A conversation with Len Bickman, Chief Augmented Intelligence Officer, Ontrak Health.
Health plans maintain a lot of member data. But often they struggle to make that data work for them. The good news? AI technologies are transforming the way we analyze data to identify and serve members living with unaddressed behavioral health needs.
To understand how these technologies work, the Editorial Staff recently sat down with Len Bickman, Chief Augmented Intelligence Officer, Ontrak Health.
What follows is a lightly edited transcript of that conversation.
Editorial staff, Ontrak Health
Ontrak talks a lot about the power of “augmented intelligence” – what is that, exactly?
Len Bickman, Chief Augmented Intelligence Officer, Ontrak Health
Augmented intelligence—in contrast to artificial intelligence—focuses on enhancing the role of the human service provider and not replacing him or her.
Although both techniques use machine learning capabilities, artificial intelligence is used without human assistance. That’s the big difference.
Augmented intelligence maintains the human aspect, using machine learning to provide humans with actionable data.
That seems conceptually clear. But can you provide some real-world examples?
Sure. If you’ve used Alexa, Siri, or another virtual assistant, you’ve used augmented intelligence. Virtual assistants don’t make decisions for you. Instead, they provide the data you need when you need it.
By contrast, examples of artificial intelligence include email spam filters, plagiarism checkers, and Google’s AI-powered search suggestions. These technologies make the decisions for you.
We use both types of intelligence capabilities at Ontrak. When used together properly, augmented intelligence and human intelligence are greater than the sum of their parts.
But we never want to lose that human element. At Ontrak one-to-one human-centered engagement is core to who we are—and what we do. So, that’s why we tend to talk about augmented—rather than artificial—intelligence.
Let’s stay on this topic of member engagement for a minute. Can you say more about how augmented intelligence supports member engagement?
Yes. First, augmented intelligence helps our member engagement team use motivational interviewing techniques. We prioritize training on motivational interview techniques because it’s shown to encourage members to make health-enhancing choices and address their behavioral health needs.
All our calls are recorded with the member’s permission. Then our AI-infused technologies analyze the conversation in real time, giving the agent feedback they can use to ensure they’re properly implementing motivational interviewing techniques.
So, basically, augmented intelligence gives agents real-time intel they can use to engage smarter and make sure members are getting the information and support they need.
We’re also building a new capability that will let us use AI to match the characteristics of available outreach specialists to potential members. In other words, our algorithms will leverage criteria—things like language style, background, experience, and so forth—to select the best outreach specialist for the member we are working with.
In this way, we can be more confident that members are getting culturally relevant attention and meaningful one-to-one engagement.
This idea of matching and identification is critical to the larger Ontrak approach, right? Like, before we can engage members, we have to identify members who are candidates for the program.
How does augmented intelligence work for that part?
That’s a great question.
We recruit our members from health claims data provided by our health plan customers.
The claims data tells us if a potential member has been diagnosed with a mental or behavioral health problem.
Previous diagnosis makes them eligible to participate in our program.
However—and this is really key—we know that many people who may have mental or behavioral health needs often do not receive a formal diagnosis.
That’s where AI-infused analysis comes in. We can identify members who may be living with unaddressed behavioral health needs—even if they’re not formally diagnosed.
We should pause there for a second. Because that’s fascinating.
You’re saying that AI analysis can potentially identify members living with unaddressed behavioral health needs—and even those members may not know they have conditions that require treatment?
That’s one way to frame it.
In the best case, yes, we can help members realize they’re living with a behavioral health condition that warrants treatment. Many times, of course, members realize they have a problem, even if they can’t name it exactly, and they are grateful that we reach out to help.
But I would say the larger point is we can use AI to find members living with unaddressed behavioral health needs who are definitely hidden to the health plan. And that matters a lot to our clients because that’s how we help them serve members and reduce overall claims costs.
By the way, this is called imputing a behavioral health diagnosis. There’s no formal, clinical diagnosis, but we can impute the diagnosis based on available data.
What kind of data?
Well, a lot of it still rests with existing claims data.
Studies have shown that persons with untreated behavioral health problems have higher health care costs overall, especially related to other chronic health conditions. There’s a lot of evidence that chronic conditions like hypertension, diabetes, and obesity are linked to unaddressed behavioral health needs.
So, our AI algorithms can scan millions of data points related to all historical claims data and with a high degree of accuracy impute behavioral health diagnoses.
It’s still not entirely clear how this works. You can’t just say, oh that person is living with hypertension, so they’re a candidate for behavioral health treatment.
Yeah, that’s right. I mean, I simplified it to some extent for illustration, but it’s certainly more complex than that.
This is where machine learning algorithms come into play. They’re calibrated to find health plan members who share similar characteristics with those who do have behavioral health claims, but who don’t have a formal diagnosis.
This results in focusing our enrollment efforts on those members who will result in lower costs for our health plan clients and better health outcomes for our members.
Thanks for that explanation. That’s much clearer. So, it’s all about obviously knowing the typical profile of optimally eligible Ontrak program members?
Yes. That’s right.
Let’s get back to member engagement before we wrap up—we’re almost at time!
Earlier we talked about how augmented intelligence supports outreach and recruitment. But what about once members have joined the Ontrak program? How does AI help with coaching and ongoing engagement?
There’s a huge risk of attrition in programs like this. It’s hard to engage members who may be lost to care. So, we’re constantly working to mitigate risk of member drop out.
By analyzing member profiles and characteristics, our AI capabilities can help us predict the probability that members are at risk of dropout. Then we can convey that information to our coaches with tailored suggestions about how to reduce the probability of dropout.
For example, AI will trigger check-in reminders based on member behavior. And recommend resources and next best steps.
In addition, AI also helps us monitor all our member interactions to ensure Care Coaches are properly using supportive conversations—and following the principles of motivational interviewing.
Every conversation is scored using validated AI scoring algorithms. Then we share the results out to our engagement team.
This keeps everyone accountable. But it also empowers us keep improving all the time. Every new interaction is an opportunity to learn, get better, and improve our training.
Every member deserves the best—and we’re going to use our technologies to deliver on that promise.