On June 30, you are a senior resident or a graduating fellow. You are surrounded by a building full of people who are obligated to answer you. An attending is assigned to your service. A program director has formal responsibility for your development. Your co-residents are in the same hallway, and they have watched you work for years. If you are unsure about something at three in the morning, there is a specific person whose job it is to be woken up.
On July 1, you are an attending.
The clinical knowledge did not change overnight. What changed is that the question "am I doing this right?" no longer has an addressee.
There is nobody whose job it is to answer. Your new colleagues are pleasant, busy, and evaluating you. Your co-residents have scattered to eleven different states. The program that shaped you for four years has, from an institutional standpoint, completed its transaction with you and moved on to next year's class. Your employer has an onboarding process, which will teach you the billing system and the EHR.
And the questions that actually arrive in that first year are almost never the ones training prepared you for:
Should I sign this contract? Is this non-compete normal? Was that complication my fault or bad luck? Why is my RVU number lower than the person down the hall? A patient wrote something horrible about me online. I have not slept properly in four months. I think I made the wrong specialty choice. I got a letter from a lawyer.
None of those are clinical questions. All of them are career-defining. And the profession has built almost nothing for them.
This is Year Zero, and the data on what happens in it should worry anyone who cares about the medical workforce.
The collapse in first-job tenure
The most striking finding is also the most recent, and it represents a genuine break from historical patterns.
According to MGMA and Jackson Physician Search data reported through the AMA:
- Physicians who completed training within the last six years average under two years in their first job.
- The comparable figure for physicians overall is about six years.
- Nearly 60 percent of physicians leave their first post-training job within three years.
- More than 25 percent consider leaving within the first year.
- Only 45 percent of administrators reported retaining at least 75 percent of new physicians after three years.
First-job tenure has fallen by roughly two thirds in a single professional generation.
And this is not merely job-hopping within medicine. Research published in The Permanente Journal in 2025 found:
- Physicians are now leaving clinical practice at a mean age of 48.1, compared with 57.1 for the 2008 cohort. That is nine years of career, gone.
- 11 percent never practised at all after completing residency.
Consider the last figure carefully. Roughly one in nine physicians completes a decade or more of training, at enormous personal and public cost, and never enters practice.
Meanwhile, 72 percent of physicians report having no professional mentor, while 56 percent report having mentored others. The profession gives more mentorship than it receives, which is arithmetically impossible to sustain and tells you something about where the effort is directed.
What the transition actually feels like
The literature has recently started naming this window properly. The phrase is "transition to independent practice," and the qualitative research is unusually vivid.
A 2025 study in the Journal of Graduate Medical Education gathered 518 responses from 186 participants including new-to-practice attendings, program leaders, and practice managers. Three themes were salient across every stakeholder group: confidence, imposter syndrome, and workload.
That cross-group consistency matters. It means this is not new attendings being fragile. The program directors see it. The managers see it. Everyone identifies the same thing and nobody owns it.
Research in the Journal of General and Family Medicine describes the "abrupt transition to independent practice" producing burnout-like episodes.
And a 2024 study in Perspectives on Medical Education captured the texture with a line that will be familiar to anyone who has lived it. A new attending, after a difficult conversation with a colleague, asking themselves:
> "Did this one 30-minute conversation undo the whole vacation?"
That is the psychological reality of Year Zero. Every interaction feels load-bearing. You have no accumulated reputation to absorb a bad day. You do not yet know which norms are real and which are that particular department's habits. You cannot tell whether the thing that just happened was normal or a disaster.
The same study documented that women who stayed at their training institution were addressed by first name in ways male peers were not, and described friendships broken by clinical disagreements. The transition is harder, and differently hard, depending on who you are.
The structural diagnosis: dissolution, not absence
Here is where most analysis goes wrong, and where the fix has to start.
The standard framing is that new attendings lack mentors, and the standard prescription is a mentorship program. Assign a senior physician. Schedule quarterly meetings. Most institutions have tried some version. Almost none report durable success, and the 72 percent no-mentor figure suggests the aggregate effect is small.
The framing is wrong.
New attendings do not lack a support network. They had an excellent one, and graduation dissolved it.
Think about what a residency cohort actually is. It is four to twelve people who watched each other work under fatigue, uncertainty, and consequence for three to seven years. They know each other's clinical character at a resolution that no assigned mentorship can approach. They are peers rather than evaluators, which means you can admit uncertainty to them. They are contemporaries, so they face the same decisions at the same time: the same first contracts, the same first complications, the same questions about children and geography and whether this was the right choice.
That is not a nice-to-have. It is the highest-functioning professional support structure most physicians will ever have, and it is delivered for free by the structure of training.
Then, on a single day, it is dissolved by geography, and no mechanism exists to maintain it.
The evidence that the cohort tie is uniquely valuable is not sentimental. Research on referral behavior found that physicians preferentially refer to people they trained with at 27.0 percent versus a 21.4 percent baseline, and that shared medical school produced no effect at all (21.6 versus 21.4 percent, p = 0.965). Only residency and fellowship co-training moved behavior. Trust built by watching someone work under pressure is categorically different from trust built by proximity.
So the prescription is not "build mentorship." It is prevent dissolution.
Why nobody owns the three-year window
Follow the incentives and the vacuum explains itself.
The residency program owns the graph and loses interest at graduation. Its accountability metrics end at completion. Whatever alumni structure exists is typically operated by the institution's advancement office, whose actual purpose is fundraising. Some programs, remarkably, outsource alumni networking entirely to third-party platforms.
The employer has a strong retention interest and no access to the graph. It cannot connect you to your co-residents; it does not know who they are. Its onboarding covers billing, EHR, and compliance, because those are what the organization needs from you. A rapid review of transition-to-practice curricula found the content dominated by finance and administration at around 37 percent.
The specialty society operates at the level of specialty, not cohort, and its early-career section is typically a discount on conference registration.
The individual is the only party with both the interest and the graph, and is precisely the person with the least time and the least idea that the network they are losing was an asset rather than a phase.
So the highest-risk three years in a physician's professional life belong to nobody at all.
What it costs
Physician turnover is expensive in ways that are widely estimated and rarely itemized. Industry estimates commonly place the cost of a physician departure between $500,000 and $1 million when recruitment, ramp-up time, and lost clinical revenue are counted. Those figures come from recruiting and consulting sources rather than peer-reviewed research and should be treated as directional.
But apply even a conservative version. With roughly 40,000 physicians completing training each year in the US, and nearly 60 percent leaving the first job within three years, the volume of early-career turnover is enormous. Reducing it by even a few percentage points represents hundreds of millions of dollars in employer value.
AAPPR reported median physician turnover of 7.3 percent in 2025. The early-career slice runs several times higher.
Then there is the cost that never appears on anyone's balance sheet: the 11 percent who never practise, and the nine years of career lost at the other end as exit age falls from 57.1 to 48.1. Society paid for that training. It is being written off quietly.
And the individual costs are real and immediate. A poorly understood first contract, with an unfavorable non-compete or a compensation structure the physician did not model correctly, can cost six figures over its term. An unsupported first complication can start a spiral that ends a career. These are decisions made once, early, usually alone, with no access to anyone who has made them before.
Why AI does not touch this
There is an irony in the current moment worth naming.
The clinical questions of Year Zero have largely been solved. A new attending in 2026 has extraordinary tools: 81 percent of physicians now use AI in practice, and a well-organized, well-cited answer to almost any clinical question is available in seconds, at any hour. The "what is the management of this condition" problem, which used to require finding a senior colleague, is essentially handled.
What remains is everything else. And everything else is the hard part.
Should I sign this? Was that my fault? Is this normal? Am I going to be all right?
No model answers those, and no model can, because the answer requires someone who has been three years ahead of you on the same road and is willing to say what it was actually like.
AI has removed the questions that a network was only incidentally good at, and left untouched precisely the questions that a network is the only possible answer to. The value of a peer cohort has not fallen with AI adoption. It has concentrated.
What actually works
Some of this is fixable by individuals, and some requires institutions to act. Both are worth stating.
If you are in Year Zero right now
Understand that the feeling is structural, not personal. Confidence, imposter syndrome, and workload were the three dominant themes across all stakeholder groups in the JGME research. The people who trained you see it. The managers see it. You are experiencing a documented transition, not a personal failure of preparation.
Rebuild the cohort deliberately, in the first six months. This is the single highest-value action available to you, and the window matters: the graph is still warm at six months and substantially cold at two years. Not a group chat that decays. A standing, scheduled thing. Four people, one video call a month, with an actual agenda: what happened this month, what decision are you facing, what do you need. It sounds slight. It is the closest available replacement for what you just lost, and peer-mentorship models for new attendings have been shown feasible in the literature and simply never scaled.
Find someone three to ten years ahead, not thirty. The senior physician who trained in a completely different era of medicine is valuable for some things and useless for the specific questions of the first three years. The person five years ahead remembers what the first contract felt like and can tell you what they got wrong.
Get the contract reviewed by someone who has signed one. Both a healthcare attorney and a physician who has negotiated in your specialty and market. The attorney tells you what the words mean. The physician tells you what is normal, which is the part you actually cannot get anywhere else.
Write down what you did not know. Keep a running list through year one of every question you had to figure out alone. It will be long. It is the most valuable document you will ever produce for the person coming after you, and it will make you an unusually good mentor at year five.
Do not make an irreversible decision in month four. More than a quarter of new physicians consider leaving within the first year. Some of those decisions are correct. Many are made in the worst possible frame of mind, during a documented transition period, without a single peer to check the interpretation against.
If you run a residency or fellowship program
Own graduation day. It is the one moment your cohort is assembled, emotional, and about to scatter. Most programs currently use it to hand graduates to the advancement office for a donation pipeline. That is an extraordinary waste of the most valuable professional graph in medicine.
Instead: establish a class covenant. A real roster with real contact details that the class maintains. A stated norm that this class answers each other. A scheduled first check-in six months out, before the network decays.
The cost is one hour on one day and someone taking responsibility for the roster. The evidence suggests it would be the highest-return hour in the entire training program.
Track your graduates for three years. Not for fundraising. To know what happens to them. Most programs genuinely do not know how many of their graduates left their first job or left medicine, which is an extraordinary gap in a training program's understanding of its own output.
If you employ new physicians
Recognize the actual retention problem. With nearly 60 percent leaving within three years and turnover costs that industry sources put in the high six figures, early-career retention is likely your largest recoverable workforce cost, and your onboarding process is teaching them the EHR.
Buy the cohort, not the mentor. Assigned mentorship has a poor track record. Funding your new hires to maintain their existing training cohort, including people who work for competitors, sounds counterintuitive and is far more likely to work, because you are supporting a relationship that already functions rather than manufacturing one that does not.
Give them someone who is three years ahead, internally. Not the department chair. The person who joined in 2023 and remembers.
Frequently asked questions
How long do physicians stay in their first job after residency? Under two years on average for those who finished training within the last six years, compared with about six years for physicians overall, according to MGMA and Jackson Physician Search data. Nearly 60 percent leave the first post-training job within three years and more than 25 percent consider leaving within the first year.
Why do so many new physicians leave their first job? The evidence points to a combination of the abrupt loss of professional support at graduation, first contracts signed without adequate guidance, and the documented psychological difficulty of the transition. Research in the Journal of Graduate Medical Education found confidence, imposter syndrome, and workload salient across new attendings, program leaders, and practice managers alike.
Is imposter syndrome normal for new attendings? It is one of the three dominant themes identified across all stakeholder groups in transition-to-practice research, reported not only by new attendings but by the program leaders and managers observing them. It is a documented feature of the transition rather than an individual deficiency.
At what age do physicians leave clinical practice? Research in The Permanente Journal (2025) found a mean exit age of 48.1, compared with 57.1 for the 2008 cohort, a nine-year decline. The same research found 11 percent of physicians never practised after completing residency.
Do physicians have mentors? Mostly not. Survey data from MedCentral's 2026 report found 72 percent of physicians have no professional mentor, while 56 percent have mentored others. The profession supplies more mentorship than it receives.
What should a new attending do in the first six months? Deliberately rebuild the training cohort while the connections are still warm, find someone three to ten years ahead rather than thirty, have any contract reviewed by both a healthcare attorney and a physician who has negotiated in the same specialty and market, and avoid irreversible career decisions during the documented transition period.
Why doesn't assigned mentorship work well? Because the problem is usually misdiagnosed. New attendings are not short of senior advice; they have lost a peer cohort of contemporaries who faced identical decisions at the same time and to whom uncertainty could be admitted without evaluation. An assigned senior mentor addresses a different need, and the referral literature suggests trust built by working alongside someone under pressure is categorically different from trust assigned administratively.
The bottom line
Roughly 40,000 physicians finish training in the United States each year. On July 1, each of them loses, in a single day, the most functional professional support structure they will ever have, and nothing is built to replace it.
Within three years, nearly 60 percent will have left the job they started. Some will leave medicine entirely, at a mean exit age that has fallen nine years in a generation. Eleven percent never start at all.
The profession's answer to this has been mentorship programs, which most physicians do not have, and onboarding, which teaches the billing system.
The cohort that made training survivable is not gone. It is scattered across eleven states, still perfectly willing to answer, and connected by nothing but a group chat that will go quiet by March.
Nobody is accountable for the three years after training. That is not because the problem is hard. It is because the program's responsibility ends at graduation, the employer never had the graph, and the only person with both the interest and the relationships is the exhausted new attending who does not yet know what they are losing.
Part of a series on the missing professional infrastructure of healthcare. Previously: The Zero-Enrolling Site Is a Search Failure
Evidence note: first-job tenure and turnover figures come from MGMA and Jackson Physician Search data reported via the AMA, and AAPPR benchmarking. Career exit age and post-residency practice rates come from The Permanente Journal (2025). Transition experience research comes from the Journal of Graduate Medical Education (2025), Perspectives on Medical Education (2024), and the Journal of General and Family Medicine (2026). Mentorship prevalence comes from MedCentral's 2026 survey, a trade survey rather than peer-reviewed research. Physician turnover cost estimates of $500,000 to $1 million per departure come from recruiting and consulting industry sources and are directional rather than measured.
