Lockout Was Not Even Attempted in Most Fatal Machine Maintenance Accidents. Most of Them Were Cleaning Jobs.

A guard is a piece of steel bolted over the part of a machine that can take your hand off. A lockout is a padlock through the power switch so the machine cannot start while you have your arm inside it. Neither is complicated, neither is expensive, and both have been legally required in American workplaces for more than thirty years.

They are also, year after year, near the top of the list of things employers get cited for. In the fiscal year that ended in September 2025, the control of hazardous energy standard — lockout/tagout — was the fourth most frequently cited OSHA standard in the country, and machine guarding was tenth. That is thirty-six years after the lockout rule took effect.

What follows is what the research actually shows about where machine injuries happen, what the two rules do and do not appear to have achieved, and which single change in a small shop was measured to move things. Some of it is uncomfortable for the regulation, and that part is in here too.

The deaths are in the servicing, not the running

The intuition is that machine injuries happen to the person operating the machine. The evidence points somewhere else. When the US Bureau of Labor Statistics analysed worker deaths from being caught in machinery, more than half of the fatalities happened during service-related activities rather than production — about 20% during general repair or maintenance, 10% while unjamming equipment, 7% during cleaning, and 4% while adjusting the machine.

That analysis is old — it covers the mid-1990s, when 189 workers died that way in a single year — and it is worth saying so plainly, because nothing comparably detailed has been published since. But the pattern it describes is the one that later, smaller studies keep rediscovering. A 2019 review summarising US figures put the fatality rate for installation, maintenance and repair workers at 9.4 per 100,000, against 2.6 for production workers and 3.6 for workers overall.

The current national totals are coarser but point the same way. In 2024 there were 5,070 fatal work injuries in the United States, a rate of 3.3 per 100,000 full-time equivalent workers. Of those, 756 were classified as contact with objects and equipment, and 213 were people struck, caught or compressed by running powered equipment. Running is the operative word.

The lock nobody reached for

The most useful single study here is unglamorous. Maria Bulzacchelli and colleagues at Johns Hopkins read the narrative summaries of OSHA accident investigations into lockout-related deaths in US manufacturing between 1984 and 1997, and simply described what had happened.

The mechanisms were what you would expect: caught in or between parts of equipment, electrocution, struck by or against objects. The scenarios were mundane to the point of being domestic — cleaning a mixer or a blender, cleaning a conveyor, installing or taking apart electrical equipment. The finding that matters is this one: in at least 58.8% of the fatal incidents, a lockout procedure was not attempted at all.

Not defeated. Not done badly. Not attempted. That distinction changes what the problem is. A story about workers cutting corners on a safety procedure implies the procedure was there to cut. In the majority of these deaths there was nothing to cut a corner on: the person went to clean a conveyor the way they had cleaned it a hundred times, and the machine was live because it had never occurred to anyone in that building that cleaning was the dangerous part.

What the shops actually look like when somebody counts

Asking whether guards are fitted turns out to be a question almost nobody had answered at scale until the National Machine Guarding Program did it. Insurance safety consultants, trained on a standardised checklist, inspected 2,632 machines in 221 small metal fabrication businesses across 31 states. Twelve machines were picked at random in each shop. The businesses averaged 30 employees.

The headline is not that the shops were lawless. On most measures they scored respectably — electrical safety 92%, work practices 89%, power transmission guards 92%. The failures were concentrated, and they were concentrated exactly where the injuries are.

Machine age mattered, and mattered steadily. Each additional year of a machine’s age corresponded to a 0.5 percentage point drop in its equipment safeguards score. The average machine in these shops was 22 years old. Milling, drilling and boring machines were the oldest class at a mean of 33 years, and had the worst safeguarding of any class at 62%.

The standard’s own record is thinner than you would guess

Everything above builds an argument for the lockout rule. The most honest thing this article can do is show you the study that tried to measure whether the rule worked, and did not find what you would expect.

OSHA’s control of hazardous energy standard took effect in 1990. The same Johns Hopkins group ran an interrupted time-series analysis on machinery-related fatality rates in manufacturing from 1980 to 2001, using the national traumatic occupational fatalities surveillance system. The crude numbers look like a success: the rate barely moved through the 1980s, then fell by 4.59% a year from 1990 to 2001.

Then they controlled for the demographic and economic changes running through American manufacturing over the same years. The regression estimate of the standard’s own effect came out as a small, statistically non-significant increase of 0.05 deaths per 100,000 production worker full-time equivalents. On this analysis, the visible decline is what you would expect from a changing workforce and a changing industry, and the rule cannot be shown to have added to it.

A second finding points the same uncomfortable way. In the Minnesota study that preceded the national programme, half the shops — 20 of 40 — had been inspected by OSHA within the previous year. Their machine guarding scores were no better than the shops that had not been inspected (p = 0.64).

None of this makes the padlock useless. The mechanism is not in doubt: a machine that cannot start does not amputate anybody, and the fatality investigations describe deaths that a lock would have prevented. What the evidence will not support is the idea that writing the rule was the same as solving the problem. Thirty-six years on, lockout is still the fourth most-cited standard in the country, and OSHA renewed its national emphasis programme on amputations in manufacturing again in June 2025.

What did move, and what was never measured

The same researchers then tried to change things. Insurance safety consultants visited each business, produced a report, and came back twice. Of the 221 shops, 160 completed the full programme and were re-audited a year later.

The outcome measured was audit scores, not injuries. No one counted whether fewer fingers were lost in these 160 shops, and the authors are candid that the magnitude of the real-world change is hard to assess. Their extrapolation — that a similar improvement across the 83,000 US metalworking establishments would touch 1.7 million machines — is arithmetic, not evidence, and should be read as such.

The more interesting result is what predicted improvement. The study also ran a full safety climate survey: 2,502 people, nine constructs, the standard instrument. Higher safety climate scores were not correlated with safety leadership (p = 0.48), machine maintenance programmes (p = 0.43), or lockout programmes (p = 0.74). Owners rated their own shops consistently better than their workers did.

One structural thing did predict improvement, and it is almost banal. Shops that started without a safety committee and added one during the year improved 21 percentage points more on the overall safety management audit and 9 points more on the machine score than shops that did not, controlling for baseline score and business size. The share of shops with a committee went from 34% to 58%.

The jam is the moment

If you had to pick a single instant to engineer around, the research points at one. Reviewing injuries involving automated machinery, Backström and Döös found that about 25% were preceded by a disturbance in the flow of material — a piece stuck, crooked, or sitting wrong. Something jams; someone reaches in.

The regulation has a hole exactly here. Lockout is required when a worker puts part of their body into the point of operation or a danger zone. But minor tool changes and adjustments, and other minor servicing that is “routine, repetitive, and integral” to production, are exempt — provided alternative measures give effective protection. Clearing a jam is routine and repetitive by definition. Whether the alternative measures are effective is decided on the shop floor, in about four seconds, by the person holding the part.

That is an argument for designing the reach out of the job rather than writing another procedure about it: interlocked guards that cut power when opened, jam-clearing designed as an external operation, controls that make the safe way the fast way. It is also the point at which this article stops being about individual behaviour, which is where a lot of workplace safety writing quietly parks the blame.

The counting is a floor, not a number

Since 2015 US employers have had to report every amputation and every in-patient hospitalisation to OSHA within 24 hours. Between 2015 and 2021 federal OSHA received 70,206 severe injury reports, of which 18,559 were amputations — roughly 10,000 reports a year, covering about half the American workforce. Manufacturing accounted for around 69% of the injuries where a body part was caught in equipment.

OSHA states in the same report that these totals are “significantly underreported”. That is not a throwaway caveat. When the Bureau of Labor Statistics funded researchers in Massachusetts, California and Washington to check its own survey against other records, the conclusion across three states and three methods was that the national survey significantly undercounts injuries even when the injury is objectively verifiable — amputations among them. Employer-reported, unaudited data has predictable failure modes.

So every number in this article is a lower bound. That cuts against the reassuring reading of the fatality trend as much as it cuts against anything else.

A recent case shows what the categories look like from the inside. In February 2026 OSHA cited an Illinois grain company after a seasonal labourer stepped through an unguarded sump hole onto a moving paddle conveyor and injured his foot. The citations included two willful violations covering machine guarding and lockout, with proposed penalties of $276,407. The hazard was a hole in a floor over a running conveyor, which is not a sophisticated engineering problem.

What to actually look at

If you work around machines, or run a shop with a dozen of them, the evidence above narrows down to a short list of things that are cheap to check and were measured to matter.

Start with the gap between hardware and instructions, because that is where the national audit found the largest hole. Almost every machine had a lockable disconnect; almost none had a written procedure at the workstation saying which energy sources that machine has and in what order to isolate them. A machine with stored pneumatic pressure or a raised ram is not made safe by turning off the electricity, and that is the sort of thing a machine-specific procedure exists to say.

Then look at the oldest machines first rather than the busiest ones. The age gradient in the data is steady and steep enough — 93% down to 71% — that a shop’s worst-guarded machine is probably its oldest, not its most used.

And notice which tasks are treated as too small to stop for. Cleaning, unjamming and adjusting are the jobs the fatality records are full of, and they are precisely the jobs that feel too brief to justify walking to the panel. If your shop has a rule about locking out for repairs but not for clearing a jam, the rule is aimed away from the danger.

Questions people ask

It is a two-minute job. Do I really have to lock it out?

Two minutes is the modal case in the fatality reports, not the exception. The scenarios that recur — cleaning a mixer, clearing a conveyor — are short jobs, and in the majority of the deaths studied no lockout was attempted at all. The regulation does carve out minor servicing that is routine and integral to production, but only where alternative measures give effective protection, and that condition is doing enormous work. If the alternative measure is that you are being careful, it is not one.

My employer says the machine is too old to guard.

Age genuinely predicts poor guarding in the data, and retrofitting an old machine can be expensive and technically awkward. It is not a legal exemption. It is also worth setting against the cost on the other side: OSHA’s own estimate of the average total cost to a business of a workplace amputation is around $133,000, against $111,000 for a crush injury and $95,000 for a fracture.

Is a tag as good as a lock?

No, and the standard does not treat them as equivalent. A tag is a warning; a lock is a physical barrier to the switch being thrown. Tagout is permitted where a machine cannot accept a lock, and then only with additional measures. Several of the fatal incidents in the OSHA narratives involve a machine restarted by a co-worker who did not know anyone was inside it, which is the failure mode a tag cannot prevent and a lock can.

Does any of this apply to my own workshop at home?

The regulations do not, but the mechanism does. Home workshop machines are old by definition, guards get taken off for an awkward cut and stay off, and there is nobody to notice. The single most transferable habit from the industrial evidence is unplugging rather than switching off before you put a hand anywhere near a blade or a belt — the domestic version of a lock is a plug in your pocket.

How do I raise a guarding problem without becoming a problem?

The finding that a safety committee predicted improvement, while surveys of how people felt about safety predicted nothing, is quietly useful here. A standing group with a remit is a route that does not require any individual to be the person who complains. Where that does not exist, workers in the US can file a complaint with OSHA and request that their name not be revealed to the employer, and retaliation for doing so is itself unlawful. Whether that protection feels sufficient is a separate question from whether it exists.

The short version

• More than half of US deaths from being caught in machinery happened during service tasks — repair, unjamming, cleaning, adjusting — not during production.
• In OSHA investigations of fatal machine maintenance incidents in manufacturing, lockout was not attempted at all in at least 58.8% of cases.
• Across 2,632 machines in 221 small shops, 87% had a lockable disconnect fitted but only 9% had an adequate lockout procedure posted. A third of point-of-operation guards were missing or inadequate.
• Safeguarding falls about 0.5 percentage points for every year of machine age: 93% on machines under ten years old, 71% on machines over fifty.
• The only published evaluation of the US lockout standard found no statistically significant effect on machinery fatality rates once demographic and economic change was controlled for.
• A year of insurer-led visits raised posted lockout procedures from 8% to 33%. Injuries were not measured — only audit scores.
• Safety climate scores predicted nothing. The presence of a safety committee predicted a 21-point gain in audit score and 9 points on machine guarding.
• About a quarter of injuries on automated machinery are preceded by a material jam. That moment, not the operating cycle, is what needs designing around.
• Every count here is a floor: OSHA says its severe injury reports are significantly underreported, and BLS-funded research found its own survey undercounts even amputations.

This article summarises published injury epidemiology, US regulatory records and one intervention study. It is not safety engineering advice and it is not legal advice. Machine guarding requirements, lockout standards and worker protections differ between countries and, within the United States, between federal and state-plan jurisdictions. Decisions about a specific machine belong with people who can see that machine.

Further reading: The Field Guide to Understanding ‘Human Error’ — Sidney Dekker (CRC Press). On why investigations that end at the injured person’s decision tend to stop one step short of the thing that could have been changed. It is written for people who investigate incidents, and it is the best short antidote to the reflex that produced the pattern in the 1980s amputation data: retraining the survivor, and leaving the machine as it was. Amazon

ROR Labs is a participant in the Amazon Associates Programme. The book link above is an affiliate link, and a purchase made through it may earn this site a commission at no extra cost to you. There are no equipment links in this article; the reason is set out in the section above. See the affiliate disclosure for the full policy.

Sources

• US Bureau of Labor Statistics, Worker fatalities from being caught in machinery, Compensation and Working Conditions. (189 deaths in 1997, a 29% rise on the previous year against an average of about 150 a year in 1992–1996; more than half of fatalities during service activities — 20% repair or maintenance, 10% unjamming, 7% cleaning, 4% adjusting; manufacturing 41%, agriculture 26%.)
• US Bureau of Labor Statistics, National Census of Fatal Occupational Injuries in 2024. (5,070 fatal work injuries; rate 3.3 per 100,000 full-time equivalent workers; 756 deaths from contact with objects and equipment, of which 213 were struck, caught or compressed by running powered equipment.)
• Bulzacchelli MT, Vernick JS, Sorock GS, Webster DW, Lees PSJ. Circumstances of fatal lockout/tagout-related injuries in manufacturing. American Journal of Industrial Medicine 2008;51(10):728–734. (Narrative analysis of OSHA investigation summaries, US manufacturing 1984–1997; lockout not attempted in at least 58.8% of fatal incidents; commonest mechanisms caught in or between, electrocution, struck by; typical scenarios cleaning a mixer or blender, cleaning a conveyor, installing or disassembling electrical equipment.)
• Bulzacchelli MT, Vernick JS, Webster DW, Lees PSJ. Effects of the Occupational Safety and Health Administration’s control of hazardous energy (lockout/tagout) standard on rates of machinery-related fatal occupational injury. Injury Prevention 2007;13(5):334–338. (Interrupted time series, 1980–2001; crude machinery fatality rate in manufacturing fell 4.59% a year from 1990 to 2001; controlling for demographic and economic factors, estimated effect of the standard was a non-significant increase of 0.05 deaths per 100,000 production worker FTEs. Figures taken from the published abstract.)
• Parker DL, Yamin SC, Brosseau LM, Xi M, Gordon R, Most IG. National machine guarding program: Part 1. Machine safeguarding practices in small metal fabrication businesses. American Journal of Industrial Medicine 2015;58(11):1174–1183. (221 shops, 2,632 machines, 31 states, mean 30 employees; point-of-operation safeguards 67%, other mechanical hazards 72%, power transmission guards 92%, lockable disconnects 87%, lockout procedures posted and adequate 9%; mean machine age 22 years; equipment safeguards score falls 0.5 percentage points per year of machine age, 93% at 10 years or under against 71% at 50 years or over; milling, drilling and boring machines oldest at 33 years and worst guarded at 62%.)
• Parker DL, Yamin SC, Xi M, Brosseau LM, Gordon R, Most IG. Findings from the National Machine Guarding Program — a small business intervention. Journal of Occupational and Environmental Medicine 2016;58(9). (160 of 221 shops completed; business-level machine score 73% to 79%, point-of-operation guards 67% to 72%, lockable disconnects 88% to 92%, posted lockout procedures 8% to 33%, all p < 0.0001; safety committees 34% to 58%; shops adding a committee improved 21 percentage points more on the safety management audit and 9 points more on the machine score.)
• Parker DL et al. Findings from the National Machine Guarding Program — a small business intervention: machine safety and safety climate. Journal of Occupational and Environmental Medicine 2016;58(1). (2,502 respondents; safety climate score not correlated with safety leadership p = 0.48, machine maintenance p = 0.43 or lockout programme p = 0.74; presence of a safety committee associated with a 16% increase in overall safety audit score, 21 points in safety leadership and 22 points in lockout programme score; owners rated their shops more favourably than workers did.)
• Samant Y, Parker DL, Brosseau LM, Pan W, Xi M, Haugan D. Profile of machine safety in small metal fabrication businesses. American Journal of Industrial Medicine 2006;49(5):352–359. (824 machines; 55% of machine-guarding items present; no single machine complied with all critical safety requirements; 20 of 40 shops had an OSHA inspection in the previous year with no difference in guarding score, p = 0.64.)
• Etherton J, McKenzie EA. The machine operator’s jammed-feedstock-clearing task: a safety design challenge. ASME IMECE 2001. (Reports Backström and Döös 1998: about 25% of injuries involving automated machinery were preceded by a disturbance in process material flow; quotes the OSHA minor servicing exception.)
• Etherton J. Industrial machine systems risk assessment: a critical review of concepts and methods. Risk Analysis 2007;27(1):71–82. (Average of 520 machine-related fatalities a year and 3.8 lost-workday caught-in-running-machine injuries per 10,000 workers, 1992–2001.)
• Illankoon P, Manathunge Y, Tretten P, Abeysekara J, Singh S. Lockout and tagout in a manufacturing setting from a situation awareness perspective. Safety 2019;5(2):25. (Summarises US figures: 9.4 fatalities per 100,000 for installation, maintenance and repair workers against 2.6 for production workers and 3.6 overall.)
• Olson D, Gerberich SG. Traumatic amputations in the workplace. Journal of Occupational Medicine 1986;28(7):480–485. (109 Minnesota amputation cases; manufacturing 43.4% of amputations with 18% of the workforce; the intervention applied after the injury was training only in 32.1% of cases and engineering controls in 10.1%.)
• Spieler EA, Wagner GR. Counting matters: implications of undercounting in the BLS survey of occupational injuries and illnesses. American Journal of Industrial Medicine 2014;57(10):1077–1084. (Studies in Massachusetts, California and Washington concluded the national survey significantly undercounts injuries, including objectively verifiable ones such as amputations.)
• OSHA, Severe Injury Report — A Seven-Year Lookback, 2015–2021. (70,206 reports under federal jurisdiction; 56,696 hospitalisations and 18,559 amputations; about 10,000 reports a year; manufacturing about 69% of body-part-caught-in-equipment injuries; OSHA states the totals are significantly underreported.)
• OSHA, Top 10 Most Frequently Cited Standards, fiscal year 2025 (1 October 2024 to 30 September 2025). (Control of hazardous energy, 29 CFR 1910.147, ranked fourth; machinery and machine guarding, 29 CFR 1910.212, ranked tenth.)
• US Department of Labor, news release, 26 June 2025. (Renewal of the National Emphasis Program on amputations in manufacturing.)
• US Department of Labor, news release, 24 February 2026. (Alliance Grain Co., Gibson City, Illinois: seasonal labourer stepped through an unguarded sump hole onto a moving paddle-style unloading conveyor; two willful violations covering machine guarding and lockout among the citations; proposed penalties $276,407.)
• OSHA Safety Pays cost estimates, cited in Parker et al. 2015. (Average total cost to a business of a workplace amputation about $133,000; crush injury $111,000; fracture $95,000.)