Epidemiology of civilian blast injuries inflicted by terrorist bombings from 1970-2016

Dnyl Mgnus,Mnsoor A.Khn,Willim G.Proud

aInstitute of Shock Physics,Blackett Laboratory,Imperial College London,London SW7 2AZ,UK

bDepartment of Surgery and Trauma,Imperial College London,St Mary's Hospital,London W2 1NY,UK

Keywords:Terrorism Bombings Blast injury Epidemiology Blast mitigation

ABSTRACT An upsurge of terrorist activity has occurred in the past two decades.As part of this,explosive devices continue to be extensively deployed against civilians in wide-ranging environments.Bombings remain the leading worldwide cause of civilian fatalities due to terrorism.This demands an understanding of modern terrorist bombing trends to inform mitigation strategy.The objective of this study was to identify the occurrence and severity of bombings against civilian targets in diverse attack settings,and to establish corresponding blast injury profiles.Data was obtained from analysis of the Global Terrorism Database(GTD)and a meta-analysis of blast injury data derived from the PubMed database.Closed environment explosions were associated with significantly greater(p＜0.05)mortality than in open spaces.The injury profiles were found to be influenced by attack setting,with higher rates of primary injury on trains and buses,and secondary injury in open space.

1.Introduction

Addressing the threat of terrorism has been a long-standing priority of the international agenda,particularly following the events of 9/11.Terrorism remains the subject of intense analysis by a broad range of research disciplines,studying changes in terrorist tactics,recruitment,operational scope,ideology and weaponry[1].Though attacks may vary greatly in their execution,from lone wolf to coordinated incidents,a hallmark of modern terrorism has been psychological warfare achieved through indiscriminate and ostensibly random attacks on the public with the intention of inflicting maximum morbidity and mortality.For this purpose,Improvised Explosive Devices(IEDs)have been utilised globally by terrorists with devastating effect.

A wide range of epidemiological studies exist reporting blast injuries sustained by military populations[2-6],with relatively less focus on the civilian environment.For blast mitigation research,the civilian setting may be of special interest due to victims largely being unequipped with Personal Protective Equipment(PPE),allowing the unmitigated impact to be assessed.The civilian case raises challenges in the nature of the available data.Patients will often experience a varied medical response and transit to trauma centres.In terms of clinical blast injury data,different trauma registries or investigators provide wide-ranging degrees of medical record detail.

The objective of the present study was to establish the incidence of blast attacks on civilian targets and how this has changed in recent years.The effect of the blast environment was examined by determining the severity of these events according to attack setting.Finally,epidemiological data for blast injuries sustained in these events were analysed with respect to attack setting.

2.Methodology

The study was comprised of two parts.The first was to address the statistical occurrence of civilian blast events.This data was obtained from analysis of the 2016 Global Terrorism Database(GTD)[7].The GTD is a database of global terrorist events since 1970 and is managed by the University of Maryland and the National Consortium for the Study of Terrorism and Responses to Terrorism(START).It includes data for 163 countries,covering 99.7%of the global population.For each event,the GTD provides a detailed account of variables including time,location,type of threat,casualty rate and mortality rate.Against civilian targets,the GTD accounts for 23,352 such events that resulted in a total of 78,772 deaths during the period 1970-2016.The parameters of the database search included worldwide incidents of bombing or explosion type where the IED functioned as intended by the perpetrators.To assist with identification of recent changes to attack patterns,two time periods were identified:1970-1999 and 2000-2016.

The second arm of the study concerned the collation of blast injury data according to attack setting.A variety of studies have reported blast injuries sustained as a result of isolated incidents ora series of incidents in a certain region during a short time period(see Table 3).A meta-analysis of these studies was performed to compile casualty figures.In open literature,analogous metaanalyses have been carried out infrequently,with the most recent studies being published in 2004[8]and 2015[9].The eligibility of studies for inclusion in the meta-analysis was assessed through a Preferred Reporting Items for Systematic Reviews and Metaanalyses(PRISMA)search[10].The PRISMA search was performed on results returned by the PubMed database,managed by the United States National Library of Medicine(NLM)[11].The database was referenced according to the keywords “blast”, “injury”and“bombing”, filtering for papers during the period 1970-2016 for consistency with the GTD analysis.The PRISMA search incorporated both primary and secondary exclusion criteria as part of screening and eligibility checks,respectively(see Fig.6).Screening exclusion criteria included studies considering non-civilian events,non-bombing/explosive events and review articles.To remain eligible following screening,studies were required to provide sufficient attack setting information and blast injury observations.The rates of injury reported in these remaining studies were extracted according to Table 1 classification of blast injuries.

This classification represented the blast injuries most commonly reported by the studies meeting the inclusion criteria,spanning primary to quaternary pathophysiology.The granularity of the injury definitions was chosen such that comparisons between studies providing disparate levels of detail could be drawn(e.g.a study reporting “penetrating extremity injuries”and another providing a detailed breakdown according to limb).Table 1 classification therefore represents a compromise between minimising redundant blast injury data and enabling collation of data across multiple studies.

The severity of attacks included in the GTD analysis were delineated by defining two categories of event:an “incident”and a“Mass Casualty Incident”(MCI).Incidents were events resulting in a minimum of one casualty,where a “casualty”was an individual who was either wounded or killed.MCIs were more severe incidents with a minimum of ten casualties.Since a formal number of casualties defining an MCI does not exist,this definition was arbitrary.In general,an MCI is an event where emergency response is overwhelmed by the number and severity of casualties.For the purposes of the present study,a numeric threshold was instructive.

To interrogate the GTD according to attack environment,eachevent was assigned a setting.These settings were defined for consistency with the settings typically identified in retrospective analyses of blast casualties and are shown in Table 2 with example environments as defined by GTD terminology.These settings distinguished between enclosed,partially enclosed and unenclosed spaces,where the physics of blast wave propagation are expected to be distinct.Settings of undetermined environment(e.g.“political party rally”etc.)were excluded.

Table 1 Blast injury classifications for review of epidemiological data.

Table 2 Attack setting definitions.

3.Results and discussion

3.1.Analysis of the global terrorism database

Against civilian targets,bombings were responsible for 38.5%of fatalities inflicted by terrorism,the highest of any attack type.Fig.1 shows time series depicting the evolution of terrorist bombing frequency and Global Terrorism Index(GTI)score.GTI score is a metric of terrorist activity published annually by the Institute for Economics and Peace(IEP)that examines trends in modern terrorism[11].It accounts for the relative impact of terrorism by encoding not only the number of events but also the corresponding numbers of casualties,expressed as

where Miis the number of events and Mkand Mwwere the number of people killed and wounded in those events,respectively.The wi,wkand wware corresponding weights for these parameters.A fourth term accounting for property damages was omitted to concentrate upon human rather than economic cost.The GTI places an emphasis on the impact of fatalities,with weights set as wi=1,wk=3 and ww=0.5.The GTI score shown in Fig.1 was computedover all events worldwide,differing from the figures quoted by GTI reports[12]due to the focus on bombing events only.

Table 3 Studies included in epidemiological meta-analysis.

Fig.1(a)trend of terrorist activity demonstrates a marked rise since the 1970 level.Accounting for casualty statistics to improve the validity of year-to-year comparisons,it is clear that the trend is closely reflected by the evolution of the GTI score in Fig.1(b).The gradual increase of attacks from 1970 to 1992 may be broadly attributed to geopolitical instability during the Cold War and the activities of separatist or extremist groups in Western states such as Spain and Northern Ireland.Following the collapse of the Soviet Union,a general decrease and plateau can be observed up to 2004.It should be noted that this worldwide trend does not always reflect the experience of an individual region.For example,the formation of terrorist groups in former Soviet states resulted in a proliferation of attacks in Eastern Europe.The most distinct feature of these results is the substantial growth of terrorist activity in the aftermath of 9/11.Analogous to the general shift of terrorist motivation away from nationalist or separatist ideologies following the end of the Cold War,9/11 may represent a further paradigm shift of terrorist activity towards motivation by “fundamentalist”aims[13].Subsequent increased and direct Western military involvement in the Middle East,combined with the fact that the region has been beset by multiple and ongoing episodes of civil unrest,continues to contribute to the significant volatility seen since 2004.During the more recent study period of 2000-2016,the majority(60%)of global fatalities inflicted by terrorist bombings have occurred in the Middle Eastern&North Africa region alone(see Fig.2).

The worldwide geographical distribution of terrorist bombings is shown in Fig.3.It can be seen that certain regions,in particular the Middle East&North Africa and South Asia,are consistently areas of significant terrorist activity.The MCI distributions of Fig.3 also illustrate that the relative predominance of terrorist activity in Western Europe is diminished when concentrating upon these more severe events.Overall,a significant growth of terrorism is observed in the more recent time period.Comparing the period 1970-1999 with 2000-2016,average incidents per year increased by a factor of 8.5 while MCIs per year increased by a factor of 6.4.

A breakdown of these incidents and MCIs according to attack environment is shown in Fig.4.The SCS environment was prominent due to the fact that a large number of environments met its broad description.The proportion of MCIs that occurred in enclosed spaces(trains and buses)was greater than overall incidents in both 1970-1999(20.4%compared with 13.0%)and 2000-2016(5.9%compared with 4.3%).A more severe definition of MCI with a greater casualty threshold emphasises this destructive capability of IEDs in enclosed spaces further.The damaging effects of blast are exacerbated in enclosed spaces where the blast wave undergoes multiple reflections,increasing the peak stress,loading duration and inflicting greater morbidity.The number of casualties and injuries also tend to be escalated by the inherently high density of public transport users.These factors have been exploited by terrorists on several occasions in recent history(e.g.Madrid,2004,London,2005 etc.)and render such environments attractive targets[14].

The damaging potential of IEDs in confined spaces is further highlighted by quantifying severity as the mean casualties per event(see Fig.5).The severity of train incidents is apparent,a common feature across both time periods.Over all incidents that occurred in the period 1970-1999(Fig.5(a)),there is a statistically significant(p＜0.05)difference in casualties per incident between at least partially enclosed spaces(train,bus and SCS)and open spaces.Explosions in open spaces are associated with the least morbidity,attributed to rapid attenuation of the blast wave in free space and,potentially,more dispersed targets.During the same period,there is no significant difference(p＞0.05)in the severity of MCIs across different environments(Fig.5(b)).The key change to this picture in the more recent time period is the dramatic increase of train attack severity.This change is statistically significant with respect to all other attack environments.Also notable is the fact that over all incidents,bus and SCS attacks were no longer significantly different in severity from open space attacks in 2000-2016.Such observations could be due to a variety of unknown factors including the proximity of the targets for a given incident,or modern trends in IED design.For example,since the blast wave dissipates quickly in open space,terrorists may attempt to overcome this by incorporating substantial shrapnel with the intention of instead maximising secondary injury[15].

3.2.Blast injury meta-analysis

The initial PubMed literature search returned a total of 197 articles.Of these,18 articles were included in the study for quantitative analysis.This is summarised by Fig.6 PRISMA flow diagram.The exclusion criteria for the screening and eligibility stages are shown in descending order of occurrence in Fig.6.Screening criteria were predicated upon the nature of the document.Almost half(48%)of studies were screened out due to being reviews discussing the theoretical background to blast,based on non-terrorist incidents or being non-civilian focussed.Other commonplace documents that were excluded were those providing guidelines for clinicians or disaster response,detailing psychiatric effects and those that were not based upon a real-world explosive event.

In the eligibility assessment,articles were excluded on the basis of their informational content,only retaining those documents with specification of both the attack settings and blast injuries.Such studies derive their data from retrospective reviews of trauma registry records.A lack of environment information was the most common reason for non-eligibility,though this was not relevant or unknown for the majority of casualties in many studies.Studies examining longer term complications(“other injuries”),such as Heterotopic Ossification(HO)or candidemia,were excluded.It was necessary to eliminate studies reviewing casualty populations from trauma records over a common time period and location,preventing casualty figures for a singular event contributing to the statistics more than once.In such overlapping studies,the study providing data most appropriate for Table 1 classification was included.The included studies are summarised in Table 3.

The objective of the quantitative analysis was to determine the predicted proportion of casualties presenting different injuries as a function of attack setting.Since the overall casualty population was stratified into separate study populations,it was necessary to account for the fact that greater confidence must be attached to statistics from study's providing a larger study population.Stratified sample theory was invoked where the studies were treated as strata.A schematic representation of this is shown in Fig.7.

There were K=18 strata where the kthstratum had a population of NKcasualties.For a given study,the total number of casualties for the event(s)were taken as the sum of the number admitted to hospital and the number of pre-hospital deaths.In some cases,this was an underestimation due to the fact that it was clearly possible for those who were not hospitalised to have sustained injury.It is also challenging to retrospectively quantify the exact number of people affected by a blast in a crowded public space.From each study's total population,NK,a sample of size nKwas taken,from which the distribution of blast injuries was reported.In a majority of cases(89%),the sampling rate was less than 100%,primarily due to a subset of hospital records being unavailable to the investigators.It was assumed that Simple Random Sampling(SRS)applied such that the sample population could be treated as having been randomly drawn from the study's total casualty population[33].With this formulation,unbiased estimators of the proportion of casualties sustaining each injury were computed for each attack setting and are shown in Table 4.

A statistically significant difference of injury rate exists for each injury,demonstrating that the attack environment is an important factor in the resulting injury profile.The highest contrast comparison was seen between the most enclosed environments(trains,buses)and open space.This supports the circumstantial severity indicated by the GTD analysis in Fig.5.Targets in open space were less likely to suffer injuries induced by the blast wave or fireball with significantly fewer casualties presenting BLI,TMP and intra-abdominal rupture.This can be attributed to the rapid decay of the blast wave in open-space,as opposed to the reflections that prolong the loading duration in enclosed spaces.This is especially clear when considering intraabdominal rupture since it is the injury that occurs at the highest pressure threshold [34].It was observed at a rate of approximately 5%in all environments except open space,where the rate was extremely low(～1%).Conversely,casualties in open space were more susceptible to penetrating trauma which may be a result of the uninhibited trajectory of fragments compared to partially enclosed environments.In terms of tertiary injury,a rationale connecting the incidence of fracture with setting is not clear since bodily displacement against rigid objects can occur in any environment.It is possible that the mobile nature of trains and buses(and the subsequent rapid deceleration of the vehicle)is also an important factor here.The significantly higher level of traumatic amputation on trains may be due to the combination of high pressure magnitude and the seated posture of many of the targets,leaving lower limbs in closer proximity to a threat detonated at ground level[23].It is apparent that burns are suffered to a greater extent in enclosed environments compared to open space due to transient confinement of the fireball.Considering the overall distribution of injures,TMP is a prevalent blast injury,owed to the fact that it occurs at a low peak pressure threshold of approximately 100 kPa for any loading durations longer than 10 m s[33].

There are several caveats to this analysis.In comparing injury rates across attack settings,it is implicit that there are no other independent variables.In actuality,the nature and TNT equivalent of the threat are unknown,as well as the proximity of the targets.This is the principal source of uncertainty.As an example,data from Israel strongly contributed the statistics for bus bombings(see Table 3).A hallmark of these attacks in the past decade has been the prevalence of penetrating trauma,with investigators finding that intra-abdominal trauma was predominantly caused by fragment impact rather than the blast wave[15].This is in contrast to earlier incidents,such as the 1989 bus bombing reported by Katz et al.where primary blast was the most significant cause of mortality[20].This was linked by investigators to modern changes in IED design including a shift towards higher mass shrapnel[15].In terms of deployment,it was also found that IEDs were increasingly detonated at chest height rather than under a seat[15],which may be a rationale for the high incidence of penetrating head,neck and face trauma observed on buses.A second caveat to note is that there are inherent differences in how comprehensive and timely the diagnosis and subsequent documentation of blast injuries are for different trauma centres,particularly in the wake of an MCI that may overwhelm services.As such,it was necessary to treat injuries dichotomously(a casualty either had the injury or did not)due to injury severity being unreported in most cases e.g.open/closed fractures,number of fractures,degree of BLI etc.

Table 4 Casualties presenting blast injuries by attack setting.

4.Conclusions

The large casualty population over the studies included in the meta-analysis enable conclusions to be drawn that may assist with informing blast mitigation strategy.Enclosed spaces are associated with enhanced primary blast injury,placing greater emphasis on the need to mitigate the blast wave.In open space,protection from fragment impact must be prioritised.Mitigation material design is fundamentally different in both cases.Blast wave mitigants may incorporate a deformable component for impulse energy absorption.In the case of secondary injury,armour is conventionally designed for defeating ballistic or penetrating threats,typically utilising combinations of hard ceramic plates and aramid fibres.For civilians,structural rather than personal solutions must be sought.The greatest potential for structural design to provide mitigation is in the enclosed environment.In particular,there is a need here for effective primary blast injury mitigation.Though terrorist activity has increased across all attack settings,a key finding of the GTD study was the upsurge of bus and train attack frequency and severity.This is supported by the blast injury meta-analysis,with train incidents resulting in the greatest morbidity.In response to this,O'Neill et al.identified the structural components of trains and rail systems that must be addressed for mitigating the effects of the blast wave[14].Areas such as the walls,doors and roof are of interest for introducing mitigation layers.An analogous need also exists in buildings where retro- fitted cladding may be preferred.Such applications demand cost-effective,practical and lightweight solutions.

Acknowledgements

The authors acknowledge the support of the Institute for Security Science and Technology and The Royal British Legion Centre for Blast Injury Studies at Imperial College London.The Institute of Shock Physics also acknowledges the support of Imperial College London.