The Person You Labeled Toxic May Be the Last One Still Trying
Every leader standing in a hospital that has gone bad arrives at the same conclusion first. There are two or three people making this place unbearable, and if they were gone the culture would recover.
Sometimes that is true. More often it is the fastest available explanation, and it is wrong in a way that costs a great deal. The people you have flagged as toxic are frequently the people who have not yet stopped raising the problem. They have raised the crash cart that keeps coming back unrestocked four times. They have raised the handover that keeps failing, and the doctor who does not answer calls. Nothing changed, so the register changed. Now they are sharp in rounds, they are audible in the treatment area, and they have an opinion about leadership that they no longer bother to keep quiet.
Years of watching emergency and specialty hospitals have convinced me that what you are looking at is usually not a character problem. It is what advocacy looks like on its fourth attempt. If you are the medical director, the practice manager, or the lead DVM who has looked around and decided this is no longer acceptable, that decision is the right one. What follows is the order of operations, because running these steps out of sequence is why most culture resets in veterinary hospitals fail by week six.
What the research actually says
Culture is not a soft variable in attrition. Donald Sull, Charles Sull and Ben Zweig, writing in MIT Sloan Management Review in January 2022, analyzed 34 million employee profiles alongside 1.4 million Glassdoor reviews and found that a toxic corporate culture was by far the strongest predictor of industry-adjusted attrition, roughly ten times more important than compensation in predicting turnover. Their follow-on work identified the five attributes that actually make a culture toxic: disrespectful, noninclusive, unethical, cutthroat, and abusive. In a later analysis synthesizing eleven meta-analyses, the strongest predictors of toxic workplace behavior were toxic leadership, toxic social norms, and poor work design, with leadership the single strongest factor.
The veterinary numbers are worse than most leaders assume. The Merck Animal Health Veterinary Team study of nonveterinarian practice team members, published in JAVMA in October 2024 from 2,271 completed surveys, found 27% experiencing high to very high burnout and 71% scoring high or very high on exhaustion specifically. Twenty percent were suffering serious psychological distress, against 6.3% of employed, college-educated US adults in the comparable national survey. Working in a negative clinic culture came out as one of the predictors of higher burnout, alongside long hours and student debt.
There is now veterinary-specific evidence for the mechanism underneath all of that. A 2026 study of 367 veterinary professionals in Australia and the UK, published in Veterinary Record, found that psychologically safe teams have lower intention to leave, and that the link runs through the quality of the supervisor's feedback rather than through co-worker support. Peer support is not a substitute. The pathway runs through the leader, and feedback skill is trainable.
Then there is the behavior itself. A 2025 study in Veterinary Sciences surveying 632 people working in Dutch veterinary clinics found that 69.6% had experienced at least one form of transgressive behavior in the prior year. Clients were the most common source of both aggression and bullying, but a substantial share of the bullying came from inside the building, with 44.2% attributed to colleagues and supervisors. Discrimination was overwhelmingly internal, 72.3% of it attributed to colleagues and supervisors. Support staff reported the highest prevalence of any group at 73.4%, ahead of veterinarians at 67.1%.
The last piece is the one that should stop you. Ethan Burris, in the Academy of Management Journal in 2012, ran a set of field and experimental studies on how managers respond to employees who speak up. Employees who used challenging voice, meaning they questioned how things were being done, were rated by their managers as poorer performers, and their ideas were endorsed less often, than employees who used supportive voice. Perceived loyalty and perceived threat were the mechanisms doing the work. The people identifying problems were being scored down for identifying them.
That is the bias you are working against when you sit down to write your list of toxic people.
Step one: separate the toxic from the exhausted advocate
Leaders systematically over-attribute a team member's behavior to that person's character and under-attribute it to the conditions they are working in. This is the core finding of Green and Mitchell's attributional model of leader-member interactions, reviewed across three decades of research by Martinko, Harvey and Douglas in The Leadership Quarterly in 2007. You see the behavior, and you do not see the four unanswered emails behind it, so you code the person. Before you act on your list, run each name through five questions.
Does the behavior track to a specific unresolved issue, or does it show up regardless of topic? An advocate is repetitive about one or two things. A genuinely corrosive person is negative across everything, including the things that were fixed last month.
Do they escalate to you, or only sideways to peers? Someone still bringing it to you has not given up on the system. Someone who only runs it in the treatment area has, and that is a later stage of the same problem rather than a different problem. The conversation avoided laterally often becomes a complaint vertically, and the reverse is also true: when the vertical conversation stops paying, everything moves sideways.
When you did fix something they raised, did the behavior change? This is the single most useful test and almost nobody runs it, because it requires being able to point at something you fixed.
Is the complaint about the work, or about a person's standing? "The crash cart has come back short twice this month" is about the work. "She should not have been made a lead" is about standing.
Where do junior staff go? Watch a new nurse with a problem at 2am. Toxic people get avoided, advocates get consulted, and your team has been sorting for that difference without you for months. There is a real category of person who is genuinely destructive, who enjoys the friction, and who recruits against you in the parking lot. That person exists and needs to be managed out. You cannot identify them reliably, though, until you have separated them from the three people who are simply out of patience. When you do not have structured visibility, the loudest signal starts to feel like the truth.
What this costs you if you skip it: exit the wrong person and you remove the last internal signal you had. The quiet that follows will feel like improvement for about eight weeks.
Step two: name your own contribution before you name anyone else's
Detert and Burris, studying 3,149 employees and 223 managers, found that managerial openness predicted employee voice more consistently than transformational leadership did, mediated by whether people felt safe enough to speak. What moved voice was something plainer than vision or charisma. People spoke up when they believed the manager would take in what they said and do something with it. The effect was strongest among the best-performing employees, which is to say the ones you can least afford to lose are the ones most sensitive to whether the door is real.
Belief like that does not get restored by announcing a new culture. It gets restored when the person at the top goes first and is specific. "I know things have been hard" is a weather report. Specific means naming the decision, the cost, and the change.
The structure is three sentences: what I did or did not do, what it produced for you, and what changes by when.
"For eight months I knew treatments were getting missed on the busy shifts and I never changed how they were assigned. What that produced is that doses went in late, people charted them afterwards because they knew they were late, and two of you told me you had started looking for other jobs. Starting on the fifteenth, treatments are assigned by name at rounds and the sheet gets checked at shift change. I will tell you at rounds every Monday whether it held."
That is the whole move, and it works because it is falsifiable. The team can check you on the fifteenth.
The failure mode here is vulnerability with no commitment attached. A leader who admits fault and then changes nothing has taught the team something worse than silence, which is that even confession produces nothing in this building. Do not open this door unless you are prepared to be measured on the other side of it.
What this costs you if you skip it: every step that follows reads as a management initiative, and your team has seen management initiatives.
Step three: let the team write the non-negotiables and the consequences
Colquitt and colleagues, in a meta-analytic review of 25 years of organizational justice research across 183 studies published in the Journal of Applied Psychology in 2001, found procedural justice correlated with organizational commitment at .57 and with trust at .61, and negatively with withdrawal at -.46. Procedural justice means having a voice in how a decision gets made, and people comply with rules they helped write at a different rate than rules that arrived by email.
Run this as a working session rather than a survey. Two rules govern it.
First, non-negotiables must be observable behaviors and not values. "Respect" cannot function as a non-negotiable, because nobody can be held to it and everybody believes they are already doing it. These are non-negotiables:
→ Rounds happen at shift change, out loud, patient by patient, even when we are two hours behind.
→ The surgery suite gets cleaned after every surgery. The anesthesia nurse owns it.
→ If you disagree with a plan, you say it in the moment to the person, and not in the treatment area afterward.
→ When a case goes badly, we debrief it before anyone goes home, and we ask the people who were in the room how they are doing.
→ A mistake does not become a story told later to people who were not in the room.
Five to seven of these, and no more, because a list of fifteen is a list nobody remembers.
Second, and this is the half that gets skipped, the team writes the consequence ladder too. What happens the first time, the second, and the third. Ask it directly: if someone breaks the handover standard, what should happen, who says it, and when? Teams routinely write a stricter ladder than leadership would have, and they stop treating accountability as something done to them once they have specified it themselves.
A workable ladder usually lands close to this. First time, a peer names it directly, that shift. Second, a documented conversation with the MD. Third, a written expectation with a review date. The team will argue about who owns step one, and that argument is the most valuable twenty minutes of the session, because step one is where every hospital's accountability actually breaks.
Then tie it to the strategic plan out loud. If the plan for this year is to open a second overnight shift or bring specialty referral in-house, say so, and ask what has to be true about how the team operates for that to work. People behave differently when they know what they are being asked to hold together.
What this costs you if you skip it: rules that arrive by email get followed by the people who were already following them, and nobody else changes.
Step four: give the loudest people a job
This is the step leaders resist, and the evidence against them is strong. A Cochrane review by Flodgren, O'Brien, Parmelli and Grimshaw, updated in 2019 across 24 studies involving 3,005 healthcare professionals and more than 29,000 patients, found that interventions using local opinion leaders produced a 10.8% absolute improvement in compliance with evidence-based practice. An opinion leader in that research is defined purely by who the group already listens to, which in most hospitals has very little to do with the org chart.
Every veterinary hospital has these people and they are usually not on the org chart. The senior tech everyone pages before they page the doctor. The CSR who knows which clients are going to escalate. The associate the new grads text at midnight.
They are also, reliably, the people with the strongest opinions about what is broken, and that is not a coincidence. Being trusted by the floor and being vocal about the floor are the same trait pointing in two directions.
Give them a defined scope rather than a title. "Culture champion" is nothing. "You own the handover standard for the next ninety days, you define what a complete handover contains, you tell people when it did not happen, and you and I review it on the fifteenth of October" is a job, bounded, with a date on it.
The mechanism matters more than it looks. A person handed ownership of the thing they were complaining about stops being a critic of your system and becomes an author of it. If they refuse the ownership while continuing the criticism, you have learned something important, and you have learned it cleanly.
What this costs you if you skip it: their influence does not disappear when you leave them without a role. It runs in a direction you do not control.
The clinical consequence nobody puts in the culture conversation
In a hospital where speaking up is expensive, the information that prevents an error stops moving. That is the entire mechanism. The nurse who thinks the potassium rate looks wrong does not ask, because the last three times she asked, the answer had an edge on it. The new grad does not call for help on a case beyond her, because the culture has taught her that asking is a status event. The overnight team does not flag the deteriorating patient at 3am because they have learned that waking that particular doctor produces more risk than the patient does.
In human medicine, communication failure is one of the most frequently identified contributing factors in sentinel events, and The Joint Commission has been publishing that finding for two decades. Veterinary medicine has no equivalent reporting body, which means the same failures happen and are never counted. They get absorbed as a bad outcome, a difficult case, or a client complaint.
Amy Edmondson's psychological safety research has held on one point for three decades: teams that report more errors are frequently the safer teams, because reporting is happening at all. If your incident log has gone quiet since the culture got bad, read that as the signal going out rather than as fewer errors.
The team felt it before you did
By the time a leader concludes the culture is toxic, the team has been operating inside that conclusion for months. The JAVMA data puts exhaustion at 71% among nonveterinarian team members, and the Dutch study found support staff experiencing transgressive behavior at a higher rate than veterinarians. Your nurses, assistants and CSRs are absorbing the culture at a higher rate than you are, and they have less control over any of it.
This is why discretionary effort breaks well before performance does. People still do the job, and they stop doing the extra thing: the callback that was not required, the mentoring of the new tech, the thirty seconds of extra history that changes the plan. That withdrawal is invisible on every metric you track, and it is the earliest reliable indicator you have.
It is also worth naming what exhaustion is actually made of here. In the New Zealand retention research published in The Veterinary Nurse in 2026, 80% of allied veterinary professionals still in clinical practice had considered leaving it, and when they were asked what their workplaces most lacked, the top answer was not pay or staffing. It was good management and leadership with good communication skills. The people you are worried about losing have already told the profession what would keep them.
People do not burn out because the work is difficult. They burn out because their effort never leads to improvement.
The cost leadership does not calculate
A peer-reviewed study in Frontiers in Veterinary Science put the all-in cost of replacing one veterinarian at roughly $100,000, and with today's higher per-DVM production a prolonged vacancy pushes that well past $150,000. Replacing one credentialed technician runs somewhere between $20,000 and $60,000 depending on how long the seat stays open. Layer the MIT Sloan finding on top, where culture outpredicts pay by a factor of ten, and the retention bonus you are considering is competing with a cause it cannot reach.
The harder number is capacity. A hospital that loses two experienced nurses does not simply pay to replace them. It runs fewer hospitalized cases safely, it holds fewer overnight beds, and it pushes work onto the doctors, who then produce less per shift. Revenue per DVM falls because the support structure underneath the DVM fell first.
The reason that rarely gets diagnosed is structural rather than anyone's oversight. Attrition is not a line item. It arrives on the P&L already converted into something else, usually doctor productivity, and by then it no longer looks like a culture number at all. A bad system will beat a good person every time, and it will do it quietly enough to show up in the wrong column.
What this looks like in practice
The sequence is not optional. Run it in this order.
Week one. Write your list of the people you believe are the problem, and run each name through the five questions in step one. Do not act on any name yet.
Week two. Do step two in front of the whole team, at a real meeting, with a date attached to the commitment. One specific failure, one specific cost, one specific change.
Weeks three and four. Hold the working session for the non-negotiables and the consequence ladder. Do not pre-write the list and ask for approval, because the team will read that instantly and the exercise will be worth nothing.
Week five. Assign ownership. Two or three people, one standard each, a ninety-day scope, and a review date on the calendar.
Week six. Report on your own commitment publicly and say whether you held it. If you did not, say so and say why. This is the checkpoint where the reset either becomes real or becomes another thing that was announced.
Weeks ten to twelve. Now revisit your original list. Some of those names will have changed behavior because the conditions changed. The ones who have not, with a fixed environment and a defined standard and a peer ladder in place, are your actual problem, and you can now address them with evidence instead of impression.
Cultures go toxic when a set of behaviors stays survivable long enough to become normal, and when the people who said so early get coded as difficult until they stop saying it. In leadership, silence is almost always read as permission, and the standard you walk past is the standard you accept. The behavior you are seeing now is the record of what leadership did not answer.
Which means the fix does not start with a personnel decision. It starts with a leader deciding to be measurable in front of the people who stopped believing they would be. When a problem becomes predictable, it stops being random, and predictability means structure is involved. If structure created it, structure can fix it.
Every month you spend managing the symptom is another month of advocates converting into cynics, and that conversion runs one direction. The person still willing to be sharp with you in rounds is still engaged. When they go pleasant and quiet, they are already gone, and they are working out the notice period in their head. Find out what is actually breaking before you decide who is breaking it.
The Stability Audit is a free ten-minute diagnostic that scores your hospital across five domains: Role Clarity, Accountability, Operational Reliability, Team Culture Stability, and Leadership Structure. You get an overall score, a breakdown by domain, and the three priorities your lowest scores point to, all of it immediately. It will tell you whether what you have is a people problem or a structure problem, and most of the time it is the second one.
If you want to go further into the material itself, there is a free preview of TRIAGE Leadership's Module 1, which is the module on toxicity.
