Your Next Fatality Has Already Happened

What SIFp means, why it matters, and what most safety programs miss about modern fatality prevention.

You just got lucky.

Somewhere on your jobsites in the last 90 days, a worker likely stepped into a condition with the energy to kill them and walked out without a scratch. No incident report. Maybe a near miss filed. More likely nothing at all. That event has a name in the research literature: a Potential Serious Injury or Fatality, abbreviated SIFp or PSIF. It is one of the most valuable signals you can collect about your safety system, and most teams throw it away.

"You got lucky" is not a turn of phrase. Modern fatality prevention starts from a hard assumption: not every worker hour without an injury is safe work. Sometimes the work was done unsafely and the energy went somewhere else.

This is the case for working with SIFp as part of a modern serious injury and fatality prevention strategy, what the research has done to make it operational, and where most operators still get it wrong.

Heinrich was wrong

Modern injury prevention is built on a 1931 paper by Herbert Heinrich. The pyramid model says one fatality for every 29 serious injuries and 300 near misses. That ratio has shaped a generation of observation programs and the assumption that reducing minor injuries reduces serious ones proportionally.

Two decades of data say otherwise. Recordable injury rates in most heavy industries have trended down. Construction fatality rates have stayed roughly flat, hovering between nine and 10 per 100,000 workers per BLS. A 2021 study analyzing more than three trillion worker hours found that TRIR has no statistical relationship to SIFs. Past TRIR does not predict future TRIR. Reducing recordables and reducing fatalities are two different problems.

Research over the last decade has shown that only about a quarter of recordable incidents have SIF potential. Most minor injuries are not warnings of bigger ones to come. They are their own category of event.

If Heinrich's pyramid worked, your fatality rate would be tracking your recordable rate down. It is not.

What is SIFp, exactly?

SIFp is a specific kind of event, not a synonym for near miss. The modern classification framework defines it through four conditions. High energy was present. The energy released and the worker came into contact with it or within six feet. No serious injury occurred. No direct control was in place.

High energy means a level of physical energy above which the most likely outcome of contact is a SIF rather than a recordable. Thirteen high energy hazard categories cover roughly 85% of SIFs in the literature: work at height, suspended loads, mobile equipment near workers, electrical above 50 volts, vehicles at speed, steam, fire, explosion, excavations, heavy rotating equipment, high temperature, and toxic chemicals or radiation. If any of these are on your job, you have high energy on site.

A direct control is the critical concept. It is a control specifically targeted to the high energy source, that mitigates exposure when installed, verified, and used, and that remains effective even when a worker makes an unintentional mistake. LOTO. Machine guarding. Hard barriers. Engineered fall protection.

Direct controls come in two forms. Absolute controls eliminate exposure: removing the energy source, full isolation, complete machine guarding. Mitigating controls reduce exposure below the SIF threshold: fall protection that limits free fall distance, thermal insulation, structural barriers that absorb impact energy. Neither relies on the worker doing the right thing in the moment.

What does not count: a toolbox talk. Training. A warning sign. A hazard discussion before the shift starts. These shape behavior. They do not stop energy.

A SIFp is high energy released, worker exposed, no direct control, no serious injury. Same conditions as a fatality. Different outcome.

Scheduled and unscheduled SIFp exposures

A useful split in the research separates SIFp exposures into scheduled and unscheduled. A scheduled exposure is high risk work you knew was coming. LOTO on a live system. A planned confined space entry. Work at height with rigged fall protection. You prepare for it. You engineer controls into it. You write a permit for it.

An unscheduled exposure is the surprise. A worker turns a corner and finds energy where none was supposed to be. Conditions change while the work is happening. The plan stops matching the reality on the ground. The last line of defense becomes a crew member who chooses to stop the job.

Both produce SIFp. They require different prevention strategies. Scheduled exposures respond to engineered controls and structured planning. Unscheduled exposures depend on safety climate, worker authority to stop work, and whether a crew sees a hazard before they walk into it. Most safety programs spend almost all their attention on the scheduled half and leave the unscheduled half to luck.

The category nobody tracks

The classification framework recognizes seven categories of safety event, not one. Most attention goes to SIFp itself, which is the dangerous one. The category nobody tracks well is the inverse: high energy present, direct control in place, no incident. The research calls this Success.

Success is underreported because it looks like nothing happening. But it is the most informative event in the system. Predictive models of safety performance require studying success alongside failure. Without it, you cannot tell which direct controls are actually working in the field versus which are nominally installed and quietly defeated.

If you start tracking anything new this quarter, track Success. The percentage of high energy hazards on a job with a verified direct control in place is unambiguously directional. More is better. It moves with the risk you care about and cannot be gamed by underreporting.

What to do with a SIFp, and what not to

When you classify an event, follow the discipline of the research. Don't ask "could this have caused a SIF?" Imagination makes any event qualify. Ask "was the most likely outcome a serious injury or fatality?" That forces evidence over intuition.

When you investigate a SIFp, treat it like the fatality it almost was. Same depth, same root cause discipline.

But don't stop at the event. The same missing direct control almost certainly exists somewhere else in your operation. Investigation is also a search across crews and sites for the same gap. The point of SIFp work is to find the dozen others before they pick a different outcome.

When you measure, do not make PSIF rate a headline metric. A high rate can mean strong reporting or poor performance. A low rate can mean strong performance or underreporting. Incentivize it and you compromise the learning. Use the classification for learning first, before you turn it into a number.

For implementation, the work fits inside any standard Plan, Do, Check, Act cycle. The principles underneath should be the same ones driving modern operational safety: everyone makes mistakes, context drives behavior, blame fixes nothing, and improvement happens through learning.

The SIFp capture problem

The conditions that make a SIFp visible exist for minutes. The worker saw it. The foreman saw it. Nobody had time to write it down. By the end of the shift, it lives only in someone's head. This is the bottleneck. The science to identify, classify, and learn from SIFp has been on the shelf for years. None of it matters if the field event never becomes data.

StepoAI was built to turn what a worker sees, hears, or describes in the moment into structured data the classification framework can actually use. Not to digitize the form. To keep the SIFp from disappearing before anyone learns from it.

The next fatality on your sites is already visible to someone. The work now is putting the science in the hands of the people who actually see the energy.

By the numbers

• 1:29:300. Heinrich's 1931 fatality to serious injury to near miss ratio. The original foundation of modern injury prevention thinking. The ratio does not hold for SIFs.
• 9 to 10 per 100,000. The construction fatality rate range, per BLS, for more than a decade. Recordable rates have trended down across the same period.
• About 25%. The share of recordable incidents with SIF potential. The remaining three quarters are their own category of event.
• 6 feet. The proximity threshold for considering a worker exposed to a high energy release.
• 5 feet. The excavation depth above which a trench is high energy.
• 50 volts. The electrical threshold for high energy.
• 30 mph. The vehicle speed above which impact crosses into high energy territory.
• 13. The number of high energy hazard categories that cover roughly 85% of documented SIFs.
• About one third. The share of SIFp events that involve multiple high energy sources, each requiring independent classification.
• About two thirds. The rate at which safety professionals agree on incident severity classification without a structured framework. With a structured framework, agreement rises to about 95%.

Frequently asked questions

Is SIFp the same as PSIF?

Yes. SIFp (lowercase p, for "potential") and PSIF (capital P, for "Potential Serious Injury or Fatality") are two notations for the same event type. Some frameworks use one, some use the other. An older term, SIF precursor, points at a related but slightly broader concept: events or conditions that precede a SIF, rather than the specific four-condition classification SIFp describes.

Is every near miss a SIFp?

No. A near miss is any unplanned event that could have caused harm. A SIFp is a specific kind of near miss where high energy was released, the worker was exposed, and no engineered control was in place to stop the energy. Most near misses don't clear that bar. The point of SIFp work is to find the small subset of near misses that share causal DNA with fatalities, not to relabel every one.

Does PPE count as a direct control?

Almost never. A direct control has to remain effective when a worker makes an unintentional mistake. Most PPE depends on the worker wearing it correctly, in the right size, at the right moment. That dependence on behavior is what disqualifies it. The exception is engineered fall protection, where the system of anchor, lanyard, harness, and deceleration device works whether or not the worker is paying attention at the moment of the fall. PPE has its place. It just isn't what stops the energy.

Do we have to abandon TRIR to do SIFp work?

No. TRIR is still useful for tracking recordable injury performance, which is real work and worth measuring. The argument is that TRIR alone is not enough, because it does not predict fatalities. SIFp work complements TRIR. You keep what you have. You add a layer that actually moves with fatality risk.

How is this different from BBS (behavior based safety)?

BBS assumes workers cause injuries through unsafe acts, and that observing and correcting those acts reduces injuries proportionally. SIFp work makes a different assumption: that energy causes serious injuries, and that the presence or absence of engineered controls determines outcomes more than worker behavior. Both can coexist in a program. If you have to choose where your next dollar goes, the data suggests it should fund verifying direct controls in the field, not more observations of worker behavior.

Where do you start with SIFp if you have not done it before?

Three moves. First, audit your last twelve months of incidents and reclassify them using the four condition test. Most teams find that the events with actual SIF potential are not the ones they spent the most time investigating. Second, identify the high energy hazards your crews routinely face and name the direct control for each one. If you cannot name it, you don't have it. Third, start tracking Success: the percentage of high energy hazards on the job with a verified direct control in place. That single metric will tell you more about your fatality risk than your last decade of TRIR data.