Confronting the Variance in Cancer Care Cost: An Interview With Donald M. Berwick, MD, MPP

Spending more to care for one patient than for another, even when those patients have similar clinical characteristics, is among the widely acknowledged cost drivers in health care.^1,2 Both CMS and commercial payers tackled this problem in cancer care with alternative payment models, including the Oncology Care Model (OCM); these models put a spotlight on outlier cases or identify physicians who stray too often from recognized pathways.^3,4

Now, new research has shown how much the cost of cancer care can vary within a multihospital system, something that must be avoided as health systems expand. In December, authors associated with the health care technology company COTA, Inc, published results that used the digital classification system the COTA Nodal Address (CNA),⁵ to reveal large differences in total cost of care for common cancers.⁶ Each CNA phenotype is developed through a peer-review process and includes evidence-based patient elements and disease-specific parameters.

For this study, authors used data from 4032 patients across the 16-hospital Hackensack Meridian Health system from January 2013 to January 2016. Costs were based on schedules from Medicare and Horizon Blue Cross Blue Shield of New Jersey—the state’s largest insurer—and validated against other insurers. Cancer cases with the same CNA have the same clinical characteristics, and results showed that cases with the same CNA had the following ranges:

• Breast cancer care costs varied from $16,000 to $88,000
• Colorectal cancer care costs varied from $25,000 to $222,000
• Lung cancer care costs varied from $44,000 to $424,000

Variation in cost among cancers with the same CNA was seen in both academic and community-based settings, and the use of clinical pathways did not eliminate variation. Results showed that the less common the cancer, the more variation increased.

Findings also show that the greatest opportunities for improvement were (1) making sure patients get on the right clinical pathway at the outset, based on comorbidities and genetic profile, and (2) eliminating unnecessary ancillary care.

Do the findings bolster the case for payer coverage of genetic or genomic testing? In an email, the paper’s lead author, Andrew L. Pecora, MD, FACP, CPE, a founder of COTA and chief innovation officer, professor and vice president of cancer services at Hackensack Meridian Health and John Theurer Cancer Center, wrote: “We must know exactly what the patient is from a disease perspective. There can be no compromise on this, so genetic testing to assign the correct CNA is a must.”

For insight into the results, Medical Economics (ME) spoke with the paper’s senior author, Donald M. Berwick, MD, MPP, the former CMS administrator who is president emeritus and senior fellow at the Institute for Healthcare Improvement in Cambridge, Massachusetts. Berwick, a member of the faculty at Harvard Medical School, is well known for coining the term, the “triple aim,” which refers to simultaneously improving the health of populations and the health care experience of patients while lowering the cost of care^.7 The following has been edited for length and clarity.

Medical Economics: Eliminating unnecessary variation in care is at the heart of what we talk about in value-based care. And yet, the argument is made that cancer care is different. Often, there’s not a less expensive version of “the best” therapy. How do your findings on the use of the COTA Nodal Address speak to this issue?

Berwick: I do think we want oncologists and other clinicians to be able to make adjustments in individualized patient care that they feel are needed. We need to trust the professional discipline and scientific integrity of the clinician. None of [these findings] should be converted into handcuffs or undue restrictions on clinicians. However, by studying variation, we can reveal variability in practice that they may not know about—and that can open the door to inquiry. We can imagine people saying: “Why do you do it that way? This is the way I do it.” And we open exploration that’s otherwise not possible, if we haven’t turned the lights on to that variation.

Second, if there is justification for the variation based on underlying scientific evidence, and there are several different evidence-based pathways to care, but one of them consumes fewer resources or is easier to implement for patients and families and clinicians—all other things being equal—we should know that. If the claim is the outcomes are the same, then why don’t we choose the approach that is less resource intense?

Third, it’s good to link the results, the resource use—which the COTA system allows—to outcomes. And although this is associational evidence, not experimental evidence, it still can be suggestive. If we begin to notice that certain channels or certain pathways of care seem to be associated with better outcomes for patients, we should notice that and use that to induce more disciplined experimental research. So, there’s a lot of different ways that this can be put to use.

ME: What accounts for so much differentiation in costs for the different pathways?

Berwick: It’s intensity of resource use. Basically, there are habits or patterns of utilization that do more testing or intervention, use a different assortment of medications, different frequency of visits. In general, we’re looking at different intensity of patterns and resources based on clinical habits and local procedures.

We know from prior research [that] when I’m in a community, my practice tends to look like others in that community—even if it’s very different from people in other communities, so we have these different local habits. Clinicians tend to stay local and learn from each other.

In oncology, that’s less true; the oncology community is very organized together, using national and international data, but these local work patterns and habits still really do matter. And a lot of the variation we’re seeing reflect these local behaviors. Dr [John] Wennberg called it the surgical signature, which means it was the characteristic of the neighborhood.⁸

ME: The findings show that there’s less variation in cost in the more commonly seen cancers, such as breast or lung cancer, and more variation in the rare cancers—the rarer the cancer, the more variation we see. Is there a case to be made for greater data sharing in these rare cancers, so that we can close this gap?

Berwick: Absolutely, yes. First, the phenomenon could be partly statistical, because the higher the sample size in a particular cell, just statistically, the less variation [we will see] among the entries in that cell. That’s just a statistical fact. We would expect more variation in the lower sample size cells in the COTA system. That said, another important factor is scientific certainty: The less experience I have with the management of particular cancer, the more chance there is that variability and knowledge will begin to operate. And I’ll be developing patterns that aren’t based on large data sets. So yes, the ability to aggregate data for small or smaller incidents, and smaller prevalence, cancers or any disease, is a real asset. It’s like what we discussed in the first question: It’s a chance to raise some questions that otherwise couldn’t be raised because we wouldn’t have the data.

ME: We often hear the about the need to balance too much variation with the need to treat the patient in front of me, kind of going back to the first question. What did the CNA findings say about the need to strike this balance?

Berwick: Overall, the CNA findings suggest we are out of balance right now. Because I’ve been in this research field for so long, the amount of variation is not surprising, but it doesn’t look ideal. I mean, this amount of variability in the care of cancers that are relatively common, for example, isn’t scientifically easily defensible. So, number one, I think it does raise some questions as to whether the oncologic community and clinicians in general need to reconsider their commitments to simply using the best practice they can find.

Second, there is the need to adjust care to the local circumstance, to every individual patient. At the Institute for Healthcare Improvement, we always say, we need to practice medicine not only about what’s the matter with you, but what matters to you.

Even if a person has the identical form of breast cancer or colon cancer, the individual’s needs and preferences and what matters to the patient must be honored. That’s going to be sensed locally, and we would expect that as clinicians adjust to individual needs, a very important and valuable form of variation will (and should) enter into the picture.

Also, although the COTA Nodal Address uses a lot of variables—and this is a data-intense way to explore patient variation (variation at the patient level)—it doesn’t cover absolutely every variable that may be known. I suspect that some variation occurs in response to knowable characteristics about individual patients that don’t happen to be variables in the in the coded data set. But knowing the size of the COTA data set and the number of things that are being that are being explored, [this] should raise questions for any clinicians who notices that they are outliers with respect to their peers.

ME: So what are the next steps? Where are we and where do we go from here for closing the variation gaps?

Berwick: I’d say 2 to 3 things. One is the power of the…COTA Nodal Address system. Any large registry grows with the size of the database, and so the aspiration would be that the amount of information in the database grows as fast as possible because the more data we have, the faster we can learn. So a growth of use is my hope.

A second would be partnership with the oncologic community—your readers. This should not be felt to be surveillance or enforcement or anything close to regulatory at this point. What’s important is partnerships so that the curiosity and the scientific bias of the scientific intent of clinicians get raised. I hope that the COTA system can become better known to the oncologic community, so they can make use of it and ask the questions they probably want to be able to ask.

Third, there’s probably a patient side of this—that patients deserve voice and control over their lives and care. There are, of course, the mathematical characteristics of the CNA that are a little complicated, but they’re not insanely complicated. And I think patients and families could come to understand the COTA system as a way to illuminate variation in their own care and raise questions about, “Well, why do you do it that way, as opposed to this way?”

There may be very important uses for research. There’s an obvious potential nexus between the research community and the COTA system, as COTA begins to reveal unexpected variability that should raise the curiosity of researchers, who could say, “We’re doing this 4 different ways. Which is the best?” I think it could induce some serious research that would be to the benefit of patients.

On this cost side, the amount of variation observed does suggest that very likely a lot of money is being wasted. And it’s not just money, it’s time—of patients and families and clinicians. And it’s also out-of-pocket costs, because a lot of the health care costs even in the insured population are coming out of patients’ pockets. If it turns out that there’s variation in resource use, but not in outcomes, that should give us an opportunity to actually look at ways of doing what we should do, which is to reduce health care costs without harming anyone—in fact, making things better.

Disclosure

Berwick has an equity interest in COTA, Inc.

References

1. Diaz A, Paredes AZ, Hyer JM, Pawlik TM. Variation in value among hospitals performing complex cancer operations. Surgery. 2020;168(1):106-112. doi:10.1016/j.surg.2020.03.010

2. Ellis P, Sandy LG, Larson AJ, Stevens SL. Wide variation in episode costs within a commercially insured population highlights potential to improve the efficiency of care. Health Aff. 2012;31(9):2084-2093. doi: 10.1377/hlthaff.2012.0361

3. Oncology Care Model. CMS website. Updated September 14, 2020. Accessed January 10, 2021. https://innovation.cms.gov/innovation-models/oncology-care

4. Baumgardner J, Shahabi A, Zacker C, Lakdawalla D. Cost variation and savings opportunities in the Oncology Care Model. Am J Manag Care. 2018;24(12):618-623.

5. Pecora AL, Norden AD, Hervey J, et al. Development of a precise, clinically relevant, digital classification schema for cancer. JCO Clin Cancer Inform. 2018;2:2-10. doi:10.1200/CCI.18.00006

6. Pecora AL, Ip A, Wang CK, et al. The COTA Nodal Address model: a novel digital classification system identifies variances in cancer care cost to support value based care delivery. J Prec Med. 2020;6(4): https://www.thejournalofprecisionmedicine.com/the-cota-nodal-address-model-a-novel-digital-classification-system-identifies-variances-in-cancer-care-cost-to-support-value-based-care-delivery/

7. Berwick D, Nolan TW, Whittington J. The triple aim: care, health, cost. Health Aff. 2008; 27(3):759-769. doi:10.1377/hlthaff.27.3.759

8. Weinstein JN, Bronner KK, Morgan TS, Wennberg JE. Trends and geographic variations in major surgery for degenerative diseases of the hip, knee, and spine. Health Aff. 2004;23(suppl 2):VAR-81-VAR-89. doi:10.1377/hlthaff.var.81