Illuminating data through AI, advanced analytics and machine learning
“It’s an exciting time to be a data scientist here,” said Dr. Elizabeth Parker, principal data scientist for Arkansas Blue Cross and Blue Shield. “Technology is allowing us to analyze data in fresh ways that improve the service we can offer groups and members.”
As data scientists, Parker’s team is heavily involved in the organization’s use of its cloud computing, proprietary artificial intelligence (AI) and machine learning.
“AI is really just an automation of human behavior. It makes some people nervous to talk about AI, but in reality, it’s not going to take over people’s jobs,” Parker said. “It’s going to make things way more efficient so we can make things better and allow our staff to be even more effective.
While we’re doing more with AI, it’s important to us that humans are always involved in our operations for quality control and to provide the best experience possible.”
Governance of AI at Arkansas Blue Cross
Parker said Arkansas Blue Cross requires parameters and oversight for its AI use with member and group data. Arkansas Blue Cross’ AI governance board oversees all applications of AI and machine learning within the company. Parker is a founding member.
“Our purpose is to review how we can ethically and efficiently make more effective use of AI to improve our members’ health and lower our customers’ health costs,” she said. “We see enormous potential for AI. There are ways to save time and money, to better target members with timely interventions before they experience costly complications and to free up our staff for work that only humans can do. But it takes keen oversight and active discussion to ensure our use of AI improvements are ethical and compliant.”
New regulations increasingly require documentation of precisely how AI is being used in health insurance. “That’s a good thing,” she explained. “Whenever you’re using computing to make assumptions about people, that needs governance. It needs human checks and balances, so it’s exciting that we’re building that out.”
Sky-high tech capabilities
Arkansas Blue Cross is starting to reap the benefits of another transformative technology it’s been building out for several years: cloud computing.
“We are almost fully functional in the cloud now. That puts us technologically ahead of a lot of health plans,” Parker said. “Cloud computing changes everything from a data science perspective, including how fast we can process huge amounts of data and how creatively we can apply analytics to make clinical and operational improvements.”
She explained that AI is especially good at doing tedious, repetitive tasks that used to eat up people’s time. “Now that we’re operating in the cloud, we have programmed bots to do repetitive tasks we used to do manually. It’s been wonderful to free us up to focus on more meaningful work.”
Empowering groups with self-service analytics
Another competitive differentiator for Arkansas Blue Cross is the launch of Illume, a self-service analytics portal employers can use for anytime-access to their health plan’s reporting and to run hundreds of custom reports. Besides being highly configurable and accessible 24/7, the new reporting platform can go beyond providing results. In a few clicks, users can download their reports into PowerPoint-ready data visualizations that come with key insights extracted and highlighted. Parker said it’s proven highly popular with groups.
“I’m very confident when I say our reporting platform is a leader in the industry. What we’re providing today through Illume is superior to anything our competitors are doing, without question,” Parker said. “The accessibility, intuitive design and customizability for our customers is unlike anything else on the market.” She added that a future use of Illume is to deliver custom modeling and analytics from her team directly into a large group’s Illume dashboard.
Combining clinical predictive modeling with prescriptive interventions
Parker’s data science team also developed an innovative predictive risk model that allows the organization to focus case management interventions on those who have the most need. The model can pinpoint not just who has the highest risk today, but which members have the highest predicted risk over the next 12 months. It flags combinations of health concerns that are likely to lead to future problems, helping case managers identify members who most need immediate help.
“If our nurses and social workers can reach them sooner, we can often help them avoid more costly complications and connect them to resources that fill gaps,” Parker said. She said the data can be shared with a group’s partnering health solution vendors that focus on areas like diabetes management or maternity care. “Getting those insights into which members would most benefit from their help means they can be more effective at reaching them.”
The model Parker’s team developed, called Tapestry, has endless applications. Some of them focus on determining not just who would benefit from interventions, but what interventions are likely to be the most effective for them. Parker said her team has been working on classifying different segments of membership into identifiable “buckets” the algorithms can sort into profiles, so case managers can make educated, informed assumptions based on shared characteristics.
For example, Tapestry could identify a 71-year-old member as being in a remote area, having a high school education and an average of three comorbid conditions. Based on that data, that member may need additional care but not have easy access to a hospital within 30 miles and may not have broadband. In that case, the Arkansas Blue Cross case managers would know that offering telehealth services probably isn’t the best approach for that member, Parker explained, and they can instead work to connect them to more local resources and telephonic help.
Analyzing what’s driving costs
Other applications of data science are operational, analyzing claims to understand trends. For example, Parker said, when groups see their medical costs rising, they often want to know why.
“Is it because the same medical expenses now cost more, or more members are getting sick, or is it mainly from a few participants with very expensive illnesses? We can look at the data and see.” Evaluating all the data that comes in and drilling down into it to figure out why costs are rising is often the first step to the account team strategizing how to offset those costs.
Empowering better living
Parker said she works with the best people, during an exciting time for her field.
“Our team is amazing. I could brag on them all day,” she said, adding that in September one of her team’s data models received national recognition from an industry organization.
“We really are putting our best minds on ways to help people live healthier lives,” she said. “I understand why people get nervous about how their data is being used,” Parker said, “but the intent behind anything that comes out of my team’s shop is to increase our members’ quality of life equitably and increase equity through these models.
“Our entire purpose is to meet those needs and identify those people who might not have adequate access to healthcare, to find out where we can find assistance for people who need it. But in order to do that, it takes a lot of data, a lot of information. My team’s job is parsing and understanding that data so that others can act on it.”