Charting a new path to help cancer patients

The Wayfinding project, funded by the TraCS Innovation to Impact Award program, wants to prevent emergencies before they happen—and demonstrate a new method for outcome-oriented research.

When someone is diagnosed with cancer, that's often just the beginning of their time in the medical system. For many cancer patients, the months to come involve long and intensive treatments like chemotherapy, which come with plenty of unpleasant side effects. Some of these side effects, like dehydration or infection, can be serious if left untreated, causing patients to end up in the emergency room.

But if healthcare providers could somehow know ahead of time which patients were most at risk for these health emergencies, they could make sure those patients received extra care and specialized attention. And by doing so, doctors and nurses might even be able to prevent those health emergencies from ever taking place—well before a patient's health took a turn for the worse.

One way of predicting which patients might be at risk is through prognostic modeling, using statistics to see which traits (like age or pre-existing medical conditions) are associated with a higher risk of health emergencies. But many of these statistical models aren't really useful for healthcare providers. Maybe, for example, the traits associated with emergencies aren't diagnosable until months into a patient's chemotherapy treatment. Or maybe there's no infrastructure available for doctors to screen for patients with these traits.

So a new project led by Jennifer Elston Lafata, a health services researcher at the UNC Eshelman School of Pharmacy, and Jake Stein, an oncologist at the UNC School of Medicine, is now trying to create a new kind of prognostic model. The project, funded by an Innovation to Impact Award at the North Carolina Translational and Clinical and Sciences (NC TraCS) Institute, is combining statistical modeling with insight from healthcare providers and cancer care teams through a new "wayfinding" process.

With this project, the researchers hope to create a robust and comprehensive model that clinicians can use to assess a cancer patient's risk of health emergencies and take action to reduce that risk. And by tackling this question with their new "wayfinding" approach, the team is also hoping to pioneer a new way of looking at translational science—charting a path toward more functional, informed, and human-centric medical research.

"What we're trying to do," Lafata says, "is move beyond the type of models that sit on a shelf, and toward something that can actually be used."

With some basic statistics and the right data, anyone can create a model that predicts how a patient might fare during their cancer treatment. That's the easy part. But for people working on the ground with patients, these numbers often mean very little from a practical standpoint.

Maybe, for example, a model shows that patients who live in places with more air pollution are at a higher risk of health emergencies. But in many hospitals, there may be no way to identify who lives in high-pollution areas. Or maybe patients who experience frequent nausea during their treatment are at a higher risk of becoming severely dehydrated, but there's no system for doctors to check in on patient symptoms and understand who is frequently nauseated.

In addition, these statistical models don't always take into account the non-medical influences on how people navigate the healthcare system. "You have to account for the messiness of the real world, right?" says Stein. "Thinking about payment, transportation, social work, parking—these are things that make a big difference in people's lives."

Lafata and Stein aimed to capture some of these nuances through the new wayfinding process. Their project started like any other model would—throwing a bunch of data together with some statistics and seeing what popped out. But this time, that was just the beginning. The next step was taking that model to a panel of healthcare providers to ask: "What should we change to make this model more useful to you?"

Using that feedback, the team could then update the model to better reflect the reality of patient and provider needs. But they didn't just stop there. Instead, the wayfinding process involves getting new feedback and updating the model repeatedly, slowly fine-tuning it into a powerful tool that can both determine a patient's risk and empower providers to help those patients, all in real-time.

Stein says that this project would not have been possible without support from NC TraCS. The Innovation to Impact Award is designed to provide researchers with the long-term funding and flexibility needed to address a significant translational science roadblock, with the support of NC TraCS.

And the team's wayfinding process has led to some interesting discoveries. Originally, Stein says, the team planned to have the model run on the day a patient began treatment. But after discussing this with healthcare providers, they realized that's actually too late—the doctors, nurses, and navigators need to know whether or not a patient is at high risk for an emergency at least two weeks before their treatment begins. That way, they're able to alter the treatment plan or provide that patient with extra resources as needed.

"Whether you're changing the pre-medications, setting them up for more fluids, setting up some support environment at home like somebody to check in on them, or addressing transportation challenges—whatever the case may be, you actually need to know about it before day one," Stein says.

This back-and-forth process can also change how the model is created and deployed. For instance, Lafata notes that instead of simply estimating the risk of health emergencies for each patient at the beginning, their model now calculates a continuously updated risk score every time there's new information about that patient, such as when they start a new medication or get new lab results.

After plenty of tweaking, the project team has a workable model to predict the risk of health emergencies in cancer patients. And now, they plan to test it with a group of patients who have gastrointestinal, blood, or thoracic cancer. The team will run their model on this group to identify which patients are most at risk. This way, clinicians can provide them with extra resources or care to try and prevent any health emergencies. The team hopes that at the end of their study, they'll be able to demonstrate a reduction in the overall frequency of healthcare emergencies among these patients.

Lafata and Stein are keen to point out that their model is only meant to provide patients with additional care, never to take care away. The goal is to make sure that no one slips through the cracks of the healthcare system without the care they need, and that providers can spot people at high risk who might have otherwise gone unnoticed.

Through their wayfinding process, Lafata and Stein are demonstrating a new way of thinking about translational science research. To be sure, their decision to incorporate feedback on their model before they ever tested it meant they had to spend more time working on that model before it went out into the world. But the goal of medical research isn't to simply get a paper published quickly. The goal is to develop new ways of keeping people healthy, to empower doctors and nurses to support their patients, and to make sure no one misses out on the care they need.

By reflecting on these outcomes at every step along the way, the wayfinding process keeps the needs of real people—patients, providers, caregivers, loved ones—at the very forefront of the medical research process.