Cancer patients often end up in the emergency room for a variety of reasons, but in many cases there are better alternatives. One of the reasons for this is the expense: At $2,032, on average, ER visits add $32 billion to the country’s annual healthcare costs, according to the National Library of Medicine.
Nearly 66% of ER visits are avoidable, and, with the shift to value-based reimbursement models, avoidable ER visits can result in financial penalties for providers at a time when 40% are at risk of closing due to the COVID-19 pandemic.
At the Center for Cancer and Blood Disorders in Texas, oncologists are using clinical AI to help patients avoid unnecessary ER trips, saving time and money for both providers and patients. The AI enables care teams to predict which patients are likely to go into the ER in the next 30 days and recommends actions oncologists and case managers can take to keep patients stable and out of emergency care altogether.
Often, the recommendations can be as simple as educating patients to alternatives. In other cases, the AI may predict that social determinants of health play a role in the patient’s risk – for instance, that they may not have access to transportation to fill an important prescription, in which case use of a mail-order pharmacy may be appropriate, or that they lack the means to afford their medication, in which case financial benefits would help the patient avoid more costly ER visits.
Dr. Ray Page, an oncologist, and president and director of research at CCBD, said that, while the main goal is to keep patients healthy and out of the healthcare system, the financial benefits are difficult to ignore.
“With legislation through the years, starting with the ACA and going forward with MACRA legislation and alternative-payment models, the methodologies for payment are gradually changing for us,” said Page. “We’ve always been under a fee-for-service payment structure, but with MACRA there’s alternative payment models where you can go from fee-for-service to a methodology of getting compensated for the total comprehensive care of the patient.
“Oncology has its own alternative payment model,” he said. “We’re responsible for the total cost of cancer care. We get compensated and rewarded based on how well we’re able to care for these patients.”
CCBD knew that, in the interest of comprehensively managing patients, it had to identify a mechanism for identifying high-risk patients through risk stratification. At the start of the oncology model, the team came up with 25 different variables to identify these patients, which resulted in the task of managing those variables and then getting them to the case managers.
Along came artificial intelligence technology from Jvion, which can perform, among many other things, risk stratification. Page knew this would be a boon for oncology, and CCBD became one of three practices to do an initial pilot.
Much like IBM Watson, AI is able to use complex algorithms to gather data from about 4,000 different variables. And it encompasses social determinants of health, which tend to identify patients at risk of adverse outcomes. That’s something that can’t always be achieved during an office visit.
Now, through a variety of factors, CCBD can pinpoint those who are high-risk and direct them to the appropriate resources. Along with the financial benefits, there has also been a positive impact on efficiency and in ameliorating physician stress.
“We all have tremendous EHR fatigue,” said Page. “We were over-the-top stressed about the administrative burdens we had, all the boxes we had to check. The doctors put me on warning: ‘If I do one more thing where I have to check one more box, I’ll go postal.'”
With AI in place, patient information gets channeled into a weekly report looking at several different vectors, including decline in condition, risk of going to the ER, risk of admission, risk of dying within 30 days, and depression. The physician receives a weekly report in the form of a one-page dashboard that includes a list of patients identified as being potentially problematic in one or more of those vectors. It also provides the reasons for why a patient may be at risk.
If someone falls into the high-risk category, they’re automatically referred to a service line that can help with their particular issue.
“It’s an opportunity to give them an epiphany, an insight,” said Page. “It gives us a moment to take pause and have an awareness that maybe there’s a particular issue going on with a patient. That goes out to teams of people to comprehensively manage these patients.”
The technology was first implemented about three years ago. During the performance period, two practices in the CCBD umbrella were able to save about $3 million in Medicare costs. And the ROI didn’t end there, as improving its benchmarks allowed the organization to receive bonuses for its pay periods.
Data from one practice detailed evidence of the improved benchmarks, and, anecdotally, Page has seen an improvement in outcomes, a better identification of depression, and an enhanced ability to keep people out of the hospital. CCBD has also increased its referrals for pain and depression management.
“Do a data dump and strengthen your communication,” Page advised other practices. “The most difficult thing practices have to think about is taking information and implementing it in service lines and in practice. You need to have strong case management and methodologies in place to direct people where they need to be. That’s the more challenging part.”
Page believes AI will continue to play an additive role in the physician’s management of a patient.
“We know in a typical doctor-patient relationship, there’s a lot of inadequacy there,” he said. “There’s no way we can cover all the variables with a patient in a 10-minute visit. To have that technology running in the background, and to integrate that into the EHR … I think it’s going to continue to develop into being a huge tool.”