One formula, four rates
An OSHA incidence rate normalises injury counts by exposure so that a 40-person shop and a 4,000-person plant can be compared. Count the cases, multiply by 200,000, divide by hours worked. The constant is not arbitrary: 100 employees × 40 hours × 50 weeks = 200,000 hours, so the answer means cases per 100 full-time workers per year.
Four rates come out of the same arithmetic with different numerators. TRIR uses every case on the OSHA 300 log — the broadest measure of how often people get hurt badly enough to require more than first aid. DART uses cases involving days away, restricted duty or job transfer, so it counts only the injuries serious enough to change what someone does at work. The lost-time case rate narrows further to cases with at least one day away. The severity rate abandons case counts entirely and puts days in the numerator, measuring how much time injuries cost rather than how many there were.
You need at least two of them. TRIR alone cannot distinguish twelve minor cases from twelve amputations, and severity alone cannot distinguish one catastrophic injury from a pattern of moderate ones. Read frequency and severity together, and the shape of the problem becomes visible.
These rates have consequences beyond the safety office. Contractor prequalification systems, insurance underwriters and many client bid packages ask for three years of TRIR, DART and EMR before allowing a contractor on site, and a rate above the industry benchmark can disqualify a bid outright. That commercial pressure is also why the recordability decisions underneath the number have to be defensible.
Getting the numerator and denominator right
Hours worked means hours actually worked, from payroll. Vacation, holiday, sick leave and any other paid-but-not-worked time is excluded, and including it inflates the denominator and flatters every rate. If you cannot get actual hours, 29 CFR 1904 permits an estimate, and OSHA's own instructions describe estimating from the number of employees and their scheduled hours.
Recordable cases are those meeting the criteria in 29 CFR 1904.7: death, days away from work, restricted work or transfer to another job, medical treatment beyond first aid, loss of consciousness, or a significant injury or illness diagnosed by a licensed health care professional. The line that generates most disputes is medical treatment versus first aid, and 1904.7(b)(5)(ii) gives an exhaustive list of what counts as first aid — if a treatment is on that list it is first aid no matter who administered it, and if it is not on the list it is medical treatment.
DART cases are the subset with days away, restricted work or transfer. Every DART case is also a recordable case, so DART can never exceed TRIR, and the calculator treats a violation of that as an input error. Days-away cases are in turn a subset of DART.
Days for the severity rate are counted as calendar days, begin the day after the incident, and are capped at 180 days per case for both days away and restricted days. That cap matters: without it, one catastrophic case dominates the metric for years.
One structural point about the denominator. Because the rate divides by hours, a site with few hours swings wildly. At 20,000 hours — ten full-time equivalents — a single recordable case is worth 200,000 ÷ 20,000 = 10.00 points of TRIR. At a million hours, the same case is worth 0.20. Small-site rates are noise unless aggregated over several years, and treating a small site's zero as evidence of safety performance is a statistical error.
Worked example: 210 full-time equivalents with eight recordable cases
A manufacturing site worked 420,000 hours last year. The OSHA 300 log shows 8 recordable cases, of which 4 involved days away, restricted duty or transfer, and 2 of those involved at least one day away from work. Days away and restricted duty totalled 65 days. The corporate target is a TRIR of 2.00.
- Full-time equivalents. 420,000 ÷ 2,000 = 210 FTE.
- TRIR. (8 × 200,000) ÷ 420,000 = 1,600,000 ÷ 420,000 = 3.8095.
- DART rate. (4 × 200,000) ÷ 420,000 = 1.9048, exactly half the TRIR because half the cases were DART.
- Lost-time case rate. (2 × 200,000) ÷ 420,000 = 0.9524.
- Severity rate. (65 × 200,000) ÷ 420,000 = 13,000,000 ÷ 420,000 = 30.9524 days per 100 workers per year.
- Days per DART case. 65 ÷ 4 = 16.25 days.
- Cases allowed at the 2.00 target. (2.00 × 420,000) ÷ 200,000 = 4.20 cases. With 8 actual cases you are 3.80 cases above target.
Read the set together. The TRIR of 3.81 is nearly twice the target, but half the cases were minor enough not to change anyone's work assignment, and the average DART case cost 16.25 days — moderate, not catastrophic. That profile points at a frequency problem with routine injuries rather than a single severe event, and the improvement plan should look completely different from one built for a site with a TRIR of 1.0 and a severity rate of 90.
Now check what would have to change to hit target. You need 8 − 4.20 = 3.80 fewer cases, which is a 47.5% reduction in case count at the same hours. Alternatively, at 8 cases the hours needed to reach a 2.00 rate would be 8 × 200,000 ÷ 2.00 = 800,000 — nearly double the site's actual hours. That second calculation is worth doing whenever someone proposes to improve a rate by growing the denominator.
What counts as a good rate
There is no universal good number, because incidence rates vary enormously by industry. The Bureau of Labor Statistics publishes annual incidence rates by NAICS industry code from its Survey of Occupational Injuries and Illnesses, and that table — for your specific industry, not for manufacturing or construction as a whole — is the only defensible external benchmark. A rate that would be excellent in nursing homes would be poor in professional services.
Internally, the trend matters more than the level. A rate is a ratio of a small integer to a large one, so year-to-year movement at a single site is dominated by chance unless the workforce is large. Look at three-year rolling figures, and use the calculator's per-case sensitivity — 200,000 ÷ hours — to judge whether a change is bigger than one case.
Watch the relationship between TRIR and DART. A DART rate close to TRIR means most injuries are serious enough to change work assignments; a DART rate far below TRIR means most cases are minor. Both patterns are informative, and neither is inherently better: a site can drive TRIR down by managing minor cases while leaving severe-injury exposure untouched.
Be alert to the incentive problem. Because these rates feed bids, insurance and bonuses, there is real pressure to classify marginal cases as first aid or to arrange light duty specifically to avoid a days-away entry. OSHA prohibits discouraging employees from reporting injuries, and a suspiciously low rate at a site with high near-miss counts or high employee turnover deserves scrutiny rather than praise. Leading indicators — near-miss reporting rate, hazard closure time, training completion — do not carry the same incentive and are worth tracking alongside.
What one recordable case is worth, by workforce size
| FTE workers | Hours worked | TRIR value of one case | Cases allowed at a 2.00 target | Cases allowed at a 1.00 target |
|---|---|---|---|---|
| 10 | 20,000 | 10.00 | 0.20 | 0.10 |
| 25 | 50,000 | 4.00 | 0.50 | 0.25 |
| 50 | 100,000 | 2.00 | 1.00 | 0.50 |
| 100 | 200,000 | 1.00 | 2.00 | 1.00 |
| 250 | 500,000 | 0.40 | 5.00 | 2.50 |
| 500 | 1,000,000 | 0.20 | 10.00 | 5.00 |
| 1,000 | 2,000,000 | 0.10 | 20.00 | 10.00 |
Hours are FTE × 2,000. The third column is 200,000 ÷ hours; the last two are target × hours ÷ 200,000. Note what the first row means: a 10-person site cannot record a single case in a year and stay under 2.00, so its rate is either 0.00 or 10.00 — there is nothing in between.
Errors that make a rate wrong or incomparable
- Including paid non-worked hours. Vacation, holiday and sick hours are not hours worked, and including them lowers every rate by the same proportion.
- Counting first aid as recordable. 29 CFR 1904.7(b)(5)(ii) lists first aid exhaustively. If a treatment is on that list it is first aid regardless of who administered it.
- Letting DART exceed TRIR. Every DART case is a recordable case. If the counts disagree, the log has been miscategorised.
- Forgetting the 180-day cap. Days away and restricted days are each capped at 180 per case, and ignoring the cap lets one severe case dominate the severity rate.
- Comparing across industries. Benchmark against your own NAICS code in the BLS survey, not against a national average.
- Reading a small site's year-on-year change as a trend. At 25 FTE one case moves TRIR by 4.00 points, so almost any change is within noise.
- Excluding temporary workers you supervise. Where you provide day-to-day supervision, their injuries and their hours both belong on your log.
Which regulation actually governs the count
The recordkeeping rules are 29 CFR Part 1904. Section 1904.4 sets the basic recording criteria, 1904.7 defines the general recording criteria including the first-aid list, 1904.29 covers the forms — the 300 log, the 301 incident report and the 300A annual summary — and 1904.32 requires the annual summary that most employers use as the source for these rates. Certain establishments must also submit data electronically under 1904.41. Rates computed from anything other than a Part 1904-compliant log are not comparable with anyone else's, which is the reason the recordability decision matters more than the arithmetic.
Where incidence rates fit in a safety programme
Incidence rates are lagging indicators: they count what has already gone wrong. They are indispensable for benchmarking and legally required for recordkeeping, but they cannot tell you what will go wrong next, and a site with a low rate and no leading-indicator programme is often just lucky.
The exposures that generate the cases are what a programme actually manages. Heat illness is a recordable-case source with a well-defined measurement standard behind it — the WBGT heat stress calculator converts environmental conditions and workload into a work/rest schedule that prevents cases rather than counting them. Powered industrial trucks are another consistent source, and confirming a lift is inside the truck's derated capacity with the forklift capacity calculator removes an exposure before it becomes a log entry.
On the operations side, safety and reliability are usually the same problem seen from two angles. Equipment that fails frequently generates unplanned intervention, and unplanned intervention is when people are most exposed — which is why an MTBF and MTTR analysis often predicts injury exposure better than a safety audit does, and why sites with poor OEE frequently have poor safety records too.
Finally, be careful with the arithmetic of improvement claims. A move from 3.81 to 2.00 is a reduction of 47.5% in the rate, not 1.81 percentage points, because these rates are not percentages — they are cases per 100 workers per year. Reporting them as percentage points is a common and confusing error.