Heinrich's safety triangle rests on two tenets: that lower-severity occupational incidents outnumber fatalities in a stable pyramid, and that similar causes underlie events of all severities. A 13-year panel of more than 25,000 U.S. mining establishments confirmed the first tenet's descriptive form while showing that it depends on how severity is coded, but the second, causally load-bearing tenet was left untested. The open problem is to determine, at the level of the individual mining establishment, whether the root causal factors and event etiologies of reportable noninjuries (near misses), reportable (no-days-lost) injuries, days-lost injuries, and permanent disabling injuries are shared with those of fatalities — and therefore whether reducing lower-severity events can be expected to interrupt the causal chains leading to deaths — with conclusions that are demonstrably robust to the severity-delineation scheme, the temporal lag structure, and surveillance-reporting quality.

Background

In 1931, the industrial-safety pioneer H. W. Heinrich proposed the accident (or safety) triangle: for every serious or fatal workplace injury there is a large, roughly fixed count of minor injuries and no-injury incidents, often quoted as 29 minor injuries and 300 no-injury events per major one. Later elaborations, notably Frank Bird's injury pyramid, turned this ratio into a management doctrine: because high-severity events are statistically rare at any single workplace, organizations are told to record and correct the frequent low-severity events and near misses (reportable hazardous events that did not harm anyone), on the assumption that this also removes the causes of fatalities. The doctrine therefore depends on two logically distinct tenets: (1) an empirical ratio between severity tiers exists (the triangle's shape), and (2) the same underlying causes produce events of every severity, so that acting on low-severity events prevents high-severity ones. Tenet (2), not tenet (1), carries the preventive logic, yet it has rarely been tested directly.

The best establishment-level test of tenet (1) to date is the study of Yorio and Moore (2018), which used two public databases maintained by the U.S. Mine Safety and Health Administration (MSHA): the Mine Address and Employment (AE) database, a yearly census of every U.S. mine with its employment and hours, and the Accident, Injury, and Illness (AII) database, which records every MSHA-reportable incident (fatalities, injuries, and mandated no-injury events such as unplanned ignitions, roof falls, and entrapments) linked to each mine by a fixed identification code. Regressing a mine's fatality indicator in one year on its counts of severity-delineated incidents in the previous year with longitudinal logistic models fitted by generalized estimating equations (GEE, a regression method for repeated observations per unit that adjusts for within-mine correlation), they found that more low- and lower-severity incidents significantly raise the probability of a subsequent-year fatality, with effects up to hundreds of percent per additional permanent disabling injury — but also that the apparent triangle shape depends heavily on how severity is delineated: the MSHA injury-degree categories yield weak evidence, while days-based groupings of lost or restricted workdays recover the classic pattern. The authors were explicit that they tested only the first tenet; whether the causes of lower-severity incidents and of fatalities are shared within mining establishments was left as a question for future studies.

Prior and adjacent work frames what such a test must confront. Analyzing 23,030 serious reportable Dutch occupational accidents (1998–2009) coded into 36 hazard-specific causal models (bow-tie structures centered on one loss-of-control event each), Bellamy (2015) found that direct and underlying causes are shared across severity levels only within the same hazard type — a result implying that causal similarity must be assessed hazard by hazard, not pooled. In the serious-injury-and-fatality (SIF) literature, Cooper (2019) showed that U.K. fatality rates stagnated while minor-injury rates fell by two thirds, an indirect challenge to the common-cause assumption. Companion analyses of the same MSHA data by the same research group (Moore et al. 2020; Yorio et al. 2020) confirmed the predictive association and uncovered a cyclic reporting dynamic — fatalities themselves increase subsequent near-miss reporting — while stating explicitly that no empirical examination of the reasons or mechanisms behind the association was made. Tests of triangle ratios in other domains continue to return mixed verdicts: null in general aviation (Onwuka 2020) and in a four-company mining sample (Rocha et al. 2025), partially supportive in Korean traffic data (Cho et al. 2023), in power-law analyses of Chinese industry data (Zhou et al. 2025), and in nuclear operating-experience data (Šimić et al. 2022). None of these examined causes. The MSHA AII database contains, for each incident, narrative text and coded fields (accident type, activity, occupation, and nature of injury) from which causes and event etiologies can be reconstructed, and mining-specific root-cause analyses exist for fatalities alone (e.g., Quintero et al. 2024). What no study has done is compare the cause distributions of the several severity tiers within the same establishments, on which the entire triangle-based prevention argument in mining rests.

Problem Statement

Using establishment-level U.S. mining surveillance data (the MSHA AE and AII databases, or an equivalent full-population panel with incident-level causal information), determine whether the root causal factors and event etiologies of lower-severity occupational safety incidents — reportable noninjuries (near misses), reportable (no-days-lost) injuries, days-lost injuries, and permanent disabling injuries — are shared with those of fatalities within the same mining establishments. The analysis must (a) reconstruct or code the causes of incidents in each severity tier, with an explicit, validated cause taxonomy and coding procedure; (b) quantify the degree of causal similarity across severity tiers, assessed within hazard categories rather than only pooled across them, in view of Bellamy's (2015) finding that common causes are hazard-specific; (c) state whether, and to what extent, reducing lower-severity incidents can be expected to interrupt the causal chains leading to fatalities at the establishment level, i.e., whether Heinrich's second tenet holds in mining; and (d) demonstrate that the conclusion is robust to the severity-delineation scheme (including days-based thresholds), to the temporal lag used to relate incidents to subsequent fatalities, and to plausible underreporting or misclassification in the surveillance data. All four parts (a)–(d) are required, and the verdict on Heinrich's second tenet must be stated per lower-severity tier (near misses, reportable no-days-lost injuries, days-lost injuries, and permanent disabling injuries, each against fatalities) and per hazard category, with an explicit statement of whether the shared-cause finding holds for every lower-severity tier or only some and of the rule that aggregates the per-tier, per-hazard results into the overall verdict. A finding of 'shared' is a comparative claim about the cross-tier cause distributions, judged against a stated similarity criterion — a metric together with a null or baseline comparison — and hazard categories too sparse to compare fall under that same criterion. The problem is the establishment-level causal question in U.S. mining; analogous tests in other industries, regulatory systems, or international surveillance schemas are a separate open extension and not part of this problem's completion boundary.

References

  1. Patrick L. Yorio and Susan M. Moore (2018), "Examining Factors that Influence the Existence of Heinrich's Safety Triangle Using Site-Specific H&S Data from More than 25,000 Establishments", Risk Analysis 38(4):839–852, doi:10.1111/risa.12869, https://doi.org/10.1111/risa.12869
  2. Susan M. Moore, Patrick L. Yorio, Emily J. Haas, Jennifer L. Bell, and Lee A. Greenawald (2020), "Heinrich Revisited: a New Data-Driven Examination of the Safety Pyramid", Mining, Metallurgy & Exploration 37(6):1857–1863, doi:10.1007/s42461-020-00263-0, https://doi.org/10.1007/s42461-020-00263-0
  3. Patrick L. Yorio, Emily J. Haas, Jennifer L. Bell, Susan M. Moore, and Lee A. Greenawald (2020), "Lagging or leading? Exploring the temporal relationship among lagging indicators in mining establishments 2006–2017", Journal of Safety Research 74:179–185, doi:10.1016/j.jsr.2020.06.018, https://doi.org/10.1016/j.jsr.2020.06.018
  4. Linda J. Bellamy (2015), "Exploring the relationship between major hazard, fatal and non-fatal accidents through outcomes and causes", Safety Science 71:93–103, doi:10.1016/j.ssci.2014.02.009, https://doi.org/10.1016/j.ssci.2014.02.009
  5. M. Dominic Cooper (2019), "The efficacy of industrial safety science constructs for addressing serious injuries & fatalities (SIFs)", Safety Science 120:164–178, doi:10.1016/j.ssci.2019.06.038, https://doi.org/10.1016/j.ssci.2019.06.038
  6. Raoni Rocha, Eugênio Paceli Hatem Diniz, Mário Parreiras, and Ildeberto Almeida (2025), "The illusion of safety indicators in the mining industry", International Journal of Occupational Safety and Ergonomics (published online), doi:10.1080/10803548.2025.2580782, https://doi.org/10.1080/10803548.2025.2580782
  7. C. Onwuka (2020), "A quantitative test of the predictive validity of Heinrich's accident pyramid", African Journal of Health, Safety and Environment 1(1):1–10, doi:10.52417/ajhse.v1i1.18, https://doi.org/10.52417/ajhse.v1i1.18
  8. Sunghwan Cho, Dohyeong Kim, Hiba Khan, and Chang Kil Lee (2023), "Heinrich's Law for Traffic Incidents – Using the Digital Tachograph Data to Identify Traffic Accident Hotspots", Promet – Traffic&Transportation 35(6):829–837, doi:10.7307/ptt.v35i6.207, https://doi.org/10.7307/ptt.v35i6.207
  9. Fang Zhou, Xiling Liu, and Fuxiang Wang (2025), "Size–Frequency Distribution Characteristic of Fatalities Due to Workplace Accidents and Industry Dependency", Mathematics 13(12):2021, doi:10.3390/math13122021, https://doi.org/10.3390/math13122021
  10. Zdenko Šimić, Miguel Peinador Veira, and Reni Banov (2022), "Correlation between events with different safety significance in nuclear power plants", Nuclear Engineering and Technology 54(7):2510–2518, doi:10.1016/j.net.2022.01.034, https://doi.org/10.1016/j.net.2022.01.034
  11. Martin Anderson and Michael Denkl (2010), "The Heinrich Accident Triangle—Too Simplistic A Model for HSE Management in the 21st Century?", SPE International Conference on Health, Safety and Environment in Oil and Gas Exploration and Production, doi:10.2118/126661-MS, https://doi.org/10.2118/126661-MS
  12. Maria Fernanda Quintero, Vladislav Kecojevic, and Deniz Tuncay (2024), "Analysis of the root causes of fatal injuries in the United States surface mines between 2008 and 2021", International Journal of Mining, Reclamation and Environment 38(6):460–479, doi:10.1080/17480930.2024.2308966, https://doi.org/10.1080/17480930.2024.2308966