Thought Leadership
How do you identify members eligible for behavioral health intervention? It’s in the data.
By Esteban Nell
Vice President, Production Data & Member Operations, Ontrak Health.
In this blog Esteban explains the technologies and processes behind Ontrak Identify, a new standalone data analytics solution designed to help health plans identify members living with unaddressed behavioral health needs. These data insights help to better engage members and drive program participation.
Many health plans are investing in care programs designed to provide better treatment and support for members living with unaddressed behavioral health needs.
Too often, however, health plans struggle to find eligible candidates to participate in these programs. This is because many members living with behavioral health needs are considered ‘lost to care.’ That is, they’re either undiagnosed or not actively getting treatment. And so it’s difficult for health plans to identify these members to conduct targeted outreach and engagement.
In addition, member outreach is difficult because Case Managers and care coordinators don’t necessarily have all the resources and information they need to meaningfully engage members and encourage them to get the treatment they need.
AI-infused data tools are changing the game.
This is where AI-infused data analytics really make an impact. Ontrak Health recently unveiled a new standalone solution, Ontrak Identify, that works to help health plans more efficiently find and engage members eligible for behavioral health treatment.
Finding the right eligible members significantly aids member outreach, enabling health plans to focus resources where they’re needed most. In other words, pinpointing members eligible for treatment helps target engagement to drive better program participation.
Among other things, Case Managers and care coordinators can use data insights to prioritize members most likely to benefit from intervention. They can also gain critical insights that can help them engage more effectively, build trust and empathy—and ultimately guide members to the right kind of care.
How it works
Health plans send us their member list, including claims data for the past 12 months (sometimes longer). Then our algorithms get to work, interpreting 21 different factors to locate a pool of members who are eligible for care. Factors include everything from cost and claims data to geography and income (if available).
Our systems are extremely accurate when it comes to finding members eligible for care—including members who are not diagnosed. More on that below.
Cost is a key factor.
First, we sort and filter members who are driving up costs for our health plan clients. That’s the first selection criteria. We understand that for programs to achieve ROI, they ultimately need to support cost reduction.
So, our initial pool will be comprised of members whose claims exceed a given threshold specified by our clients.
Formal diagnosis
The next major bucket includes members who have a formal behavioral health diagnosis like depression, anxiety, or substance use disorder. Among members we identify as eligible, around 75% are formally diagnosed.
Of course, formal diagnosis is only one piece of the puzzle. It doesn’t mean that members are getting the treatment they require.
Members go into the eligibility pool when claims reveal a formal diagnosis, but inadequate treatment. We all want members to get the care and treatment they deserve.
Our algorithms also analyze the extent to which members receive appropriate care or treatment. Sometimes members living with behavioral health needs drive higher healthcare costs even though they’re getting the most efficient treatment.
Psychiatry and some pharmaceuticals are simply expensive. And our systems may determine that there’s no meaningful intervention likely to reduce those costs. In these cases, our systems exclude those members from the eligibility pool.
Plus, we obviously don’t want to interrupt members who are getting the care they need.
By contrast, other members may be driving higher claims through myriad Emergency Room visits or by failing to regularly visit a Primary Care Provider. Our algorithms can instantly identify patterns of inefficient, unsustainable healthcare consumption.
And these are the members who are usually most eligible for intervention and outreach.
The secret sauce: Our AI-infused algorithms can also impute behavioral health diagnoses.
On average, around 25% of members we identify as eligible for treatment do not have a formal behavioral health diagnosis.
So, how do we know they’re eligible for behavioral health treatment? Well, this is the secret sauce of our AI technology.
Basically, our algorithms can triangulate myriad data points to uncover patterns typically associated with unaddressed behavioral health needs. Within a small margin of error, our algorithms predict members eligible for treatment.
This is called an imputed behavioral health diagnosis. In other words, it’s an inference based on available claims data.
Over time our machine learning technologies have honed and refined that analysis to pinpoint the kind of healthcare consumption consistent with members we know do have a formal behavioral health diagnosis.
Sharpen identification. Drive program participation.
Taken together, our proprietary analyses can yield enormously valuable insights for health plans. By deriving a highly targeted eligibility pool, we help make member outreach so much easier—and so much more effective.
The algorithms do the hard part, which is find and validate in advance members most likely to benefit from care or treatment. This enables health plans to prioritize outreach. And also maximize their ROI, since interventions target members receiving expensive, unsustainable treatment.
It also makes it easier for Case Managers and care coordinators to engage members. They can draw on a wealth of insight to understand member needs and build empathy and trust. This ultimately helps drive program participation and achieve everyone’s goal: Long-term wellbeing and durable healthcare consumption.
