Predicting patients at highest risk
But, given that some cancer readmissions are unavoidable, how might healthcare organizations better predict who is at the highest risk of a readmission?
Carl Schmidt, MD, a surgical oncologist formerly at Ohio State University College of Medicine who is now based at West Virginia University Cancer Institute, hoped to develop a model that could estimate the risk of readmission for individual patients. With such a model in place, oncologists could better determine where to best put care management resources after discharge.
“In our original analysis, we estimated that probably about 1 in 5 readmissions was preventable,” he says. “So, in that 20%, there may be things we can do to stop them from happening. Those patients may just need more interventions. But since some of those interventions are costly, you don’t want to give them to everyone. You want to figure out who really needs them.”
He and a team of colleagues developed a logistic regression model using high-risk factors including abnormal sodium levels, low white blood cell count, solid malignancy, and emergency department visits, among others. When they validated the model, it could reliably predict which patients would go on to be readmitted with about 70% accuracy. He believes future models could use artificial intelligence (AI) tools to improve that percentage.
“Our model was pretty simple, but it still allowed us to flag patients who might need some additional support,” he says. “But healthcare organizations, particularly academic centers with the resources, could come up with a better predictive model using neural networking or other AI methods that offer more precise predictions.”
Finding the right path forward
Justin Bekelman, MD, a radiation oncologist at the University of Pennsylvania’s Perelman School of Medicine who has studied how to be reduce unplanned acute care for patients with cancer, including hospital readmissions, says there is a huge opportunity for healthcare stakeholders to think creatively about how address cancer readmissions. He agrees with Schmidt that predictive analytics will likely play an integral role.
“Today, there are not truly validated predictive analytics that can identify patients who are at the highest risk of readmission,” he says. “But in addition, healthcare organizations can also use those kind of big data approaches to simple find ways to dramatically improve the care of patients with cancer—that will reduce readmissions, too.”
Bekelman, as well as Schmidt and Khorana, all argue that addressing this issue isn’t something that healthcare organizations can do alone. There is a role for payers, too—either by funding larger scale research efforts or by providing predictive models and tools for their provider partners to use to improve care for this patient population. Jessica Saba, PharmD, director of Value Based and Population Health at Highmark, Inc., a large Blue Cross Blue Shield plan serving Pennsylvania, West Virginia, and Delaware, says it’s certainly something that health plans like hers are actively working on.
“Cancer care, traditionally, represents quite a large bit of healthcare spend—and there are many direct and indirect ways to try to manage that,” she says. “And, certainly, analytics is an emerging area for Highmark and we are looking at ways that data can help. We are working quite hard to provide more tools and insights to our physician partners so they can more effectively manage these cases.”
For his part, Schmidt says he is buoyed by the amount of academic research being done on this issue. And he hopes that healthcare organizations, both of the provider and payer variety, are not only paying close attention to the studies—but looking for ways that they can get involved to help.
“This is about a lot more than just readmission numbers,” he says. “Trying to reduce them in payer contracts or with quality metrics that say something like, ‘Oh, you need to drop your readmission rate from 14% to 13.2%,’ isn’t going to work. Healthcare stakeholders need to collaborate and come up with better, smarter models so we can understand and undertake evidence-based measures that will provide the best quality care for our patients—and, with that, reduce readmissions along the way.”
Kayt Sukel is a science and health writer based outside Houston.