It’s a holiday week (and we’re taking our 5 month old on his first beach trip!) so I’m re-upping a popular post I wrote in 2022 on the subject of care coordination.
Even three years later and with the advent of consumer-friendly LLM products, it still reads to me as accurate. Sure, people are trying to use AI to decode unstructured data like PDFs, but it’s not available at scale, and it hasn’t changed care coordination yet.
One thing I neglected in this piece is the role of EHRs and interoperability — frankly, because it’s not my area of expertise and some people (Brendan Keeler, Health API Guy) are already covering it far better than I ever could.
With that, here’s my take on care coordination:
***
When we talk about patient care, especially care for patients who are older, more medically complex, or less health literate, a unifying necessity is care coordination.
In an ideal scenario, care coordination could include such diverse tasks as making appointments with all the patient’s doctors, ensuring all the doctors have access to the patient’s records (and each other’s care plans); keeping the patient’s insurer on the same page; sharing back any acute events, like an unexpected trip to the emergency room (information that is often siloed unless the patient verbally shares it at their next appointment); organizing any home health visits and/or delivery of durable medical equipment; and coordinating with a pharmacist to make sure all prescriptions can be taken in tandem.
(I’m leaving out all of the additional parts that are just as high-touch but aren’t as straightforwardly “coordination,” including getting the patient to appointments, organizing prescriptions and finding a way to remind the patient to keep taking the medications, encouraging diet/exercise/therapy as appropriate, and more.)
In the real world, though, patient care coordination isn’t nearly as straightforward and personalized. Part of the problem is that, except in certain closed systems, there is often no one person or entity that has been deemed responsible for coordinating.
Closed systems
There might be a better phrase out there than “closed” vs. “open,” but this is what I’m trying to get at: Modern health systems are built to be self-contained healthcare entities. They include the emergency room for acute care, onsite specialists for outpatient chronic care, and suites for both inpatient and outpatient procedures. Health systems often also have onsite lab services, imaging services, infusion services — in other words, all of the pieces brought together as one.
Taking it a step further, health systems like UPMC have their own health insurance offering. While there are notable downsides to this highly concentrated, vertically integrated model (not least higher prices), the idea is that patients never have to leave the system, all the records are in one place, and the care can be more easily coordinated.
Again, reality is more complicated, and while concentrated health systems tout their efficiencies, research evidence suggests that said efficiencies haven’t really materialized. And even verticalized health systems often don’t do care coordination well. Part of the reason is that there’s limited incentive (except with health system-wide value-based reimbursement models) to invest in coordination, part of the reason is that closed systems aren’t entirely closed (and even in the case of a large health system, multiple entities in the network might be using different EHRs and workflows), and part of the reason is that systems often rely on providers to do much of the coordinating ad hoc, trusting these overworked staff members to manage the pieces.
The pandemic also made coordination more challenging: Healthcare workers, spending time together at the hospital, used to have more opportunities to informally coordinate care across patients.
All of this being said, in a perfect closed system, care coordination seems to be about processes more than partnerships. In other words: Has the system invested in the manpower and/or technology needed to accomplish coordination?
Open systems
Given the health system example, what I’m thinking of as open systems is obvious — it’s everything else. Outside of a unified health system, care coordination is far more complicated. Individual providers, LabCorp/Quest, infusion centers, and smaller hospitals might all have different electronic health record providers, different communications systems, and different interpretations of HIPAA.
To share patient information and even begin to coordinate care, all of these entities have to be on the same page, which they’re not. For open systems, care coordination depends far more on partnerships up front, before healthcare leaders can even begin to start working on processes.
A relatively new entrant to the space is making things even more complicated: Digital health companies, particularly those that provide direct patient care. These entities also have their own providers, methods, and, sometimes, their own EHRs.
(So what will happen when people using point solutions for one condition need more coordinated care as they age and develop comorbidities? I suspect companies will have to grow alongside them. More mature digital health companies like Omada and Livongo have developed models that coordinate around often-comorbid conditions like diabetes and heart disease, which suggests to me that newer companies may follow the same route as they mature and increasingly partner with employers and insurers. Without dedication to care coordination, though, digital health runs the risk of fragmenting everything further.)
Conclusion
The most striking part of care coordination is that no one is in charge. Moreover, there’s no clear responsibility for any one entity to take charge of coordination.
As risk-based reimbursement models become more common, more people are working on the problem of care coordination, so this is changing. [2025 ed note: I’m not sure this part is still accurate. I’m sensing burn out on the idea of value-based care and risk-based reimbursement. I don’t think it will go away, but it might get stuck in this current half-implementation state.]
To really see change in fragmentation, though, either value-based care will have to become a lot more widespread, and/or interoperability will have to become an actual thing, and/or we’ll have to come up with a better way to support PCPs (or another dedicated group of individuals) as care coordinators.
Excellent article. Since working in the interoperability space since the late '80s, I see the sharing of data within "Closed" systems and between competing systems improving, but it is still not the panacea we thought would happen. Introducing AI into the mix only creates more uncertainty. TEFCA and data blocking rules are a good start, but don't open the full spectrum of data housed within proprietary systems since that "proprietary" data that is crucial to the patient in that system, is often meaningless to other systems. QHINs and HIEs provide much more data than ever before and improve the availability of "meaningful" data but still, systems that do not take into consideration social determinants of health miss 80% of the reasons why the patient is ill to start with, since ALL data is health data.
How do you possibly coordinate a patient's health without all the data? Until we can create a personal health record with a patients complete health history, with duplications eliminated and SDoH included, and provide that record on an as needed basis at the POC, we will continue to struggle with chronic diseases and an aging population wondering why technology is passing them by.