment that's the key challenge because now the losses are not hypothetical. They'll be under two-sided risk, and so those losses would be real. Given that and given the tight margins that they're facing, that's a very urgent problem and is one that is causing at least some academic practices that participated in OCM to bow out of EOM. Other elements of the model I think are a little bit more nuanced. For example, CMS really wants to start collecting information about patient demographics and health related social needs in order to use that data to address health equity as an element of quality in the larger model. The AAMC is strongly in support of that as a goal, and we recognize that data collection is the first step in making that possible; however, it just presents operational challenges to make sure that you as the practice are in fact collecting that data, that you get those workflows down, that you figure out how to implement it in a way that doesn't contribute to burnout. This would be an example of just a new program requirement under EOM that is grounded in good reasoning but might be difficult to operationalize.
What are the future opportunities for oncology value-based care models?
Theresa Dreyer: I think that to the extent that clinical data is available, it could really revolutionize these models. To the extent that you can pair clinical and claims data, you can more accurately assess whether the appropriate treatment regimen is being pursued for each patient, and that will appropriately link your financial outcomes under the model to the patient and the risk of the patients that you're seeing. It does, however, represent a large burden to the providers to be reporting all of that data, and so there's two ways that I think all payers should be thinking about this. First, if they want to use clinical data, which in oncology particularly makes sense in a way that it might not be necessary to the same extent for something like joint replacement, but in oncology, cancer stage is essential, the treatment history of the patient is essential, as well as molecular mutations and biomarkers.
Cancer registries already collect a great deal of this data, so one approach that payers could consider is to partner with the registries in order to obtain the data from them and eliminate the duplicative reporting requirements. To the extent that that's feasible, great. If it's not feasible, another way to approach it would be to collect the minimum viable number of clinical data elements from the practices that are being used in the risk adjustment. If clinical data is not available, there are also tweaks you can make to the model design to better align someone's performance under the model with the care they're actually delivering. One way would be to narrow the model to be specific to a given cancer. Say to yourself, "Okay, we can't boil the ocean here, so we're really going to narrow the number of elements that are under consideration." That could be in the number of cancers. It could also be in the type of cost you could consider. For example, if you carved out drugs and made this about ED visits and hospitalizations and avoidable use in that space, I think you'd get much stronger buy-in from many organizations.
Another way that you could think about it is to use treatment regimen as a proxy for clinical risk, and you could group treatment regimens into different categories like chemo alone, chemo and radiation, et cetera, and then use that as a risk adjustment element from claims data to try to tighten up some of the financial methodology and better match it to patient risk. Last but not least, another model that is being explored by the private sector right now is aligning care with clinical pathways. Instead of making it specifically cost based, make it pathway based and look for a certain degree of affinity with published pathways, say 80% alignment or something along those lines, so that you could look at the appropriateness of care rather than making the primary metric of interest cost.