It counts the injuries its recordkeeping rule requires employers to log, which in 2023 came to 2.4 recordable cases per 100 full-time workers in private industry (BLS Survey of Occupational Injuries and Illnesses, 2023). Knowing what sits inside that number, and what does not, is the first step to reading any safety claim an employer or union makes.
Union Times Today publishes information, not professional advice. This article explains how federal injury recording and reporting works; it does not tell anyone what their own case requires.
What does a 'recordable' injury actually mean?
A case is recordable when it is work-related and involves death, days away from work, restricted work or job transfer, medical treatment beyond first aid, loss of consciousness, or a significant injury diagnosed by a physician (OSHA recordkeeping rule, 29 CFR Part 1904). First aid, as the rule defines it in a closed list — a single dose of nonprescription medication, a cleaning and bandaging, and similar — does not count.
The definition does the heavy lifting. The difference between a logged and an unlogged case is often the difference between a prescription and an over-the-counter bottle, or one stitch and a butterfly bandage.
What forms does an employer have to keep?
Three, and they work together. The OSHA 300 log lists each recordable case through the year. The 301 incident report captures details for each entry. The 300A annual summary totals the cases and hours worked, and covered employers must post it in the workplace from February 1 through April 30 each year (OSHA, 29 CFR 1904.32).
- Record each qualifying case on the 300 log within seven calendar days of learning of it.
- File a 301 report for the same case.
- Tally the log on the 300A summary after the year closes.
- Post the 300A from February 1 to April 30.
- Establishments covered by electronic reporting submit the 300A data through OSHA's portal, most by March 2.
Why does the rate matter so much in disputes?
Because the rate, usually called TRIR, is the currency of the safety argument. It divides recordable cases by hours worked and multiplies by 200,000, the annual hours for 100 full-time workers. A company with a rate below the industry average can market itself as safer; a union can use a rate above average to press for staffing changes. Both claims rest on the logged data, with all its definitional edges.
Government buyers and large contractors use the same rate to screen bidders, which gives employers a financial reason to argue for the narrowest defensible recording. Safety professionals have debated that incentive openly for years.
What does the record undercount?
The survey itself carries qualifications. The Bureau of Labor Statistics notes that the survey measures recorded cases, and researchers have long documented gaps around musculoskeletal disorders, long-latency illness and cases shaped by incentive systems that reward low counts (BLS SOII methodology notes; academic literature summarized there). Illness categories in particular are hard to attribute to a workplace.
Undercounting does not mean falsification, and this article does not allege it. It means the number is a regulated artifact: it counts what a rule defines, collected from the employers the rule binds. Contract and gig workers, who fall outside a host establishment's log, are a structural blind spot.
Which injuries must be reported immediately, not just logged?
Three triggers stand apart from the logging system. Employers must report a work-related fatality to OSHA within eight hours, and an in-patient hospitalization, amputation or loss of an eye within 24 hours (OSHA, 29 CFR 1904.39). These reports feed enforcement, not just statistics.
Penalty amounts adjust annually for inflation. The maximum penalty for a serious or other-than-serious violation was $16,131 per violation in 2024, and willful or repeated violations carried a maximum of $161,323 (OSHA civil penalty amounts, 2024). Those figures apply per violation, not per case.
Two limits of the survey deserve their own paragraph. First, industry rates vary enormously, so the national 2.4 figure is close to useless for bargaining or bidding in a specific sector; heavy trucking, warehousing and nursing homes sit well above it, offices far below (BLS SOII, 2023). Second, the survey is an estimate with a sampling frame, not a census of the private economy, and small-establishment cells carry wide confidence intervals that rarely make it into the quotes.
Can employees see the employer's log?
Yes, within limits the rule itself sets. Current and former employees, and their representatives, have a right to review the 300 log and the 300A summary, and the employer must provide copies by the end of the next business day (OSHA, 29 CFR 1904.35). The rule also bars the employer from discouraging reporting, directly or through incentive schemes that punish reported cases — a provision written precisely for the safety-bonus programs that quietly priced cases out of the log.
Retaliation for reporting is separately enforceable. Section 11(c) of the Occupational Safety and Health Act protects workers who report injuries, and complaints run on short administrative deadlines through the department, not the courts (DOL/OSHA, Section 11(c) procedures). The two protections work together: the recordkeeping rule makes the data available, and Section 11(c) makes it costly to suppress.
For employers, the compliance calendar is annual and unforgiving. The log runs on the calendar year; the summary posts through April 30; electronic submissions land by March 2 for covered establishments, and the size and industry triggers for electronic reporting have themselves changed by rule in recent years (OSHA e-recordkeeping rule updates). A safety claim made in May about the prior year, therefore, can be checked against a posted document in the workplace itself — which is exactly what the posting requirement was designed to make possible.
How should a reader use any published injury rate?
Ask three questions. What year does the rate cover, since the figures are annual and lagged. Which industry baseline is the comparison, since a warehouse rate against an office average tells you nothing. And is the rate a log-based measure or a survey estimate, because the two come from different systems with different margins of error.
The 2023 private-industry rate of 2.4 recordable cases per 100 full-time workers is a real, sourced anchor (BLS, 2023). Everything a reader builds on it inherits its definition, and its limits.
Related: Algorithmic management at work · Reading employer health data.
For more context, read Automation on the warehouse floor.
For more context, read Reading employer health data.
For more context, read Workplace surveillance and the law.