Violence: A Public Health Issue in the United States
Community violence has long been recognized as a major public health burden both nationally and abroad (Cooley-Strickland et al., 2009 ; Rosenfeld, 2016 ; World Health Organization, 2014). Most victims of violence survive the initial assault but face a 44% reinjury rate and increased odds of subsequent violent death within 5 years (20%), underscoring the need and opportunity to intervene in the cycle of violence that costs society as much as $70 billion annually (Corso, Mercy, Simon, Finkelstein, & Miller, 2007 ; Haider et al., 2014 ; Morrissey, Byrd, & Deitch, 1991 ; Sims et al., 1989). Community violence is a challenging public health problem that unique disciplines (e.g., health care, law enforcement, public health) cannot adequately address when operating in silos. Comprehensive and collaborative population health solutions are required to address this public health burden.
Incomplete Data Are Used to Guide Violence-Related Decisions
Communities often rely solely on law enforcement data to understand the nature and scope of community violence. In the United Kingdom, studies found that law enforcement data alone are often an incomplete record of all community violence that occurs and that there may be as little as 11% overlap between law enforcement data and hospital data as they relate to community violence (Sutherland, Sivarajasingam, & Shepherd, 2002). This may not be unique to the United Kingdom, as evidenced by one major U.S. city where as many as 13% of the nonfatal shootings that were medically documented in emergency departments (EDs) did not appear in citywide law enforcement records (Kellermann, Bartolomeos, Fuqua-Whitley, Sampson, & Parramore, 2001). De-identified hospital ED assault data sources are situated as useful complements to law enforcement data collection systems that could enhance community violence data as they are open 24 hr daily, are ubiquitous in distribution, and collect relevant data from each patient (Barthell et al., 2002).
The Cardiff Model
The Cardiff Model enhances the overall understanding of community violence by combining de-identified ED-based data regarding assault-related injuries with other data sources. Combined data are then shared with public health stakeholders including law enforcement and community organizations to develop more fully informed violence intervention strategies. To date, the Cardiff Model has been externally validated (Boyle, Snelling, White, Ariel, & Ashelford, 2013) including in a systematic review of evidence of effectiveness (Droste, Miller, & Baker, 2014), has reduced violence-related hospitalizations in Cardiff by 42% relative to control cities (Florence, Shepherd, Brennan, & Simon, 2011), and is cost beneficial (Florence, Shepherd, Brennan, & Simon, 2014). This violence prevention model has been widely implemented in the United Kingdom (Analytical Unit, Social Care, Local Government and Care Partnerships Directorate, Department of Health and Social Care, 2015) and is endorsed by the World Health Organization (Butchart, Phinney, Check, & Villaveces, 2004). Data collected from patients presenting with an assault-related injury include the precise location of the event, weapon(s) used, age of assailants, and day/time of the violent event (Figure 1; Shepherd, 2007). On a regular basis, de-identified patient data are combined with law enforcement data, the locations of violent incidents are geocoded and mapped, and a summary report is prepared, shared, and discussed with stakeholders, the community, and law enforcement partners at regular intervals. This surveillance process allows for tracking of overall violence incidents, trends in violence “hotspots,” trends in types of weapons used by assailants, and precise times (e.g., day, week, year) when violence is more prevalent.
The Cardiff Model was originally developed in the United Kingdom where the benefit of a national health system and a uniform data collection system facilitates implementation, analysis, and dissemination of public health initiatives (Rothman, Greenland, & Lash, 2008). In the United States, the complexities of the health care system pose challenges to the implementation of a collaborative and comprehensive violence-related injury surveillance system.
The purpose of this study was to assess the feasibility of translating the Cardiff Model's data collection procedures into the electronic medical record (EMR) at a pediatric Level I trauma center in a large urban Midwestern city. Feasibility was evaluated with domains related to implementation, integration, practicality, acceptability, and demand (Figure 2; Bowen et al., 2009) The study, funded by the National Institute of Justice, also evaluated the quality of data gathered in the ED and piloted methods to prepare the data for future linking with law enforcement data, as required by the Cardiff Model.
Study Site and Population
The study site is a stand-alone pediatric Level I trauma center in a large urban Midwestern city. The ED serves youth who present for a multitude of medical issues including violent trauma. As the intervention was implemented as a screening process, there were no exclusions based on language barriers, chief complaint, or illness acuity.
Implementation and Integration: Emergency Department
All patients presenting to the study site ED are evaluated initially by a triage nurse who determines clinical stability and chief complaint and initially screens patients and families for various safety issues. As this was the most consistent initial contact point for arriving patients, the study team and hospital leadership determined that implementation of the Cardiff Model's data collection procedures would fit most naturally in the triage portion of the ED visit. Prior to implementation, the study team, in conjunction with an identified nurse champion from the pediatric ED, met with nurse leadership to discuss insertion of the questions into the triage portion of the EMR (EPIC, Verona, WI). Changes were made to the EMR to include the Cardiff Model's data collection procedures, which included screening and incident-level questions essential for translating the Cardiff Model (Figure 1). Following the design and testing of the new triage insert, the study team presented the new triage questions to all nurses working in the pediatric ED at their required quarterly staff meetings. Nurse training included the assessment of the study questions in the EMR, imputing the required data into the EMR during patient screening, and if asked, assuring patients that these data are de-identified before merging with police data. Between January 2015 and July 2017, screening in the EMR using the Cardiff Model was performed for all patients reporting to the pediatric ED. Incident-level data were collected from patients who provided positive answers to the two screening questions. Implementation and integration were measured by the percentage of eligible patients who were screened. All cases of domestic or child abuse discovered through this process were handled as per the site's standard policies, with appropriate referrals and social work management.
Implementation: Data Quality and Preparation
To further assess the extent of the implementation of study procedures in the ED, the quality of the data collected, and the feasibility of the Cardiff Model, the study staff engaged in a small-scale demonstration of data cleaning and preparation for future linking with law enforcement data. In conjunction with the Information Services Department at the site hospital, a reporting mechanism was developed to extract the Cardiff Model data and other relevant de-identified data fields from the EMR and to monitor nurse utilization of the screening questions. On a monthly basis, a dedicated research assistant extracted a report from the EMR, cleaned the data to remove all identifiers and to clarify exact locations when able, and forwarded this to the team's data manager. The study team then determined the spatial and temporal specificity of each ED injury, in addition to whether or not the injury was the result of an assault. Geocoding of location information provided in the monthly reports, as well as geospatial analysis, was conducted using ArcMap for Desktop Version 10.3.1 (ESRI, Redlands, CA). As the data were de-identified and the research team had no influence over the data collection procedures by the hospital staff and law enforcement, waivers of assent and consent were obtained from the institution's institution review board, which approved all procedures. The initial results were shared with the ED nurses to encourage feedback, assist with preliminary data interpretation, and continue engagement with study procedures.
Logistic regression models were used to determine whether having a missing injury location address in the data set was associated with patient- and injury-level factors. The dichotomous variable for whether a patient's injury location address was missing or not (1, 0) was tested on a number of variables, including gender, race, insurance payer status, age in years, injury time of day, injury type, and means of arrival to the trauma center. Logistic regression analyses were conducted using IBM SPSS Statistics 21 software (Version 21.0).
Practicality, Acceptability, and Demand
Nurses who completed data collection using the Cardiff Model were recruited to participate in a survey to further assess integration and implementation of the questions, assess practicality and acceptability of implementation, and provide open-response qualitative feedback on how to further streamline implementation into the ED workflow. The survey questions included 6-point Likert-type scales (e.g., strongly agree to strongly disagree), which were recoded to reflect “agree” (“strongly agree” and “agree”) and “disagree” (“strongly disagree” and “disagree”), and “satisfied” and “dissatisfied.” De-identified online survey administration with SurveyMonkey occurred from August 2015 to October 2015 (Time 1) and a second time from March 2017 to April 2017 (Time 2). Although the survey instrument itself was not validated, the questions within the survey were based on the feasibility domains outlined previously (Bowen et al., 2009). Results of the survey were disseminated back to the nurses and were used to assess study procedure and EMR question format.
An exploratory focus group of ED providers, including nurses and physicians, was held 1 year following the initial survey. The purpose of this focus group was to gather qualitative data on the feasibility domains described earlier, as well as to gather data regarding participant perceptions of Cardiff Model fit with community and hospital initiatives. A total of five providers—three nurses and two physicians—participated in the focus group. This focus group was audio recorded and transcribed. Results were summarized and grouped into initial themes by individual study team members. The study team then reconciled the initial themes into overarching themes, which inherently fit with the study's identified feasibility domains.
Implementation and Integration: Emergency Department
Nurses successfully screened 98.2% of patients over the study period (Figure 3). Average triage times during the initial 2-month pilot implementation (January–February 2015) increased by 30 s (4.2 min vs. 3.9 min). Correspondingly, nurse respondents self-reported that collecting this information did not require a large investment of time—at survey Time 1, 50% of survey respondents indicated that the data collection took 1 min or less to complete after a positive assault screen and 69% reported less than a minute during Time 2 (50/69%). Most survey respondents across the two time points (62/72%) reported that screening patients for assault took less than 15 s.
Table 1 shows results of the nurse feasibility survey after initial implementation at Time 1 and again at Time 2 following changes suggested by nurse survey respondents at Time 1. At both times, the majority of respondents reported that data collection was satisfactorily integrated into the workflow within the ED (88/90%) and that patients were cooperative when asked the questions (86/88%). In open-ended questions, a few respondents indicated that at times, the questions felt repetitive and the addition of radio and other buttons may streamline data collection in the EMR. In addition, by gathering provider input at regular intervals and making changes to the data collection module, the study staff had been able to further streamline questions, making the data collection process less of a burden to nurses. Focus group participants reiterated the importance of consistent feedback on data collection for successful implementation and integration of Cardiff Model questions into the EMR (Figure 2).
Implementation: Data Quality and Preparation
During the study period, 1,160 patients screened positive for assaultive injury using the EMR modules. Table 2 provides the distribution demographic information regarding the participants, as well as the various types of injury sustained by participants. A majority of participants were male, Black, and covered by public insurance (e.g., Medicaid); the most prevalent injury type was blunt injury. Of the 1,160 cases, the nurses acquired sufficient address data to geocode and map 769 cases (66.3% of the total; Figure 3). Cases that were not able to be mapped had either missing or incomplete address data. During initial data cleaning and preparation, the study team found that 46% of cases were unable to be mapped because of missing or incomplete address information. This improved to greater than 75% by providing training to nurses on the appropriate data to include in the various fields and dedicating a research assistant to clean data (Figure 4). Table 3 shows results of regression analysis comparing cases that were geocoded versus those that were missing an address for the assault. The overall logistic regression model was significant, χ2 (df = 7, n = 1,160) = 79.433, p = .000, indicating that the one or more of predictors reliably distinguished between those who did have missing location data and those who did not have missing location data. Specifically, compared with White participants, participants who were Asian were less likely to have their injury location address missing. Compared with participants who were covered by private insurance, participants who were covered by public insurance (e.g., Medicaid) were 2.202 times as likely to have missing location information. In addition, participants who were either injured after school hours or at unknown times, and participants who sustained a bite injury, were significantly less likely to have had missing injury location than the corresponding reference groups.
Practicality, Acceptability, and Demand
Few respondents agreed that data collection interfered with their work (21% first survey/15%), and most were satisfied with their participation (93%/92%). Most respondents felt that data collection was congruent with the ED and hospital goals and mission (93%/98%), that it should continue (88%/90%), and that patients and their families found it acceptable when they asked study questions (88%/90%). In addition, survey respondents reported that the data collected made them feel they were better positioned to impact violence in their community, as reflected in the following open-ended statements: “We are not only helping one patient but also helping the community as a whole”; “I like feeling like I'm contributing to making the city safer for our patients”; and “I feel the information is useful to the community.”
A majority of respondents (63%/69%) reported that they wanted to see summary reports of the data generated through module questions, and 66%/72% were interested in collecting assault and injury data in their everyday work. In terms of patient demand, 55%/69% of respondents agreed that patients want the ED staff to ask them about assaults and 43%/67% agreed that patients want the ED staff to help them with reporting assaults to the police. Finally, most respondents felt that the data collected were useful for clinical care in the ED (79%/70%). A significant portion (44%/61%) reported that they do not know how often their completion of module questions led to a report to police on the patient's behalf, and 52%/59% indicated that their asking of the module questions led to a referral to social work only some of the time. The second round of surveys showed improvements on most items, suggesting that the implemented changes that were based on nursing feedback were well received.
Addressing the theme of “Internal Workflow,” focus group participants noted some factors that may influence data quality. Lack of privacy was a concern for participants, who indicated that, “...it's not really private out there (in the triage area) ... so in terms of gathering accurate information, I find myself having to talk quiet, or maybe I won't ask a question as candidly as I might have in a private room....” Also, the lack of a “hard stop” was perceived to be an issue: “There's no hard stop, so if the (Cardiff Model) questions aren't asked for some reason, the system doesn't force you to enter anything,” which would result in incomplete population health data.
Focus group participants further felt that the culture of the ED setting and the hospital as a whole supports the acceptability and demand for violence-related data collection. For this “Culture” theme, participants indicated that they feel a responsibility to ensure that children who are victims of assault receive the needed resources, both within and outside the hospital setting, and they felt a responsibility as a provider to collect Cardiff Model data. They agreed that part of being healthy is being safe, stating that “...if we make kids feel a little bit safer, we improve their overall health.” In addition, in the context of the theme of “External Factors,” providers felt that collecting Cardiff Model data is part of their responsibility to the community—“Our mission at [the hospital] is more community-based, providing the best care for children in the area, not just in the hospital ... our role in gathering these data [regarding assaultive violence involving children] is so important.” Participants indicated that they felt they were making a difference in the community by providing these data and that health care can be a source of information for communities. Finally, participants indicated that the Cardiff Model is a good fit with the Affordable Care Act and that it “...fits a greater good, and has a public health piece to it.” This sense of responsibility and desire among participants to make a difference in their community resulted in a culture that allowed for the integration of Cardiff Model questions. Although focus group participants vocalized demand and responsibility to gather Cardiff Model data, they were unsure whether they were able to actually affect change in their community that would lead to the prevention of violence.
Implementation and Integration: Emergency Department
We report the results of a partnership between academics, hospital leadership, nursing, and information technologies to implement the Cardiff Model's data collection procedures into the EMR of a busy Level I trauma center pediatric ED. To our knowledge, this is the first attempt in the United States to capture de-identified assaultive injury information in a pediatric setting to further public health efforts to address and prevent violence. This study's qualitative and quantitative analyses suggest that the data collection procedures were successfully integrated into the EMR without disruption to patient flow. The majority of nurses also reported a less than 1 min increase in time required to gather the data, indicating that integration of these questions was not perceived to be unduly burdensome. Future implementation steps could include the collection of additional nurse feedback, which could further streamline data collection procedures and clinical workflow, or the collection of feedback from patients themselves.
Implementation: Data Quality and Preparation
Nurse completion of location fields increased over time as their feedback was integrated into the EMR module. For example, many nurses completing the information collection initially included pertinent data in the wrong field, specifically when there were “subjective” text boxes as options. Providing continuing education to nurses around the various fields in the Cardiff Model's data collection module and tailoring EMR module changes to nurse feedback facilitated easier and more standardized data collection and completeness. Qualitative analysis indicated that ensuring a private setting for data collection and making the questions mandatory may also increase the quality of the data collected in the ED. However, a 98.2% screening rate over the 2-year evaluation period suggests that the nurses are engaged without mandates. Our experience differs from early study of similar questions in Cardiff, Wales, where triage nurses did not feel as though they had enough time to complete the questions and the questions caused unacceptable increases in triage times (Goodwin & Shepherd, 2000 ; Shepherd, 2007). The number of cases that were able to be geocoded also improved after data cleaning was handled by dedicated research assistants embedded in the study site ED (Figure 4). This suggests that the quality of data cleaning and not nurse input alone may influence data quality. Significant differences in present injury addresses for race, insurance payer, injury time, and injury type shown in Table 3 may suggest variation in nurse data collection procedures or in what information is self-reported by patients and their families. Further research is needed to determine the potential causes of differential address availability and to identify potential strategies to decrease missing information for these groups.
Practicality, Acceptability, and Demand
The possibility that changes will be fully integrated into a system may increase if those directly affected feel that the changes are acceptable and practical. We found that it is also crucial to obtain the support and buy-in of key system partners (e.g., administration and users) prior to the implementation of systems change in order to maximize participation and enhance the quality of the data that are collected. Results of our feasibility survey and focus group indicate that ED nurses participating in Cardiff Model data collection believed that not only this project was feasible for implementation in their work environment but also the information gathered through asking these questions was useful for clinical care, professionally satisfying, and potentially impactful to community health and safety. The majority of survey respondents in this study were satisfied with their participation in the data collection process and believed that the collection of these data should continue in their work environment over the next year.
With the understanding that regular feedback to the nursing team may enhance and sustain nurse acceptability, demand, and buy-in for the project, our study team began to provide nurses with summary reports in early 2016 as part of their regular staff meetings in order to show nurses the progress of the project, to allow nurses to view and understand our preliminary findings, and to ask questions and provide feedback to the study team. The nurses were appreciative of the opportunity to review the data and felt that their input was integral to maintaining proper surveillance to better inform public health initiatives to reduce violence. Summaries of the data gathered through the Cardiff Model, including maps generated through a GIS mapping process, were and will continue to be shared with nurses so that they can observe how the data they are collecting are being used to inform prevention strategies in their community.
Youth who suffer penetrating injury often have previous minor near-miss events seen in the ED and are at greater risk for fatal outcomes than those with no prior ED visits (Caputo et al., 2012 ; Cunningham et al., 2009 ; Fein, Mollen, & Greene, 2013 ; Melling, Lansdale, Mullassery, Taylor-Robinson, & Jesudason, 2012). At present, the ED is an underutilized resource in efforts to reduce assaultive injury among our nation's youth and the ED visit is a missed opportunity for the collection of information to facilitate the development of population health strategies to address and reduce violence. The results of this pilot study—reported here and elsewhere (Hernandez-Meier, Xu, Kohlbeck, Levas, & Hargarten, 2017)—have already led to the inclusion of additional hospital sites in the region and to the recruitment of community partners. The results also suggest that the Cardiff Model offers an opportunity to integrate place-based information that can provide the real-time development of prevention programs and policies that may prevent future injuries and deaths. The recruitment and convening of multidisciplinary stakeholders from communities served by participating health and safety entities are hypothesized to lead to the development of behavioral, environmental, and policy- and law enforcement-oriented violence prevention strategies, which could mirror those implemented in the United Kingdom. For example, U.K. bars adopted the use of plastic bottles to reduce injuries related to fights with broken bottles, sidewalks were expanded to reduce crowding, and law enforcement targeted their street patrols and increased closed-circuit television cameras (Sivarajasingam, Shepherd, & Matthews, 2003).
It is both feasible and meaningful to implement assault-related injury-specific questions into the EMR, with little perceived disruption of workflow and triage times. Nurses are receptive to participating in the collection of data that may inform clinical care and community violence prevention activities.
Our study was limited to a single ED at a pediatric Level I trauma center and therefore our results may not be generalizable to other pediatric facilities or to adult facilities or community hospitals. However, we feel that it is generalizable, given that all trauma centers (Levels I–III) are required to engage in injury prevention activities, and the Cardiff Model is consistent with this requirement.
The Cardiff Model implementation in the United Kingdom included collection by the reception staff (Goodwin & Shepherd, 2000 ; Shepherd, 2007). Our implementation deviated from the initial implementation by utilizing nurses to collect the data. As this is the only study that we are aware of in the United States that is implementing the Cardiff Model in a pediatric Level I trauma facility, we did not adhere to a specific model that accounted for the unique challenges posed by our nation's health care system. Further replication and implementation at other hospital sites will further our understanding of how this model can be fully integrated into EDs and be able to inform community antiviolence efforts across communities in the United States
- Nurse collection of assault-related surveillance data in a busy ED is feasible, is meaningful to participants, and adds timely input to the public health narrative surrounding violence and injury reduction.
- The Cardiff Model of data collection to inform violence-reduction initiatives is feasible in the United States and may enhance total knowledge of where and when assaults are occurring in our communities.
- Emergency department nurses, as key members of the health care team, should continue to be involved in public health initiatives, as they are often the first point of contact for patients, are often trusted by the community, and offer unique insight into both data collection and intervention implementation.
This project was supported by Award No. 2014-IJ-CX-0110, awarded by the National Institute of Justice, Office of Justice Programs, U.S. Department of Justice.
Analytical Unit, Social Care, Local Government and Care Partnerships Directorate, Department of Health and Social Care. (2015). Information sharing to tackle violence
: Audit of progress. Washington, DC: Author.
Barthell E. N., Cordell W. H., Moorhead J. C., Handler J., Feied C., Smith M. S., Collins M. A. (2002). The Frontlines of Medicine Project: A proposal for the standardized communication of emergency department data for public health
uses including syndromic surveillance
for biological and chemical terrorism. Annals of Emergency Medicine, 39(4), 422–429.
Bowen D. J., Kreuter M., Spring B., Cofta-Woerpel L., Linnan L., Weiner D., Fabrizio C. (2009). How we design feasibility studies. American Journal of Preventive Medicine, 36(5), 452–457.
Boyle A. A., Snelling K., White L., Ariel B., Ashelford L. (2013). External validation of the Cardiff Model of information sharing to reduce community violence
: Natural experiment. Emergency Medicine Journal, 30(12), 1020–1023.
Butchart A., Phinney A., Check P., Villaveces A. (2004). Preventing violence
: A guide to implementing the recommendations of the world report on violence
and health. Geneva, Switzerland: Department of Injuries and Violence
Prevention, World Health Organization.
Caputo N. D., Shields C. P., Ochoa C., Matarlo J., Leber M., Madlinger R., Waseem M. (2012). Violent and fatal youth trauma: Is there a missed opportunity? The Western Journal of Emergency Medicine, 13(2), 146–150. doi:10.5811/westjem.2011.6.6765
Cooley-Strickland M., Quille T. J., Griffin R. S., Stuart E. A., Bradshaw C. P., Furr-Holden D. (2009). Community violence
and youth: Affect, behavior, substance use, and academics. Clinical Child and Family Psychology Review, 12(2), 127–156.
Corso P. S., Mercy J. A., Simon T. R., Finkelstein E. A., Miller T. R. (2007). Medical costs and productivity losses due to interpersonal and self-directed violence
in the United States. American Journal of Preventive Medicine, 32(6), 474–482. doi:10.1016/j.amepre.2007.02.010
Cunningham R., Knox L., Fein J., Harrison S., Frisch K., Walton M., Hargarten S. W. (2009). Before and after the trauma bay: The prevention of violent injury among youth. Annals of Emergency Medicine, 53(4), 490–500. doi:10.1016/j.annemergmed.2008.11.014
Droste N., Miller P., Baker T. (2014). Emergency department data sharing to reduce alcohol-related violence
: A systematic review of the feasibility and effectiveness of community-level interventions. Emergency Medicine Australasia, 26(4), 326–335. doi:10.1111/1742-6723.12247
Fein J. A., Mollen C. J., Greene M. B. (2013). The assault-injured youth and the emergency medical system: What can we do? Clinical Pediatric and Emergency Medicine, 14(1), 47–55. doi:10.1016/j.cpem.2013.01.004
Florence C., Shepherd J., Brennan I., Simon T. (2011). Effectiveness of anonymised information sharing and use in health service, police, and local government partnership for preventing violence
related injury: Experimental study and time series analysis. BMJ, 342, d33313. doi:10.1136/bmj.d3313
Florence C., Shepherd J., Brennan I., Simon T. R. (2014). An economic evaluation of anonymised information sharing in a partnership between health services, police and local government for preventing violence
-related injury. Injury Prevention, 20(2), 108–114. doi:10.1136/injuryprev-2012-040622
Goodwin V., Shepherd J. P. (2000). The development of an assault patient questionnaire to allow accident and emergency departments to contribute to Crime and Disorder Act local crime audits. Journal of Accident & Emergency Medicine, 17(3), 196–198. doi:10.1136/emj.17.3.196
Haider A. H., Young J. H., Kisat M., Villegas C. V., Scott V. K., Ladha K. S., Efron D. T. (2014). Association between intentional injury and long-term survival after trauma. Annals of Surgery, 259(5), 985–992. doi:10.1097/sla.0000000000000486
Hernandez-Meier J., Xu Z., Kohlbeck S., Levas M., Hargarten S. (2017). 30 Enhancing violence surveillance
: The contribution of paediatric emergency department and paramedic assault incidents to police administrative data. Injury Prevention, 23, A11–A12.
Kellermann A. L., Bartolomeos K., Fuqua-Whitley D., Sampson T. R., Parramore C. S. (2001). Community-level firearm injury surveillance
: Local data for local action. Annals of Emergency Medicine, 38(4), 423–429.
Melling L., Lansdale N., Mullassery D., Taylor-Robinson D., Jesudason E. C. (2012). Penetrating assaults in children: Often non-fatal near-miss events with opportunities for prevention in the UK. Injury, 43(12), 2088–2093. doi:10.1016/j.injury.2012.03.020
Morrissey T. B., Byrd C. R., Deitch E. A. (1991). The incidence of recurrent penetrating trauma in an urban trauma center. The Journal of Trauma, 31(11), 1536–1538.
Rosenfeld R. (2016). Documenting and explaining the 2015 homicide rise: Research directions: A report to the National Institute of Justice. Washington, DC: Department of Justice.
Rothman K. J., Greenland S., Lash T. L. (2008). Modern epidemiology. Philadelphia, PA: Lippincott Williams & Wilkins.
Shepherd J. (2007). Preventing violence
—Caring for victims. The Surgeon, 5(2), 114–121. doi:10.1016/S1479-666X(07)80066-7
Sims D. W., Bivins B. A., Obeid F. N., Horst H. M., Sorensen V. J., Fath J. J. (1989). Urban trauma: A chronic recurrent disease. The Journal of Trauma, 29(7), 940–946; discussion 946–947.
Sivarajasingam V., Shepherd J. P., Matthews K. (2003). Effect of urban closed circuit television on assault injury and violence
detection. Injury Prevention, 9(4), 312–316.
Sutherland I., Sivarajasingam V., Shepherd J. P. (2002). Recording of community violence
by medical and police services. Injury Prevention, 8(3), 246–247.
World Health Organization. (2014). Global status report on violence
prevention 2014. Geneva, Switzerland: Author.