A department chair needs a pediatric cardiologist. The position has been open for seven months.
He has done everything the process prescribes. Posted to the boards. Engaged a search firm at 25 percent of first-year compensation. Reviewed candidates. Two declined offers. One accepted and then withdrew.
Meanwhile, somewhere in his own medical staff, there are probably four physicians who trained with someone who would be perfect for this job, who is not looking, who would take a call from a friend, and who will never see the posting because they are not searching.
Nobody in that building can query that. Not because the information is secret, but because it exists only in individual memory and has never been written down anywhere.
The market already knows the answer
Here is the striking part. Physician hiring already runs on referrals. This is not a proposal; it is a description.
From a CompHealth survey of 592 early-career physicians:
- 51 percent searched through referrals.
- 48 percent searched through networking.
- 40 percent obtained their first job through referral or networking.
- 12 percent were offered a position at their training site.
And the general hiring literature is consistent that referred candidates are substantially more likely to be hired and to stay longer, with commonly cited estimates in the range of three to four times more likely to be hired.
So the highest-yield channel in physician recruiting is the one that runs on personal relationships, and it accounts for a plurality of actual hires.
Now look at what the market spends money on.
- Median time to fill for a physician: 118 days (AAPPR, 2025), with 77 days for advanced practice providers and up to 332 days for oncology.
- Search firm fees: 20 to 25 percent of first-year compensation on contingency, 25 to 35 percent retained.
- Median physician turnover: 7.3 percent, across roughly 130 organizations running something like 15,000 searches a year.
And the trend line is deteriorating sharply in one specific place. Offer acceptance fell from 83 percent to 71 percent in a single year.
Read that as a market signal. Nearly three in ten physician offers are now declined. Something is failing late in the process, after all the sourcing spend, which is exactly where trust deficits show up.
The cost of an empty chair
The economics of the vacancy are what make this urgent for employers.
Physicians generate roughly $2.4 million a year in average revenue for their organizations by industry survey estimates. Translated to monthly vacancy cost, AMN Healthcare's survey work put family medicine at roughly $176,000 a month in forgone revenue and orthopedics substantially higher.
At a 118-day median time to fill, a single family medicine vacancy runs to something like $690,000 in forgone revenue before search fees. For higher-revenue specialties it is multiples of that.
So the arithmetic facing every recruiting organization is: months of enormous carrying cost, a fee of a quarter of first-year compensation, and a nearly 30 percent chance the eventual offer is declined.
The 72.8 percent nobody can reach
Here is the structural fact that explains why sourcing spend performs so poorly.
Doximity survey work found that 72.8 percent of physicians describe themselves as curious about opportunities but not aggressively seeking, with only 10.5 percent actively looking.
Roughly three quarters of the physician workforce is passive. They are not on job boards. They do not respond to recruiter outreach, which now arrives in industrial volume and increasingly appears machine-generated. They are, however, entirely reachable by one mechanism: a text message from someone they trained with.
So the recruiting industry spends heavily to reach the 10 percent who are actively searching, while the 72.8 percent who represent the best available candidates are accessible only through a channel nobody has built.
The best candidate for almost any physician job is not looking, and is two edges away from someone already on your medical staff.
The missing object
Be precise about what does not exist, because it is a single, definable thing.
A verified vouching edge. A recorded, checkable statement of the form: "I trained with this person, or operated alongside them, or covered for them, and I would hire them again."
Consider what each existing system holds instead.
Job boards carry postings, which are advertisements, not endorsements.
Search firms hold candidate lists and relationships, which are their asset and are deliberately not shared. When a recruiter changes firms, those relationships walk with them.
LinkedIn endorsements are unverified, self-selected, reciprocally inflated, and universally understood to mean nothing.
References are collected at the very end of the process, from people the candidate chose, after everyone has already decided.
Nothing anywhere records, in a queryable form, who has actually worked with whom and would vouch for them.
And this is not a hypothetical construct. The referral research in medicine indicates co-training produces a real, durable, behavior-changing trust edge, with physicians referring patients to co-trainees at 26.2 percent versus a 21.4 percent baseline. The graph exists. It has measurable effects. It is simply not written down.
An employer cannot ask "who on my medical staff trained with this candidate." A candidate cannot ask "who at this organization trained with someone I trust, so I can find out what it is really like there." Both of those questions are answerable in principle from data that exists in people's heads, and neither is answerable in practice.
Backchannel checks are the real hiring process
Anyone who has hired or been hired in medicine knows what actually determines outcomes, and it is not in the formal process.
A candidate receives an offer. Before accepting, they text three people from residency: "Do you know anything about this place? Is the chair reasonable? Why did the last person leave?"
An employer considers a candidate. Before extending, someone calls a friend: "You worked with her, right? What is she actually like?"
These conversations decide the outcome and are entirely invisible to the formal process. They happen through personal networks, which means their availability is distributed by who you happen to know.
A candidate with a well-connected residency cohort gets accurate intelligence about an employer. A candidate without one signs blind. That asymmetry is a substantial part of why offer acceptance is falling and why first-job tenure has collapsed to under two years for recent graduates, a topic covered earlier in this series.
The backchannel is the real hiring process. It is entirely unsupported infrastructure.
Why AI outreach is making this worse
There is a development in the last two years that has accelerated the failure, and anyone with a physician inbox has already noticed it.
Recruiter outreach has been industrialized. Personalized-looking messages arrive in volume, generated at near-zero marginal cost, referencing your specialty and your city and occasionally a detail scraped from a profile. A physician now receives more recruiting contact than at any point in the history of the profession.
The predictable consequence is that the entire channel has lost its signal. When a message costs nothing to send, receiving one tells you nothing about whether the sender considered you specifically. Physicians have responded exactly as economics would predict: they ignore all of it.
Which means the cost of reaching a passive candidate through outbound channels is rising sharply at the same time that the response rate collapses.
And it produces an inversion worth naming. As machine-generated outreach becomes free and worthless, a human vouch becomes scarce and valuable. A message that begins "my colleague Sarah, who you trained with, suggested I contact you" now cuts through in a way it did not need to five years ago, because it is the only message in the inbox that could not have been generated automatically.
The recruiting industry is responding to falling response rates by increasing volume, which accelerates the collapse. The alternative, routing through the trust graph, requires a graph nobody has built.
What would actually work
Verified co-training and co-practice edges, queryable. The residency and fellowship graph, recorded, so an organization can find who among its own physicians has a genuine tie to a candidate, and a candidate can find who at an organization can tell them the truth.
Vouches recorded at the time, not solicited at the end. A reference collected after the decision is theater. An attestation recorded when two people actually worked together is data.
Two-sided by design. Employers want vouches for candidates. Candidates want honest accounts of employers. A system that serves only the employer reproduces the existing information asymmetry, and physicians will not participate in it.
Reachability for the passive candidate, on their terms. The 72.8 percent will not respond to recruiters and will respond to a peer. Any mechanism that respects that distinction, where the approach comes through someone the candidate actually knows, addresses the largest structural failure in physician recruiting.
And a caution about incentives. Anything resembling payment for candidate referrals in healthcare requires careful legal review, and a mechanism designed around professional reciprocity rather than transactional bounties avoids that hazard entirely while working better.
What you can do now
If you are hiring
Ask your own physicians first, systematically. Not a casual "let me know if you hear of anyone." A structured request: who did you train with, in this specialty, who might be persuadable, and would you make the introduction? Most organizations have never done this deliberately, and it reaches the 72.8 percent that nothing else does.
Map your medical staff's training history. Where did each of your physicians train, and when? That single spreadsheet is a map of the trust edges your organization already contains, and virtually no organization has built it.
Move references to the front. If a candidate is two edges from someone on your staff, that conversation is more informative in week one than a formal reference in week ten.
Look at your offer acceptance rate honestly. With the national figure at 71 percent, a substantial share of recruiting spend is being consumed by late-stage failures, which are typically trust failures rather than sourcing failures.
If you are looking
Use your cohort before the boards. Forty percent of first jobs come through referral and networking. Your residency class is the highest-yield search tool you have, and most physicians underuse it out of a sense that asking is imposing.
Backchannel every offer. Before signing, find someone who has worked there. If you cannot find anyone, that is itself information about how connected the organization is.
Ask why the last person left. Then ask someone who does not work there.
If you are early career
Your classmates are about to scatter into every health system in the country. In five years, that group will collectively have direct knowledge of dozens of employers. Maintaining it deliberately is the single highest-return professional investment available to you, and it costs a group video call every few months.
If you lead a specialty society
You hold the cohort data. Societies know who trained where and when, across institutions, at a scale no employer can assemble. That is the raw material of the trust graph, and it is currently used for dues renewal and conference mailing lists.
Members would use a verified peer-introduction service. It serves the member rather than the institution, which is the thing societies are uniquely positioned to do and mostly do not.
Frequently asked questions
How do physicians actually find jobs? Primarily through relationships. A CompHealth survey of 592 early-career physicians found 51 percent searched via referrals and 48 percent via networking, with 40 percent obtaining their first job through referral or networking and 12 percent being offered a position at their training site.
How long does it take to fill a physician vacancy? Median time to fill was 118 days for physicians according to AAPPR 2025 data, compared with 77 days for advanced practice providers, and stretching to 332 days for oncology. Offer acceptance rates fell from 83 percent to 71 percent year over year.
What do physician search firms charge? Industry sources indicate contingency arrangements typically run 20 to 25 percent of first-year compensation, with retained searches at 25 to 35 percent.
What does a physician vacancy cost? Substantial forgone revenue. AMN Healthcare survey work estimated roughly $176,000 per month for a family medicine vacancy, with higher-revenue specialties considerably more, against average physician revenue generation of roughly $2.4 million a year. At a 118-day median fill time, a single vacancy commonly runs into the high six figures.
Why is it hard to reach the best physician candidates? Because most of them are not looking. Doximity survey data found 72.8 percent of physicians describe themselves as curious about opportunities but not actively seeking, with only 10.5 percent actively looking. Passive candidates do not respond to job boards or recruiter outreach but will respond to a peer they trained with.
Do referred candidates perform better? The general hiring literature consistently finds referred candidates are substantially more likely to be hired and to stay longer, with commonly cited estimates around three to four times more likely to be hired. In medicine specifically, co-training has been shown to produce durable, behavior-changing professional trust, with physicians referring patients to co-trainees at measurably higher rates.
The bottom line
Physician hiring already runs on vouching. Half of early-career physicians search through referrals, forty percent get their first job that way, and every consequential decision on both sides of the table is made after a backchannel text message to someone who trained with someone.
Meanwhile the formal market spends 20 to 35 percent of first-year compensation and 118 days to arrive at outcomes that are declined nearly three times in ten.
The gap between those two facts is the entire opportunity. The profession owns a large, verified, behavior-changing trust graph, and pays intermediaries handsomely to approximate it from the outside.
The chair with a seven-month vacancy is probably four conversations away from the right hire, standing in a building full of people who could make the introduction, unable to find out which ones.
Part of a series on the missing professional infrastructure of healthcare. Previously: The 112-Day Dead Zone
Evidence note: search behavior figures come from a CompHealth survey of 592 early-career physicians (2019). Time to fill, offer acceptance, and turnover figures come from AAPPR 2025 benchmarking. Passive candidate figures come from Doximity survey work. Vacancy cost and physician revenue generation figures come from AMN Healthcare survey research and are industry-published estimates. Search firm fee ranges come from industry sources and represent typical rather than universal terms. Co-training referral effects come from Pany and McWilliams in Health Services Research (2021). Referral hiring performance figures come from general recruiting industry research rather than healthcare-specific study.
