Skip to main content

Getting health care right

Living Health Dynamic Platform: Meaningful, Actionable Data

“The challenge we are taking on with the Living Health Dynamic Platform is not just the integration of data, but the interpretation and presentation of data to make it meaningful for each individual and their health team,” explains Richard Clarke, Highmark Health's chief analytics officer.

When Highmark Health announced a six-year agreement with Google Cloud to build the platform, Clarke and Omar Khawaja discussed the importance of data security, privacy and choice. In this follow-up interview, he provides updates on the project, and offers insights on how the platform will use data and analytics to support a more holistic, proactive, and personalized approach to health, including a focus on social determinants of health.

Updates on the Living Health Dynamic Platform

Don Bertschman: Let's start with the collaborative agreement with Verily Life Sciences, announced in March. Where does Verily fit into the big picture, and what’s exciting about working with them?

Richard Clarke: Verily is a sister company to Google Cloud. That’s helpful for lots of reasons, not the least of which is that Google Cloud is the underlying technology for what we’re building with the Living Health Dynamic Platform. We’re excited about the synergies there, the shared mindset toward transforming the industry, and the ability to take solutions we develop to the world. As part of Alphabet Inc., the capacity to scale and expand solutions is vast.

Richard Clarke, chief analytics officer, Highmark Health

Richard Clarke, chief analytics officer, Highmark Health

In terms of specifics, Verily is helping with two major elements. The first is collaboration on Living Health’s personalized health solutions — digital-first health solutions like the Onduo diabetes product we have already been piloting. There will be a series of solutions, focusing largely on chronic conditions at first. Some solutions Highmark Health and Verily will build jointly, some Verily may build separately, but all will be tested rigorously within our ecosystem and then customized and configured specifically for Highmark Health.

Second, Verily is helping us extend our next-best-action capabilities — data-driven recommendations delivered to an individual, clinician, or member of a person’s health team. We’ve been working on this within my department for quite a while. We have data “listening posts” throughout our ecosystem where we listen for any indication that someone might benefit from one or more of our interventions — the “next best action” — and then we reach out immediately to navigate them to those interventions. Verily will help us continue to expand and evolve that approach to be even more integrated with providers, to be informed by new sets of data, such as biometrics and wearable device data, and to leverage the powerful artificial intelligence tools available on the Google Cloud Platform.

We’re excited about what Verily's technical capabilities bring to the work itself, and also their team's commitment to transformation. We have this big, bold vision — a learning health system that uses data and analytics to inform action, gauge whether the action had the right impact, and then adjust so it’s continuously getting better. That’s as exciting for them as it is for us, so there’s a fun sense of, OK, let’s lock arms and go transform an industry.

Don Bertschman: Can you describe in broad terms what work is started and what the timeline looks like for the Living Health Dynamic Platform?

Richard Clarke: The build is big and complex — that’s why these are six-year agreements. We’re hard at work on critical elements, starting with building the secure, scalable infrastructure. This is definitely, “measure five times, cut once,” because we want it to be built appropriately, automated and scalable so we can accelerate as we have more things we want to do in the future.

We hit an important milestone in April, when the initial internal sandbox was created and we loaded the data necessary to support our first analytic use case. The rest of 2021 will be spent continuing to build that secure and scalable infrastructure, loading more data, migrating internal use cases for testing, and doing design for the future. In 2022, as we move from that sandbox test environment to a production environment and have live data and the design elements of the new customer and clinician experience, we want to make sure we have enough in place to drive interaction, because that’s how the platform will keep learning and improving.

As an added note, the COVID-19 pandemic obviously created a long list of challenges — personal, societal, economic, and others. However, it also increased use of virtual care and use of data and analytic insights in decision-making. We have never worked so closely with our provider partners as we have during COVID-19. Right from the start we were learning what was happening from a data and analytics perspective, not only about disease progression, but who would be at greatest risk if they got infected, and what could be done to help them. Same with vaccination — data and analytics have been key to understanding who needs the vaccination most, where vaccination penetration is lower than expected, and what socially vulnerable groups need more support. Those connections will serve us well as we pivot to apply data and analytics in other areas.

Supporting a more holistic, proactive approach to health

Data and analytics can help determine when the root cause of health challenges involves unmet social needs, and then match someone with the appropriate support.

Data and analytics can help determine when the root cause of health challenges involves unmet social needs, and then match someone with the appropriate support.

Don Bertschman: That’s a good bridge to discussing social determinants of health — why are they important to the platform?

Richard Clarke: The intent of Living Health is to provide a more holistic, proactive, personalized plan for health. Social determinants influence a large array of health outcomes, so our platform must deeply understand those determinants and then match them to appropriate solutions.

The value of this approach is not just theory, we’ve seen it in practice. Take our Community Support Platform, powered by Aunt Bertha. This online social care network provides a national directory of free or reduced cost social services and an online referral tool to connect people with those services. Our platform is available to internal staff and to the people we serve and community members. In the first week it launched, during the pandemic, we saw about 12,000 searches on the platform. That’s evidence of the need.

Even pre-pandemic, our team was working really hard on anticipating needs and supporting proactivity — to say, how can we better understand the path that someone is on and get them support earlier? Part of that is understanding the root causes, including social determinants. One tactical example we’ve worked on involves people who are frequently showing up in the emergency department for non-emergent needs. With high fidelity, data can predict who is likely to end up coming back multiple times because of certain underlying drivers, so let’s identify them early, engage them at that first visit, and navigate them to appropriate support. Then they won’t need the emergency department to be their food pantry, or primary care, or place to get shelter.

With COVID, social determinants were also important drivers of some of the predictions about who was at risk, and how we could help. It’s one thing to say that a person is at high risk if they get infected. The next thing is to say, how can we help them shelter in place so they reduce their risk of infection? People with food insecurity or housing instability, for example, obviously have more difficulty staying home, so it becomes a question of how to engage them, including getting community organizations involved, to provide solutions and appropriate support.

The pandemic also has accelerated existing underlying trends. We worked with the Penn State Clinical and Translational Science Institute on a research study where our data showed increased prevalence of diseases of despair between 2007 to 2018. Diseases of despair are three behaviorally-related medical conditions — drug overdose (including alcohol), suicide, and alcoholic liver disease — that increase among groups of people who experience despair due to a sense that their long-term social and economic outlooks are bleak. You can imagine the accelerating effect the pandemic could have on what was already a disturbing trend, so we’re thinking about how to identify appropriate resources to address those needs.

Don Bertschman: Dr. Tony Farah sometimes talks about going from “the analog way” to “the digital way” in expanding and scaling improvements to care. Can you give an example of what that looks like with social determinants?

Richard Clarke: Sure, our organization’s provider network has a number of great programs for behavioral health, healthy eating, and other supports. The main way that people get into those programs is through manual referrals — for example, a PCP identifies an issue during a visit and then refers the patient to a program. These referrals are really good, because they were run through the filter of a dialogue with a physician, and rate of enrollment is usually high because the patient got recommended by their physician. The challenge, though, is that this way — the analog way — only reaches a small percentage of people who should be in a program, because it has to be the perfect combination of the person coming to the right place at the right time and having the right dialogue with the right clinician who knows about that program.

The digital way, which we are already working on, is to learn from those referrals and create look-alike models. If we can understand which people look like the people who get those “analog” referrals, then we can go find them — we don’t have to wait and hope they stroll into a doctor’s office. And then a big part of the thinking behind Living Health is not just finding them, but understanding them well enough to know the most effective channels to engage them and get them whatever support will help. It might be in person through their physician, it might be phone outreach, it might be digital or an app. We are using data to create scalable methods to identify and reach more people, and then also to expand the ways we contact those people, so we are not relying only on face-to-face office visits.

We have many predictive models running to help us understand where a population is heading. The addition of social determinants of health variables improves the predictive power of those models because social determinants underlie many outcomes. Even more importantly, since those variables give us a better understanding of the root causes driving health outcomes, the resulting interactions and solutions can be made more effective.

Don Bertschman: It’s easy to see how the more data we pull together, the better we can understand and support a person’s health, but does that approach also raise its own challenges?

Richard Clarke: The need for more data and better ways to analyze data presents a really important technical element to consider. It requires interoperability among systems, and information needs to be shareable and understandable. To enable that, we are betting on interoperable standards, either those we define as an industry, or those coming out of regulation — in fact, not just betting, we are active leaders there. When it comes to social determinants, for example, we have people with leadership roles in the Gravity Project, which aims to establish a national consensus about how we are going to define these elements so they can be shared and aggregated in the ways we want to achieve with Living Health.

Keeping it personal

Don Bertschman: What about from the individual’s perspective? Some social determinants data you get by asking me, other data may come from my zip code or being a certain demographic, but in both cases I can imagine challenges with how some people feel about that.

Richard Clarke: Social determinants data can be challenging to collect directly, because it depends on what people are willing to share. We have standard screening and survey forms, but we have to combine direct collection with indirect estimation to build a robust social determinants of health infrastructure.

We're working to improve direct collection both in terms of the frameworks, to make sure what is collected is consistent and can be combined, and in terms of doing it in the most patient-friendly way, so it's a good experience and people are comfortable sharing information. When we use estimation, we recognize the challenges that we need to think about and overcome — including aspects of privacy and the possibilities of estimation not being accurate — but because it is so important to understand those root causes, it's worth all the work to figure it out and get it right.

An important thing to keep in mind is how we think about the ethical applications of not only social determinants data, but all data. We are carefully examining each use case to determine the potential applications of these data and how we can make sure they are used in ways that are fair, appropriate, and aligned with our ethical principles. Once we have a use case we feel good about, then we use quantitative approaches to understand how it might play out, to make sure that we are not having any unintended consequences.

We are also counting on a rapid feedback loop with our end users. The analytics identify that next best action we want to recommend, but if the people getting that recommendation tell us something is ineffective, uncomfortable, or leading down a path they don’t like, we can quickly modify recommendations to get to a better place.

Don Bertschman: And that “better place” would vary with different people, right?

Richard Clarke: Hugely. And that’s in our core principles — “simpler, proactive, personalized.” We have to get better and better at personalization in terms of appropriateness, and personal preferences. We have to deeply understand what people want and what they’re comfortable with, so we do not design something for one big blob of a population, but instead get more precise, preferably down to a segment of one where the solution is perfectly tailored.

We will be doing just that with the generous financial support of $5 million which Highmark Health received from the Richard King Mellon Foundation. We will use those grant funds to leverage our partnerships with Google Cloud and Verily to develop and deploy holistic, digitally enabled, personalized health support programs, and increase access to support services. The focus will be on the whole person, considering the presence of multiple chronic conditions, with a particular emphasis on behavioral health — a major need in the community that has been exacerbated by the pandemic.

The other part worth emphasizing is that we also have to make sure information is useful to clinicians and health teams. If all we do is inundate clinicians with more information, the response will probably be, “hey, we already had too much to deal with, so this is not helpful,” and they won’t engage with the platform. Clinicians will engage if we are really good at curating information and presenting it to the right person at the right time. Whatever data we have, we are always going to hear Dr. Farah’s voice saying, “and make it actionable.”

Follow Highmark Health on social:

Highmark Health and its subsidiaries and affiliates comprise a national blended health organization that employs more than 35,000 people and serves millions of Americans across the country.

Questions or comments?