A primary care physician at a community health center is looking at a patient's rash that has not responded to two rounds of topical treatment. It is not urgent. It is not straightforward either. She has three options open on her screen: write a standard dermatology referral that will put the patient on a waitlist measured in weeks, tell the patient to try a third cream and come back, or click the eConsult button sitting in the same EHR module she has clicked past for two years.
She does not click it. She writes the referral.
Not because the eConsult tool does not work. Her health network's eConsult program has been live since before she started here, it is staffed by real dermatologists on a rotation, and the data her own administrators could pull, if anyone asked, shows that a meaningful share of dermatology eConsults never need to become a referral at all. She has simply never used it, does not know which specialist answers it, does not know how fast the answer comes back, and has never had a reason to find out, because nobody has ever told her.
Multiply her by roughly twelve thousand colleagues across the same network, and multiply that network by every other community health system running the same infrastructure, and you have the shape of one of the stranger failures in this series: a tool with strong, published, replicated evidence of working sits almost entirely unused, not because it is broken, but because the people it was built for have no reason to trust the person on the other end of it.
The tool works. The published record is not close
Start with what eConsult actually does when a clinician uses it, because the evidence base here is unusually strong for ambulatory care.
Ontario's provincial eConsult program, run through the BASE eConsult service out of the Bruyère Research Institute and the University of Ottawa, has published more than seventy-five specialty-specific outcome papers. The pattern across them is consistent: a formal in-person referral was avoided in 32 percent of plastic surgery eConsults, 44 percent of ophthalmology eConsults, and 81 percent of eConsults for patients in correctional facilities. A separate cancer-genetics program found a referral "contemplated but then avoided" in 34 percent of cases (68 of 200). In the ophthalmology study, more than 88 percent of cases were rated at least 4 out of 5 in clinical value by the referring provider.
This is not a single pilot with a favorable write-up. It is a large, sustained, multiply-replicated body of evidence, across specialties as different as plastic surgery and correctional medicine, all pointing the same direction: a structured question sent to the right specialist resolves a meaningful share of what would otherwise become a waitlisted referral, in days rather than weeks.
Nobody is arguing that eConsult fails to work when it is used. The entire problem is that it is barely used.
The adoption number that should stop everyone in this field
Here is where the story turns, and where a genuinely well-evidenced intervention meets an adoption wall that three years of live infrastructure could not move.
Larson, Der-Martirosian, Boston and Gold published, in Telemedicine and e-Health in 2025, the largest study of eConsult adoption in US community health centers to date. Across 437 sites in 18 states, covering 13,769 providers, only 10 percent ever submitted a single eConsult between April 2021 and March 2024. Of those who did try it, 73 percent used it fewer than ten times in three years. Across a network serving 1,690,788 patients, there were 20,066 total eConsults, which means the tool that avoids a referral in a third to four-fifths of cases, depending on specialty, actually touched roughly 1 percent of the patient population it was built to serve.
Read those two facts side by side. A tool this effective, live for three years, across 437 sites, funded, integrated into the EHR, staffed by real specialists on the other end, and nine in ten providers who had access to it never used it once.
That is not a rounding error in an adoption curve. That is a structural failure, and the instinct in most health IT departments is to misdiagnose exactly what kind of failure it is.
The misdiagnosis: treating this as a button problem
The default response to a stalled adoption curve in health IT is to add friction reduction: put the button somewhere more visible in the EHR, run a webinar, send a reminder email, offer a small incentive for the first ten uses.
None of that addresses what the Larson data actually implies. Three years is enough time for a genuinely frictional workflow problem to resolve itself, at least partially, through habit and word of mouth, the way most EHR features eventually do. A 10 percent lifetime adoption rate after three full years of live availability is not the signature of a UI problem. It is the signature of a discovery and trust problem: providers do not click the eConsult button because they do not know, and have no way of finding out, who is going to answer it, how fast, or how good the answer will actually be.
Consider what a PCP has to trust, silently, every time she chooses eConsult over a standard referral. She has to trust that a real specialist, not a triage nurse, will actually read the question. She has to trust that the answer will come back inside a window that matters to the patient. She has to trust that the specialist answering is someone whose judgment she would actually want, rather than whoever happened to be assigned the eConsult queue that week. None of that trust is visible to her anywhere in her EHR. It exists, if it exists at all, only inside the memory of colleagues who happen to have tried the tool and happen to have mentioned it to her.
The tool asks a provider to substitute her own judgment about a specialist's reliability for a referral pathway everyone already understands, and gives her no information to make that substitution with.
Why nobody has built the missing signal
This is where the structural gap becomes visible, because the organizations closest to this problem are each positioned to fix only a fragment of it.
OCHIN and Epic operate the technical infrastructure that makes eConsult possible at scale, 437 sites and counting. But they are vendors serving one institution's data at a time. Neither has a product, or a business reason, to surface a specialist's response time or referral-avoidance rate across institutions, because their contracts are with the network, not with a cross-network trust layer.
The Ontario BASE program itself is the strongest evidence generator in the field and publishes prolifically. What it has not built, and was never funded to build, is a live, queryable directory that tells a referring provider outside its own network which specific specialist reliably answers fast and well. The knowledge exists inside the program's own internal quality data. It does not travel.
Doximity Dialer carries secure messaging between physicians, but it has no structured clinical-question workflow, no reciprocity obligation on the receiving end, and nothing resembling a responsiveness track record attached to a name.
Each eConsult program, in other words, is a well-run island. The specialist who answers eConsults quickly and well at one CHC network builds a real track record, and that track record is invisible to every referring provider outside that network's own analytics team. The primitive that is missing is not another eConsult portal. It is a cross-institutional graph: which named, verified specialist answers, how fast, and with what referral-avoidance record, queryable by any referrer regardless of which network or EHR she happens to sit inside.
Why this is getting more expensive to leave broken
Three trends are compounding the cost of the gap rather than shrinking it.
CHC network consolidation already proves the technology scales. OCHIN alone spans 437 sites across 18 states. The technical case for "this cannot be built at scale" is closed; the adoption ceiling sitting at 10 percent after three years is not a scaling limitation, it is evidence the missing input was never technical.
Post-2020 telehealth normalization removed the last excuse. Asynchronous specialist interaction is now a familiar workflow pattern to most clinicians in most other contexts. The barrier to eConsult adoption is not unfamiliarity with the format.
AI-drafted eConsult questions are about to raise volume on the sending side, which will increase pressure on the receiving side precisely where there is currently no verified reliability signal to route that volume intelligently. More questions arriving at specialists with no visible track record makes the discovery gap worse, not better, unless something is built to sort them.
Meanwhile, the cost of the status quo is not abstract. This series has already established what a bounced or delayed referral costs in staff time, travel, and patient wait (see the companion piece on referral wait times). eConsult is the tool that was supposed to absorb a meaningful share of that cost. At 1 percent penetration of the patient population it was built to serve, it is absorbing almost none of it.
The structural failure, stated plainly
Every incentive in the current system rewards building more eConsult infrastructure and none of it rewards building the trust layer that would make the existing infrastructure get used.
A CHC network that spends its optimization budget on a better EHR integration can point to a concrete deliverable: a new button, a new workflow, a training module completed. A CHC network that tried to build a cross-institutional responsiveness ledger would be building something that, by definition, extends past its own walls, requires cooperation from competitor networks, and produces a benefit that is hardest to capture for the institution that paid to build it. Nobody funds the thing whose value accrues mostly to people outside the paying organization.
That is the same shape of failure that recurs across this series: a genuinely valuable coordination asset does not get built, not because nobody has thought of it, but because no single existing player captures enough of the benefit to justify paying for it alone.
What would actually work
A response-time and referral-avoidance ledger that spans institutions, not just one network's internal analytics. The Ontario BASE program alone has published response and avoidance data across more than seventy-five specialty papers. That data exists. It has never been assembled into something a referrer outside the program can query before she asks a question.
A named, verified specialist attached to every measured response, not an anonymous queue. Providers do not build trust in a queue. They build trust in a person. The responsiveness data has to be attached to an identifiable, reachable specialist, the same way a good referral relationship is built on a name, not a department.
A pledge model with a real response window, not an open-ended queue. Part of what makes a referral feel more reliable than an eConsult today is that the referral pathway has an implicit, if slow, guarantee that someone will eventually see the patient. An eConsult responder pool needs an explicit, visible commitment: this specialist answers within this window, tracked and shown honestly.
Coverage deep enough to matter in a given specialty and metro before it is marketed as broad. A cross-network graph that is thin everywhere is not useful to anyone. A graph with real density in a handful of high-referral specialties in a given region, dermatology, endocrinology, cardiology, is immediately usable by a referrer in that region.
Outcome logging on every answered question, honestly, including the ones that still needed a referral. A responsiveness ledger loses credibility fast if it only shows favorable outcomes. Tracking both avoided and converted referrals is what makes the signal trustworthy enough for a provider to change behavior on the strength of it.
A pathway that connects to, rather than competes with, CPT reimbursement for interprofessional consultation. Codes already exist for this work (99446 through 99452). A responsiveness layer that helps providers find who to send billable eConsults to strengthens the economic case for every party rather than adding a new cost.
Specialty society involvement without specialty society control. CME credit tied to answered eConsults could meaningfully raise specialist participation, but the responsiveness data itself has to remain independent of any single society's membership interests to stay credible across institutions.
What you can do now
If you are a referring clinician
Ask your own network's eConsult program directly who answers fastest and best in the specialties you refer to most. That data almost certainly already exists inside your network's own quality reporting, even if nobody has ever surfaced it to you. Asking once is often enough to get it.
Try it once on a genuinely low-stakes question. The adoption data suggests most providers who never try eConsult are extrapolating from zero experience. A single low-risk use is the fastest way to build the personal trust signal the system currently fails to provide institutionally.
Tell a colleague when it worked. The informal word-of-mouth trust network is real, even if it is small and slow. You are part of the only mechanism currently capable of closing this gap for someone else.
If you administer or fund an eConsult program
Publish your own specialists' response times and avoidance rates internally, by name, before you spend more on integration. The Larson data suggests the ceiling is not being hit because of workflow friction after three years of live access. Surfacing the reliability signal you already collect is cheaper than another EHR module and addresses the actual barrier.
Talk to a peer network about sharing responsiveness data, even informally. The value of this asset compounds specifically because it spans institutions. A bilateral data-sharing arrangement with one comparable network is a meaningfully useful first step toward the cross-institutional graph that does not yet exist anywhere.
Track and report the referral-avoidance rate as a headline metric, not a background statistic. Ontario's published body of work shows this number, done well, ranges from roughly a third to over four-fifths of cases by specialty. Most CHC administrators could not tell you their own network's figure without a special request.
If you build health IT or specialty-network infrastructure
Build the ledger as a cross-network product, not a single-EHR feature. The structural reason this has not been built is that every existing candidate builder, EHR vendor, single network, single society, only captures value inside its own walls. A product designed from the start to span institutions is solving a problem none of them can solve alone.
Design outcome logging in from day one. A responsiveness score without honest negative cases attached to it will not survive contact with a skeptical referring provider for long.
Frequently asked questions
What is an eConsult? A structured clinical question sent electronically by a referring provider to a specialist, answered asynchronously, often within days, without requiring the patient to be seen in person. Published outcomes across Ontario's BASE eConsult program show a formal referral avoided in 32 to 81 percent of cases depending on specialty.
How is an eConsult different from a referral? A referral schedules an in-person visit, often with a multi-week wait. An eConsult routes a specific written question to a specialist who answers asynchronously, frequently resolving the clinical question without any visit at all. A 2025 Telemed J E Health study found only 10 percent of 13,769 US community health center providers ever tried an eConsult in three years, despite this evidence.
Why is eConsult adoption so low? The leading evidence points to a discovery and trust gap rather than a technology gap. A 2025 study (Larson et al., Telemedicine and e-Health) found only 10 percent of 13,769 providers across 437 community health center sites in 18 states ever submitted an eConsult over three years, and 73 percent of those who did used it fewer than ten times, despite the tool being fully integrated and staffed. Providers have no visible signal of who will answer or how reliably.
What percentage of eConsults avoid an in-person referral? It varies by specialty. Ontario's BASE eConsult program found referrals avoided in 32 percent of plastic surgery eConsults, 44 percent of ophthalmology eConsults, and 81 percent of eConsults for correctional-facility patients. A cancer-genetics program found a contemplated referral avoided in 34 percent of cases.
How fast do eConsults get answered? Response-time data exists inside individual eConsult programs' internal reporting but is rarely published or made visible to outside referrers, which is a core part of the adoption problem this article describes. No cross-network public benchmark of specialist response time currently exists.
Do financial incentives fix low eConsult adoption? CPT codes for interprofessional consultation (99446 through 99452) already exist and provide a reimbursement pathway, but the Larson et al. data shows adoption stalled at 10 percent despite three years of live, reimbursable, integrated access, suggesting payment alone has not been the binding constraint.
The bottom line
The dermatology eConsult in that PCP's EHR is not broken. It has never been broken. Ontario's program alone has published more than seventy-five papers showing this exact category of tool avoids a formal referral in anywhere from a third to over four-fifths of the cases it touches, depending on specialty, and rates as clinically valuable in the overwhelming majority of uses where it is tried.
What is broken is that three years of live, funded, EHR-integrated access across 437 sites and nearly fourteen thousand providers produced ten percent lifetime adoption. That is not a training problem or a button-placement problem. It is what a discovery failure looks like at scale: a tool works, and the people it was built for have no way to know who is answering it or whether to trust the answer.
Every organization positioned to fix this owns a piece of the infrastructure and none of them owns the missing signal. OCHIN and Epic run the pipes. Ontario's BASE program generates the strongest evidence in the field and does not travel outside its own network. Doximity carries messages with no reliability record attached. The specialist who answers eConsults quickly and well anywhere in this system builds a track record that is invisible to every referrer who is not already inside her own institution's walls.
So the PCP writes the referral, the patient waits weeks for something an eConsult could plausibly have resolved in days, and the specialist staffed to answer that exact question sits with an empty queue, unknown to the one person who needed to find her.
Part of a series on the missing professional infrastructure of healthcare. Previously: Trusted By, Not Followed By
Evidence note: US adoption figures are from Larson AE, Der-Martirosian C, Boston D, Gold R, "Patient, Provider, and Clinic Characteristics of eConsults in Community-Based Health Centers," Telemedicine and e-Health, 2025 (PMID 40828006), covering 437 community health center sites across 18 states and 13,769 providers from April 2021 to March 2024. Referral-avoidance figures by specialty are drawn from the Ontario BASE eConsult program's published outcome studies: Goulet et al. (correctional facilities, PLOS ONE 2024), Britton et al. (ophthalmology, Clinical Ophthalmology 2025), and a plastic surgery study (Plastic Surgery, 2026, whose full URL was not independently re-verified in this pass). The cancer-genetics referral-avoidance figure is from Rusnak et al., Genetics in Medicine Open, 2025, whose full URL was also not independently re-verified in this pass. These are program-published outcome studies rather than independently audited figures, and adoption behavior outside the specific CHC networks and Ontario program studied has not been separately confirmed. Nothing in this article should be read as an endorsement of any specific commercial eConsult vendor.
