MSHA publishes a fatality alert and a final report for the deaths it investigates, and both are free on msha.gov. Reading them one at a time answers one question: what happened at that mine. A safety manager usually needs a different one answered: how many deaths per year, in which accident class, in which commodity, in which state.
Those count questions are answered from MSHA's open government data, not from the report PDFs. The reports tell you why a death happened. The data tells you how often, and where your own work sits in the pattern.
For the counts themselves, refreshed every month from MSHA's data, see MSHA fatalities this year: deaths by year since 2000, by month, coal against metal/nonmetal, and by accident classification.
Worked example
A published table in the SME Mineral Processing and Extractive Metallurgy Handbook lists 49 conveyor-related fatalities reported by MSHA from 1995 to 2006. You want to check the table's percentages from its own counts before you quote it in a toolbox talk. The counts are 28, 16, 2, 1, 1 and 1. The table prints the same label, failure to provide adequate maintenance procedure, on its first two rows. Treat them as two separate hazard categories until you read the original study.
| Step | Calculation | Result |
|---|---|---|
| Add the counts to confirm the total | 28 + 16 + 2 + 1 + 1 + 1 | 49 fatalities |
| Share of the first row | 28 / 49 x 100 | 57.1% (table prints 57) |
| Share of the second row | 16 / 49 x 100 | 32.7% (table prints 33) |
| Combine the two maintenance-procedure rows | (28 + 16) / 49 x 100 = 44 / 49 x 100 | 89.8% of the 49 deaths |
| Everything else in the table | 49 - 44 = 5; 5 / 49 x 100 | 5 deaths, 10.2% |
Answer: 44 of the 49 conveyor deaths, 89.8%, fall in the two rows that carry the maintenance-procedure label, and the printed percentages match the counts to rounding. That is a 1995-2006 snapshot, so run the same split on current years before you rely on it.
Where MSHA publishes the reports
MSHA posts fatality information on its data and reports pages. Two document types matter, and they arrive at different times.
The fatality alert comes first. It is short, usually a page or two, and written so other operators can act before the investigation finishes. It states what happened, lists best practices, and sometimes shows a sketch of the scene.
The final report comes later, once MSHA closes the investigation. It is the long one. It tells you why the accident happened, not only how.
Browse by year, commodity or accident class rather than in date order. A site that runs conveyors should read every conveyor alert from the last few years. Reading every report is a full-time job, and most of them do not apply to your site.
What a final report contains
Reports differ in length, but they follow a recognizable shape. Expect these parts:
- What happened. A timeline of the shift, who was where, and what the victim was doing.
- The investigation. Interviews, the scene, the equipment, the records MSHA reviewed.
- Root causes. The conditions that let the accident happen, such as a missing procedure, a defeated guard or a training gap.
- The standards cited. The 30 CFR sections MSHA found violated, and the orders and citations issued.
Read the root-cause section first. It shows whether the accident was a one-off or a pattern you could have at your own site.
Enforcement is worth knowing about too. The MSHA Program Policy Manual, Volume III, in its section on determination of penalty (100.5, p. 102), lists fatality and serious-injury violations as requiring a Special Assessment Review (MSHA Form 7000-32). That applies at every negligence level, from none to reckless disregard. A citation tied to a death gets a higher level of review than an ordinary one.
The same manual (49.6, p. 65) says MSHA has the option to run a Part 50 reporting audit at a mine where a fatal accident occurs. It may do so after complaints of unreported accidents, other reason to suspect underreporting, or injury rates that differ sharply from similar mines. A fatality can bring a records audit as well as an investigation.
Why one report is not enough
A single report covers one death. It cannot tell you whether that death is typical, rising or rare.
Take conveyors. The SME Mineral Processing and Extractive Metallurgy Handbook (5.2 Belt Conveyors, p. 698) reproduces a table of 49 conveyor-related fatalities reported by MSHA from 1995 to 2006, drawn from Kecojevic et al. 2008. The two rows labelled with maintenance procedure failures hold 28 and 16 of those deaths. Read the reports one by one and you would never see that pattern.
You find it by counting. That is why the worked example above matters: the table gives you counts, and the arithmetic turns them into a share you can say out loud to a crew.
The table is also old, and it covers conveyors only. Your own question might be about haul trucks, or ground fall in your state, over the last five years. Nobody has published that table for you, so you build it from the data.
The questions a safety manager actually asks
The questions are plain, and they all reduce to counts with a filter:
1. How many fatalities per year over the last 10 years? 2. Which accident classes cause the most of them? 3. How does my commodity compare with the others? 4. What does my state look like against the national count? 5. What did the narratives say about the deaths in a class I care about?
Questions 1 to 4 are group-by queries on MSHA's open data. Question 5 is a text search. None of them requires opening a report.
The open data is published through the same MSHA data and reports section. It includes accident records, violations, inspections and mine information. Spreadsheets work for a first pass, but the files are large and the codes need a lookup table.
Read the filters before you quote a count
Part 50 accident data records more than work-related traumatic deaths. A few real narratives show why:
- A 2016 Wyoming entry reads that the employer did not believe the fatality was work related, and that verbal findings pointed to a cardiac episode. It is filed under "Injuries due to natural causes".
- A 2014 Florida entry describes a truck driver who suffered a fatal heart attack while reporting to work. Same class.
- A 2017 Utah entry describes a fatal cardiac arrest from a personal illness. Same class.
- A 2007 Kentucky entry describes a refuse truck driver struck by a train at a crossing. The note says MSHA had not determined whether the accident was reportable, because the location belonged to the railroad.
A count of "fatalities" that ignores these will not match a count of investigated accidents. Neither is wrong. They answer different questions. Write down which one you used, because the first person to challenge your slide will ask.
A 2025 West Virginia entry in the hoisting class is different. It records an employee with fatal crushing injuries while assisting an MSHA inspection of an elevator inside a preparation plant. That narrative belongs in the count.
Asking the questions in plain words
You can build all of this in a spreadsheet or a database. If you would rather type the question, Big Dog Mining's free mining AI tool is one way to do it. It is one connector you add to Claude or ChatGPT.
It queries MSHA accident, violation, inspection and mine data, and returns counts, rates, trends and charts. It also searches MSHA's own accident narratives in plain words. A third function searches 22 mining engineering handbooks and quotes short passages with the book, chapter and page. It is free, it needs an email sign-up, and it is read only.
Here is how the five questions look typed out:
- "How many fatalities per year for the last 10 years, excluding natural causes?"
- "Break fatalities down by accident class for metal and nonmetal."
- "Compare fatalities by commodity: sand and gravel versus crushed stone."
- "Show fatalities by state for the last 5 years."
- "Search accident narratives for conveyor deaths during cleaning."
Treat the answers as research aids built from public data and published references. Check any number that goes on a slide against the source record. The tool does not replace the final report's root-cause discussion, and it does not decide anything on a live operation. That belongs to the competent person on site.
Reading data lag and reporting rules
Open data trails the event. The MSHA Small Mine E-Guide (Top 20 Citations by Mine Type, p. 100) lists the reporting checklist. Form 7000-1, the Mine Accident, Injury, and Illness Report, goes to MSHA within 10 working days of an accident. The quarterly employment report on Form 7000-2 is due within 15 days after the quarter ends.
So the most recent weeks of any trend are incomplete. A dip in the latest month is not an improvement. Leave off the current quarter, or label it provisional.
Contractors should also know their own reporting duty. The Program Policy Manual (49.6, p. 65) says independent contractors performing any of the nine service or construction types listed under Section 45.3 report accidents, injuries and illnesses under 30 CFR 50.20, keep records under 50.40, and file quarterly employment reports under 50.30. Other contractors are not required to comply with those sections unless the district manager decides otherwise. The text of 30 CFR Part 50 is free on eCFR. Read it against your own scope of work.
Putting the numbers next to your own site
A national count shows the overall picture. It does not tell you whether your site is safe. For that you need a rate, which divides incidents by hours worked. The MSHA incidence rate calculator does that arithmetic for your own hours and your own injuries.
Your daily report is the other input. The crew list, equipment, hours and hazards are already on it. A superintendent who records them consistently, as in what a superintendent should put on the daily report, has a record that can be matched against a pattern in the fatality data. A hazard noted on Tuesday that matches a conveyor-cleaning narrative from MSHA is worth a conversation on Wednesday.
Training is a related lever. The SME Mining Reference Handbook, 2nd edition (Chapter 23: Health and Safety, p. 561) lists the minimum plan content MSHA requires of each U.S. operator. It includes 40 hours of basic training for new underground miners with no underground experience, 24 hours for new surface miners with no surface experience, 8 hours of annual refresher training for all miners, and task training for new jobs. If a fatality pattern points at a task, the annual 8 hours is where it can be taught.
Any record is only as good as what the crew writes in it. A hazard that never makes it onto paper never reaches a trend line.
A short routine you can run each quarter
1. Pull the fatality counts for the last 10 years, with your filter written down. 2. Split by accident class, then by your commodity. 3. Pick the top two classes that match work you do. 4. Read the alerts and final reports for those classes, root causes first. 5. Search the narratives for your task, such as cleaning, tire work or travelways. 6. Bring one finding to the next toolbox talk, with the source on the slide.
That routine takes a morning, not a month. It also leaves a paper trail: the query, the filter, the date and the report you read.
If you need to report a hazard rather than study one, MSHA's hazard line is (800) 746-1553, per its Miners' Rights and Responsibilities publication (OT 2). You do not need to identify yourself. Include the mine name, location, state and MSHA ID if you have it.
Read the reports for the why, count the data for the how often, and write down your filter. Then take one finding to the crew.
Run your own numbers
Rather than 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 I find MSHA fatality alerts and final reports?
Both live on msha.gov under its data and reports section. The alert comes first and is short. The final report follows after the investigation closes and carries the full findings. You can browse by year, mine type or accident class, so start there before reading anything in full.
Why do fatality counts differ between sources?
Some deaths are reported under Part 50 but are not work-related, such as cardiac events. Others sit outside MSHA jurisdiction. A count that includes everything will not match one limited to investigated accidents. State your filter every time you quote a number.
Does a contractor have to report a fatality to MSHA?
Independent contractors who perform any of the nine service types listed under Section 45.3 must report accidents under 30 CFR 50.20, keep records under 50.40, and file quarterly employment reports under 50.30. Other contractors are generally not required to, unless the district manager decides otherwise. Check your own work against Part 45.
Can an AI tool replace reading the final report?
No. A tool can count, trend and search narratives quickly, which tells you which reports deserve your time. The root-cause discussion in a final report is still where the lessons are. Decisions on a live operation belong to the competent person on site.
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.
