Every injury and accident an operator files under Part 50 comes with a written narrative: a few lines in the operator's own words about what the person was doing and what went wrong. That text is the useful part of the record. The classification codes only tell you which bin the incident landed in.
So search by what happened. "Caught in a conveyor belt while cleaning spillage" finds incidents your crew can learn from. A code search returns every conveyor injury, including the ones that have nothing to do with your task.
The work is turning the matches into a toolbox talk or a job hazard analysis: the pattern across incidents, the task being done, and the control that was missing.
Worked example
You are building a rollover toolbox talk. You pull 6 Part 50 narratives from the public record and read them all before deciding what the talk is about. The narratives are real; the point is the reading method, not a statistic.
| Step | Calculation | Result |
|---|---|---|
| Count the narratives that match the task in plain words ("overturned") | 2006 truck overturned hauling stone + 2009 992 loader turned over into a pit = 2 matches out of 6 narratives | 2 of 6 narratives |
| Express the match as a share of what you pulled | 2 / 6 = 0.33 | 33% of this reading pile (a pile of 6 is a reading list, not a rate) |
| Check the gap between an injury and the date it became reportable in the slip narrative | 21 Sep 2006 to 31 Jan 2007 = 9 days left in Sep + 31 Oct + 30 Nov + 31 Dec + 31 Jan | 132 days between the event and the doctor's call that made it reportable |
| Convert the 10 working day Form 7000-1 deadline into calendar days for an accident on a Monday, with no holidays | Tue to Fri = 4 working days, next Mon to Fri = 5 more (9), following Monday = 10th working day; 14 calendar days after the accident | 14 calendar days |
Answer: Of the 6 narratives, 2 (33%) are rollovers, so the talk is about mobile equipment turning over. One narrative took 132 days to become reportable, so the date on a record is not always the date of the event.
What a Part 50 narrative holds
Part 50 requires an operator to report each accident, occupational injury and occupational illness at the mine. Form 7000-1 is due within 10 working days after an accident or injury occurs or an illness is diagnosed (MSHA Small Mine E-Guide, Part 50 Accident Reporting, p. 39). The form carries a free-text description of what happened.
MSHA's own privacy notice says the information is collected to help determine the cause of accidents and injuries, and to help prevent and reduce future ones (MSHA Small Mine E-Guide, Part 50 Accident Reporting, p. 44). The narrative is where that cause is written down in plain words. The category fields beside it are a filing label.
The SME Mining Engineering Handbook makes the same point about the wider record. MSHA reports contain detailed information and are designed so that such tragedies do not reoccur (SME Mining Engineering Handbook, 3rd edition, 15.1.6 Information Sharing, p. 1560). That passage covers investigation reports, which are a separate and larger document than a Part 50 narrative. The idea still carries over: the written account is the part meant for learning.
The full rule text is free on eCFR, and MSHA publishes its accident, injury and illness data at msha.gov. If your question is about a death rather than an injury, the fatality investigation reports are a different set, covered in this guide to finding and questioning fatality reports.
Writing the search the way a miner would say it
Codes are a poor starting point for a toolbox talk. A code like "machinery" or "handling of materials" covers everything from a pinched finger to an amputation. You would read hundreds of records to find the 5 that match your job.
Plain words work better. Describe the task, the equipment and how it went wrong, the way a shifter would at the pre-shift meeting:
- "Caught in a conveyor belt while cleaning spillage"
- "Loader turned over on a pit ramp"
- "Pushed up out of a work area and hurt a shoulder"
- "Foot caught between conveyor car and glides"
Each one is a task plus a mechanism. Write it that way and the nearest narratives come back first, even when the filer used different words, such as "tail pulley" for one and "return idler" for another.
Keep three habits:
- Search the task, not the injury. "Cleaning spillage" is a task. "Crushed hand" is a result.
- Run two or three phrasings. One crew says "walking out a stuck truck", another says "recovering a bogged haul truck".
- Read at least 5 to 10 narratives before you decide what the talk is about. One narrative is an anecdote. A repeat is a pattern.
Reading each narrative for three things
Pull the same three items from each narrative, on a sheet or in a spreadsheet.
The task. What was the person doing in the minute before the injury? Skip the job title. "Cleaning spillage at a tail pulley" is a task. "Laborer" is not.
The pattern. What does this narrative share with the others? Look for the same equipment, the same step in the task, the same kind of recovery or cleanup, or the same time pressure.
The missing control. What would have stopped it? This column is your judgment, not MSHA's. Common answers are lockout not applied, guard removed for the job, no spotter, no berm, no stop-work call, no second person.
The narratives rarely state the missing control outright. They describe the event. You infer the control by asking what a well-run version of the same task looks like on your site. A competent person should check that inference.
Building the toolbox talk
A good talk runs 5 to 10 minutes and covers one thing. Use the sheet to build it in this order:
1. Name the task. "Today we clean up under the tail pulley on number 2." 2. Read one narrative, short and unedited. Strip names and mine details. Read what the filer wrote. 3. Say what repeated. "We found 3 narratives with the same step: cleaning with the belt still moving." 4. State the control you use, or the one you are adding. One sentence. 5. Ask the crew one question. "Where on our belt is this hardest to do the right way?"
The last step matters most. The crew knows where the guard gets pulled and where the lockout point is a long walk away. Their answer is often the real hazard finding.
Keep the talk to one task. A talk that tries to cover conveyors, rollovers and shoulder strains teaches none of them.
Turning the pattern into a job hazard analysis
A JHA breaks a task into steps, names the hazard at each step, and lists the control. The narratives feed the middle column.
| Task step | Hazard seen in narratives | Control on our site |
|---|---|---|
| Stop the belt and isolate | Cleaning started before isolation | Lockout point named, tagged, verified |
| Remove spillage near the pulley | Foot or hand drawn into a pinch point | Guard in place, cleaning tool with reach |
| Restart | Person still in the area | Walk-down and call-out before restart |
The left column comes from your own task. The middle column comes from the narratives. The right column comes from your site and your competent person. Narratives can show you hazards you had not listed. They cannot supply your controls.
Then put the control where the work is recorded. If your crew already fills in a shift report, add the check there. A daily report template for mining and civil crews is a natural home for a line like "belt isolation verified before cleaning." The control then leaves a trace the same day.
Where the narratives mislead
Read them with a sceptical eye. Four limits matter.
Quality varies. Some narratives are detailed. Others are a line. A thin narrative tells you little about the missing control.
Some are about the paperwork, not the hazard. Among the narratives pulled for the example above, one operator states that the company disagrees with MSHA's position that the event required a Form 7000-1, and is filing to provide the information MSHA believes it must. Another notes that a violation was issued for failing to file a report. These tell you about reporting disputes. They say nothing about preventing injury.
Dates are not always event dates. In one slip narrative, the event was in September 2006, care was requested in November, and a doctor's call on 31 January 2007 made it reportable. That is 132 days. A trend chart by report date can push an incident into the wrong period.
The match may be near, not exact. A search for conveyor cleaning can return "foot caught in between the conveyor car and glides" from a coal operation. The equipment differs, but the lesson about a pinch point and a moving machine holds. Say so in the talk instead of presenting it as your exact task.
Still read the narratives. Read several, and say plainly what each one is.
Doing the search with the Big Dog Mining tool
You can do all of this by reading the public data yourself. The slow part is the first pass: finding the narratives that match a task in everyday words.
The free Big Dog Mining mining AI tool can speed that up. It is a single connector you add to Claude or ChatGPT. It queries MSHA accident, violation, inspection and mine data, returning counts, rates, trends and charts. It searches MSHA's own accident narratives in plain words. It also searches 22 mining engineering handbooks and quotes short passages with the book, chapter and page. It is free with email sign-up and read only.
A prompt for the conveyor example:
> Find MSHA accident narratives where someone was caught in a conveyor belt while cleaning spillage. List the task in each, what the narratives have in common, and any handbook passage on guarding or lockout.
You get the matching narratives and a page-referenced passage to check. You still decide what the missing control was. The tool is a research aid built on public data and published references. Decisions on a live operation belong to the competent person on site.
Sizing your own crew against the record
A pattern in other people's incidents tells you what to talk about. It does not tell you how your crew compares. For that, you need your own recordable count and hours worked. The incidence rate calculator turns those two numbers into a rate you can track quarter to quarter.
For the industry's worst outcomes, the MSHA fatality numbers are counted from the same Part 50 reports every month, by accident classification, commodity and state.
Keep the near misses too. A report of a close call is the cheapest narrative you will ever get. If your crew stops writing them, your own record goes quiet before your injuries do. The shift supervisor's log is the natural home for them, and this note on what a superintendent should record daily covers how to keep it usable.
Reporting duties still apply
Learning from other sites does not change your own duties. An immediately reportable accident needs a call to MSHA at 1-800-746-1553, no later than 15 minutes after you know or should know it happened (MSHA Small Mine E-Guide, Part 50 Immediate Notification, p. 35). The operator must also investigate each accident and write its own report. That report has to include a description of the steps taken to prevent a similar occurrence (MSHA Small Mine E-Guide, Part 50 Accident Reporting, p. 39).
That last item is what your toolbox talk produces: a named control, written down, for a specific task. Run the talk before an accident and your investigation report starts from a better place.
Run your own numbers
You don't have to work it by hand every time. The msha incidence rate calculator does this arithmetic and shows its working. The rest of the set is on the mining calculators page.
Common questions
Where do the narratives in MSHA accident data come from?
They come from the operator or contractor who files Form 7000-1 under Part 50. The form asks for a description of what happened, so the text reflects the filer's wording and care. Some are detailed, some are one line. Read several before drawing a conclusion about any single task.
How fast must an accident be reported to MSHA?
For immediately reportable accidents, call 1-800-746-1553 right away and no later than 15 minutes after you know or should know it happened. The written Form 7000-1 is due within 10 working days after the accident or injury. The full wording is in 30 CFR Part 50 on eCFR.
Can I use a Form 7000-1 as my own accident investigation report?
Generally no. Under 30 CFR 50.11(b), the operator must investigate each accident and write a separate report, and cannot use Form 7000-1 for it. The exception is a mine with fewer than 20 miners, for an occupational injury not related to an accident. Check the rule text for your case.
Is a search of narratives enough to write a JHA?
No. Narratives show what went wrong elsewhere. A JHA still has to describe your task, your equipment, your site conditions and your controls, and the competent person on site decides what applies. Use the narratives to ask better questions during the walk-through, not to replace it.
Getting this off paper
Most contractors already know these numbers. The cost is in re-keying them into a report, then a timesheet, then an invoice. If that is where your week goes, book a 20-minute call and we will look at your actual paperwork.
