A restaurant back injury costs 20x what a kitchen burn costs

Ask a restaurant owner what they worry about on the workers' compensation side and you'll hear about the fryer. Burns, knives, the things that make you flinch.

AmTrust studied roughly 130,000 restaurant workers' compensation claims from 2018 through 2023. Cuts, punctures and scrapes were the most frequent injury and averaged $1,798. Burns and scalds were second most frequent and averaged $4,326.

Then the other end of the distribution:

A back disc claim costs roughly twenty times a burn and forty-seven times a laceration. Frequency and severity point in opposite directions, and almost every restaurant safety program is built around frequency — which is the cheap end.

Lifting, carrying, and repetitive strain are the expensive exposure. That is a scheduling and equipment problem before it is an insurance problem: how heavy the stock deliveries are, whether there are carts, whether one person breaks down the walk-in alone at 11pm.

Your workers' comp rate is mostly geography

The Oregon Department of Consumer and Business Services publishes a biennial comparison of workers' comp rates across all 51 US jurisdictions. The calendar-year 2024 edition, released June 2025, puts the median index rate across all classes at $1.09 per $100 of payroll — the lowest since the study began in 1986, and roughly a quarter of the 1994 peak.

For restaurant class codes specifically:

Same class code, same operation, seven-fold price difference. For a multi-state operator that makes the blended rate a geography decision before it is a safety decision — and it means any "average restaurant workers' comp rate" figure is arithmetic performed across incompatible regulatory regimes.

Two California-specific notes, because it is the largest and priciest market. The Insurance Commissioner approved advisory pure premium rates averaging $1.65 per $100 of payroll effective 1 September 2026, a 6.6% increase — moving against the national trend, where workers' comp premiums fell 3.7% in the first quarter of 2026. And for policies incepting on or after 1 September 2024, California retired the old combined restaurant classification and split it into six, covering full-service, fast food, bars, food trucks and others. All six currently share a single advisory rate until enough data accumulates to differentiate them. Today's classification exercise may produce a very different rate in two or three years with no change to your operation.

The soft market isn't for you

The Council of Insurance Agents and Brokers reported the first overall commercial property and casualty premium decline since 2017 in the first quarter of 2026 — down 1.2% across all accounts, with property down 5.5%.

Underneath that headline: small accounts still paid +1.1% while large accounts got −2.7%. And commercial auto — the line governing delivery — posted +5.8%, its 59th consecutive quarterly increase.

An independent restaurant that delivers is in the two worst buckets at once. Budget accordingly, and treat "insurance rates are falling" coverage as being about someone else's renewal.

Now the part that decides claims: exclusions

Coverage lists are easy to write and nearly useless. What determines whether a policy responds is the exclusion schedule, and restaurants have four recurring gaps.

The bar fight

Many liquor liability and general liability policies carry an assault and battery exclusion, or sublimit assault and battery to something like $25,000 to $300,000 inside a $1M policy. The sublimit is frequently inclusive of defense costs — meaning the defense lawyer is paid out of the same bucket as any settlement.

The structural problem is that a bar fight can be pleaded two ways: improper service of alcohol, which points at the liquor policy, or negligent security, which points at the GL policy. If either policy carries the exclusion, the plaintiff's lawyer will simply plead into the other one. Neither policy should carry it. Check both.

The wage-and-hour claim

Employment practices liability insurance is sold as the answer to employment claims. It routinely excludes wage-and-hour claims outright, or covers defense costs only under a sublimit — commonly $100,000 to $500,000 — with no indemnity for the settlement or judgment.

For an industry organized around tip credits, split shifts, off-the-clock prep, and side work, the wage-and-hour claim is the modal employment claim. It is also the one your employment policy is least likely to pay. If you buy EPLI for one reason, know before you sign whether it covers this and whether "covers" means defense or indemnity.

The delivery driver

Your employee delivers in their own car. Their personal auto policy excludes business use — delivering food for pay is business use. You have no commercial auto policy because you own no vehicles.

Hired and non-owned auto coverage is the only thing that responds, and it covers your liability, not damage to the employee's vehicle. It also does not cover third-party platform couriers, who are not your employees; that exposure sits with the platform. If you run your own delivery, this is not optional, and it sits in the one line of coverage that has been rising for fifteen straight years. I covered the operational side of running your own delivery in the delivery operations post.

The walk-in

Standard commercial property covers fire, theft, and physical damage. It does not cover mechanical or electrical breakdown. A compressor failure, motor burnout, or power surge needs an equipment breakdown endorsement. The spoiled inventory needs a spoilage endorsement. If the cause was an off-premises utility failure, that may be a third thing — utility service interruption coverage.

One event, three coverage parts, and the default policy has none of them.

What COVID actually settled about business interruption

The common summary is that the courts closed the door. That is not what happened, and the real lesson is more useful.

The Pennsylvania Supreme Court held unanimously for the insurer in Ungarean v. CNA in September 2024: "direct physical loss" is unambiguous and requires physical damage, destruction or alteration. Appellate courts in a substantial number of states reached the same conclusion.

Three months later, the North Carolina Supreme Court held unanimously for the policyholders in North State Deli v. Cincinnati Insurance — a case brought by restaurants. The decisive fact was that the policy contained no virus exclusion. Vermont's high court and a Louisiana appellate court also found for policyholders.

So the cases turned on the exclusion schedule, not on the phrase. The ISO virus and bacteria exclusion has been available since 2006 and is now close to universal, so the pandemic-specific question is largely settled by contract. But the same "direct physical loss" trigger governs business interruption claims for smoke, water intrusion, and contamination events — which is why equipment breakdown and utility interruption endorsements are worth more attention than re-litigating 2020.

Liquor liability by jurisdiction

Eight states provide no dram shop cause of action at all: Delaware, Kansas, Louisiana, Maryland, Nebraska, Nevada, South Dakota, and Virginia. Even there, liability for service to a minor generally still attaches. I'm giving you the eight rather than a full fifty-state table deliberately — the only comprehensive survey I could get to was inaccessible, and every other list I found was a personal-injury firm's lead-generation page. This is a subject where a legislature can change the answer in one session, so verify your own state against its statute.

Two states publish hard caps worth knowing as reference points:

Illinois caps dram shop recovery at $90,411.55 per person for personal injury and $110,503.00 for loss of means of support — as of 20 January 2026. Those figures are indexed to CPI-U and change every January; the 2026 adjustment was 2.68%. A cap that specific makes limits selection straightforward, as long as you look up the current year's number rather than the one in an article.

Utah caps at $1,000,000 per person and $2,000,000 aggregate per occurrence.

Texas offers a training safe harbor under §106.14 of the Alcoholic Beverage Code: if you require employees to attend a Commission-approved seller training program, the employee actually attended, and you didn't encourage the violation, the employee's actions aren't attributed to you. The nuance most articles omit: the safe harbor reliably protects your liquor license from administrative action, but courts have rarely applied it as a defense in civil dram shop suits. Certification is cheap and worth doing. It is not civil immunity.

Two numbers on what claims cost when they go wrong

Marsh and Oliver Wyman's restaurant loss-cost work found general liability loss rates rising 5% annually since 2019 on an inflation-adjusted basis — while falling 2% annually unadjusted. Slips, trips and falls grew by five percentage points as a share of losses. And losses from claims over $100,000 rose 10% in a single year as litigation conversion rates doubled. That is social inflation with numbers attached: not more claims, but more claims becoming lawsuits and more lawsuits becoming large.

The same work found mandatory in-person safety training associated with a 5–10% reduction in both claim frequency and severity, and formal cost-allocation systems with 10–12% lower severity. Those are the two interventions with measured returns, and neither is an insurance purchase.

On foodborne illness, the only peer-reviewed cost model I could find is a Johns Hopkins simulation putting a single outbreak between roughly $4,000 and $2.6M depending on service format and outbreak size, with lawsuits and lost revenue as the dominant drivers — and, notably, subsequent insurance premium increases modeled as a named cost category. I'd flag that the study is from 2018 and simulation-based rather than an observed claims distribution. There is no current equivalent, which is itself worth knowing before someone quotes you a confident number.

One correction to something I wrote

In the onboarding post I cited restaurant turnover as running north of 75% a year. The directly comparable figure fell to 65.8% in 2024 from 75.6% in 2023, so the number I used was roughly two years stale. I've corrected that post. The argument in it doesn't change — quits in accommodation and food services still run about double the private-sector rate — but the figure was wrong and it's worth saying so here rather than editing quietly.

Disclosure: I work at Katalyst. We sell restaurant software, not insurance, so I have no stake in which policy you buy. What I'd do with this post is narrower than shopping: pull your liquor policy and your GL policy and search both for "assault," pull your EPLI and search for "wage," and find out whether anything in your file responds when a server rear-ends someone on a delivery run. Three searches, one afternoon, and they tell you more about your actual coverage than the declarations page does.