Hovden, Størseth and Tinmannsvik (2011) proposed criteria for multilevel learning from accident investigations — learning at the individual, organizational, and societal/regulatory levels and between these levels — derived from three Norwegian public transport accident investigations and a pilot expert workshop, and they explicitly flagged that the criteria set was unvalidated and that no data existed on whether following the criteria improves safety. The open problem is to turn this pilot criteria set into validated empirical knowledge: refine and validate the criteria through more comprehensive data and in sectors beyond the original pilot (e.g., offshore, healthcare), operationalize the notion of a 'culture for learning', and establish by longitudinal tracking of objective safety performance (accident, incident, and near-miss rates) whether and which learning criteria are associated with measurable safety improvement rather than merely reported change. No multi-sector longitudinal validation of these criteria, or of their link to objective safety improvement, is known to date.

Background

After a serious accident, transport and industry authorities typically appoint an official investigation whose report analyzes causes and issues recommendations intended to prevent recurrence. The premise is organizational learning: the organization, the regulator, and the wider sector are supposed to change so that similar events become less likely. 'Multilevel learning' as used by Hovden, Størseth and Tinmannsvik (2011) combines two ideas: the level at which learning is supposed to take place — micro (individuals and front-line crews), meso (companies and their regulators), and macro (society, legislation, and the political environment) — and how learning takes place within and between these levels, for instance when an investigation report changes regulation or industry practice rather than only local procedures. The authors derived a set of criteria (indicators of factors that potentially support such multilevel learning) by analyzing three Norwegian public accident investigations, one each in rail, aviation, and maritime transport — the maritime case being the 1999 sinking of the high-speed craft MS Sleipner, officially investigated in NOU 2000:31 (the numbered series of Norwegian official public inquiries) — together with a pilot expert workshop drawing on participants from the three transport sectors. A 'near-miss' is an event that could plausibly have resulted in an accident but did not; accident, incident, and near-miss rates are the standard lagging indicators of safety performance (outcome measures recorded after the fact), and in high-reliability industries — sectors such as rail, aviation, and chemical processing where failures are rare but potentially catastrophic — accident rates are low enough that measurable change requires long observation windows.

The broader literature on learning from incidents had, by the time of the source paper, produced frameworks for how organizations learn from unwanted events (e.g., Lukic, Littlejohn and Margaryan's workplace learning-from-incident framework), and assessments of how well the steps of the learning cycle (investigation, analysis, decision, implementation, follow-up, and sharing) function in practice — Jacobsson, Ek and Akselsson's method for assessing the effectiveness of that cycle, and the Akselsson et al. project, found shallow incident analyses, purely local fixes, late implementation, and weak corrective solutions, with strong correlations between learning and safety-climate variables (safety climate being employees' shared perceptions of how much their organization values and prioritizes safety). Later reviews (Le Coze 2013; Stemn et al. 2018) describe the field as fragmented and repeatedly find that organizations fail to learn effectively from incidents. What this literature does not establish is the link the source paper asks about: whether fulfilling learning criteria, or conducting investigations that satisfy them, is actually followed by measurable reductions in accidents, incidents, or near-misses. The source study itself, being a pilot (one workshop, one case per sector, cases more similar than expected), explicitly did not collect such outcome data, left the meaning and operationalization of a 'culture for learning' unexplored, and noted that some root causes — such as what actually caused the Sleipner navigation error — remained hidden.

Problem Statement

Hovden, Størseth and Tinmannsvik (2011) proposed a set of criteria — indicators of factors that potentially support multilevel learning — for learning from accident investigations, where multilevel learning means learning at the micro level (individuals and front-line crews), the meso level (companies and their regulators), and the macro level (society, legislation, and the political environment), and the transfer of learning within and between these levels. The criteria were derived from three Norwegian public transport accident investigations (one each in rail, aviation, and maritime transport, the maritime case being the 1999 sinking of the high-speed craft MS Sleipner) and a pilot expert workshop with participants from the three transport sectors. The problem is to convert this pilot criteria set into validated empirical knowledge — determining whether, how, and under what conditions multilevel learning from accident investigations produces measurable, lasting improvement in safety performance. This includes refining and validating the criteria through more comprehensive data collection than the pilot and testing their applicability in sectors beyond transport (such as offshore and healthcare), identifying sector-specific differences in the conditions for learning; operationalizing and empirically examining the hitherto unexplored notion of a 'culture for learning' and the role of organizational learning capabilities, which the pilot workshop could not address in depth; and establishing the outcome link the source study left open: by longitudinal tracking of objective safety performance after accident investigations (accident rates, incident and near-miss frequencies), determining whether and which of the learning criteria, and follow-through of investigation recommendations, are actually associated with measurable reductions in accidents and incidents rather than merely reported change. At the core is an empirical test and refinement of the claim that multilevel learning from accident investigations, as specified by the criteria, improves safety. That claim is a single question whose parts are posed jointly rather than as independent sub-questions: criteria refined and validated across sectors, the instruments that operationalize them, a 'culture for learning' operationalized together with organizational learning capabilities, and the outcome link between criteria fulfillment and measured accident, incident and near-miss rates. Because no ready-made multi-sector, multi-year record of accident, incident and near-miss rates after investigations exists for the sectors concerned, that outcome link is to be adjudicated from the best available evidence. A finding of no association between criteria fulfillment and safety improvement is as substantive an outcome of that test as a finding of association.

References

  1. Jan Hovden, Fred Størseth, Ranveig Kviseth Tinmannsvik (2011). Multilevel learning from accidents – Case studies in transport. Safety Science 49(1):98–105. DOI 10.1016/j.ssci.2010.02.023. https://doi.org/10.1016/j.ssci.2010.02.023
  2. Anders Jacobsson, Åsa Ek, Roland Akselsson (2012). Learning from incidents – A method for assessing the effectiveness of the learning cycle. Journal of Loss Prevention in the Process Industries 25(3):561–570. DOI 10.1016/j.jlp.2011.12.013. https://doi.org/10.1016/j.jlp.2011.12.013
  3. Roland Akselsson, Anders Jacobsson, Marcus Bötjesson, Åsa Ek, Ann Enander (2012). Efficient and effective learning for safety from incidents. Work 41(S1):3216–3222. DOI 10.3233/WOR-2012-0661-3216. https://doi.org/10.3233/WOR-2012-0661-3216
  4. Dane Lukic, Allison Littlejohn, Anoush Margaryan (2012). A framework for learning from incidents in the workplace. Safety Science 50(4):950–957. DOI 10.1016/j.ssci.2011.12.032. https://doi.org/10.1016/j.ssci.2011.12.032
  5. Linda Drupsteen, Jop Groeneweg, Gerard I.J.M. Zwetsloot (2013). Critical Steps in Learning From Incidents: Using Learning Potential in the Process From Reporting an Incident to Accident Prevention. International Journal of Occupational Safety and Ergonomics 19(1):63–77. DOI 10.1080/10803548.2013.11076966. https://doi.org/10.1080/10803548.2013.11076966
  6. Jean Christophe Le Coze (2013). What have we learned about learning from accidents? Post-disasters reflections. Safety Science 51(1):441–453. DOI 10.1016/j.ssci.2012.07.007. https://doi.org/10.1016/j.ssci.2012.07.007
  7. Jan-Erik Vinnem (2013). Use of accident precursor event investigations in the understanding of major hazard risk potential in the Norwegian offshore industry. Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability 227(1):66–79. DOI 10.1177/1748006X12468670. https://doi.org/10.1177/1748006X12468670
  8. Linda Drupsteen, Frank W. Guldenmund (2014). What Is Learning? A Review of the Safety Literature to Define Learning from Incidents, Accidents and Disasters. Journal of Contingencies and Crisis Management 22(2):81–96. DOI 10.1111/1468-5973.12039. https://doi.org/10.1111/1468-5973.12039
  9. Linda Drupsteen, Peter Hasle (2014). Why do organizations not learn from incidents? Bottlenecks, causes and conditions for a failure to effectively learn. Accident Analysis & Prevention 72:351–358. DOI 10.1016/j.aap.2014.07.027. https://doi.org/10.1016/j.aap.2014.07.027
  10. Francisco Salguero-Caparros, Manuel Suarez-Cebador, Jesús A. Carrillo-Castrillo, Juan Carlos Rubio-Romero (2018). Quality evaluation of official accident reports conducted by Labour Authorities in Andalusia (Spain). Work 59(1):23–38. DOI 10.3233/WOR-172666. https://doi.org/10.3233/WOR-172666
  11. Eric Stemn, Carmel Bofinger, David Cliff, Maureen E. Hassall (2018). Failure to learn from safety incidents: Status, challenges and opportunities. Safety Science 101:313–325. DOI 10.1016/j.ssci.2017.09.018. https://doi.org/10.1016/j.ssci.2017.09.018
  12. Eric Stemn, Carmel Bofinger, David Cliff, Maureen E. Hassall (2019). Investigating the Maturity of Incident Investigations of the Ghanaian Mining Industry and Its Effect on Safety Performance. Safety 5(1):3. DOI 10.3390/safety5010003. https://doi.org/10.3390/safety5010003
  13. Anastacio Pinto Goncalves Filho, Adonias Magdiel Silva Ferreira, Magna Fernandes Ramos, Anderson Rogério Albuquerque Pontes Pinto (2021). Are we learning from disasters? Examining investigation reports from National government bodies. Safety Science 140:105327. DOI 10.1016/j.ssci.2021.105327. https://doi.org/10.1016/j.ssci.2021.105327
  14. Albert P. C. Chan, Junfeng Guan, Tracy N. Y. Choi, Yang Yang, Guangdong Wu, Edmond Lam (2023). Improving Safety Performance of Construction Workers through Learning from Incidents. International Journal of Environmental Research and Public Health 20(5):4570. DOI 10.3390/ijerph20054570. https://doi.org/10.3390/ijerph20054570