The Cost of Workplace Violence in Healthcare:
Building the Business Case for Culture Change

 By Casey Goldschmidt, CEM | Chief Operating Officer, Vistelar 

Most healthcare leaders already know they have a workplace violence problem. What they usually cannot tell you is what it costs them.

The instinct is to go looking for a national figure and scale it down. There is one, it is large, and it is quoted everywhere. It is also, for this purpose, the wrong number, and it is wrong in a way that nobody has bothered to check.

I have sat on both sides of this budget conversation. From the operations side, the problem looks like a program that cannot get funded because nobody can size it. From the finance side, it looks like a request built on a statistic the requester cannot defend when asked a second question about it. Both are true at once, and they reinforce each other. The safety leader brings a national aggregate. The CFO asks what it means here. The conversation ends.

This article is about ending it differently.

The $18.27 Billion Figure, Taken Apart

The most-cited estimate of workplace violence costs in U.S. healthcare includes more than workplace violence. Here's what the data actual shows.

 This is the number every article about workplace violence costs quotes. This is also, by the American Hospital Association's own classification, what is actually in it. 

The report does not normalize its own total, and the obvious move, dividing by the number of hospitals, produces a misleading answer. American healthcare is concentrated. Most hospital employees work for large systems, so an average across every hospital in the country is an average of a distribution dominated by small facilities, and it will understate what a real organization is carrying.

So this article uses one unit throughout: dollars per hospital employee per year.
It is the unit that survives the industry's structure, it is the unit that scales with the risk, since violence happens to people rather than to buildings, and it is the one number every reader already knows about their own organization.

The Same Money, Per Employee

The $18.27 billion national figure, carried down to a single person on the payroll (6.2 million hospital employees).

The national figure carried down to a single person on the payroll. We have not found this division published anywhere else, which is the more interesting fact.

The denominator is the 6.2 million hospital workers AHA's own report uses, so numerator and denominator come from the same document. That figure runs above current federal hospital employment data, which puts it nearer 5.8 million, so every per-employee figure in this article is conservative by roughly 7%. Use the lower denominator and they all rise.

The workplace violence lines in the most authoritative national dataset available come to about $107 per employee per year. For a three-thousand-person hospital that is $321,000. It is not a number that funds a program. It is a number that gets a program declined.

Nobody is hiding the cost of workplace violence. It simply does not sit anywhere that a single person can see it.

It lives in four places, and each place has a different owner, a different system, a different reporting cycle, and a different reason to care:

Workers' compensation claims sit with risk management or the third-party administrator, coded by injury type rather than by cause.

Lost and restricted work days sit in the OSHA 300 log with occupational health or safety, counted as days rather than dollars.

Clinical turnover sits with HR and nursing leadership, reported as a rate, with cause captured in an exit interview that is read once and filed.

Secondary and backfill costs sit in the staffing budget with the unit or with finance, where agency premium and overtime appear as line items with no cause attached to them at all.

Four owners, four systems, four vocabularies. In every system I have worked in, no one owned the total. Not because anyone was hiding anything, but because aggregating it is nobody's job, and the person who most needs the sum is the person furthest from all four sources.

Four Streams, One Number

Four owners, four systems, four vocabularies. No one owns the total. 

 Four owners, four systems, four reporting cycles. The sum is nobody's job, which is why almost no hospital has it. 

This is also why the problem is structural rather than analytical. The arithmetic is not hard. The organizational design is the obstacle, and that is a leadership problem, not a spreadsheet problem.  

Building the number for your own hospital

For scale before the arithmetic: on 2023 to 2024 data, hospitals record 18.3 violence-by-others injuries involving days away from work per 10,000 full-time equivalent workers, against 2.6 across all industries. The health care and social assistance sector as a whole sits at 11.9, general medical and surgical hospitals at 14.9, and psychiatric and substance abuse hospitals at 138.8. If your organization runs a behavioral health service line, its numbers are not the same as the rest of the building's and should not be averaged into them.

AHA reports 16,990 hospital workers with violence-related injuries involving days away from work in 2022, plus 8,740 with restricted duty or job transfer, against an estimated 6.2 million hospital workers. Applied to 3,000 employees, that is 8.22 cases with days away and 4.23 with restricted duty.

Now, how to value them. Most models get this wrong in a way that matters.

The instinct is to value the missed days as lost wages. Seven days at a clinical wage looks like the cost of the absence. It is not. When a nurse is injured and off work on a compensable claim, the hospital is generally not paying that nurse's salary. The insurer or the self-insured fund is paying indemnity. Counting the salary as a loss counts money that was never spent.

The two things the organization actually pays are the claim, meaning indemnity plus medical, and the coverage, meaning whoever works the shift instead. Those two do not overlap, so they add.

The claim. NCCI's severity figures give the bracket, and the coding matters:

 Code 74 is where patient-inflicted injury usually lands, and patient-inflicted injury is most of healthcare workplace violence, so it is the conservative base. 8.22 claims × $24,378 = $200,400. Code the same events as criminal assault and the same 8.22 claims come to $354,300.

Stream two: $227,700, and up to $381,600 if the events are coded as criminal assault. 

Three caveats, running in both directions. NCCI's severity is drawn across all industries for those cause codes rather than healthcare specifically. Treating every days-away case as one indemnity claim is an approximation. And against those, AHA says its own work-loss estimate "is likely to be an underestimate and does not account for reduced productivity or time loss due to untreated elements of injury, such as psychological harm, work missed due to post-traumatic stress disorder, or unreported injuries." That exclusion list is larger than the thing being measured. 

Everything above rests on recorded events, which means everything above is a floor.

The claim that circulates here is that fifty to seventy percent of healthcare workplace violence goes unreported. That range is not published anywhere. The nearest published statement is a 2019 American Nurses Association issue brief noting that research "has variously found that only 20 to 60 percent of nurses report incidents of violence," which describes non-reporting of 40 to 80 percent. How that became fifty to seventy is anyone's guess.

What is published is more useful, and more interesting.

The strongest study is Arnetz and colleagues in 2015, which matched what hospital staff said had happened to them against what the incident system actually recorded. Of 275 employees who reported experiencing a violent event in the prior year, 88% had no corresponding record in the electronic incident reporting system, and 48% had not reported the event through any channel at all.

Those two numbers are both correct and they mean different things, which is exactly why this gets quoted badly. The 88% describes the formal system. The 48% describes whether anyone was told. A great deal of violence in hospitals is known to supervisors and colleagues and invisible to the organization.

More recent data says the same thing from the other direction. In a 2026 survey of 2,090 nurses, of the 1,204 who had experienced an incident, 54% reported it formally and 46% did not. The leading reason given for not reporting was that they did not believe anything would change. Of those who did report, 40% saw no action taken and 9% felt supported by leadership.

What Happens After a Report

Of every 100 staff who experience an incident

Reporting rates are not a measure of how much violence occurs. They are a measure of what staff have learned to expect from reporting it.

Read those together and the reporting rate stops looking like a data problem. It is a rational response to observed organizational behavior. Staff are not failing to report. They are declining to spend twenty minutes on a form that has historically produced nothing.

Two consequences follow, and the second one is the one that gets missed.

First, every cost figure you build is conservative by an unknown but large margin, and you should say so when you present it rather than waiting to be asked.

Second, fixing the reporting system without fixing the response is worse than doing nothing. A 2023 study found that a strong violence-reporting culture was associated with a stronger relationship between violence and burnout, not a weaker one. That is the same finding as the 9%, arrived at independently: asking people to report into a system that does not respond teaches them something about the organization that they did not previously know.

Which means the first move in a workplace violence program is not a reporting campaign. It is deciding, in advance and in writing, what happens after a report.

Two organizations can run the same curriculum, hit the same completion percentage, produce the same binder for the same surveyor, and get opposite results. The difference is not the content. It is everything the organization does around it.

A compliance program asks whether the training happened. A culture asks whether anything changed, and it answers that question through what leadership funds, governs, and holds people accountable for.

This is measurable in a way that surprises people. A cross-sectional study of 3,132 hospital staff with patient contact at one academic medical center found that higher patient safety culture was associated with substantially lower odds of workplace violence, with teamwork across units and quality of handoffs among the strongest individual contributors. It is a single site and an association rather than a demonstrated cause. But the organizational conditions associated with fewer clinical errors are the same ones associated with less violence, and that is not usually how either problem is budgeted.

Culture also shows up in mechanisms nobody thinks of as safety mechanisms. A rounding survey that asks whether staff were "courteous and responsive," with no field for what actually happened in that interaction, is quietly working against staff safety while everyone involved has good intentions. Staff safety is a prerequisite for patient safety rather than a competing priority, and an organization that treats them as competing has already told its own people where they rank.

The principle underneath

Vistelar's entire system rests on one foundational principle: treat people with dignity by showing them respect. It sits underneath all six phases of the 6 C's of Conflict Management rather than inside any one of them.

In a cost conversation, that line does more work than it looks like it does. It is not a values poster. It is the mechanism. A patient who feels respected, heard, and given a real explanation, treated as a person having a bad day rather than a problem to be managed, de-escalates on their own far more often than one who feels handled. Every avoided escalation is an event that never enters any of the four cost streams.

The principle contains a distinction that makes it usable under pressure: showing respect does not require respecting. Respect is earned and rests on your values. Showing respect is a professional behavior you perform regardless, because it works. That is what makes it available to a nurse in hour eleven of a twelve-hour shift dealing with someone who has just insulted her.

These are learnable, trainable behaviors rather than personality traits, which is exactly why they belong in a curriculum and not only in a leadership talk track. The skills themselves are Pillar 1 territory. If you want the operational side, start with the skills your staff need.

The obligation does not stop at the main campus

The same obligation, and the same exposure, extends to every place your name is on the door.

Urgent care centers with two staff and no dedicated security officer. Medical office buildings where a billing specialist handles an angry patient at the front desk with no duress system and no escalation pathway. Freestanding emergency departments. Standalone behavioral health clinics. Usually the same risk with less infrastructure, and almost always excluded from the program that was designed for the main campus.

"A safety culture that stops at the hospital entrance is not a safety culture. It is a policy with a geographic limit."

Every one of those sites generates workers' compensation claims, lost days, and turnover that land in the same four cost streams. They are simply harder to see, which means the number you build without them is lower than the number you actually have.

We took the measurement argument seriously enough to apply it to ourselves.

Vistelar contracted with the University of Wisconsin–Milwaukee to evaluate our program at Milwaukee County Behavioral Health Division and to follow the results for two years. The study was jointly funded by the county and by Vistelar, and we did not conduct it or control its findings. A summary of the methods and results was published in the Journal of Healthcare Protection Management under the title "An evaluation of an all-staff violence-reduction training program."

The results were positive overall. The finding that changed our business was less comfortable: conflict management skills improved a month after training and had declined by the end of two years.

That result, along with feedback from customers, is why we moved away from a traditional training model built on a packaged curriculum, train-the-trainer delivery, and a workbook purchase requirement, toward content licensing with a flexible curriculum, digital courseware, a partnership relationship, and an explicit focus on sustainment.

The relevance to a cost conversation is direct. If capability decays, then a one-time training expenditure is not an investment. It is a subscription you forgot to renew, and the second-year numbers will tell you so.

"We commissioned a study, it told us something we did not want to hear, and we changed the business because of it."

The arithmetic above is the easy part. I want to be straight about which part is not.

An article can give you a method, a set of published inputs, and every assumption named. What it cannot do is get four department heads to hand data to someone who does not report to them. It cannot sit in the room when the CFO asks the second question. And it cannot keep the measurement running four quarters later, when the person who cared about it has moved on and the reporting has quietly reverted to a completion percentage.

Those are not analytical problems. They are the same organizational problem this article opened with: the cost is spread across four owners and the sum is nobody's job.

Building the number once is an afternoon's work. Building an organization that keeps it current is the actual project. That is why the governance section is in here at all, and it is the part nobody can do for you from the outside.

The bottom line

The business case for workplace violence prevention is not weak. It is unbuilt.

The national figure everyone reaches for measures something else. The per-claim benchmark everyone quotes does not exist. The effect sizes that circulate in this category cannot be sourced. And the one number that would settle the whole question for your organization is being collected in your exit interviews and thrown away.

None of that is a reason to wait. It is a description of a decision that is currently being made on borrowed evidence when it could be made on your own.

Four numbers. Four owners. One afternoon to put them in the same spreadsheet. At the end of it you will have a figure that survives being taken apart, which is the only kind that ever gets funded.

"A number your CFO can take apart is worth more than a bigger number they can only take on faith."

If you want to work through it against your own figures, our workplace violence cost and ROI calculator is ungated, shows its formula, and lets you set your own reduction assumptions rather than accepting ours. And if you would rather have the conversation with someone who has built these cases from the inside, that is what a discovery call is for.