Workplace injuries: reconcile case severity and exposure hours
Fewer reported injuries do not necessarily mean a lower injury rate. Exposure hours may also fall, and the severity distribution can worsen while the total rate stays unchanged. Research should reconcile the cases, their classifications and actual hours worked before interpreting a safety improvement.
The deliverable is a US-scope establishment ledger with separate columns for total cases, DART cases, days away, exposure hours and fatalities. The example is fictional, uses no medical conclusions and does not determine an employer’s legal reporting obligations. Evidence is read through October 25, 2026; a company-specific recordkeeping question requires its applicable rules and records.
Locate the rate and case definitions
BLS’s Employer-Reported Workplace Injuries and Illnesses in California, 2024 identifies its figures as nonfatal Survey of Occupational Injuries and Illnesses estimates. Table A’s first footnote gives the incidence formula: cases divided by employee hours worked, multiplied by 200,000. The Private industry injury and illness case types section distinguishes DART from other recordable cases.
The two-hundred-thousand-hour base represents one hundred equivalent full-time workers under the specified reference convention; it is a scaling constant, not a claim that the observed establishment actually employed one hundred people. The source’s case-type footnotes also clarify that days-away cases can involve transfer or restriction as well.
These definitions support the arithmetic and the need to avoid overlapping case counts. They do not turn a state or industry estimate into a list of companies, an OSHA enforcement finding or a universal cross-country benchmark. Keep sector, geography, year and the source’s exclusions beside any comparison.
Rebuild the two-period main case
Assume a fictional establishment reports one million actual hours worked in Period 1. It has ten total recordable nonfatal cases, including four distinct DART cases and six other cases. For this example no fatalities occur and the company uses the label TRIR for total cases per 200,000 hours. The numerator is explicitly stated rather than inferred from an acronym.
Total rate is 10 ÷ 1,000,000 × 200,000 = 2.0. DART rate is 4 ÷ 1,000,000 × 200,000 = 0.8. In Period 2 hours fall to eight hundred thousand, total cases to eight and DART cases remain four. Total rate is still 2.0, while DART rises to 1.0.
| Metric | Period 1 | Period 2 |
|---|---|---|
| Actual hours worked | 1,000,000 | 800,000 |
| Total recordable nonfatal cases | 10 | 8 |
| Distinct DART cases | 4 | 4 |
| Other recordable cases | 6 | 4 |
| Total rate per 200,000 hours | 2.0 | 2.0 |
| DART rate per 200,000 hours | 0.8 | 1.0 |
The case count fell 20%, but exposure fell by the same percentage. Saying the total rate improved 20% would confuse counts with rates. DART’s unchanged numerator spread over fewer hours produces a 25% increase in its rate. Preserve both outcomes instead of selecting the more favourable headline.
Close categories before calculating rates
Assume the four DART cases in each period comprise two days-away cases and two transfer/restriction-only cases. Count each case once in the DART union. A days-away case that later involves restricted duty remains one case in this ledger, not one days-away case plus another restriction case.
The category identity is total cases = DART + other recordable cases, under the expressly defined classification. If the company’s tables do not close, request the case rules, revisions and scope before inserting a residual. Some reporting systems present overlapping descriptive flags rather than mutually exclusive case categories; their columns cannot simply be added.
An injury count is also not necessarily the number of days or people affected. One worker can have more than one distinct event, and one case can involve several days. Use the source’s event and person identifiers when available instead of treating all three objects as equivalent.
Retain severity beside frequency
Suppose the two days-away cases in Period 1 involve a combined eight recorded days away, while the two in Period 2 involve forty. Under this fictional, consistently measured case cohort, days-away severity has increased even though the total recordable rate remained 2.0.
These days are not eight or forty additional cases. Keep case count and days in different columns, and document whether the days are calendar days, scheduled workdays or another company-defined measure. Do not combine incompatible day conventions to infer a severity change.
Fatalities deserve an explicit separate field. Both periods in the main example show zero. The BLS source used here concerns nonfatal estimates; do not infer a fatality count from its nonfatal rate or use that table alone to assess fatal risk. If a fatal event is disclosed, record it with its actual scope and case treatment rather than allow a low rounded frequency metric to obscure it.
Test aggregation and exposure boundaries
For a separate fictional two-site comparison, Site A has two cases in one hundred thousand hours, a rate of four. Site B has eight in nine hundred thousand hours, a rate of approximately 1.78. The combined rate is ten divided by one million times 200,000, equalling two; it is not the simple average of four and 1.78.
The correct combined denominator is summed comparable exposure. Do not average site rates without weights. Likewise, headcount is not a direct substitute for actual hours worked when overtime, part-time work or shutdown time changes. If only employment is available, mark the hour-based rate unverified instead of inventing every employee’s annual hours.
Contractor boundaries need explicit matching. An employer-only case numerator divided by hours including contractors could make a rate appear lower without any change in underlying events. This worksheet does not decide which entity is legally responsible for a case; it asks whether the issuer’s chosen numerator and denominator cover the same population.
Compare company evidence with survey context
The California release gives industry and sector context, with methodology, exclusions and rounding notes. A company’s global disclosure may include other countries, contractors or a different reporting year. Do not rank it against a California private-industry number without first establishing comparability.
A rounded 0.0 is not necessarily a count of zero. Request unrounded data or the event count before treating a displayed rate as proof of no incidents. For small establishments, one extra case can move a rate substantially; report the count and exposure rather than imply precision from decimal formatting.
Use the exposure verification sheet
| Input | Obtain | Check |
|---|---|---|
| Cases | Event register and classification rules | Distinct cases and revised outcomes |
| DART | Days-away and transfer/restriction records | Union, not overlapping flags added twice |
| Hours | Actual work-hour records | Same population, site and period as cases |
| Severity | Days and fatal-event disclosures | Units and scope retained separately |
| Aggregation | Site numerators and denominators | Sum cases and hours before forming rate |
| Benchmark | Dated BLS table and notes | Nonfatal, geography, industry and exclusions |
When reports disagree, compare event dates, later severity reclassification, ownership changes and contractor scope. Preserve the earlier version and the reason for revision. Do not force an unreconciled case difference into a safety-performance narrative.
If hours or classifications are missing, provide the counts that are supported and identify the absent denominator. A lower injury count is a valid observation, but the rate conclusion remains unknown. No arithmetic ratio by itself establishes why incidents occurred, whether all were recorded or whether controls are effective.
Update after revised cases, recorded days, actual hours, boundary changes or a new comparable BLS release. The conclusion should state frequency and severity together, specify the exposed population and explain what changed in counts versus hours, without converting a low rate into a guarantee of safety or a legal compliance conclusion.
Preserve case revisions and the day-count clock
A case may later acquire days away or restrictions, changing the severity information attached to the original event. Record the issuer’s case-year convention and revision treatment rather than assign the later update to whichever period makes the trend look better. Keep original event date, classification date and published revision date distinct.
Days can continue accumulating after a reporting cutoff. Compare the same case cohort at a comparable follow-up stage, or state the different observation windows. Otherwise a recently reported group may look less severe simply because less time has elapsed. This is a limitation of the available measurement, not a medical prediction about the individuals involved.
For consolidated data, preserve acquisition and disposal dates with the hours and cases actually covered. A period-end employee count does not reconstruct partial-year exposure. If the company changes the perimeter, provide an explicitly supported comparable series or leave the trend conditional; do not mix old-scope cases with new-scope hours.
FAQ
Can fewer cases leave the injury rate unchanged?
Yes. In the main case, both total cases and exposure hours fell 20%.
Is a company’s global TRIR directly comparable with a BLS state estimate?
Only after population, period, case definitions and other scope differences are reconciled.