Even as health plan leaders work more collaboratively together with provider leaders than ever before in order to lower costs and improve outcomes, challenges of all kinds continue to pose obstacles to progress. One health plan that has made major progress in this area is the Boise-based Blue Cross of Idaho, which began work in this area a decade ago, with that work ramping up and progressing in the past several years.
Consider the following: leaders at Blue Cross of Idaho have applied advanced data analytics to 27 different specialties and more than 1,000 episodes of care, on behalf of plan members. The results?
• 3,100 providers and 635 clinics now have participated in this pay-for-performance program.
• This advanced analytics work has saved Blue Cross Idaho about $6.5 million annually.
Orthopedic surgeons are among the groups most likely to use this information. Blue Cross shared findings with an orthopedic surgeon who was shocked that his costs were far higher than his peers.
• He immediately switched to a high-quality surgical center that significantly lowered his costs.
• Overall, Blue Cross of Idaho has seen a 3-percent decline in orthopedic costs.
The data showed this payer that members with osteoarthritis who had no physical therapy were six times more likely to have a knee or hip surgery than members who had physical therapy.
• Physical therapy is far more cost-efficient and may have a better health outcome than major invasive surgery.
• Blue Cross of Idaho designed benefit enhancements with a zero-dollar out-of-pocket cost to members, easing access to a physical therapist.
In doing this work, the Idaho Blue Cross leaders have partnered with the Ann Arbor, Mich.-based Merative Health, for support in building their analytics capabilities.
Per all this, Healthcare Innovation Editor-in-Chief Mark Hagland spoke recently with Marc Roberts, senior vice president and chief actuary & analytics officer at Blue Cross of Idaho, about the progress being made at his plan in this area. Below are excerpts from their interview.
What led you and your colleagues to initiate this program?
Everything we’ve done around analytics has been for the purpose of making healthcare more affordable. We’re a not-for-profit organization, and we try to make healthcare affordable. Our stakeholders are telling us how unaffordable healthcare is becoming now. And we felt that we had to have analytics as a big part of that, first and foremost, to solve the problems we need to solve. We don’t want to do what health insurers are sometimes accused of, getting in between our policyholders and providers; but we wanted to lower the total cost of care.
Why was there so little valuable data out there for orthopedic surgeons and other specialists to use to reach cost-effective clinical decisions around care?
The healthcare industry, as you know, everybody’s got their own lens. An orthopedic surgeon knows information about their patients, prior to surgery. They’ve run tests, done scans, biological screenings; but what they really value from the insurance side is the overall picture of the member. How many times has that member seen a primary care physician, or visited the ER? As the insurance company, we’re paying the claims and we know which drugs they’re taking, if they’ve seen other specialists, tried physical therapy, etc. So there are two lenses there, ours and theirs. And so much can be done in terms of partnering between insurers and providers. So it’s effective cross-pollinate that information.
What were the mechanics of creating this analytics platform?
There are a lot of technical challenges there. You need to do some pretty sophisticated mathematical maneuvers. We thought that some of the hard work would be around how to do those computations. But we were actually able to solve those challenges pretty quickly, using the people we had in place and adding others. It wasn’t that difficult to predict with relatively high accuracy who would need an orthopedic surgeon; the challenge has been getting the right information to the right person. We’ve got a lot of data scientists, but if only the data scientist is aware of information, it doesn’t help others, if it’s just sitting on their desk. So the challenge is how we get that information into the hands of a clinician, either within our walls, or a practicing provider who could use that information and connect with a member.
So how did you do that?
There are a lot of hurdles. Once we develop a predictive model or algorithm, the first thing we’ve got to do is to figure out who can use it. Some clinicians in our walls work directly with members; they can use that information. And then provider partners, especially those with value-based arrangements with us, can use that information. And nobody wants to hear from their health insurer, we’ve analyzed your data and predict you’ll need a surgery soon. That sounds very “Big Brother.” So we have to help clinical experts understand why predictions are being made. It’s not as though we say, Mark had this cereal rather than that this morning; instead, Mark is the right demographic and just got a diagnosis of osteoarthritis. Then it feels very different if a clinician reaches out and says, hey, I see you’ve got a condition and you’re in physical therapy, and it looks like you might need a surgery soon. So yes, we have arrangements under value-based care, with either a PCP or a specialist. If your doctor can say, hey, it looks like you’re getting regular knee injections; can we get you in to see you? And we do have trained clinicians on staff. For us, it often comes down to what the right care for the right person from the right provider, is important; and it’s what really lowers the cost of care.
What were the challenges involved in rolling out the program?
It’s along the lines of what we’ve talked about: how do we get valuable information from the data scientists to the clinicians? The clinicians are all very, very smart, but they’re not data scientists; so they see a lot of data and information; but it became very challenging to get the information into the right systems, given the intricacies and difficulties. We needed that information to show up in EHRs or whatever clinical information systems the clinicians were using. Logging records. We can’t just email them information and hope that it goes into the normal workflows. We have to put the information into the normal workflows. So who do we reach out first, and who’s most likely to respond?
What timeframe was involved here?
If we want to predict a health outcome, for example, how likely is it that Mark will be on the path towards a back surgery, we can build one of those predictive outcomes now within a matter of days; but to get it fully integrated into provider partners’ workflows—that overall process took about a year. We did the first version of that in 2019. We now have about 15 models that predict different types of health events? The first was a very general model predicting needs for healthcare. Now, who’s going to need surgery, for example; and during the pandemic, we could predict who might deal well with COVID or not, and our clinical team could focus on those people who could use that information. We can also look at who’s likely to develop a behavioral health episode or be admitted to the hospital for a behavioral health episode. Our biggest cost is around orthopedics, so back, hip, and knee surgeries: we do predictions all three types of those surgeries; also, diabetic episodes, who’s on the path to diabetic episodes. And end-stage renal disease is a rare by very costly disease. There are about 15 altogether.
How have physicians come to accept and embrace the program?
Especially for those with a similar mindset—we want to help serve Idahoans through value-based care arrangements—the program is really helpful information for them. And again, it’s a view of the world that they don’t have. They might know weight, height, screenings, of a patient; but they might not know all the drugs the patient is on, or other doctor visits. So we can say, here are the people on a verge of a major health event; it makes all the difference in managing patients’ care.
What have the biggest lessons learned been so far on this journey?
Some of the ones I’ve hit already: how do we create a better predictive model, and dial up the accuracy? But the real value comes from shifting from not using the information to actually using it and putting it to practice use. As I talk to others in other organizations, so much of the focus is on the analytical rigor, which is important; but what’s really important is how you’re going to put it to work. It doesn’t help the community if the information is staying on data analysts’ desks.
What does the near future look like for the program?
Technology in healthcare is just continuing to advance at a really rapid rate, from computing power to speed of computing. We’ve sort of figured out some of these solutions, and have started partnering with other health plans interested in doing this, and have started selling these services to other plans.
To other Blue Cross plans, or general?
We’ve had several plans, Blues and non-Blues, interested.
What is the number of patients who can be impacted?
The people who need the most care are usually our top 1 to 2 percent of our members, and we have 600,000 members, so you’re talking about maybe 12,000 or so members annually.
In your view, what is the optimal way to get a predictive analytics program going?
Meeting non-analytic, non-technical folks where they’re at, is important. If we had told providers to figure out how to use this by themselves, we wouldn’t have gotten any traction. Getting this into clinical workflows, and translating the data and meaningful, is what’s important.
Tell me about your partnership with Merative?
One of the things we noticed at the outset is that—how a claim comes in and gets paid is based on where a service takes place, but it’s hard to translate that into a type of member episode. So in the raw data, around, say, diabetes, we don’t necessarily have a clear view. So the really productive partnership with Merative is that they help us talk those raw claims and translate them into episodes that we can analyze. If someone had a bunch of procedures leading up to a knee surgery, those specific procedures don’t necessarily get grouped together; that’s where Merative’s solution has really helped us, to help us look at things from a clinician’s standpoint or a patient’s standpoint.
