Jay Deady is CEO of Jvion of Suwanee, GA.
Tell me about yourself and the company.
I’ve been in health IT for 30 years, having started in 1989 with Cerner. I’ve had a series of opportunities and roles on both the clinical revenue cycle and analytics sides of the business. Mostly focused on providers, but with some exposure to payers along the way and keeping my career focused solely on health IT.
Jvion is an industry-leading prescriptive AI company. Our mission is to drive down preventable harm to patients, both clinical and cost-related harm, however we can. That has been company’s mission since Day One. The co-founders have done a great job bringing the company forward over the last eight or nine years. I was fortunate to have the opportunity to join a few months ago as CEO.
Are health systems interested in how AI and predictive analytics work under the covers, or are they just looking for solutions that can deliver the results they need?
They are definitely looking at some of the details. The reason is that over many years, certain terms in healthcare and healthcare IT tend to get somewhat abused and therefore misunderstood. It was “workflow” and “analytics” back in the day and now everybody seems to be an “AI” company. Health systems, ACOs, and payers want to understand how Jvion is different from some other company that claims to be in the space. They are clearly interested in the outcomes and benefits that current clients are achieving and they want to understand how our approach is different.
Do health systems, and particularly clinicians, struggle to trust AI that functions as a black box with hidden proprietary algorithms?
It’s a balancing act. We have proprietary technology and methods, and other companies might say the same. Under an NDA, we will go to a certain depth to explain how it is that we do what we do. Fortunately, we have a relatively large number of clients that have been using Jvion for a while, so those documented outcomes and references help in those conversations. Details about how we approach the data science and strong peer references help. We also use a model control study versus just a benchmarked pre- and post-analysis. We have a lot of rigor around documenting the outcomes we have helped clients achieve.
Will AI become another example where technology companies try to solve problems they don’t understand because they don’t know healthcare?
There is some of that. There’s another side as well. On one hand, you have AI companies that don’t understand how healthcare works. They don’t understand the triangle between a patient / member, a payer, and a provider and how you add value to each constituent by understanding their alignment. On the other hand, AI draws a lot of different correlations and can provide a lot of different solutions for a company that does healthcare, but understands that healthcare is complex and needs help with a lot of questions. It’s challenging, from a corporate perspective, to narrow the focus so that you can efficiently scale versus answering one question for one client and trying to multiply that.
How important is it when training a model to avoid amplifying existing biases and to resist the urge to overstretch the model’s capabilities?
One of Jvion’s differentiators is that we have 33 million lives with between 2,500 and 4,000 data points within our machine. We don’t take in a large volume of data for one particular client, which will be biased to their capture solely, and then run the analysis only against that. Our scale and our nine-plus years of experience allow us to leverage the underlying clusters across those 33 million to even out any regional or local biases that might come from a single data source or data from a single region.
What information from outside the EHR can help identify patients who could benefit from an intervention?
Beyond EHR data, the machine uses publicly available data from the federal government, such as community vulnerability and social determinants of health. There are various capabilities around lab data and claims data. EHR-specific data makes up less than one-third of the data that we have in the machine.
What do clients most commonly learn when they apply a broader set of analytics capabilities to data that extends beyond their Cerner and Epic systems?
There’s a lot of additional data that isn’t contained within the EHR. Cerner and Epic are clearly trying to go down the path of balancing, however they describe it, between analytics and AI. But there’s additional behavioral data — environmental data, lifestyle data, transportation data, and even weather. These have impact on the health of a population and on the health of an individual in a specific area, but they aren’t within the EHR. That is one way that we significantly differentiate our offering from the nuanced early capabilities of what Cerner and Epic are doing.
Is social determinants of health information useful other than recognizing that an individual has a problem that goes beyond the health system’s ability to fix it?
Our clients aren’t just hospitals. While source data for SDOH does in some cases come from health systems, we gather information from other sources.
We break our market down into three segments. We have health systems on the provider side. We have population health entities on the provider side, where on their own or in conjunction with maybe a payer joint venture. There are ACOs or other initiatives where some level of risk is being taken around the defined population, whether that is the hospital’s employee base if they are really large or expanded into a provider-sponsored health plan. We have more than hospitals as clients and sources of SDOH.
What opportunities have arisen from helping customers address COVID-19?
It certainly was an unexpected impact for the industry, the nation, and for Jvion. I started as a new CEO three days after Georgia locked down, and multiple months into my career at Jvion, I think I’ve met 18 of my colleagues in person. I just went on my first in-person client visit in Georgia two days ago, wearing masks and socially distancing. Otherwise, it has been a virtual engagement, and that has had a big impact on general business operations.
At the solution level, the hospital provider segment has been impacted the most. Their economics have been fairly devastated. They were a 2-3% margin business, generally not for profit. They lost 30-60% of their high-margin business for a period of time. Our average health system client will probably be off 20, 30, or 40% of the financial operating numbers they had expected for the calendar year, and that is massively impactful from the operations side. From the caregiver side, the daily onslaught of delivering care in this COVID world versus a multi-service line clinical care delivery system is very different.
We initiated a COVID map that we pushed out for free. We worked with Microsoft on it. It’s available online. We’ve had 4 to 5 million hits and uses of it, everybody from the Pentagon and the White House Task Force to the CDC and others. We mapped down to the actual block area to show the vulnerability of a particular community, which is more beneficial – particularly for health systems – than looking at government data that’s at a county level. We expose that for our clients as well as anybody that would care to use it. We’ve been happy with the massive use.
For our clients, we took a look at their current patient lists, applying both the COVID map and other data we created and something we do for our normal solutions. We don’t just create a list of folks who might be susceptible to a negative quality event coming up and predict that. We do that, but we also put that in rank order based on the ability to intervene with a suggested intervention that could make a positive trajectory change and improve the potential outcome based on what the current trajectory is. A number of our clients are using that to outreach to those in their capture who might be the most susceptible and vulnerable from a COVID perspective to make sure patients are getting assistance.
We created a triage select solution, which we refer to as a vector. It works both for COVID and for any type of potential respiratory-impacting areas or diseases, such as basic flu, where you may need to make triaging decisions around the right time and appropriateness to ventilate. How do you prioritize that as the patients are presenting? That helps our clients deal with the onslaught of folks coming in.
I’m really proud of the team here at Jvion and appreciative of the feedback that we got from our clients in critical, overwhelming times. We were able to take that input, understand their needs, and bring our resources, assets, and capabilities to assist.
Do you have any final thoughts?
The US health system environment has faced challenges in my 30-year career and in the past, but they were more financially market oriented, where hospitals had reduced access to the bond market during the financial crisis, for example. But I’ve never seen anything that was so impactful to the actual operations of the health system itself. We will move through this at Jvion.
We are also looking at our prescriptive AI, which historically has been solely clinical in nature, to understand the challenges of our health system clients. In those parts of the country that are post-COVID or in a lesser COVID world, how do they start getting a return to care? One client’s research found that 68% of community patients are reluctant to seek care because of fear of going to a medical facility related to COVID.
That deferment of care is having a major impact on the providers and the services that they can provide to patients. They will have higher acuity and more severe illness and disease state based on the deferment of that care. If they’re a commercially insured patient or member, payers have an influx of money today based on all the deferment of care, but there’s a tsunami coming of that care having to be delivered, and it will be more expensive later than right now.
It’s an interesting alignment period, with patients getting the care they need sooner than later, providers needing those types of patients back into their health system, and payers wanting them to get the care now versus deferring it and it being more expensive later. We’re focused at Jvion on how we can help drive that alignment across those three constituents whose interests are aligned with a single incentive.