A cardiologist is walking from the parking garage to the cath lab. Somewhere between the third floor and the elevator, a hospitalist she vaguely recognizes falls into step beside her and says the eight words that structure a meaningful share of American clinical decision-making:
"Hey, quick one for you if you've got a second."
She has about ninety seconds. She gets maybe sixty percent of the relevant history, delivered in the order the hospitalist happens to remember it rather than the order she needs it. She does not see the images. She does not see the medication list. She gives an answer, because that is what colleagues do, and because the alternative is to be the person who says no.
She will not document this. She will not bill for it. She will not remember it by Thursday. Nobody will ever check whether her answer was right.
And she will do this roughly three or four more times this week.
This is the curbside consult, and if you want to understand the professional infrastructure of medicine, you have to start here. Not because it is broken exactly, but because it is enormous, load-bearing, invisible, and completely ungoverned. It is the single largest clinical decision channel in the profession, and it runs on a routing algorithm that can be stated in four words: whoever you happen to know.
Just how big is this, really?
Most people underestimate the curbside because it leaves no trace. Nothing is billed, so it does not appear in claims data. Nothing is documented consistently, so it does not appear in the chart. It exists in hallways, text threads, and the space between two pagers.
The best measurement remains a study published in JAMA that surveyed subspecialists directly. The findings:
- 87.5 percent of subspecialists had fielded at least one curbside consultation in the prior week.
- The mean was 3.6 curbsides per week per subspecialist.
- Primary care physicians reported requesting about 3.2 per week.
- 80.2 percent of subspecialists said the information they were given was insufficient to answer well.
- 54 percent of infectious disease physicians agreed that the absence of reimbursement reduced their enthusiasm for answering.
Now scale it. If even 300,000 American subspecialists field 3.6 curbsides a week, and each takes roughly ten minutes of genuine cognitive work, that is on the order of nine million hours a year of expert clinical judgment. At a conservative $200 an hour, that is $1.5 to $2 billion of professional labor.
Given away free. Recorded nowhere. Credited to no one. Routed by acquaintance.
A study of the "corridor consult" published in Obstetric Medicine found the pattern holds internationally: 91 percent of hospital doctors in one Israeli sample agreed to hallway consultations when asked, while 54 percent simultaneously held negative attitudes about the practice. An infectious disease service in the same literature logged 532 official consultations against 269 unofficial ones in a single year, and only 11.5 percent of the unofficial consultations ever converted into a formal consult.
Read that last number again. Nine out of ten informal consultations never become formal. They are not a triage step on the way to real consultation. They are the final answer, for most of the questions that enter them.
The uncomfortable study
Here is where the story gets harder, and where most discussions of the curbside politely stop.
A team publishing in the Journal of Hospital Medicine did something obvious that nobody had done rigorously: they took curbside consultations and then performed a formal consultation on the same patients, and compared.
- The information conveyed during the curbside was inaccurate or incomplete in 51 percent of cases.
- Management advice differed after formal consultation in 60 percent of cases.
- When the initial information had been wrong, advice changed 92 percent of the time.
This is not an indictment of the physicians involved on either end. It is arithmetic. The asker is compressing a complex clinical situation into ninety seconds from memory, unaided by the chart, without knowing which details the answerer needs. The answerer is pattern-matching against an incomplete picture and is professionally disinclined to interrogate a colleague like a hostile witness.
The curbside does not fail because clinicians are careless. It fails because the interface is a hallway.
And the profession has never treated this as a teachable skill. Research presented at IDWeek in 2025 found that more than 70 percent of internal medicine and infectious disease clinicians had received no formal education at all in how to give or receive a curbside consultation. We train people for a decade in clinical reasoning and never spend an hour on the channel through which a large fraction of that reasoning is actually exchanged.
The channel moved to your phone and nobody updated the rules
While the profession was not looking, the curbside migrated. The hallway became the group chat.
A 2026 study in PLoS One at a tertiary center found that 98.9 percent of clinical consultations between colleagues ran through WhatsApp, and that 63.7 percent of respondents were concerned about the legal implications of doing so. This is not a fringe finding. Studies of surgical trainees have found roughly 88 percent using group texting for case discussion. Clinicians in one Australian hospital study sent a median of 12 patient-bearing messages per day, with about half believing the practice conflicted with the law.
The privacy exposure is not theoretical. In one clinical group studied, consent had been obtained for only 52 percent of posted images and 32 percent of posted laboratory results. A scoping review of 16 studies of clinical messaging found "no clear mechanisms for record keeping or data storage."
At the same time, and this is the part that makes simple prohibition the wrong answer, these channels are delivering enormous clinical value in exactly the places with the least specialist access. A WhatsApp teledermatology service in Botswana handled 811 consultation threads over four years, 62 percent of them provider-to-provider, substituting for specialist visits that would otherwise not have happened. In parts of South Africa's Eastern Cape, 166 doctors serve a population of roughly seven million with four infectious disease specialists. The group chat is not a compliance problem there. It is the health system.
So we have arrived somewhere genuinely awkward. The profession's most-used consultation channel now runs on consumer messaging that has:
- no verification that the person answering is who they claim to be,
- no consent capture,
- no de-identification,
- no archive the treating team can reference in six months,
- no routing when the group goes quiet,
- and no way to find the right group if you are not already in it.
Institutional secure messaging does not solve this, and understanding why is the whole point. TigerConnect, Epic Secure Chat, and their peers work beautifully inside an organization. But the hard case, by definition, is the one where nobody inside your organization knows the answer. Enterprise messaging follows the employer boundary. The trust graph clinicians actually use crosses employers, states, and countries. The two never meet.
Follow the money: why nobody fixes a channel nobody pays for
Every durable dysfunction in healthcare has an economic explanation. Here is this one.
The curbside is free at the point of use, and its cost is borne entirely by the answerer, in a currency nobody measures: attention, interruption, and unpaid cognitive labor. More than half of infectious disease physicians in the JAMA study said the lack of reimbursement dampened their willingness to answer.
Meanwhile, the same clinical exchange, conducted through a formal channel, has an established price and a body of evidence.
Electronic consultations have been studied for two decades and they work. Programs including Champlain BASE in Ontario, the San Francisco General model, and the Veterans Health Administration report:
- Turnaround times of roughly 1.2 to 2.6 days.
- 32 to 70 percent of face-to-face specialist referrals avoided, depending on specialty and program.
- One Ontario analysis of specialist-to-specialist e-consults found a face-to-face referral avoided in 69 percent of cases.
- A transgender-care e-consult program reported a median response time of 1.2 days, requiring a median of 18 minutes of specialist time, and avoiding 32 percent of referrals.
- Peer-reviewed VA data found face-to-face visits in a hematology unit fell 18 percent within two years of e-consult implementation.
- Payers pay real money per e-consult, and CMS maintains interprofessional consultation codes (99446 through 99452) precisely for this exchange.
Eighteen minutes of specialist time. That is what a well-structured version of the question costs. Compare that to the 26-day average wait for a new-patient specialist appointment in large US metros, or 34.5 days in dermatology, for an encounter that frequently resolves a question that eighteen structured minutes would have answered.
So the obvious question: if e-consults are faster, better documented, reimbursable, and demonstrably effective, why has the curbside not been replaced?
Because e-consults are sold to employers and payers, and they route only within a contracted panel.
That single design decision determines everything. If your health system bought an e-consult product, you can reach the specialists in that contract, for the specialties in that contract, on the terms of that contract. If your question falls outside the panel, or you work somewhere that did not buy one, or the person who has genuinely seen your problem works for a competing system three states away, the formal channel simply does not reach.
The hard question is the one that crosses institutional walls. The formal channel stops at the walls. That is not an oversight. It follows inevitably from who signed the contract.
Which leaves the hallway. And the group chat.
The liability nobody warns you about
There is one more force keeping this channel in the shadows, and it deserves a plain explanation because most clinicians have never had it laid out.
When you answer a curbside, you are generally not establishing a physician-patient relationship, and courts have largely protected informal consultation from creating a duty of care. That is the reassuring version, and most of the time it holds.
Here is the complication. That protection depends heavily on the exchange remaining genuinely informal and general. And the IDWeek 2025 data suggests the exchange has quietly stopped being either of those things:
- 95.7 percent of infectious disease respondents said protected health information is usually or always shared in curbside consultations.
- Between 61.6 and 73.9 percent said curbside advice is often documented in the medical record.
Look at what that combination creates. Patient-identifiable information is exchanged. Specific advice is given about a specific patient. The advice is then written into the legal medical record, frequently attributed by name, without the named person's knowledge or review.
The answerer has now been documented as having given patient-specific guidance on a case they never saw, based on information that the literature says was incomplete about half the time.
That is a genuinely uncomfortable position, and almost nobody who gives curbsides has thought about it explicitly. It is also, quietly, one reason senior clinicians in high-liability specialties become less generous with curbsides over time, which removes exactly the most experienced answerers from the channel.
What a governed curbside would actually look like
None of this argues for abolishing the curbside. The curbside is fast, collegial, and frequently correct, and any solution that makes it slower will simply be routed around. The relevant question is what a version with the same speed and better properties would require.
The design problem is unusually well specified, because we know exactly which properties are missing. A governed curbside would need to be:
Routed by expertise, not by acquaintance. Today's algorithm is "whoever I know." The upgrade is a match on who has actually done the thing, with the person's trust distance to you (co-trainee, prior answerer, colleague of a colleague) used to decide who to approach first. This is the entire difference between a directory and a router.
Structured at the point of asking. Half of curbside information is incomplete because the asker does not know what the answerer needs. A template per question type solves most of this at essentially zero cost, and is the highest-yield fix available today.
Explicit about what kind of question this is. General educational guidance and patient-specific advice are different acts with different legal and ethical weight. Making the asker declare which one they want, in one click, protects everyone and takes three seconds.
De-identified by default. No patient identifiers, ever, with de-identification assistance at the moment of posting rather than a policy document nobody reads.
Time-bounded with an escalation path. The fatal flaw of the group chat is silence. A question that goes unanswered for two hours should escalate beyond the group, to someone whose expertise matches, rather than dying quietly.
Recorded for the asker. Not as a consultation note that creates a duty for the answerer, but as the asker's own reference: what I asked, who answered, what they said, when. This is what makes the exchange survivable at scale and reviewable later.
Credited to the answerer. This is the piece everyone skips and it is the one that determines whether the system exists in two years. Supply in this market is currently pure goodwill, and goodwill burns out. Whether the credit is money, reciprocity, professional recognition, or an obligation ledger matters less than the fact that something must account for the answerer's time.
Escalatable to a formal pathway. When a question genuinely requires patient-specific advice, there must be a clean, consented route into formal consultation rather than a nudge to keep going informally.
Note that none of this is technically hard. The obstacles are structural: who owns it, who pays for it, who bears the liability, and whether the answerers show up. Which is why it has not been built by anyone whose business model requires selling access to physicians, or by any institution whose interest stops at its own walls.
A working curbside protocol you can use tomorrow
Until infrastructure exists, the channel still runs, and you can run it far better than the average. This is the practical section. Take it to your department.
If you are asking
1. Say what kind of answer you want, first. Open with one of three sentences: "I want your general thinking on a pattern," or "I want to know if this warrants a formal consult," or "I need patient-specific advice and I will document it." The answerer will calibrate everything to that sentence, and you have just eliminated the most common source of mismatch.
2. Lead with the question, not the story. Clinicians narrate chronologically because that is how we learn. It is exactly wrong here. State the question in one sentence, then give the facts that bear on it. The answerer can then steer you toward what they actually need.
3. Have the chart open. The 51 percent incompleteness figure exists because people curbside from memory. If you can answer follow-up questions in real time with actual numbers, you have removed the single largest defect in the channel.
4. Give the five things that change the answer. For most questions there are roughly five variables that determine management. Know what they are for your question before you ask. If you do not know what they are, that itself is worth saying out loud.
5. Ask explicitly before documenting a name. "Is it all right if I note that I discussed this with you?" is a ten-word question that prevents a genuine professional harm. If they say no, document the substance without attribution.
6. Close the loop. Tell them what happened. This costs you thirty seconds and is the single strongest determinant of whether that person answers you next time. It is also how the answerer learns whether their pattern recognition was right, which is the only feedback they will ever get.
If you are answering
1. Name your confidence level out loud. "I am confident about this," "I am reasoning from a nearby situation," and "I genuinely do not know, but here is who would" are three completely different answers, and the asker cannot tell them apart from tone alone.
2. Ask the two questions that would change your answer. Even in ninety seconds. Most curbside errors trace to a fact the answerer never asked for because they were being polite about time.
3. Say when this needs to be formal. You are allowed to say "I will give you my thinking, and this one genuinely needs eyes on the patient." That sentence protects the patient, the asker, and you.
4. If the case is outside your recent practice, say so and route. "I did a lot of that ten years ago, but Chen has been doing them all year, let me connect you" is a better answer than a confident one from stale expertise. Directing well is a skill, and it is the behavior that would make an expertise graph work if one existed.
5. Track your volume for one month. Count the curbsides you field. Most subspecialists dramatically underestimate this. The number is your unpaid, uncredited second job, and you cannot make a case for it existing formally until you can state it.
If you run a department
Make the template. For your five most common incoming curbside types, write down the fields that determine the answer. Distribute it. This alone will move your incompleteness rate substantially and takes an afternoon.
Measure the channel. Ask your subspecialists to log volume for two weeks. You will discover an enormous quantity of clinical work happening entirely outside your operational picture, some of it from people you have never counted as consultants.
Fix the documentation norm explicitly. Decide, as a department, whether curbside advice gets documented with attribution, and whether attribution requires consent. The current state at most institutions is "it varies by who is typing," which is the worst option available.
Ask where the questions go when nobody here knows. That answer will be a group chat, and the honest follow-up is what happens when that group has nobody with the right exposure. Right now, in most places, the answer is nothing happens.
Frequently asked questions
What is a curbside consult? A curbside consultation, also called a hallway consult or corridor consult, is an informal request for clinical advice from one clinician to another, given without a formal consultation, without examining the patient, and usually without documentation or billing. It is the most common form of interprofessional clinical consultation in medicine.
How common are curbside consults? Very. Research published in JAMA found 87.5 percent of subspecialists had fielded at least one in the prior week, averaging 3.6 per week, with primary care physicians requesting about 3.2 per week. In one 2026 study at a tertiary center, 98.9 percent of clinical consultations between colleagues took place through WhatsApp.
Are curbside consults accurate? Often, but less reliably than clinicians assume. A Journal of Hospital Medicine study comparing curbside advice with formal consultation on the same patients found the information conveyed was inaccurate or incomplete in 51 percent of cases, and management advice changed in 60 percent of cases after formal consultation, rising to 92 percent when the initial information was wrong.
Can you be sued for giving a curbside consult? Courts have generally held that informal consultation does not establish a physician-patient relationship or a duty of care, and this protection usually holds. The risk rises when the exchange stops being general and informal: when protected health information is shared (which IDWeek 2025 data suggests happens usually or always), when the advice is patient-specific, and when it is documented in the medical record under the answerer's name (reported by 61.6 to 73.9 percent of respondents). Ask before documenting attribution, and state clearly when a question needs formal consultation.
What is the difference between an e-consult and a curbside consult? An e-consult is a documented, reimbursable, asynchronous consultation conducted through a formal system, typically with turnaround of 1.2 to 2.6 days, and it avoids 32 to 70 percent of face-to-face referrals. A curbside is informal, undocumented, unpaid, and immediate. The decisive practical difference is reach: e-consults route only within a panel your employer or payer contracted for, while curbsides route to anyone you personally know. Neither reaches the person you do not know who has actually seen your case.
Is it a HIPAA violation to discuss a patient on WhatsApp? Sharing protected health information through consumer messaging platforms that lack a business associate agreement and appropriate safeguards creates real regulatory exposure, and studies show consent is frequently not obtained (one clinical group study found consent for only 52 percent of posted images and 32 percent of labs). The practical guidance is to de-identify rigorously, obtain consent when images are involved, and understand that your institution's secure messaging tool does not solve the cross-institution case, which is precisely why the practice persists.
Can you bill for a curbside consult? Not as an informal exchange. CMS interprofessional consultation codes 99446 through 99452 exist for documented consultative discussions between clinicians and do allow reimbursement when their requirements are met, including documentation and, in many cases, patient consent. The vast majority of curbside volume does not meet those requirements and is therefore unpaid.
The bottom line
Medicine's largest consultation channel processes something on the order of nine million hours of expert judgment a year, resolves the overwhelming majority of the questions that enter it, changes management decisions constantly, and has no router, no record, no quality layer, no training, and no compensation.
It survives on the professional generosity of people who answer because that is who they are. That generosity is real and it is one of the better things about the profession. It is also finite, unevenly distributed, and dependent entirely on whether you happen to know the right person.
The uncomfortable truth in the data is that the curbside is not failing because clinicians give bad advice. It is failing at the front door. The information going in is incomplete half the time, and the person receiving it was chosen because they were nearby rather than because they knew.
That second failure is the deeper one. Somewhere out there is a clinician who has managed your exact problem forty times. Nothing in medicine connects you to them, so you asked the person by the elevator instead.
Fixing the interface is a productivity improvement. Fixing the routing would be a different kind of change entirely.
Part of a series on the missing professional infrastructure of healthcare. Previously: Healthcare Knows Where You Work. It Does Not Know What You Know.
Evidence note: figures are drawn from JAMA, the Journal of Hospital Medicine, Obstetric Medicine, PLoS One, IDWeek 2025 proceedings, Journal of Telemedicine and Telecare, published e-consult program evaluations, and CMS coding guidance. Where a figure comes from a single-center study or a small sample, that limitation is stated in the text. Nothing here is legal advice, and liability standards vary by jurisdiction.
