Marked Decline of Sudden Mass Fatality Events in New Zealand 1900 to 2015: the Basic Epidemiology

INJURY AND HARM

Marked decline of sudden mass fatality events in New Zealand 1900 to 2015: the basic epidemiology

Nick Wilson,1 Adrienne Morales, 1 Nicola Guy, 1 George Thomson1

ass casualty events are important Abstract internationally, with one report Midentifying 749,000 earthquake Objective: To describe the basic epidemiology of sudden mass fatality events for the 1900 to deaths in the past 20 years, more than 2015 period in New Zealand (NZ). 160,000 heatwave deaths and more than Methods: Official lists and internet searches were used to identify the events. Events were 1 130,000 deaths from one storm. The number categorised, rates calculated and time trends analysed. of weather- and climate-related disasters has Results: A total of 56 sudden mass fatality events with 10 or more fatalities between 1900 and more than doubled over the past 40 years, 2015 in NZ were identified. There were 1,896 deaths in total, with the worst event being the accounting for 6,392 events in the 20-years Hawke’s Bay earthquake (258 deaths). Events were classified as transportation-related (64%), 1996–2015, up from 3,017 in 1976–1995.1 natural causes (11%), industrial (9%), war (9%) and infrastructure (5%). There were marked Even though developing countries declines in the rate of events per person-years of exposure and the associated mortality rate experience relatively high mortality burdens (both p<0.0001). Knowledge gaps were identified around: i) the basic epidemiology, e.g. non- from such disasters, no country is immune. fatal injuries and numbers of survivors; ii) the role of subsequent official inquiries in guiding In New Zealand, for example, the Canterbury preventive measures; and iii) the likely cost-effectiveness of measures to prevent harm from earthquake caused 185 deaths, and 6,659 such events. people were injured in the first 24 hours.2 The Conclusions: The occurrence and health burden of sudden mass fatality events have markedly NZ Treasury has estimated the cost of this declined in NZ over time. earthquake to be around NZ$40 billion for capital costs alone.3 In the aftermath of such Implications for public health: There remains large scope for addressing the knowledge gaps events, large resources are also typically spent in the basic epidemiology and societal responses to these events to guide primary prevention on official inquiries, e.g. the inquiry into the and appropriate disaster response. Canterbury earthquake produced seven very Key words: mass fatality events, disasters, official inquiries detailed volumes.4 These inquiries may also lead to new practices and new laws – which Methods spread out over weeks, months or years (e.g. also can involve large levels of resource use, the 1918–1919 influenza pandemic and the such as with new building regulations for Event definition 1957–1959 influenza pandemic).7 earthquake resistance. We defined a sudden mass fatality event as Despite such impacts, we could not identify one that involved 10 or more fatalities, as per Identifying events any basic descriptive epidemiology of such one component of a previously published Most events were primarily sourced from mass fatality events for NZ. Indeed, official framework for “mass casualty incidents”.6 The an official list of disasters8 but with various lists of ‘disasters’ in NZ were not systematically ‘sudden’ component of the event required additions and subtractions based on collated and appear incomplete in various that most of the deaths occurred within a 24- additional searches. Most additions arose ways. An Australasian study provided hour period. The location – NZ’s boundaries from a detailed text on NZ shipwrecks,9 but descriptive epidemiology that included NZ in 2016 – included the country’s Exclusive others came from searches of online lists, e.g. events, but that list was also incomplete Economic Zone (EEZ; 200 nautical miles out to for air transport crashes and mass homicide and there were no NZ-specific analysis sea) and its territorial claim on Antarctica (the events. Subtractions relative to the official and trends.5 To facilitate further analysis of Ross Dependency). The time period used was list arose from definitional issues e.g., the this topic, we aimed to describe the basic 1900 to 2015. We excluded epidemics and 1988 Whanganui air crash in an official list8 epidemiology of sudden mass fatality events pandemics, given that these did not meet our appears to have involved nine rather than (10 or more deaths) in NZ for 1900 to 2015. criteria for being sudden as they are typically the 10 fatalities assumed. Events outside NZ’s

1. Department of Public Health, University of Otago, New Zealand Correspondence to: Professor Nick Wilson, Department of Public Health, University of Otago, PO Box 7343, Wellington South, Wellington 6021, New Zealand; e-mail: [email protected] Submitted: June 2016; Revision requested: October 2016; Accepted: November 2016 The authors have stated they have no conflict of interest. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. Aust NZ J Public Health. 2017; 41:275-9; doi: 10.1111/1753-6405.12652

EEZ were not included, given our definition categories. Some of the early inquiries into events, a flood and a major storm. Following of an event (e.g. various ship sinkings, such as ship sinkings were held in Australia and we this were industrial causes (9%), with all four the Turakina and Rangitane in 1940, and air included these. related to coal mining; those related to war crashes of two NZ Air Force bombers on route (9%); and those related to infrastructure (5%), from Fiji to NZ in 1944). Denominator data and rates i.e. two fires and a platform collapse. However, After this initial search work, a comparison Population denominator data was sourced multiple causal pathways were often involved was made with the Australasian study by from Statistics NZ for mean population and such nuances are not captured in Table 1. Bradt et al.5 that covered the period 1900 estimates by year.13 Where these data were For example, some of the transport-related to 2012 (and included 38 events with missing for specific years (in those prior to events such as ship sinkings were also 10+ fatalities). This comparison resulted 1926), we interpolated from the five-yearly sometimes related to storms and so could be in no changes in the number of events Census values (assuming linear population partly related to natural events. Also, some but stimulated further research by us to growth). Rates for events and deaths were transport-related shipping events could also identify additional deaths for one event calculated using millions of person-years be considered due to industrial activity (e.g. (the delayed deaths from a 1943 Liberator exposure for the whole NZ population ships carrying coal or used for industrial-scale air crash). Remaining differences with the (the issues around other denominators are fishing). Furthermore, while we classified work by Bradt et al. appear to result from detailed in the Discussion Section). the Ongarue train crash and the Tangiwai various methodological differences, e.g. their train crash as primarily being transportation- related, natural events were also implicated (a inclusion of other types of disasters with <10 Results fatalities, events outside of NZ territory, and storm-related landslide and a volcanic-related what seem to have been 18 events they did Types of events lahar (flow of mud or debris), respectively). not include (mainly less well-known ship The classification of five events as primarily A total of 56 events were identified and sinkings in the early 1900s). ‘war-related’ was also complex. These events summary details on each are in the were unusual in their characteristics, e.g. Supplementary Table (available online). The enemy mine explosions in NZ waters in World Data on cause, fatalities and inquiries results of our classification of the major event War I and a riot in a prisoner-of-war camp. These data were sourced from official cause are shown in Table 1 and this suggests Three of the five related to United States sources: either Te Ara (the Encyclopaedia of that most (64%) of events involved forms 10 military aircraft or naval vessels that were in NZ), the online official site New Zealand of transportation: especially ships (n=26), 11 NZ during World War II. History, or the official inquiry into the event. then aircraft (n=6) and trains (n=3). Next in NZ has many types of official inquiries,12 frequency were events classified as having but for simplicity we took the approach of predominantly natural causes (11%), which Mortality burden and trends not distinguishing between the different included three earthquakes, two volcanic The total mortality burden was 1,896 deaths with the most severe single event being the Table 1: Primary causes of the sudden mass fatality events identified (see text for multiple causal pathways Hawke’s Bay earthquake in 1931 (258 deaths). sometimes being involved). There was an average of 16.3 deaths per year Time period Transport Natural Industrial War Infrastructure Human Total over the period studied. Natural events had agency* the highest average mortality burden per 1900-1919 18 0 2 1 0 0 21 event (Table 2), but the standard deviations 1920-1939 4 4 1 0 0 0 9 were large. Earthquakes were the most lethal 1940-1959 7 0 0 4 2 0 13 natural event and there were three of these 1960-1979 5 1 1 0 0 0 7 with 460 deaths in total (4.0 deaths per year 1980-1999 1 0 0 0 1 1 3 on average). 2000-2015 1 1 1 0 0 0 3 The rate of all sudden mass fatality events Total 36 6 5 5 3 1 56 declined significantly over the period Row% for total 64.3% 10.7% 8.9% 8.9% 5.4% 1.8% 100.0% * There are limitations with this term “human agency” to cover a mass shooting event by a person with a possible mental illness (but no better classification (p<0.0000001, see Table 3). This was also could readily be identified). Purposeful human agency was also a factor in some of the war-related events (mining of offshore waters and in the prisoner- the pattern when war-related events were of-war camp riot). For an example of a different classification system see Bradt et al 20155 which used the following categories: climatological, geophysical, excluded (Figure 1, Table 3). The same pattern meteorological, hydrological, biological, transport, industrial, residential, and terror/mass shooting. of decline was also apparent for the decline Table 2: Mortality burden from sudden mass fatality events by time period in New Zealand (1900 to 2015, ordered in the mortality rates from these events by descending mean deaths per event). (Table 3). The decline in transport-related Type of event Events Deaths per event Median deaths per event Total deaths events (especially ship sinkings) was a major (N) Mean SD (N) (N) driver of this trend but, even when these Natural 6 91.0 104.7 37 546 were excluded (along with excluding the war Infrastructure 3 30.7 14.6 37 92 events), there was still a significant decline in Transport 36 28.3 46.6 15 1020 the remaining events (p=0.012). Industrial 5 22.4 13.8 19 112 War 5 22.4 16.2 16 112 Survivors and non-fatal injuries Human agency 1 14.0 - 14 14 There were no survivors for 44.6% of the All events 56 33.9 53.2 16 1896 events (25/56). These completely fatal

events were mainly from ship sinkings and FigureFigure 1: 1: Rate Rate of suddenof sudden mass mass fatality fatality events events in New in ZealandNew Zealand by time-period by time-period (excluding with five superimposed war-related trendevents, then aircraft crashes. But, where there were line (excluding five war-related events, 281 million person-years of exposure in the NZ population). 281 million person-years of exposure in the NZ population). survivors, identifying the precise numbers 12 was found to be too problematic. For example, for events such as earthquakes and 10 large storms, there are complex definitional issues given large proportions of the whole NZ population affected by the event. Even for 8 train crashes, identifying the total numbers of Event rate per 10 million 6 people on the train was sometimes difficult person‐years exposure (but possibly could be estimated from further in-depth work). Similarly, no further analyses 4 were performed for non-fatal injuries, given that it became apparent that such data were 2 generally of low quality and sometimes not ‐ readily available – especially for events in the 1900‐1919 1920‐1939 1940‐1959 1960‐1979 1980‐1999 2000‐2015 first decades of the study period. Even for hospitalised non-fatal injuries, substantial research of media reports and official time, as has the mortality rate from such with heat waves) could become relevant to inquiries would probably be required to events. This could reflect a large number of NZ as well. But preventive measures could produce reasonable estimates. factors, such as improvements in transport act to counter such patterns, e.g. heat wave vehicle design, weather forecasting, marine deaths are partially preventable with building Official inquiries and aircraft navigation systems, and safety design and features of urban design. When considering the non-war-related systems in general. Indeed, the decline Although 1,896 deaths from these events is events, official inquiries were held after 78% observed would probably be even greater of note, there is a need to put these in a wider (40/51) of these events. This rose to 100% for for transport-related events overall if we had context of health loss for NZ. For example, if non-war events after 1936. Two of the last used a denominator of kilometres travelled the deaths from sudden mass events involved events without inquiries were the Murchison by the population for each transport mode an average loss of 40 years of life per victim, earthquake in 1929 and the cyclone in 1936. (which we did not have the data for). Yet in then this would be 75,800 years-of-life lost in The latter had the unusual characteristic of contrast to air travel, the marked decline in total. However, this is only 8% of the 955,000 wide spatial distribution, with the 12 deaths passenger ship travel and rail travel in the disability-adjusted life-years lost in the NZ occurring in 10 locations in the North Island. past half-century might partly explain the population from multiple conditions in the relative lack of such events in this period. In single year of 2006.14 Perhaps this suggests some cases new laws may also be preventing Discussion that planning for preventing and managing certain types of events, e.g. mass-shooting mass fatality events should still be a relatively Main findings and interpretation with semi-automatic weapons (with these low priority when compared to addressing weapons made illegal after the 1990 This analysis identified 56 sudden mass other more major causes of health loss in NZ Aramoana mass shooting). fatality events that occurred in NZ over (e.g. obesity and tobacco, etc). the 1900 to 2015 time period. These were The decline in sudden mass fatality events for The relatively high proportion of NZ events highly diverse events in terms of their cause NZ contrasts to that for the country’s closest followed by official inquiries appears and impact, e.g. from a bus crash to major neighbour Australia, where such events have impressive at first glance – but there are some 5 earthquakes (in Hawke’s Bay and Canterbury). been increasing, due to bushfires and floods. qualitative limitations with these inquiries. However, the rate of such events (per millions Indeed, if future climate change continues Some inquiries could not identify the cause of of person-years exposure) has declined over to be rapid then these type of events (along the event (e.g. the reason for the Holmglen to sink) and some inquiries were subsequently Table 3: Sudden mass fatality events by time period in New Zealand (1900 to 2015). shown to reach incorrect conclusions, such as Person-years of All events Events excluding war-related events the sinking of the SS Elingamite (when islands Period NZ population Events Event rate Deaths Death rate Events Event rate Deaths Death rate were later found to be incorrectly located exposed (the (N) (per 10 million (N) (per million (N) (per 10 million (N) (per million denominator) person-years) person-years) person-years) person-years) on maps).15 Other inquiries were followed 1900-1919 20,254,749 21 10.4 437 21.6 20 9.87 411 20.3 by more detailed subsequent investigations 1920-1939 29,259,880 9 3.08 373 12.7 9 3.08 373 12.7 and also legal cases – with the causes being 1940-1959 38,329,400 13 3.39 411 10.7 9 2.35 325 8.5 contested (e.g. the Erebus crash inquiries). 1960-1979 56,267,900 7 1.24 412 7.3 7 1.24 412 7.3 1980-1999 68,887,200 3 0.44 38 0.6 3 0.44 38 0.6 Study strengths and limitations 2000-2015 67,638,300 3 0.44 225 3.3 3 0.44 225 3.3 This study seems to be the most systematic Total/(overall) 280,637,429 56 (2.00) 1,896 (6.8) 51 (1.82) 1,784 (6.4) collation and analysis of these mass fatality p-value for trend* <0.0000001 <0.0000001 <0.0000001 <0.0000001 events for NZ to date. It identified 18 * Extended Mantel-Haenszel chi-square for linear trend calculated using OpenEpi (Version 3.03a). additional disasters with 10+ fatalities for

NZ than the study by Bradt et al.,5 and more the most recent decades by collating: non- not yet obtained an appropriately detailed than four times as many as those listed in an fatal hospitalisations, survivors of locality- large-scale picture of mass fatality events. international disaster database (the CRED focused events, and the role of emergency 16 database), which only had 13 such non- services in each event (numbers rescued, Conclusions epidemic mass fatality events of 10+ deaths etc). We note one event where seven listed for the 1900 to 2015 period. Relative rescuers also died (in 1939 at the Glen This study identified and classified a total to our work, the latter also had examples Afton mine). For more recent decades, of 56 sudden mass fatality events with 10 of: double counting of events, lower fatality in-depth research might be able to identify or more fatalities occurring between 1900 numbers (n=3 events), and what appear to be such characteristics of the victims as: age, and 2015 in NZ. There was a marked overall wrong years in the dataset (n=2). gender, ethnicity and socioeconomic decline in the rate of events per person-years There are potential limitations however in status (based on area-based deprivation of exposure and the associated mortality rate. terms of how the events are classified (given measures used routinely in NZ). The use of But various remaining knowledge gaps were multiple causal pathways) and some events a range of methods to determine disaster- identified around: i) the basic epidemiology 19 represent what might be regarded as related attributed mortality (e.g. Combs et al.) such as non-fatal injuries and numbers of to highly anomalous situations, e.g. those might help in some cases. survivors; ii) the role of subsequent official related to two World Wars. Future work could 2. What has been the overall value of official inquiries in guiding preventive measures; and improve the classification system and also inquiries into these mass fatality events iii) the likely cost-effectiveness of measures to identify events that may be climate change for advancing preventive interventions? prevent harm from such events. related (e.g. floods and severe storms). The At first glance it seems to us to be quite total fatality counts may also slightly under- high, given regulations and laws passed References estimate the true burden of some events. after some of them. For example, Bradt et For example, the true number of deaths al.5 helpfully detail legislative responses 1. The Centre for Research on the Epidemiology of Disasters. Poverty and Death: Disaster Mortality 1996- arising from the Hawke’s Bay earthquake may to the Seacliff fire, the Ballantyne’s fire, 2015 [Internet] Brussels (BEL): Université Catholique well have been slightly higher for various the Aramoana mass shooting, the Cave de Louvain School of Public Health; 2016 [cited 2017 Feb 2]. Available from: http://cred.be/sites/default/files/ 17,18 reasons. Indeed, if events such as this Creek platform collapse, the Pike River CRED_Disaster_Mortality.pdf caused non-fatal injuries that resulted in mine explosion and both the Hawke’s Bay 2. Ardagh MW, Richardson SK, Robinson V, Than M, Gee P, Henderson S, et al. The initial health-system response deaths months or even years later, these are and Canterbury earthquakes. Inquiries to the earthquake in Christchurch, New Zealand, in unlikely to have been included in official may also potentially inform the best use February, 2011. Lancet. 2012;379:2109-15. fatality totals. of resources in managing mass casualty 3. N Z Treasury. Rebuilding Christchurch, Our Second- biggest City (Budget Policy Statement 2014) [Internet]. As discussed above, the denominator used events when they occur, given a systematic Wellington (NZ): Government of New Zealand; 2013 review suggesting that there is still much [cited 2017 Feb 2]. Available from: http://www.treasury. in the rate analysis could be improved upon govt.nz/budget/2014/bps/06.htm 20 for transport events and potentially also for to learn. Finally, if official inquiries tend 4. The Canterbury Earthquakes Royal Commission. The industrial events (which could use estimates to effectively advance preventive effects, Canterbury Earthquakes Royal Commission [Internet]. Wellington (NZ): New Zealand Department of Internal of the size of the industrial workforce). then perhaps they should be used more for Affairs; 2012 [cited 2017 Feb 2]. Available from: http:// With extensive additional work it may be the much larger causes of health loss in NZ canterbury.royalcommission.govt.nz/ (e.g. the obesity and tobacco epidemics). 5. Bradt DA, Bartley B, Hibble BA, Varshney K. Australasian possible to at least determine non-fatal disasters of national significance: An epidemiological injuries resulting in hospitalisations for 3. Given the best estimates of health loss analysis, 1900-2012. Emerg Med Australas. 2015;27:132- 8. these events. But this exercise may still have from mass fatality events and availability of 6. Kim CH, Park JO, Park CB, Kim SC, Kim SJ, Hong KJ. limited utility, given that the threshold for preventive interventions, what then is the Scientific framework for research on disaster and mass hospital admission changes over time, and likely cost-effectiveness of such preventive casualty incident in Korea: Building consensus using Delphi method. J Korean Med Sci. 2014;29:122-8. some events (such as major earthquakes) measures? For example, the Government’s 7. Viboud C, Simonsen L, Fuentes R, Flores J, Miller MA, damaged local hospital capacity. For tightly passing of a law to limit access to semi- Chowell G. Global Mortality Impact of the 1957-1959 Influenza Pandemic. J Infect Dis. 2016;213:738-45. defined events that exclude events such automatic weapons after the Aramoana 8. Ministry for Culture and Heritage History Group. New as earthquakes and widespread storms, it mass shooting may well have been very Zealand Disasters Timeline [Internet]. Wellington (NZ): Government of New Zealand; 2015 [cited 2017 Feb 2]. might also be possible to meaningfully define cost-effective, since passing a law in NZ is Available from: http://www.nzhistory.net.nz/culture/ numbers of survivors. This could potentially relatively inexpensive.21 But what about new-zealand-disasters/timeline help define the populations for which post- building standards to reduce earthquake- 9. Ingram C, Wheatley P. New Zealand Shipwrecks: Over 200 Years of Disasters at Sea. 8th ed (revised and updated traumatic stress disorder may have been a related injuries? Measures to reduce by L Diggle, E Diggle and E Gordon). Auckland (NZ): plausible risk. masonry falling into the street might well Hodder Moa; 2007. 10. Ministry for Culture and Heritage. Te Ara – The be cost-effective, but perhaps not so for Encyclopedia of New Zealand [Internet]. Wellington (NZ): very expensive internal modifications of Government of New Zealand; 2015 [cited 2017 Feb 2]. Possible priority additions to the Available from: http://www.teara.govt.nz/en future research agenda building structures? 11. Ministry for Culture and Heritage History Group. New A modest investment in such a research Zealand History [Internet]. Wellington (NZ): Government This analysis has suggested to us the of New Zealand; 2015 [cited 2017 Feb 2]. Available from: agenda could help guide future preventive following areas that could potentially benefit http://www.nzhistory.net.nz/ efforts and optimise the use of societal 12. Simpson A. Commissions of Inquiry – Functions, Power from additional work: and Legal Status. In: Te Ara – The Encyclopedia of New resources spent on prevention. While modern Zealand [Internet]. Wellington (NZ): Government of 1. The basic descriptive epidemiology of NZ society seems very capable of resourcing New Zealand; 2015 [cited 2017 Feb 2]. Available from: these events could be improved at least for http://www.TeAra.govt.nz/en/commissions-of-inquiry/ detailed inquiries into specific events, it has page-1

13. Statistics New Zealand. Historical Population Estimates 18. Clement C. Hawke’s Bay Earthquake, Tuesday 03 Tables [Internet]. Wellington (NZ): Government of February 1931, Napier, Hawke’s Bay, New Zealand. In: Supporting Information New Zealand; 2015 [cited 2017 Feb 2]. Available from: New Zealand Disasters and Tragedies On Line Memorial http://www.stats.govt.nz/~/media/Statistics/browse- [internet]. Lehi (UT): Rootsweb.com; 2015 [cited 2017 Additional supporting information may be categories/population/estimates-projections/national- Feb 2]. Available from: http://freepages.genealogy. found in the online version of this article: pop-estimates/historical-pop-estimates.xls rootsweb.ancestry.com/~sooty/napier.html 14. N Z Ministry of Health. Health Loss in New Zealand: A 19. Combs DL, Quenemoen LE, Parrish RG, Davis JH. Supplementary Table 1: Full list of the 56 Report from the New Zealand Burden of Diseases, Injuries Assessing disaster-attributed mortality: Development sudden mass fatality events with 10 or more and Risk Factors Study, 2006–2016. Wellington (NZ): and application of a definition and classification matrix. Government of New Zealand; 2013. Int J Epidemiol. 1999;28:1124-9. fatalities occurring for the period 1900 to 15. Te Ara. Elingamite. In: Te Ara – The Encyclopaedia of New 20. Timbie JW, Ringel JS, Fox DS, Pillemer F, Waxman DA, 2015 and occurring in New Zealand territory Zealand [Internet]. Wellington (NZ): Government of Moore M, et al. Systematic review of strategies to New Zealand; 2015 [cited 2017 Feb 2]. Available from: manage and allocate scarce resources during mass (ordered by year of event). http://www.TeAra.govt.nz/en/1966/disasters-and- casualty events. Ann Emerg Med. 2013;61:677-89 e101. mishaps-shipwrecks/page-7 21. Wilson N, Nghiem N, Foster R, Cobiac L, Blakely T. 16. Centre for Research on the Epidemiology of Disasters. Estimating the cost of new public health legislation. Emergency Events Database (EM-DAT) [Internet]. Bull World Health Organ. 2012;90:532-9. Brussels (BEL): Université Catholique de Louvain School of Public Health; 2015 [cited 2016 Oct 29]. Available from: http://www.emdat.be/database 17. McSaveney E. Historic Earthquakes – The 1931 Hawke’s Bay Earthquake. In: Te Ara – The Encyclopedia of New Zealand [Internet]. Wellington (NZ): Government of New Zealand; 2016 [cited 2017 Feb 2]. Available from: http://www.TeAra.govt.nz/en/historic-earthquakes/ page-6