A friend for everyone

I'll be the first to admit I'm not the best at keeping in touch -- and apparently my last real LinkedIn post was during #JPM in January ??

In part, I've been able to proudly look on as Kendra Seavey, MPH and her extraordinarily driven team at care.coach have expanded our impact every day across our growing list of customers ?? Perhaps you see their LinkedIn posts, as they're out and about at conferences and customer launches across the country. Every new customer contract and every avatar device fulfillment ticket submitted represents another step toward improving the health & well-being of the most high-need, complex individuals in our healthcare ecosystem ???? Our end-users, Program of All-inclusive Care for the Elderly (#PACE) participants and folks with developmental disabilities, are close to our hearts. When we have the opportunity to talk personally with one of them, Kendra & I like to call it "soul work." So, it always warms the heart just to see our numbers growing, knowing that behind every number is a person whose life we are transforming for the better.

Other than that, I've been rather heads down, in full geek-mode, working with our impressively talented R&D team led by Chantal Kerssens & Deepak Sahoo, not only developing but commercializing state-of-the-art approaches to scaling our impact to millions of Americans. So, with American Society on Aging's annual #ASA2024 aging innovation conference in full swing (I'm there in spirit!), and folks having asked for updates on our AI/tech development progress, below is a quick update on what we last touched on in Healthcare Startups & Gen AI: the Stakes are High | LinkedIn.

First, a recap of what we're building at care.coach. Presuming you're familiar with common AI & healthcare terms, you can think of it as a multi-modal #generative model for all aspects of providing whole-person #biopsychosocial support:

• healthcare outcomes and cost prediction based on claims & EMR/EHR data to identify target populations for which health plans can drive maximum positive impact through behavior change (care.coach Augur? for Medicare Advantage & care.coach Augur? for PACE)
• individualized conversational engagement of these targeted populations to develop health-supportive friendships and reduce #loneliness (care.coach Engine? which powers care.coach Avatar? and care.coach Fara?)

• risk identification and care management escalation to improve access to care and reduce social isolation, health plan risk adjustment, and insights to address disparities in social determinants of health, a.k.a. #SDoH (care.coach Engine?)

What's rather extraordinary about modern generative AI is that we can train a single AI model to do all of the above! Until recently, each function would have to be a separate machine learning model, each trained just to perform its one specialized function. With #genAI, we're able to build an AI that's able to not only become a specialist at a variety of disparate skills, but also generalize across skills and figure out commonalities, patterns, and things we didn't even train it to understand??

This ability to generalize across isolated experiences and skills is what makes us so capable as humans, and is especially important when working with complex populations and older adults, most of whom live with multiple chronic conditions like diabetes, hypertension, heart failure, COPD, depression, cognitive impairment, and so on, and many of whom also face SDoH challenges due to lack of social connections, low income, housing/food/transportation insecurity, etc. #Aging, #geriatics and #gerontology are very complex fields of study, and the USA doesn't have nearly enough professionals trained to adequately support our changing demographics without e.g. shuffling someone between a family caregiver, a doctor, another specialist doc, a social worker, a geriatric care manager, another two family members who don't agree on what the care manager meant, a well-meaning neighbor, and perhaps a therapist -- not a happy experience ???? Especially if any of the formal care mechanisms were not actually needed or helpful. What many people don't realize is that health plans are not only concerned about reducing waste and avoidable utilization, but also motivated to improve the quality and cohesiveness of the member experience, and to help with many SDoH factors like ????????????? These are the daily concerns that people naturally talk about with their friends and family, but may not know to reach out to their health plan about. So many biopsychosocial factors to understand and act on appropriately... if only there were a way to unify all of this understanding into one person's brain and have that person look through health plan data to figure out which members could benefit from his/her support, then befriend every one of them... ??

What if, in addition to any family and friends an individual may or may not be fortunate to have, every American could have such a friend who is not only caring, fun, and generally someone you'd want to chat with, but also has enough essential breadth of healthcare understanding to help with care coordination by serving as the personable "face" for the entire health plan and healthcare system, introducing to the individual the most appropriate form of support for whatever their biopsychosocial needs are? And what if this ubiquitous friend also leverages its friendships and healthcare understanding to mobilize and coach the most abundant healthcare resource in America -- individuals themselves -- to live healthier lives? ??????????????

That's the future we're building, thanks to genAI and our brilliant team at care.coach, including our global team of health advocates who work diligently 24x7 to support diverse and complex people across the US while simultaneously teaching our AI to be empathic, fun, and a good friend ??

With that context, a few brief updates:

• care.coach Fara? is in production, though due to NDAs that national-scale payors take very seriously, that's about all I can say for now. ?? Fara's looking forward to making a couple million friends in the next couple of years!
• care.coach Augur? for PACE has our engineering team impressed by its performance, combining claims and EHR data via a #genAI approach to predict outcomes and costs ?? I suspect it will be in production soon, bringing to PACE leaders game-changing abilities to target scarce resources for maximum impact on highly complex participant populations, as if they had a ??. Per our #NPA2023 poster with West Health, we need to verify Augur's predictions against a dataset that they've held out from us, before we share more details and performance metrics publicly.
• Efforts to add computer vision to our unified AI model are starting out well, per our latest NIH award to digitize CDC's STEADI fall prevention program. So far, we've demonstrated the ability to harness the sight that care.coach Avatar? grants us to measure ??gait parameters relevant for physical therapists and fall prevention experts. We're productionizing it now in a privacy-preserving edge compute manner, and then we can potentially add the visually-derived gait parameters to our outcomes/fall prediction capabilities for care.coach Augur?.
• We've productionized our own AI compute infrastructure. Our crack team has set up GPU servers across care.coach headquarters in California and the care.coach Hospitality Hub in Kansas for secure, geo-redundant AI model training and production-grade inference. The blinking of lights, whirring of fans, and wafting heat of pure computational power -- it's beautiful ??

In summary, I'm super proud of the work the care.coach team has pulled off in the past couple of months. And that's it for now, as far as AI-related updates!