We have just added a third measure to every state page. Alongside forklift-involved severe injury reports and amputations, each state now carries hospitalisations. That sounds like a small addition. It is not, because it changes the question a number can answer.
Counting events is not measuring harm
A severe injury report is an event: an employer telling federal OSHA that someone was hurt badly enough to require reporting within 24 hours. Across the 21 states in this dataset, forklifts were involved in 4,928 such reports between 2015 and 2024. That number tells you how often the worst kind of forklift incident happened. It tells you nothing about what happened to the person.
The reporting rule itself supplies the missing dimension, because it requires three distinct things to be reported: inpatient hospitalisation, amputation, and loss of an eye. Those are severity categories, not incident types, and they can be counted separately. Of those 4,928 reports, 4,408 involved a hospitalisation and 854 involved an amputation. The two overlap, since one incident can be both.
The ratio is the interesting part
Roughly 17 in every 100 reported forklift-involved severe injury reports include an amputation. That is the figure worth carrying around, because it reframes what a forklift incident usually is. A forklift does not typically bruise someone. When a forklift incident is serious enough to be reportable at all, there is close to a one in six chance that a body part did not survive it.
That is a different argument for a safety budget than an incident count. A plant manager who has absorbed “we had four reportable incidents last year” has heard a number. The same manager who hears “across the states in this dataset, one in six reported incidents involve an amputation” has heard a consequence.
The two measures do not agree on who is worst
If severity tracked volume, the states with the most reports would also have the most amputations, and the ranking would be the same list twice. It is not. Among states with at least 100 reports over the decade, Kansas runs the highest amputation share at 27 per cent (31 amputations in 115 reports), well clear of the 17 per cent figure across the dataset. That lead survives a one-report change: take one amputation away from Kansas and it still sits above 26 per cent, and add one to the next state down and that state still sits below 24. Arkansas (25 of 110), Missouri (32 of 143) and Wisconsin (44 of 201) follow in a cluster between 21.9 and 22.7 per cent, and we deliberately do not order them against each other: the gaps between them are smaller than a single amputation, so any ranking within that cluster is an artefact of small numbers rather than a finding. None of the four is in the top five by raw volume, which is led by Texas with 889 reports, then Florida, Georgia, Ohio and Pennsylvania.
Read carelessly, that looks like a claim that Kansas forklifts are more dangerous. It is not. A share like that is shaped by what a state’s forklifts are doing: industry mix, what is being handled, how much of the work is at height or on uneven ground. A high amputation share is a signal about the kind of work, and a prompt to look at guarding and at where hands end up, not a league table of care.
What we are not claiming
Two caveats travel with every number here, and they are on the methodology page as well. This dataset covers states under federal OSHA jurisdiction only, so states running their own state plans are absent rather than safe. And severe injury reports are a reporting-based measure: research on the programme has found meaningful under-reporting, so these counts are a floor, not a census.
The two caveats do not touch the ratio in the same way, so they deserve separate treatment. The coverage caveat leaves the ratio alone: an absent state is absent from both the numerator and the denominator, so the 17 per cent figure describes the 21 states it covers and nothing more. Under-reporting is less tidy. Both counts come from the same snapshot, so the ratio inherits their reporting limits, and it holds only if hospitalisation-only cases go unreported at the same rate as amputations. That is an assumption, not a finding. Nothing in this dataset shows whether it is true, and an amputation is harder for an employer to overlook than a single-night admission; if that difference is real, the true share sits below 17. Read 17 in 100 as the share among reported incidents, not as a floor.
Why this measure was missing until now
The honest answer is that the site launched with what the dataset made easy. Total reports and amputations were straightforward to derive; hospitalisations needed the same treatment applied to a separate flag, and until that landed, every state page answered “how often” three times and “how badly” only once. Adding the third measure was less about new data than about refusing to let an easy number stand in for a useful one.
What to do with it
If you run a floor, the practical move is to stop tracking incidents as one number. Split your own near misses and injuries the way the federal data now splits: how often, and how badly. The first tells you whether your controls are working; the second tells you what a failure costs when it comes. Our Risk Scorer weights the inputs that drive the severe end of that split, and the state pages now show all three measures side by side so you can see whether your state’s profile is volume, severity, or both.
