A primary care physician is reviewing her afternoon chart prep and finds a medication on the list she did not put there and does not recognize the reason for.
Her patient has heart failure, stage 4 kidney disease, and a recent hip fracture. He sees a cardiologist at one hospital system, a nephrologist at another, and a home-health nurse from a third organization who visits twice a week. Each of those clinicians can, in theory, now see a version of his chart through the data-sharing improvements of the last several years. What none of them can do reliably is see each other, or reach each other, in anything close to real time.
She calls the number on file for the cardiology practice. It rings to a general scheduling line. She leaves a message asking someone to call her back about a medication change, and three days later a callback comes from a nurse who does not have the visit note in front of her either.
The patient, meanwhile, is the one actually relaying information between his own specialists, because he is the only party in this arrangement who reliably shows up to every appointment.
He mentioned the new medication to his PCP himself, at his last visit, almost as an aside. That is how she found out at all.
Even where the data has become visible across institutions, there is still no verified, reachable roster of who is actively treating this patient right now, and no accountable channel to ask them a direct question this week.
The population this fails is small and extraordinarily expensive
This is not a complaint about healthcare in general. It is a specific, quantifiable failure concentrated in a specific, expensive slice of the patient population.
A 2013 VA Evidence Synthesis Program review found the sickest 10 percent of patients account for 65 percent of total US health expenditures. That is the population this article is about: patients with three or more chronic conditions, seen by multiple specialists across competing organizations, exactly the profile of the man in the PCP's chart above.
The same review found that roughly 75 percent of Medicare readmissions are potentially preventable, and named discontinuity of care as a contributing factor. It goes further, noting that increased medical specialization has had, in its words, the unintended consequence of increased fragmentation of primary care. The more specialists a complex patient accumulates, each one an expert in a narrow piece of the picture, the harder it becomes for any single clinician to see the whole patient, let alone reach everyone else who does.
A related finding cited within the same review, drawn from earlier research on elderly patients, found that seeing a different physician at each visit was associated with significantly higher rates of emergent hospital admission compared to patients with continuous primary care. Fragmentation is not a comfort issue. It shows up in admission rates.
Interoperability solved a different problem than this one
It is worth being precise about what has actually improved over the last decade, because conflating two different problems is exactly why this gap has persisted.
Initiatives like TEFCA, Carequality, and Epic's Care Everywhere have made real, measurable progress on data visibility. A clinician at one institution increasingly can pull up records, notes, and results generated at another institution. That is genuine progress and it should not be minimized.
But data visibility is not the same thing as reachability, and it is not the same thing as accountability. Knowing that a cardiologist changed a medication two weeks ago, because you can now see the note, is different from being able to ask that cardiologist why, this afternoon, and getting an answer from a person who is expected to respond. The PCP in the opening scene could, plausibly, now see the medication change in the chart. She still could not reach the person who made the decision in any way faster than a scheduling-line voicemail.
This is the same distinction this series has drawn elsewhere between data existing and a person being findable and accountable for responding. Interoperability moved the data. It did nothing to build the relationship layer on top of it.
Nobody maintains the roster, because no single institution can
Here is the structural reason this has not been solved by any existing system, EHR vendor, or care-coordination product.
Each EHR knows its own institution's care team completely. It has no visibility into, and generally no incentive to build visibility into, who else across other institutions is actively treating the same patient. Nobody aggregates across institutions in a patient-consented, clinician-verified way. The roster of "who is on this case right now" simply does not exist as a maintained object anywhere; it has to be reconstructed, informally, by whichever clinician happens to need it, usually by asking the patient.
The disincentive is not hypothetical. Competing health systems have limited commercial reason to make cross-referral coordination frictionless, because easier coordination can mean a patient's care, and the revenue that follows it, moving to a competing system. Care-coordination products that do exist, built by payers or single health systems, are typically owned by one side of that competition, which means they see inside their own walls and stop exactly where the next organization's boundary begins.
Meanwhile the structural conditions that make this worse are accelerating rather than easing. 57.8 percent of US physicians now report being employed rather than practice owners, per AMA physician-practice data. That number matters because it quietly breaks an old assumption: being at the same hospital used to imply being on the same care team. Large employed physician groups spanning multiple hospitals and multiple competing systems have made that assumption unreliable, even within what looks, from the outside, like a single institution.
What clinicians on the ground actually say the problem is
A study in Special Care Dentistry examining referral coordination between HIV clinics, dental care, and mental health services asked stakeholders directly what got in the way. The barriers they named were stigma, disconnectedness, and inconsistent access, and the facilitator they pointed to as most valuable was, in their words, "interprofessional awareness and collaboration."
That phrase is worth sitting with. Stakeholders did not describe the problem as missing clinical data. They described it as not knowing who else was involved and not having a working relationship to lean on when it mattered. That is a reachability and awareness problem, not a records problem, and it is exactly the gap that interoperability initiatives, however successful on their own terms, were never designed to close.
Why the incumbents will not build this
EHR vendors could build a cross-institution reachability layer, and have limited incentive to. Epic and Oracle Health compete with each other, and making it trivially easy for a clinician on a rival system to reach one of their own physicians directly runs against the commercial logic of platform lock-in, even where the clinical case for it is obvious.
Payer-owned care-coordination tools are not neutral across a patient's full care team. Products like naviHealth-style platforms are built to manage cost and utilization for the payer that commissions them, not to serve as a neutral relationship layer that every clinician on a case, regardless of which system employs them, can trust equally.
Health systems will not build a tool that makes it easier to lose the patient. A hospital investing in coordination software is, quite reasonably, building it to keep referrals inside its own walls. A genuinely neutral, cross-institution roster runs directly against that incentive.
A neutral, clinician-owned reachability layer is structurally different from anything any of these parties is incentivized to build, which is exactly why, years into serious interoperability investment, it still does not exist.
What would actually work
A patient-consented, clinician-verified active care team roster. Not a static referral list, but a living record of who is currently, actively treating this patient across institutions, updated as the care team changes.
A direct, accountable reach channel attached to the roster. Seeing a name is not enough; the roster has to connect to a real channel where a message reaches an accountable person, not a general scheduling line.
Verified identity on both ends. The clinician requesting contact and the clinician being reached both need to be verifiable, so the channel cannot be used to extract information under false pretenses.
Strict, explicit consent and minimum-necessary design. This has to be built inside HIPAA's minimum-necessary standard from the ground up, with the patient's consent governing who appears on their roster and who can reach whom about their care.
Built cross-vendor and cross-competitor by design. The entire value proposition depends on working across Epic, Oracle Health, and every smaller EHR simultaneously, and on being trusted equally by systems that compete with each other, which is precisely why no single EHR vendor or health system is the right owner.
A response-time expectation, not just a directory entry. A roster nobody answers replicates the current fax-and-wait experience with better branding; the value is in a working expectation that a message reaches someone accountable within a defined window.
Measured and published, not just assumed to work. A recurring benchmark on what share of complex-patient care teams can actually be fully rostered and reached within 48 hours would make an invisible failure visible and trackable over time.
What you can do now
If you are a PCP, hospitalist, or specialist managing a complex patient
Ask the patient directly, every visit, who else is currently treating them. Until a real roster exists, the patient remains the most reliable relay available; make asking a routine part of the visit rather than a fallback when something looks off.
Build a short list of direct lines, not scheduling numbers, for the specialists you coordinate with most often. A personal relationship with two or three specialists at competing systems will get you further, faster, than any portal message, and it costs nothing but the outreach.
Document what you learn about the rest of the care team in a place the next clinician can find it. A note listing every active provider and how to reach them is a small act that compounds in value for whoever reads the chart next.
If you lead a SNF, home-health agency, or hospice interdisciplinary team
Make "who else is on this case" a standing intake question, not an afterthought. Your patients are disproportionately the sickest 10 percent this article describes; the coordination failure hits your setting hardest and most often.
Push back on care-coordination tools that only see inside one system. If your organization is evaluating coordination software, ask explicitly whether it works across competing EHRs and institutions, or only within your own, before assuming it solves this problem.
If you build health IT or lead interoperability strategy
Treat reachability as a distinct deliverable from data visibility. TEFCA and Carequality-style initiatives are solving the right problem for chart access; explicitly scope a separate reachability and accountability layer rather than assuming better data-sharing eventually produces it as a side effect.
Design for cross-competitor neutrality from day one. Any tool built or owned by a single health system or payer will structurally fail to be trusted by that system's competitors, which are exactly the other organizations a complex patient's care team spans.
Frequently asked questions
Why is care coordination hard across different hospitals? Because interoperability initiatives like TEFCA, Carequality, and Epic's Care Everywhere have improved data visibility, letting clinicians see records from other institutions, but none of them create a verified, reachable relationship between the actual clinicians treating the same patient. A 2013 VA Evidence Synthesis Program review found roughly 75 percent of Medicare readmissions are potentially preventable, with discontinuity of care named as a contributing factor.
How do doctors communicate about a shared patient at different health systems? Mostly informally and slowly: fax, portal messages that may go unread for days, phone calls to general scheduling lines, or relying on the patient to relay information between specialists. No cross-institution, verified roster currently tells a clinician who else is actively treating a given patient right now or connects them to a direct, accountable channel to that person.
What causes fragmented care for complex patients? Increased medical specialization, per the 2013 VA Evidence Synthesis Program review, has had the documented, unintended consequence of increasing fragmentation of primary care even as each individual specialist's care improves; the sickest 10 percent of patients, who account for 65 percent of total US health spending, are disproportionately affected because they accumulate the most specialists across the most institutions.
How can readmissions be reduced through better care coordination? The 2013 VA review found approximately 75 percent of Medicare readmissions are potentially preventable, with discontinuity of care as a contributing factor, and separately found that elderly patients seeing a different physician at each visit had significantly higher rates of emergent hospital admission than those with continuous primary care, suggesting coordination quality has a measurable relationship to avoidable utilization.
Does better EHR data-sharing solve the care coordination problem? Only partially. Data-sharing improvements solve visibility, letting a clinician see what happened at another institution, but do not solve reachability, the ability to quickly contact and get an accountable response from the clinician who made a specific decision. Stakeholders in referral-coordination research describe the remaining barrier as "interprofessional awareness and collaboration," not missing data.
Why doesn't Epic or another EHR vendor just build this? EHR vendors compete with each other, and a tool that makes it trivially easy for a clinician on a competing system to reach one of their own physicians runs against platform lock-in incentives, even where the clinical case is clear; similarly, payer- or health-system-owned care-coordination tools are typically built to manage cost or retain referrals for their own organization rather than to serve as a neutral layer across a patient's full, cross-institution care team.
The bottom line
The sickest 10 percent of patients drive nearly two-thirds of US health spending, and roughly three-quarters of their readmissions are potentially preventable. Interoperability investment over the last decade has made real progress on letting clinicians see each other's data. It has made almost no progress on letting them reach each other, or know for certain who else is even on the case.
Every incumbent that could build the missing layer has a specific reason not to. EHR vendors compete with each other. Payer tools serve the payer. Health systems have limited incentive to make it easier for a patient's care, and the revenue attached to it, to move somewhere else.
What is missing is not more data. It is a patient-consented, clinician-verified, cross-institution roster of who is actively treating a given patient right now, attached to a real channel that reaches an accountable person this week rather than a scheduling line that calls back in three days.
Until that exists, the patient himself remains the most reliable coordination mechanism his own care team has, showing up to every appointment so that someone, eventually, finds out what changed.
Part of a series on the missing professional infrastructure of healthcare. Previously: Ethics of One: The Allocation Decision Made Alone at 2 A.M.
Evidence note: the 65 percent spending concentration among the sickest 10 percent of patients, the roughly 75 percent preventable-readmission estimate, and the continuity-of-care admission finding all come from a single 2013 VA Evidence Synthesis Program review (PMID 27606397); the continuity-of-care finding on elderly patients is itself drawn from an older study cited within that review and should be read as dated, directional evidence rather than a current, primary measurement. The HIV-clinic referral-coordination stakeholder findings (Bhanat et al., Special Care Dentistry, 2023) come from a study of a specific clinical population and may not generalize precisely to all complex-patient coordination settings, though the coordination barriers described are consistent with this series' broader findings. The 57.8 percent physician-employment figure is drawn from AMA Physician Practice Benchmark data cited in prior internal research and was not independently re-verified this session. Nothing in this article describes a specific patient's care; the scenario in the opening section is illustrative and composite, not a real case.
