Learning Healthcare Systems

If you’re like me, when you see doom and gloom articles about US healthcare: it’s broken, it’s too big, it’s not working, it’s controlled by payers, too much money is wasted, we’re not as good as others, etc etc you just click to the next post because it feels like we’ve heard it all before and nothing is going to change. But for the first time since I moved to the US in the 1980s I actually feel like something IS changing. And here’s why: the technology and tools that healthcare needs are finally coming of age.

I have had the good fortune to work directly with some incredibly talented data scientists, clinicians, software engineers, and business people not just at Cognome, but also with our partners at hospitals, medical schools and research organizations. We are all working together to bring the pieces of our incredibly fractured and siloed healthcare infrastructure into focus. Our goals so are big and audacious, it's been difficult coming up with a name for what we’re doing. The one that I like the best is Learning Healthcare System.

Here’s a 2018 article that describes what we are up to. It’s as relevant today as it was six years ago and shows you how long it takes to make things happen in healthcare. (A global pandemic may have slowed us down a wee bit as well.)

The 2018 article shows that the promise of AI and ML Models isn’t in some distant future, it’s been happening and is happening right now. How do I know? Because I’ve seen its direct effect on patients in the real-world at The Montefiore Health System: a large, diverse urban hospital system in New York City from which Cognome is being launched NOW.

We’ve seen our mission become reality many times over at Montefiore. And it’s now time for us to bring what’s been built to the rest of the world. Hospitals are deploying Machine Learning models because, simply and humbly put, they can save lives. More completely, they give PHYSICIANS the capability to save lives and improve patient experience; while also providing operational and technology leaders with AI, Large Language Models (LLMs) and insights that will revolutionize their health system

It wasn’t until I joined Cognome that I realized physicians are forced to fly in the dark much of the time. ML models process a vast number of data points across multiple systems, in milliseconds and with computational accuracy. Physicians don’t have the bandwidth, the tools or most importantly the time, to compile these numbers on all their patients. Certainly not in real-time. Live ML models are a scalable, clinically validated extension of the physicians themselves - except they are not flying in the dark.

Once trained, ML models are faster and better at predicting a patient’s risk level. The more complex the patient, the more you need a model to monitor them. ML models don’t need breaks, they are self-learning and they get smarter over time and with scale. 

There are likely no models or LLMs deployed in your hospital today. Large Academic Medical Systems and other Large Health Systems who have been working on this for years, may have a few models deployed in limited scope situations. Next year you will start to hear more and more hospitals deploying ML models. The year after that you’ll see exponential growth.

Eventually, ML models will be everywhere serving side-by-side with clinicians as the primary form of clinical and operational decision support. This is one of the key capabilities that will transform hospitals into Learning Healthcare Systems. Together they will learn from one another and bring cohesiveness where there is otherwise entropy. Led by Dr. Parsa Mirhaji, we have had this vision for years. And we have a set of patents that describe how you can deploy an “analytic tapestry” of interconnected models that all learn from one another.

But to go back and answer the question poised at the beginning of this post: How is a Learning Healthcare System going to change US healthcare? Together, we are going to reduce physician burnout, reduce administrative burden, lower costs, improve efficiency, improve outcomes, improve patient communication, provide real time access to researchers and reduce or remove many of today's necessary evils (like prior authorizations) that drive everyone crazy.

Contact us if you’re in!