Currently Viewing:
Evidence-Based Oncology August 2019

How Data, AI Are Unlocking Secrets to Better, More Efficient Cancer Care

Allison Inserro
The Institute for Value-Based Medicine®, an initiative of The American Journal of Managed Care®, traveled to Teaneck, New Jersey, for presentations in partnership with Regional Cancer Care Associates. 
Practices taking part in the Oncology Care Model (OCM) will soon reach a point when 2-sided risk will be forced upon them,1 and innovation will be key to survival, according to a group of stakeholders who recently discussed their successes and challenges.

Innovation will hold the key to improving quality metrics, patient care, and the bottom line, healthcare executives said at a July 18, 2019, meeting of the Institute for Value Based Medicine®, a special project of The American Journal of Managed Care®. Iuliana Shapira, MD, chief medical officer of Regional Cancer Care Associates (RCCA), served as moderator and host for the meeting held in Teaneck, New Jersey, where panelists shared how they are  transforming piles of data into action items that can improve health and quality of life. Driving this transformation are new analytic tools and artificial intelligence (AI).

Barry Russo, chief executive officer of The Center for Cancer and Blood Disorders (CCBD), based in Fort Worth, Texas, described the journey the oncology practice has been on since it began working with Jvion, a prescriptive analytics company, to help it better manage the 6000 patients it sees a year.

Russo said there are 5 areas where AI can be particularly valuable: 

• Coping with a data explosion, in which medical data are projected to double every 73 days by 2020.2

• Addressing physician shortages and rising burnout

• Taming an increased number of medical images 

• Managing vulnerable populations

• Dealing with increased costs and complexity in the life sciences CCBD went through several phases as the practice worked to figure out how to best use AI, Russo said.

At the point of care, physicians are using it to riskstratify patients. “When I went through the litany of how many value-based arrangements we have, that equates to about 1000 to 1100 patients that we should be managing at any given time, meaning that they need case management oversight,” said Russo.

However, he noted that CCBD “can’t hire enough case managers to focus on all of the bodies we’re bringing into a value-based arrangement. It’s not possible, financially or even or from a people standpoint.”

In trying to determine which patients needed active management, the practice started with all patients who were in stage IV, which ended up being the incorrect move, he said. That group did not have higher costs than others. Then the practice looked at all the patients who had cancer of the lung, head and neck tumors, and pancreatic cancer, as well as those who were stage III and beyond. That group, too, was not focused enough.

CCBD realized that by participating in the OCM, they had 6 years’ worth of claims data, because practices in the OCM are responsible for the total costs of care, not just oncology costs.3 It turned out their most costly patients are cancer survivors—those who are 5 to 10 years post treatment who are still being monitored.

“They are kind of healthy,” said Russo, “except they fall. They have joint replacements. They have MIs [myocardial infarctions], and when we looked at our data, we’re like, wait a minute. We have people who are just on AI; unfortunately, they’re attributed to us because we see them for E&M [evaluation and

management] visits more often than anybody else does. And they’re costing us a fortune.”

THE LESSONS FROM JVION

Jvion ingested all of CCBD’s clinical data and mixed it with socioeconomic and other data to create a proprietary system of 7 vectors. The organization found that socioeconomic data is a huge flag that signals risk for more complex care in a patient population. Risk factors include having lower education levels, lower income, living alone, and residential instability, or not owning a home.

Data companies like Jvion buy zip code–level data from government agencies, whether it is the Census Bureau or the Department of Housing and Urban Development, or from technology companies like Amazon, which signals a person’s level of technology literacy.

“It isn’t just clinical complexity, it’s socioeconomic complexity and how you link the 2 of them together,” Russo said.

Jvion sorts patients into 7 vectors at risk for the following:

• 30-day mortality

• 30-day pain management

• 6-month depression risk

• 6-month risk for deterioration

• 30-day avoidable admission

• 30-day emergency department visit

• 90-day readmission

It also takes in historical claims data from every other provider.

CCBD soon started sending reports back to physicians based on their individual data. Some resisted the information, but others were open to the idea, Russo said. The reports offered value because they flagged which patients would never have been identified as being at risk of morbidity or mortality.

CCBD is also validating the 7 vectors internally, with varying levels of progress. For instance, the 30-day mortality risk is 35% validated, while the risk of a patient becoming an inpatient admission within 30 days is 100% validated.

Case managers can now focus on the people at the highest risk, Russo said. CCBD is still optimizing how Jvion can recommend interventions, but one thing it can do is automatically refer patients for pain management. In addition, the psychology team is proactively calling patients flagged by the system, which the practice now calls “the brain.”

Jvion also is using the system to identify patients who show up in multiple vectors, as Russo said, because “people that are at risk for more than 3 [vectors], maybe that actually would be a better way to catch the patients.”

CCBD is also identifying their own group of what Russo calls “socially challenged” patients so that social workers can make proactive calls to this group. The social workers make sure that applications for financial assistance are filed; Russo noted that the more people that file and are approved for help, the more the OCM score rises.

Another outcome: As a result of the ability to look more closely at data related to morbidity and mortality, referrals for pain management, palliative care, and hospice have all climbed.

In addition, CCBD is asking for Jvion to create a vector for inpatient fall risk; in the meantime, CCBD asks patients if they have had a fall in the past 30 days. Russo said they are “begging Jvion to move into the outpatient area, so that we can get those patients automatically referred to our pre-hab program.” This is because rehab costs are higher than those for inpatient admissions and are attributable to CCBD under the OCM.

Without AI, “there’s no way to manage these populations. We’ve got to have a better way,” said Russo.

The next steps for CCBD will be to use AI to go after issues related to electronic health records (EHRs) so that clinical, financial, molecular, pathways, costs of care, referral preferences, or requirements of accountable care organizations data are integrated and can be more than just a repository that cannot be acted upon.

The biggest issue, however, is putting all of the relevant information about a patient before the provider at the point of care, without the EHR rejecting requests and creating more work, Russo said. As an example, he pointed to UnitedHealthcare’s decision to only pay for the brand-name pegfilgrastim, Amgen’s Neulasta, and not a biosimilar as of July 1. Russo noted that Aetna uses Sandoz’s Zarxio, and other payers may insist on something else. As more oncology biosimilars enter the marketplace, this will only become more complicated for physicians as they try to enter orders, he said.

“We have to go back to the physician, start the order process again, and go back to the precert process again. We don’t have time for it. You can’t do it. Biosimilars are complicating the whole process of the point of care so much,” he said. 

USING DATA TO ENGAGE WITH EMPLOYERS, DRIVE VOLUME

When the physicians at Michiana Hematology Oncology in South Bend, Indiana, decided to get ahead of what they saw as a dawning technological  transformation in healthcare, they wanted to do so on their own. Kim Woofter, RN, who works at the practice, began using a local data analytics firm a few years ago to understand all of their disparate data sets so physicians could start acting on them.

Then South Bend awarded a $1 million grant to build their “data lake.” With that, Michiana spun that part off into a new company, the Advanced Center for Cancer Care (ACCC), which is now the data component of Michiana, said Woofter, who is now the executive vice president of strategic alliances and practice innovation at ACCC.

Michiana Hematology Oncology, a practice with 15 physicians, 9 locations, and about 4000 new patients a year, intended to use the data for its own internal purposes. But then a manufacturing firm with about 8500 employees called her and asked, “Hey, Kim, if you can look at your data, can you look at my spend data? 

While employers across the country are struggling with healthcare costs, this company was also trying to retain a skilled workforce in an area where the unemployment rate was 1.2%; their goal was to lower their costs so that they could also drastically reduce the workers’ share.

The data lake now includes data sets from oncology, employers, orthopedics, surgery centers, and multispecialty groups. The most important thing, Woofter discovered, was being totally transparent in the cost of care; she said she was shocked by the variances in payments between what self-insured employers (even her own company) paid to the network and what the network paid providers. It gave her a chance to see “both sides of the coin,” she said.

To really make a difference with employers and demonstrate value, she recommends setting up a fee schedule that bills the contracted rate. In one case, she was able to show that for a single ill employee, the employer was charged $4400 for Neulasta even though the local hospital charged $19,519. Co-pays change patterns as well, Woofter said; for example, at an outpatient center, for infusion, the copay is zero, but at a hospital, it’s $500.

“Those HR directors truly believed in their very soul that they had no control at all over the cost of healthcare that they were paying,” Woofter said. “We all agree that quality is the No. 1 denominator, but we have to marry that up with costs if we’re going to be a good steward.”

 
Copyright AJMC 2006-2019 Clinical Care Targeted Communications Group, LLC. All Rights Reserved.
x
Welcome the the new and improved AJMC.com, the premier managed market network. Tell us about yourself so that we can serve you better.
Sign Up