An OSHA recordable incident rate calculator takes two inputs, a case count and an hours figure, and returns a rate. The arithmetic is the easy half. The count is where most rates go wrong, because a calculator cannot tell whether a logged event belongs in the numerator. Give it a loose count and the output looks just as precise as one built on careful determinations, and it is wrong in exactly the same confident way.
This post runs one year of a small site through the calculation, one event at a time. The figures are illustrative, built for this example rather than taken from a customer. You can check every line of arithmetic against your own numbers.
The formula, once
The incident rate per 100 full-time equivalent workers is recordable cases multiplied by 200,000, then divided by total hours worked. The 200,000 is the base for 100 full-time workers at 40 hours a week for 50 weeks. The Bureau of Labor Statistics publishes a hypothetical example of 11 cases and 130,000 hours giving a rate of 16.9, and the National Safety Council's definition of TRIR uses the same structure. For a step-by-step walk through the formula itself, read the TRIR calculation guide.
Two conventions matter before any number goes in. Hours are time actually worked, so holidays, sick leave and vacation are excluded. The numerator is recordable cases under Part 1904 for the calendar year, which means the case count has to pass the determination first. The rest of this post is about that second point.
Set the inputs for the example
Our example site worked 124,000 hours in the year. That is 62 full-time equivalents at 2,000 hours each. Seven events were logged against the site during the year, and each one was run through the three gates in 29 CFR 1904.4: work-relatedness, new case, and a recording criterion.
| Event | Gate outcome | In the numerator? | Why |
|---|---|---|---|
| Seasonal cold reported at shift handover | Not work-related | No | Personal illness with no workplace exposure identified under 1904.5 |
| Ankle twist in the staff car park going home | Not work-related | No | Commuting exception under 1904.5(b)(2) |
| Finger cut closed with a butterfly bandage | First aid only | No | Non-prescription treatment, with no restriction or time away |
| Laceration closed with sutures | Medical treatment beyond first aid | Yes | Sutures are a treatment beyond first aid |
| Lifting strain with a five-day lifting limit | Restricted work | Yes | Lifting is a routine job function and the restriction prevents it |
| Shoulder flare-up after documented full recovery | New case from a work event | Yes | Recovery is documented, the lifting event is new, and prescribed physiotherapy is medical treatment beyond first aid |
| Hearing shift at audiometry, no baseline on file | Needs review | Not yet | The shift cannot be confirmed against the baseline |
Read the table from the top. Three rows fall away at the first gate or the first criterion, and one row cannot be decided until a fact arrives. Only three rows are settled as recordable.
Three rates from one year
Counting every logged event gives 7 × 200,000 ÷ 124,000 = 11.3. That is the figure a spreadsheet produces when the count is simply the number of rows in the log. It is wrong, because three of those seven events sit outside the numerator entirely.
Counting only the confirmed recordable cases gives 3 × 200,000 ÷ 124,000 = 4.8.
Counting the open hearing case as recordable gives 4 × 200,000 ÷ 124,000 = 6.5.
While the hearing case is open, the honest figure is a range from 4.8 to 6.5. A single number on a dashboard will be quoted, and it will be quoted as settled. A range with a named open item is the rate you can defend to a site manager, an insurer or an auditor. The gap between the top and bottom of that range is one case, and it is the case that needs an answer.
Why the hearing case stays open
Hearing is where a count most often stalls. Under 29 CFR 1904.10, a standard threshold shift is a change of 10 decibels or more, averaged across 2000, 3000 and 4000 Hz, measured against the employee's baseline audiogram and corrected for age. The test result alone does not close the case. A dip on this year's audiogram means nothing until a baseline audiogram sits beside it, and the age correction has been applied.
Without the baseline on file, the shift cannot be confirmed, so the case cannot be closed as a yes or a no. The tempting move is to count it because the test looked bad. That is a guess dressed as a determination. Put the baseline on file, apply the age correction, and the case moves to one side of the line. Our standard threshold shift guide walks through the test in order.
The calculator does not check the count
This is the point the table makes and the formula cannot. Every row changes the answer, and only one of them, the cold, is decided by a reading of the rules the calculator never performs. A calculator that multiplies whatever number you type returns 11.3 with the same certainty as it returns 4.8. The output looks identical. Nothing on the screen tells you three of the inputs are outside the numerator.
The shoulder case shows why a gut check is not enough. The worker never missed a shift, yet the case is recordable. It is a new case caused by a workplace event, and the prescribed physiotherapy is medical treatment beyond first aid, which meets the third gate on its own. The lifting strain is recordable for a different reason: restricted work is a criterion in its own right, even though it cost no days away. Counting days away alone would miss both. For the trigger list, see the recordable criteria checklist.
Job13's free recordability check runs each event through the gates in order and returns NEEDS-REVIEW when a missing fact could change the count, instead of forcing a yes or a no. The recordable decision tree sets out the same sequence on paper if you want the logic before the tool.
Practical checks before you publish a rate
- Count after the determination, not before. The log is the input and the determined list is the numerator. Keep a short decision note beside each borderline entry, so the next person can see why it was counted or dropped.
- Use one hours basis every year. Exclude leave consistently. Changing the basis between years makes the trend meaningless before the cases even arrive.
- Publish a range while a case is open. Name the open item in the same line, so the reader knows what would move the figure.
- Keep the 300A in step with the rate. The OSHA 300A summary supplies the case count that the rate uses, so the two should match before you publish either. The 300A summary guide covers the totals.
- Check the illness-only base. For illnesses alone, BLS uses 20,000,000 hours in place of 200,000, which gives a rate per 10,000 full-time workers. Confirm which base a calculator uses before you compare its output with anyone else's; the calculators compared covers the inputs to check.
Why small hours bases swing so hard
On 124,000 hours, each recordable case moves the rate by 200,000 ÷ 124,000, which is about 1.6 points. A site with a larger base absorbs a single case far better. This is the arithmetic behind the advice to read a small company's rate as a trend rather than a score. The TRIR calculation guide explains the volatility in more detail, and the DART rate explainer splits the same count into days-away and restricted cases if you need that breakdown.
Frequently asked questions
What goes in the numerator of an incident rate?
Recordable cases under Part 1904 for the calendar year. That means every logged event that passes the work-relatedness, new-case and criteria gates. A logged event is not automatically a recordable case.
Does a rate calculator tell me whether a case is recordable?
No. It performs arithmetic on the count you give it. The determination comes first, and the calculator inherits whatever the count contains.
Why did my rate jump after I added one case?
On a small hours base, one case is a large step. At 124,000 hours each recordable case adds about 1.6 points. A larger base damps the swing, so a single case reads as a smaller change.
Should I publish a rate while a case is still under review?
Publish a range that names the open case, not a single figure you will have to revise. A range that says what would close it is more useful to a reader than a precise number that changes next month.
