Letter to CMMI Detailing OCM Concerns

Author: Community Oncology Alliance (COA)

March 16, 2018

Anand Shah, MD
Chief Medical Officer, Center for Medicare & Medicaid Innovation Centers for Medicare & Medicaid Services
7500 Security Boulevard
Baltimore, MD 21244

Dear Dr. Shah:

On behalf of the Community Oncology Alliance (COA), we are submitting our concerns regarding the Oncology Care Model (OCM) to the leadership at the Center for Medicare & Medicaid Innovation (CMMI). As you know, COA is the leading cancer organization championing the OCM as witnessed by the significant resources we have dedicated to the effort and by having over 80% of the OCM participants (accounting for an estimated 90+% of the OCM patients) in a cooperative learning and information exchange network. We want to underscore the substantial commitment we are making to ensure the success of the OCM – a success that we believe is very much in doubt.

With that said, now that the first Reconciliation Reports have been released, we have some pressing

and key concerns, summarized as follows:

  1. The OCM in its current form is methodologically flawed with respect to predicted episode pricing, including significant deficits related to:
    1. Risk adjustment for breast, prostate, and bladder cancers;
    2. attribution and MEOS claims submission; and the
    3. approach towards novel therapies
  2. Complexity in attribution and delays in receiving data regarding attribution are leading to major financial, operational, and clinical issues for participants. These issues are heading towards large recoupment amounts that will need to be paid back to the Centers for Medicare and Medicaid Services (CMS) after the first true-up of MEOS claims, skewed reconciliation results, and problems in quality measurement and in the data reported to the OCM clinical registry. Planning for MEOS recoupments is especially creating financial hardships for participants given cash flow dynamics, complicating continued participation for practices.
  3. Greater flexibility is needed regarding allowed timeframes given the complexity of the program, notably related to attribution and contestation submission. A high-priority situation where greater flexibility is needed is the 12-month window for revising submitted MEOS claims, particularly due to the extended length of time before participants received their official attribution lists and attribution related data.
  4. The approach towards novel therapies requires special, immediate attention and modification. The current methodology relating to novel therapies opens up the risk of creating perverse incentives for using inferior drug treatments that could adversely impact patient care. This is especially the case given other issues related to episode pricing. We note that we are r any circumstances, about any incentives or pressures to lower costs by forcing the use of clearly inferior treatments.
  5. Total transparency by CMMI is essential to enable full reproduction of the OCM methodology by participants. This includes clear disclosure of all weights and variables needed to reproduce the methodology in full. Confusion about important variables, such as the trend factor and inconsistencies in implementation of the novel therapy adjustment, are creating an environment of chaos, which will ultimately result in model failure. In short, how can practices attempt to increase quality and cost efficiencies if they do not understand the “grading” system?
  6. For the performance multiplier related to OCM measures one (1) to three (3), the distribution of cut points separating levels of performance is very narrow. While these cut points are empirically motivated, the limited range introduces disconcerting sources of error and sensitivity to random effects.
  7. Geographic adjustments to episode costs are done at the level of the patient, rather than where services are provided. This distorts episode costs for patients who receive cancer care in a different Hospital Referral Region (HRR) than their primary residence.
  8. Learning communities created by the OCM vendor have to date been far too one-way only, with very little transparency and two-way communication.

On behalf of the COA Board of Directors, our internal OCM/OCM 2.0 team, our Oncology Payment Reform Committee, and the OCM participants in our network, we cannot underscore enough our very serious concern about the viability of the OCM. On its current path, the OCM is simply not a realistic, viable model for all care Part B oncology payments, which is the underlying purpose of implementing this demonstration project.

We sent you a memo (attached) dated February 22, 2018 about our concerns relating to novel therapies. Now, with the release of the Reconciliation Reports, we are convinced that the time for very serious modifications to the OCM are long overdue. We know of several OCM participants considering dropping out of the program right now. So many participants are frustrated and bordering on angry. We do not want this to become a watershed event.

We welcome an opportunity to have a much more detailed conversation about our concerns. Please note that we have resisted making any comments to the press, now that the Reconciliation Reports have been released, in the spirit of cooperation and, hopefully, discussion with CMMI.

What follows are some more detailed points relating to the issues/problems we summarized above.

Methodological Issues in Price Prediction Related to Risk Adjustment

Since Reconciliation Reports were released, we have become aware that a number of participants, particularly large ones, did not achieve any performance-based payment (PBP). We are deeply concerned that the delay in implementing the changes in prostate and bladder cancer risk adjustments (these known issues were discovered early on but were not applied until the third (3rd) performance period), as well as growing concerns around risk adjustment in breast cancer (as outlined to you in the attached memo), are significant contributors to this failure to achieve savings.

We note that an informal survey of eighty-six (86) OCM participants, including the largest OCM practices/facilities, found that only sixteen (16) achieved any positive savings. Although, this is sixteen percent (16%) of the participants, our rough estimate is that it is only about ten percent (10%) of the activity as measured by patients. Additionally, some of the most active and successful practices that have participated in other Medicare (COME HOME) and private insurer (e.g. Aetna, United Healthcare, Priority, others) models/pilots actually performed very poorly in terms of OCM savings.

The risk adjustment issues are causing the inappropriate loss of potential shared savings and are the foundational reasons for why: 1) so few participants achieved shared savings; and 2) those practices that did achieve shared savings were under target by comparatively small sums when compared to how far over target most practices were.

In addition, flaws in the risk adjustment models for the cancers referenced above are leading to such large-scale problems that it is likely that all of the coefficient weights in the OCM prediction model for episode costs are adversely affected, diminishing the validity of pricing globally. Together, these issues all but preclude making meaningful inferences about participant performance in performance period one (1).

We strongly suggest that the known issues in prostate and bladder cancer and errors in the breast cancer risk adjustment be addressed immediately. Financial models used to predict episode costs from the baseline period and performance period one (1) should be rerun for more accurate weight variables. Reconciliation results for performance period one (1) should be rerun to reflect the corrected risk adjustment. It simply is not consistent with clinical medicine to label existing patients as low risk when that is not the case.

These significant problems in pricing, their impact on success in the program, and the ambiguity they create in understanding actual performance means that participants will have fewer opportunities than initially anticipated to learn how to substantively improve performance from the first reconciliation results. This need for adjusting time frames is further emphasized by a statement in the Participation Agreement: “If the Practice is a Non-Pooled OCM Participant, CMS may terminate this Agreement if the Practice does not earn a PBP by the time of initial Reconciliation of the fourth Performance Period.” OCM participants are currently in the fourth (4th) performance period and are frantically trying to understand and respond to these reports. There are only three and a half (3.5) months remaining in the fourth (4th) performance period. The ability and time realistically and practically to respond to these reports in order to implement practice changes that can lead to a positive PBP is very limited, at best. In addition, CMS will need time to review these recommendations, resolve the known faults, and modify the program accordingly. Therefore, we strongly recommend that CMS not remove par CM until after they receive their Reconciliation Reports for the sixth (6th) performance period.

Attribution and MEOS Claims Submission, Quality Measurement, and Clinical Data Submission

Numerous challenges confront OCM participants related to attribution, particularly related to oral triggers. Many practices face large recoupments of MEOS funds related to complications of attribution. Gaps in attribution are likely leading to lost shared savings as delays in beneficiary identification diminishes the opportunity for care management. Recoupments create financial stressors, especially with respect to cash flow and availability of funds.

These issues are heightened by the lengthy delays prior to practices receiving official attribution lists. The burden is further magnified by the twelve (12) month time window available to modify MEOS claims. Given how long it has taken C MI to deliver official attribution reports, inadequate time is available to modify MEOS claims fully when needed.

CMMI ideally should provide definitive attribution lists early on in performance periods. If that is not feasible, at a minimum, the twelve (12) month MEOS modification period should be extended and the OCM claims files should be provided to OC an project attribution accurately, particularly for patients on orals.

Novel Therapies

In addition to the issues raised in our recent memo to you (attached to this letter), we are also concerned about problems in the implementation of the novel therapy adjustment, particularly related to the very high threshold required before the novel therapy adjustment comes to the fore, and then only at a rate of eighty percent (80%), rather than full cost adjustment. Given the limited variance at the upper end among many OCM participants, the concern holds that misaligned incentives may emerge that could adversely impact patient care. Again, greater transparency and clarification is essential.

Quality Measures

The three (3) most prominent PBP measures around risk adjusted hospitalizations, risk adjusted ED visits, and hospice have very narrow cut points between quartiles, creating a disproportionate impact on results. In addition, the first two (2) risk adjustment measures have a separate risk adjustment methodology from the episode pricing, which creates more confusion and methodological misalignment. We strongly suggest more granularity in measurement, moving away from a quartile met risk adjustment approaches to be more explicit and easier to understand.

Enhancing Transparency

We believe that the OCM team is doing its best given the complexity of one of the first ambulatory specialty payment models that CMMI has fielded. However, greater transparency is clearly needed, especially relating to the specific weights and variables essential to OCM participants understanding the model, their performance, and how CMMI arrived at its results. Given the complexity of OCM as a model, being able to simulate the methodology in full is critical to deepening understanding of the model by making payment model design more concrete.

Additional light is especially needed regarding the trend factor and experience adjustments so that practices can better understand how these issues impact their performance. Lack of transparency in this area is especially troubling. We’ve counted four (4) distinct sets of coefficients that are used in the OCM prediction model for different analyses, but OCM has only published one set: the coefficients that, together with the experience adjustment, constitute an episode’s baseline price. At least three (3) sets of coefficients have not been made available, which go into the calculation of the trend factor and experience adjustors themselves. The variance around the weighting for HRRs is another example of unnecessary opacity. Our team noted that in the last two (2) OCM learning sessions and webinars there were requests for such factors to be revealed, but the answers were not provided and to date no follow up has been acknowledged.

In a similar vein, releasing OCM claims data at a faster rate – monthly rather than quarterly – is critical to deepening understanding of the model and increasing the probability of participants reaching shared savings. Further, there are currently significant gaps between the patients included in quarterly feedback data and OCM-eligible patients that participants include in their data submissions to CMMI. Giving participants further information about patients who they are treating is necessary to improve attribution by CMMI and patient identification by practices.

Our extensive work with these models on the private insurer side has revealed key learnings. Practices need to clearly understand how they are being graded – what they are doing right and what needs to be improved. After all, the OCM should not be focused on an operational grade. It should be about practice transformation. That starts with a focus on the patient and changing clinical and operational processes to deliver the highest quality cancer care in the most cost-efficient manner. That change, and continuous improvement, is only possible with timely, accurate, and meaningful feedback, including “grading” that is totally transparent and understandable. Unfortunately, the OCM as it is currently being implemented is sorely lacking along those lines.

Geographic Adjustments

In an effort to adjust prices to reflect differences in health care costs in different markets, the OCM prediction model uses weights associated with each HRR. However, episodes are assigned to the patient’s primary residence zip code, which may not coincide with the primary point of service. We urge OCM to adjust the model so that prices reflect the HRR for the attributed practice/facility, which would more accurately reflect differences in the cost of care for patients who travel or maintain multiple residences. This change would be easy to implement and would improve precision in controlling for geographic disparity of health care costs.

Communication

The OCM team is trying to be responsive and we sincerely appreciate the interaction with the team, especially as led by Lara Strawbridge. However, there are so few opportunities to engage with the team in real time that it reinforces a pervasive fear that nothing will change from a program perspective. Additionally, there is also a palpable frustration that care transformation, while costly and labor-intensive, does not correlate at all to PBP savings. As stated previously, the most experienced practices with oncology payment reform models — that have truly and dramatically transformed their practices – are among the least successful and most frustrated. On that front, our team is very concerned that the weighting of these practices essentially competing against themselves is an exercise in futility given how much they have already transformed their practices.

It is our desire to work closely with CMMI on solutions, but to do so we must move beyond webinars that are too crowded for real dialogue and advance into serious and immediate policy changes that can improve the program. On its current course, we are concerned that the OCM will fail, bot ect and as a model that can realistically transform the Medicare Part B payment system.

Our recommendation is to convene a meeting ASAP with our team and a s participants. We now have arrived at a serious and critical juncture. Real action is needed!

Sincerely,

Bruce Gould
Chair, Oncology Payment Reform Committee

Ted Okon
Executive Director