Your copy of The Clinic Retention Guide is on its way

So here is a model that can be reconstructed. Every input is visible and every input is arguable. Change any of them and re-run it against your own clinic; that is the point of showing the working.

The assumptions

Where the money actually goes

This is a stated-assumptions model, not an industry statistic. It is built from the inputs in the table above and nothing else. Quoted without its assumptions it becomes the same kind of unfootnoted number this section exists to replace.

Run the same model on the version where an agency is used and the ramp takes twelve months rather than thirty weeks, and the figure lands closer to $162,000.

The finding that matters

Look at where the cost sits. The vacancy is small. The agency fee, in the bad version, is small. Roughly three-quarters of the cost of losing a clinician is the ramp — the months the replacement spends below full caseload while being paid in full.

That relocates the whole problem. Turnover looks like a recruitment problem, so owners respond by getting better at recruitment. But the expensive part is not finding someone. It is the speed at which the person you found becomes productive, and that is a mentoring and role-design question.

Which gives the argument a commercial edge that runs the other way, too. A senior clinician paid a mentoring component of, say, $10,000 a year, who moves a new clinician from a twelve-month ramp to a fifteen-week one, returns something in the order of $70,000 of contribution. The mentoring cost is funded several times over out of the ramp it produces. The leadership structure in section 5 is not an overhead. It is the thing that pays for itself first.

Three pieces of Australian evidence, read together, say something more useful.

Across nine regulated health professions, pay is not in the top five reasons people leave. In a 2025 study of 25,752 completed responses, the top five were mental burnout (32.9%), retirement (30.5%), lack of recognition or feeling undervalued (28.5%), lack of professional satisfaction (27.9%) and work no longer feeling fulfilling (25.1%). The reasons people gave for staying were enjoyment of the work (59.0%), the work feeling meaningful (53.9%) and flexibility (50.8%).

Two honesty notes on that study, because they matter if anyone challenges you with it. Physiotherapy is not in the dataset — the correct phrasing is always across regulated health professions, never physios. And respondents over sixty were over-represented (20.7% of respondents against 9.0% of registrants), which inflates retirement as a reason.

The physiotherapy-specific research says something different about money, and it should be reported honestly. A 2026 study run by Ahpra with the Physiotherapy Board and the Australian Physiotherapy Association names poor financial sustainability first among the reasons practitioners give for leaving, followed by limited career progression pathways, then workplace conditions and burnout. Reasons to stay: colleagues, community, and career advancement. The retention strategies practitioners themselves named were improved remuneration, more career progression opportunities, and burnout prevention. It is qualitative — twenty-one interviews plus open survey responses — so it reports what practitioners say, not a rate. It is also the regulator’s own peer-reviewed research naming career progression as a retention strategy, which is a different order of evidence from a consultant’s opinion.

And roughly half of mid-career physiotherapists cannot see a career pathway in their own profession. The Australian Physiotherapy Association’s 2025 Workforce Census puts the figure at 52% who believe a viable pathway exists. The same census reports average earnings of $102,000 in 2025, up 7.4% since 2023, and finds that 93% of physiotherapists who feel fairly paid report job satisfaction, against 44% of those who do not.

Reconciling the two

The honest reading of those three sources together is not “it isn’t about the money”. It is narrower and more useful:

Meaning is why people are in the profession. Money is why they leave it. And money is the simplest structural lever available — within reason.

Within reason is doing work in that sentence. Pay that sits below what the market and the award will bear is a live retention risk and should be fixed before anything else in this document is attempted. Pay that is competitive stops being the lever. Past that point, more money buys weeks, not loyalty, because the thing the person is actually short of is a direction.

That is the reframe the whole workshop runs on: retention is a pathway problem. The question is not whether your people are happy. It is whether every one of them has somewhere to go next, and whether they know what it is.

Why paying your own people is the lowest-risk money you spend

There is a commercial asymmetry here that most owners have never articulated, and it is worth stating plainly because it changes what a pay conversation feels like.

Losing a good team member over money is therefore a decision, not an accident. A weak one, maybe — and that is a different and entirely acceptable outcome.

The mechanism underneath

Self-determination theory, developed by Deci and Ryan over roughly forty years of research, holds that people do not need motivating in the way the word usually implies. They have three basic psychological needs that a workplace either meets or does not: autonomy, competence and relatedness.

The reason this matters for retention is what it predicts about promotion. The standard reward for a high-performing clinician is a leadership title. Consider what that does to all three needs at once:

• Competence resets. They were excellent at the thing they have just stopped doing full-time, and they are a beginner at coaching, difficult conversations, and holding a standard.
• Autonomy shrinks, despite appearing to expand. More decisions land on them, more of those decisions have consequences for other people, and the range of things they can act on without checking usually gets narrower, not wider.
• Relatedness shrinks. A person who becomes a leader has materially fewer peers inside the practice than they had the week before. This one is rarely spoken about and universally felt.

A promotion that shrinks all three needs simultaneously is not a reward. It is a well-intentioned demotion in everything except title and pay, and it is one of the most reliable ways to lose a good clinician within eighteen months of promoting them.

Sections 4, 5 and 6 exist to stop that happening.

What it is

Psychological safety is a shared belief, held by the people on a team, that the team is safe for interpersonal risk-taking. The construct was defined by Amy Edmondson of Harvard Business School in a 1999 study of hospital teams.

In a clinic it means something narrower and more useful than the phrase usually suggests: whether a clinician will tell you about a problem while it is still a problem, rather than after it has become a resignation, a complaint or a patient issue.

It is not about being nice. It is not the absence of hard conversations, it is not everyone getting along, and it is not the same as a friendly team. A clinic can be pleasant, loyal and quiet and have very little of it. Psychological safety is compatible with — and in the evidence, dependent on — high standards and direct feedback.

The single sentence that carries the argument:

If people don’t feel safe, they don’t stop having concerns. They stop telling you.

Which is why silence reads as contentment right up until the resignation lands. The failure sequence runs in a specific and boring order: someone raises something, nothing visibly happens, they update their model of what raising things is worth, they stop raising things, and the owner reads the last step as everything being fine.

The four phrases

Psychological safety is difficult to measure directly and reasonably easy to observe indirectly. Four sentences do most of the work, because each one costs the speaker something to say in front of a person who controls their roster, their pay and their progression.

• “I made a mistake.”
• “I need help.”
• “I don’t know.”
• “I disagree — there’s a better way.”

A team that says these routinely has priced interpersonal risk at close to zero. A team that never says them has not run out of mistakes, uncertainty or disagreement. It has run out of reasons to voice them.

Two observations that make the diagnostic sharper:

• Measure it on your seniors, not your grads. A new graduate in week one has complete permission to not know things, and generally uses it. The question that tells you something is whether a clinician six years in feels able to say they need help. That is the one where the professional cost of saying it is real.
• The fourth phrase is the ceiling. Disagreement offered with an alternative attached is the difference between a team that thinks and a team that nods. Most clinics never get there, and most owners never notice, because a room full of agreement is a comfortable room to be in.

The diagnostic question: when did you last actually hear one of those four sentences from someone on your team? Not whether they would say it — whether they did. This week, this month, or can’t remember.

“Can’t remember” is the most common honest answer.

One structural note on how safety gets built, because it is the part owners skip: nobody takes an interpersonal risk in a room where the person with the power never has. If the leader says “I got that wrong” about something real and recent, before anyone else does, mistake-making becomes a normal event rather than an exposure. If the leader never says it, the four phrases are a poster.

The matrix: two dials, not one

The most common error in how psychological safety gets discussed is treating it as a single dial where more is always better. Edmondson’s own model, set out in The Fearless Organization (2019), uses two axes: psychological safety, and accountability for high standards.

Two of these quadrants are routinely misread.

Apathy is the worst position, not anxiety. In anxiety there is at least a force acting on the person. In apathy nobody is holding them to anything and nobody is catching them, and there is no mechanism inside the quadrant that produces movement.

Comfort is the trap most good clinics sit in, and it is the reason this section exists. It looks like success. Tenure is decent, the team is friendly, nobody is upset, and “we’re a family here” is usually true. What is missing is the second dial. A comfortable, unchallenged high performer is a bored high performer, and boredom in a high performer is a resignation with a lead time on it. Comfort also decays: without challenge it slides toward apathy rather than holding position.

The route out of comfort is accountability, not pressure. Adding challenge to a high-safety environment moves a team toward learning. Adding challenge while removing safety moves them toward anxiety, where the observable result is not better performance but better concealment. Anxiety is also a learning inhibitor in its own right — a clinician worried about their caseload numbers has less capacity available for the skill development that would fix their caseload numbers.

What actually builds it — and it is not the leader’s tone

This is where most published advice on the topic becomes unreliable.

The largest meta-analysis available — Frazier and colleagues (2017), 136 independent samples, over 22,000 individuals — ranks the antecedents of psychological safety. The order is the finding.

Role clarity is the largest single antecedent, and it is larger than anything the leader says or how they say it. That is a design finding, not a communication finding. It points at who owns what, how the work is built, and what people are measured on — not at the owner’s warmth.

The mechanism is visible in any clinic past four or five clinicians. Where a role boundary is ambiguous, two people end up with the work sitting between them. Who owns rebooking. Who chases the patient who dropped off. Whose job it was to tell the front desk. That produces friction, and friction between people who work together every day does not stay on the task — task conflict bleeds into relationship conflict at ρ = .54. A relationship problem that has been running six months does not present as a role-design problem. It presents as “a culture issue”, which is where it becomes unfixable, because the diagnosis is wrong.

The sequence, stated once:

Unclear role → friction → conflict → they leave. And it is usually your best people who leave, because they have the most options and the least tolerance for it.

What the evidence does not support

We state the limits because a document that carries only the flattering half of the evidence is not worth quoting.

• Psychological safety cannot be installed. No intervention has been shown quantitatively to raise it. Two independent evidence reviews reach the same conclusion: it remains unclear how to definitively increase it. Anyone selling a psychological-safety intervention is selling ahead of the evidence. The defensible position is that it behaves like a condition, moved by role clarity and work design — the two largest antecedents — rather than a deliverable you buy.
• The correlations do not prove causation. Frazier and colleagues report that only 13% of extracted correlations came from different sources, and that effect sizes ran 27% higher where data came from the same source at the same time. The authors name common-method bias themselves. The literature is mostly cross-sectional.
• In healthcare, the link to objective patient-safety outcomes is equivocal. A 2025 review found only nine qualifying studies, five significant, with no pooled effect available.

What the evidence does support robustly is that psychological safety correlates with learning behaviours (ρ̂ = .62 individual, .52 group), information sharing (.52 / .50) and task performance (.43 / .29). Those are real, and they are worth having. They are also not the same claim as “run a workshop and your team gets safer”.

The Australian layer

Most writing on this topic is American and stops at culture. In Australia it has a statutory dimension. Safe Work Australia’s Model Code of Practice: Managing psychosocial hazards at work (July 2022) names conflict and poor workplace relationships among fourteen common psychosocial hazards, and every Australian jurisdiction now has psychosocial regulations in force. Victoria’s Occupational Health and Safety (Psychological Health) Regulations 2025 go further and provide that where controls are combined, training must not be the predominant control — sending two people on a communication course is expressly not the answer where redesigning the work is reasonably practicable. There is no headcount threshold; a five-person clinic is inside the regime in full.

Which is a striking convergence: the meta-analysis and the regulation point the same direction. Fix the role, not the conversation.

General information, not advice. This section summarises Australian statute and regulator guidance current to 30 July 2026. It is not legal advice and does not account for the facts of any particular practice.