Designing Effective Value-Based Oncology Models: Lalan Wilfong, MD

The success of value-based care models in oncology hinges on getting the underlying assumptions right from the outset, according to Lalan Wilfong, MD, chief medical officer at Navista. Even small nuances or variations in those assumptions can significantly skew a model's outcomes, he said, so designers must deeply understand what a model measures and how it measures it to ensure practices are held accountable for the right things.

Wilfong moderated the session “Are You Just Playing With the Numbers?” at the Community Oncology Alliance Payer Exchange & Innovation Summit earlier today, on September 15. Alongside panelists Jason Altieri, ASA, MAAA, and Andrew Mackenzie, FSA, CERA, MAAA, they explored how actuarial methods, risk adjustment, cohort design, and model decisions influence performance, along with where optimization ends and "playing with the numbers" begins. 

A central theme Wilfong highlighted was cohort design. Outcomes are only as good as the patient population fed into the model, he noted.

“Cohort designs influence the outcomes of value-based care models, because what you put in the funnel is what you get out of the funnel at the end of the day,” Wilfong said.

Therefore, cohorts that are too narrow exclude patients whose variation could affect financial or quality performance, whereas cohorts that are too broad risk comparing dissimilar patients, or "apples to oranges," as Wilfong put it. As an example, he cited early lessons from the Oncology Care Model roughly a decade ago, when HER2-positive and HER2-negative breast cancer were initially bundled together despite meaningfully different costs and outcomes; this grouping could distort how a practice appeared to perform.

Wilfong also explored the tension between legitimate model optimization and manipulating the numbers. While clinical and patient-specific factors genuinely warrant adjustment, he explained that organizations can exploit the same flexibility by using risk adjustment for comorbidities, strategic cohort placement, or shifting drugs from one billing category to another to make the numbers seem better. Balancing fair accountability with resistance to gaming is a "chicken and egg" problem, Wilfong said, requiring careful thought about measurement, benchmarking, and adjustment methods.

Lastly, he turned to the broader challenge of keeping value-based oncology models transparent and effective amid rapid clinical change, as new molecular diagnostics, drugs, and treatment algorithms can render a model's assumptions outdated almost as soon as it is implemented. Wilfong said he did not know the right answer to this challenge, describing oncology today as both exciting and difficult: major advances in patient care are paired with model design struggles to keep pace with how quickly the standard of care is evolving.

“Anyone will tell you that oncology is difficult right now, fascinating right now because [of] how much better we're taking care of patients and managing them with all the advancements that are occurring, but challenging, then, to think about what we should be doing,” he concluded.