Study Finds Flaw in Method for Key Oncology Value-Based Care Model

When the Enhancing Oncology Model (EOM)¹ assigns a patient to a cancer type for purposes of episode attribution and financial accountability, it relies on an administrative method: the plurality of predefined Evaluation and Management (E&M) visits. A research team from McKesson and The US Oncology Network set out to quantify something they believed was true from anecdotal reports: this approach often misses the mark.

Based on results presented at the 2026 American Society of Clinical Oncology Annual Meeting (ASCO), the gap can be substantial, particularly for patients receiving oral therapies.² The findings could be significant for oncology practices that signed up for the EOM in July 2023, and for those practices that remain in the model.

Presented by Jillian Hellmann, MSN, value-based care transformation lead for McKesson, which supports The Network, the study drew on EOM performance period data from July 2023 through June 2025 across practices from The Network. The team identified initiating episodes using Part B claims for injected agents and Part D claims for oral agents, then matched each episode to the electronic health record (EHR) by patient, drug, and date. The cancer type documented in the EHR at the time of treatment initiation was then compared against the EOM-attributed cancer type.

The differences by route of administration were striking. For injection-initiated episodes, totaling 13,803, EHR-matching rates ran approximately 90%, with only 5% lacking a treatment-based cancer diagnosis and 4.5% showing a discordant cancer type. Oral-initiated episodes told a very different story. Of the 5441 oral episodes, only about 40% matched an initiating event in the EHR, with 32% missing a cancer diagnosis entirely and 24% carrying a mismatched diagnosis.

Breast, lung, prostate, and small intestine/colorectal cancers showed particularly poor EHR concordance for oral initiations. Cancers with a higher share of oral-initiated episodes including chronic leukemia and prostate cancer, correspondingly showed the lowest overall matching rates.

In an interview with The American Journal of Managed Care® (AJMC®), Hellmann said the data largely confirmed what her team had suspected based on individual reports. "It was really a validation of what we had anecdotally observed," she said. The rich data infrastructure EOM has made available to participants—an expansion on what was offered under the Oncology Care Model (OCM)—made an analysis of this scope possible.

Significance of Findings to EOM Accuracy

Why does attribution matter? The EOM, like its predecessor, operates around episodes that are triggered when cancer treatment starts. If the start date for treatment is not properly recorded, that can throw off results.

The trouble with the EOM is that practices did not have a period to test drive the model to uncover such problems before they had to operate at full risk—a key change from the OCM. This difference is widely believed to have led to the low uptake in the EOM relative to its predecessor.³

“A key driver for the day one, ‘Should we or shouldn’t we be in the EOM?’ was really driven by our appetite for 2-sided risk on day one in a model like this,” Hellman said in the interview.

As AJMC explained in a recent look back at the 10th anniversary of the OCM, there were other differences between the 2 models: there were only 7 cancers in EOM, and the operating window between getting a bonus payment and owing Medicare money was far narrower than it was under the OCM.³

Because of lags in the way EOM data are reported, if something in the way the model is attributing a patient’s cancer type is off—and consistently so—practices would face a problem but not realize it right away. The EOM started out with 44 practices and by May 2026 was down to 28, in part because many practices that received data in the summer of 2025 could not make sense of the results.³

Can the Problem in the EOM Be Fixed?

Asked whether the problem had been raised with CMS, Hellmann could not offer specifics but praised the agency's receptiveness. "The Network has been incredibly appreciative of EOM's approach to feedback," she said, describing multiple channels the agency has opened for practices to share concerns. Just having the opportunity to present findings at ASCO, she suggested, lends credibility and context to that ongoing dialogue.

The attribution challenge is likely to worsen, the authors noted, as oral agents proliferate and patients increasingly carry concurrent malignancies. Hellmann described the core issue as one of perspective: attribution has historically been viewed through an administrative lens, and the question is whether it can be brought closer to a clinical one— ensuring that financial accountability, quality metrics, and outcome tracking attach to the correct patient cohort.¹

For this abstract, Hellman said, the immediate question is discrete: "Could we reconcile attribution with the treatment data?" But she sees broader implications for how future value-based models might handle surveillance, survivorship, and episodes outside active treatment—care moments where, she observed, significant opportunity may lie.

“The sky’s the limit,” she said.

References

1. EOM (Enhancing Oncology Model). CMS.gov. May 4, 2026. Accessed June 3, 2026. https://www.cms.gov/priorities/innovation/innovation-models/eom
2. Liu H, Mann K, He B, Hellmann J, Neeb J, Albaugh J. Rethinking cancer attribution in oncology VBC models: E&M vs treatment-based approach. J Clin Oncol. 2026;44 (suppl 16): abstr 1580. doi: 10.1200/JCO.2026.44.16_suppl.1580
3. Caffrey M. Special report: the Oncology Care Model: 10 years later. Am J Manag Care. 2026;32(Spec No. 7):SP308.