We have demonstrated that a national ED monitoring system can usefully identify individual and community risk factors for assault and changes in service pressures with calendar, celebration, and sporting events. Routine analyses of assault data often uses police recorded crime to examine the impact of calendar events on assaults. However, such data can be confounded by both levels of police activity (number of individuals working in any area), policing policy (e.g., which violent events warrant warning and which arrests) and where they take place (e.g., detection of assaults in public vs. private spaces). ED data are not directly impacted by such confounders, provide a measure of health harms relating to nighttime assaults and include events that are not reported to police [11
Using such ED data this study identifies that nights preceding work-free days see more than double levels of assaults presentations (Figure
a, b). Assault levels peak in summer months falling to a low in January (Figure
), when alcohol consumption can also reach its nadir [39
]. Although violence has been linked with warmer weather [40
], a concentration of individuals’ personal holidays in the summer period may also be a contributing factor despite many individuals holidaying abroad [41
]. Constraints that employment places on the length of nights out and alcohol consumption are removed not only by holidays but also by unemployment [42
]. Thus, the most deprived communities showed the highest assault rates and a greater proportion of assaults on Sunday-Thursday nights; consistent with more individuals having no employment pressures midweek. Further, while deprived and affluent males both showed peaks in assaults rates in their late teens, rates reduced more rapidly in the most affluent (Figure
a). Movement into employment in post-adolescence can reduce excessive alcohol use [43
] – although how this impacts on exposure to violence is less well studied. We also identified deprivation-related differences in assault presentation at early ages. Critically, by age 15
years males in the most deprived quintile had exceeded the peak presentation level achieved in the most affluent quintile at age 19
a). Worse, for females assault presentations in the affluent quintile peaked at age 20
years and rates in the most deprived quintile exceeded this peak by age 13
Globally, social inequality, poverty and youth unemployment have been associated with increased violence [44
] and even rioting in some countries, including the UK [45
]. Internationally, attention has focused on both immediate policing measures to prevent further violence and, increasingly, the need for longer-term multidisciplinary life course approaches to improve young people’s prospects and reduce their overall propensity for violence [1
]. On the former, some local ED data systems have already been used to record assault location and inform the targeting of police activity [11
] often in nightlife areas. On the latter however, potential roles for ED data remain largely underdeveloped. Results presented here identify a much earlier escalation in violence in the poorest communities and a peak at a much higher level (Figure
a,b). Early life exposure to violence represents a direct risk to children’s immediate and long-term physical and mental health [48
]; in some circumstances resulting in permanent disability. Moreover, such exposure also leaves individuals more likely to engage in violence later in adolescence and adulthood [48
]. Early life exposure to assaults can be reduced through parental support, pre-school enrichment, and social development programmes [16
]. Several of these programmes, such as Nurse Family Partnerships [16
], have already begun to be scaled up in a number of countries including Canada [51
], Australia [52
] and England [53
]. Here we have identified how a national ED data system can provide a benchmark; identifying areas most in need of such interventions.
As well as a role in targeted long-term prevention, we have shown that a national ED data system identifies peaks and troughs in violence that are strongly associated with events such as celebrations and sporting events. In England Halloween, Guy Fawkes Night and St Patrick’s Day are now heavily commercialised events with themed alcohol promotions, organised public events (such as club nights and bar crawls) and private parties. All three were associated with significant increases in assaults (Figure
). However, Valentine’s Day and St George’s Day showed no significant increases. Sporting events also varied in impact on assault presentations. Presentations increased dramatically on nights when the national team played in the World Cup but not with other football or rugby fixtures (Figure
). The association between sport and public violence has been examined elsewhere [54
]. However, this study identifies how ED data can measure the impacts of violence beyond that typically observed around city centres and gatherings such as sporting events. Thus, broadcast access to the Beijing Olympics was associated with a small but significant reduction in overall assault presentations in England (Table
). While understanding such patterns exposes expected pressures on ED departments, they are also pertinent to other frontline services such as ambulance and police. Currently, there is little information on how well emergency staffing levels are attuned to demand and national ED data, with local intelligence, could help inform the efficient distribution of staff and other resources on a calendar basis.
Our analysis only examined public holidays, national celebrations, and some major national and international sporting events. In planning holidays and events nationally more thought should be given to how selection of specific times, days, and months could be used to minimise any resultant increase in violence. Moreover, health and other agencies should consider such intelligence when timing campaigns to reduce binge drinking and related violence, stipulating license requirements, and enforcing critical legislation (for example, no sales of alcohol to those underage or already drunk).
The ability of ED data to provide intelligence on nighttime assaults relies on individuals reporting violence as the reason for their presentation. Such reporting may be affected by issues of confidentiality. More work is required on protecting confidentiality by establishing optimal levels of data access for different organisations and at different geographical levels [56
]. While some local ED systems in England collect and share information on assault location, few share information on residence; despite this being routinely collected in the ED services. This combination of data is urgently needed to understand trends in and relationships between public (e.g., city centre) and private space (e.g., homes) violence. Together, these data would enable an effective multiagency response both nationally and locally. However, even the experimental data utilised here exposes some important gaps in our understanding of nighttime violence. Thus, some events are violence promoters (such as St Patrick’s Day and England games in the football World Cup), while others are nonbelligerent (such as St George’s Day and the Rugby Six Nations) or perhaps even protective (for example, the Olympics). The relative impact of different holidays and events may vary with locality and nation. For example in Cardiff, Wales (where rugby is often considered to be the national sport), international rugby matches involving the Welsh team have been associated with increased ED assault attendances [55
]. Research is needed to understand the factors protecting relatively peaceful celebrations, and the roles commercialisation and linkage with alcohol promotions play in coupling celebrations with violence.
This study has a number of important limitations. English national HES A&E data are still incomplete. Although a full audit of data quality is not available, comparison with the QMAE suggests HES A&E represent 74% of all presentations regardless of cause [28
]. HES A&E though covers all major EDs with much missing data arising from other emergency service providers such as walk in centres, which only accounted for 1.2% of nighttime assault presentations in this dataset. However, although emergency services are the principal resource for urgent assault treatment at night not all individuals assaulted, or even requiring treatment will present to them. England also has a general practice on call system where doctors can be asked to attend individuals’ places of residence. Further, injured individuals may also attempt to treat themselves or to delay treatment until the next day when there is a greater range of treatment options and their attendance time may fall outside of this study’s inclusion criteria. The study cannot quantify how frequently such options are exploited by those injured in assaults; although they are unlikely to be options for those requiring immediate attention.
In ED data, reason for attendance was coded as unknown in 4.7% of cases but data coding relies on patients revealing that their injuries have been sustained through violence and this being accurately coded in busy EDs. Although the absence of any violence related presentations from nine providers suggests under-recording of assaults, currently it is not possible to quantify the scale of such miscoding across all EDs. These issues will inevitably affect any calculation of rates. However, our findings focus on comparative risks; largely between different days or different demographics. We are not aware of any calendar, deprivation, or age/sex related bias in missing data that could confound our results, although this cannot be entirely discounted. Our focus has been on levels of emergency presentations for assault and therefore we have excluded ED attendances for follow ups relating to a previous ED attendance. We have not attempted to remove multiple presentations by the same individual for different assaults (see methods). Consequently, demographic analyses relate to probabilities of presentations being from a particular demographic. However, across the three-year study period only 6.7% of individuals presented for nighttime assault more than once and analysis of individuals, rather than presentations, would be unlikely to substantively affect results.
Sporting events included were a convenience selection based on those best known and highly promoted. There are a wide range of other local events that might have been included in this analysis and the impact of even national events (such as a football cup final) may vary with locality; if, for instance, a local team are involved. The analyses undertaken should be considered a proof of concept for the utility of ED data, which could be implemented much more widely with a complete national data set. We could not distinguish assault locations (e.g., home or city centre bar), and thus we have made no assumptions about whether assaults took place in public places or private residences. While the national ED system does not currently collect location of assault, the collection and sharing of such data at local level is increasing [11
]. Finally, while this study has examined the utility of a national ED dataset in measuring calendar and demographic risk factors for nighttime assaults further analyses are now possible. ED data allows additional exploration of the residence of those involved in violence (e.g., by population density, urban vs. rural locality, etc.). Data on alcohol consumption by those presenting to EDs is not currently available nationally but, routinely collected even from a subset of EDs, could provide important intelligence on the impact of alcohol on nightlife assaults [11