Outcomes, Data, and Indicators of Violence at the Community Level

The NVDRS is an active state-based surveillance system designed to collect information on violent deaths including suicide, homicide, unintentional firearm, legal intervention, child maltreatment, and other undetermined injuries (CDC, 2014i). Colorado, Maryland, Michigan, North Carolina and Virginia are among the 32 states that are part of the NVDRS. These data provide accurate information on violence that leads to fatalities as the basis for conducting injury surveillance and for developing and evaluating violence prevention programs. All types of homicides, including variables such as the victim-suspect relationship (e.g., intimate partner, child-parent) are captured. Additionally, since NVDRS includes and links all deaths that occur as part of the same incident, we can identify homicide-suicides and multiple victim incidents. NVDRS also captures legal intervention deaths (individuals who are killed by law enforcement officers and officers who are killed in the line of duty). Data on alcohol and drug involvement based on autopsy results are also collected. The data available for analysis are robust and include the date, time and manner of death, demographic information, and narratives from the coroner or medical examiner and law enforcement officers.

A major benefit of the NVDRS is that it links data from multiple sources, including death certificates, coroner or medical examiner reports, crime incident reports, and crime laboratories (CDC, 2014j).The NVDRS provides unique insight into potential risk factors for both single and multiple victim events and can assist in the development and evaluation of programs and policies designed to prevent deaths from violence. NVDRS data are useful for informing decision makers and program planners about the magnitude, trends, and characteristics of violent deaths in their state or community that can then serve to inform and evaluate prevention efforts. For instance, JHU is drawing upon NVDRS data about alcohol-involved homicides to augment its alcohol policy intervention research, and the VCU maps the data to identify areas with high rates of mortality.

The NVDRS data are also available for query through the Web-based Injury Statistics Query and Reporting System (WISQARS). WISQARS is an online database that can be queried to provide fatal and nonfatal injury, violent death, and cost-of-injury data from a variety of sources (CDC, 2014i). WISQARS is operated by the CDC to provide summary statistics, including frequencies and rates by specified search criteria in the United States. NVDRS WISQARS only reports state-level data. However, smaller geographic level data are available in the Restricted Access Database (RAD) which can be accessed with special request. All of the YVPCs use WISQARS data to determine and monitor violence rates in their respective states and occasionally to evaluate policies (CDC, 2014k).

Despite its strengths, the NVDRS data are not made available for utilization in real time. In most cases, the data are about 3 years old. However, the data provide useful information to examine rates and the impact of implemented prevention approaches.

YVPCs use mortality data from their state’s vital registry to monitor youth violence locally. Mortality data are publicly available in most states and can be obtained from the state vital record offices. The type of data available may vary from state to state. Typically, the data provide information on cause of death and the victims’ characteristics.

The death certificate data are helpful in determining and monitoring violence-related mortality rates in any community. Data can be used for small area analysis to guide targeted interventions. Largely, death certificate data are accurate and reliable; however, they only capture the basic demographic characteristics of the deceased. The data can be linked to other data sources using social security number, date of birth and other identifiers for further analysis. Linked datasets allow for the examination of determinants as well as risk and protective factors that are not readily available for researchers (Putnam-Hornstein, Cleves, Licht, & Needell, 2013). Because administrative datasets may not accurately document identifiers, linking data may be challenging. Probabilistic matching (Clark, 2004; Jaro, 1995), a method that uses data across multiple fields to link cases from separate datasets, may be a useful and necessary tool when linking large administrative databases. Nonetheless, obtaining personal identifiers such as social security numbers and date of birth requires permission from vital statistics and health departments. While data that include personal identifiers may be of interest to some researchers, they may not be necessary for stakeholders seeking to develop, plan, implement, and evaluate youth violence prevention programs and policies. YVPCs use local and national death certificate data to evaluate interventions and policies. For instance, the YVPC at Johns Hopkins used the national death certificate data from the National Center for Health Statistics to evaluate firearm laws concerning adolescent suicide rates (Webster, Vernick, Zeoli, & Manganello, 2004).

Moreover, mortality data measure the most severe consequence of violence. However, these data do not provide an overall estimate of the prevalence of violence, which is much more common than fatal injuries. The data only contain basic demographic characteristics and cause of death and do not support the examination of other contextual factors that may be important in examining violence data. Further, death certificate data include a number of primary and secondary variables for cause of death, which make the data difficult to code. The complexity of coding this variable may lead to the misclassification of specific forms of mortality.

Another source of data on lethal violence available at the national, state, and local levels is the Federal Bureau of Investigation’s UCR and SHR systems that collect data on murders and non-negligent manslaughters. The SHR provides incident-level data on important factors not available from public health sources including the suspects’ age and relationship to the victim, and the circumstance category (e.g., robbery, drug-related). The UCR and SHR provide data that can be used for smaller geographic areas. The UCR is discussed in great details below.

All of the YVPCs use crime incident data to monitor youth violence rates and measure the impact of their interventions. Crime incident data are the standard measure of crime used by law enforcement, media, citizens and academics to characterize the level of violent and non-violent crime in a geographic location. Local data can be collected from law enforcement agencies including crime incident and arrest reports. These data generally consist of a number of characteristics of the crime incident, including date, time, location, and classification of the nature of the criminal incident. The YVPCs use the standard definitions and mechanism for recording and reporting crime incidents provided by the Federal Bureau of Investigation (FBI) Uniform Crime Reports (UCR; FBI, 2014). These classifications include definitions of violent crimes (e.g., murder, non-negligent manslaughter, aggravated assault, aggravated battery, criminal sexual assault, and robbery) as well as other, non-violent crimes (e.g., theft, burglary, and vandalism). Crimes may also be categorized as index (i.e., more serious offenses) versus non-index crimes. A hierarchy among crime categories exists so incidents that, for example, involve an assault, a robbery and rape are categorized based on the most severe offense, with the level of severity descending from rape, to robbery, to assault.

Crime incident data have a number of advantages as indicators of community violence in a surveillance system. First, these data generally have very high face validity due to their frequent use to report crime statistics in the media, by law enforcement, and in academic journals. Second, reporting systems have been in place for decades, thus providing useful opportunities to examine trends in violent crime over time. Third, with the increasing collection of detailed incident information, including geocoded location, these data can be easily aggregated to provide statistics for a defined geographic area. This allows for relatively easy integration with other sources of data (e.g., census, community surveys, or other location-based data). The ability to use longitudinal geocoded data on incidents of violent crime and arrests using standard definitions is incredibly valuable for describing variations in youth violence over time and place, providing the capability to evaluate the efficacy of neighborhood or place-based interventions.

The Johns Hopkins YVPC has used Baltimore’s Police Department data on murders, non-negligent manslaughters, and nonfatal shootings as well as arrests for weapon and drug violations to evaluate youth violence prevention interventions such as Safe Streets (based on the Cure Violence model) and law enforcement programs focused on violent gun offenders (Webster, Whitehill, Vernick, & Curriero, 2012). The Hopkins YVPC has also used local police data to examine the relationship between violent crime incidents, arrests for alcohol-related crime, and the location of alcohol outlets to advance policies to curtail problem alcohol outlets (Jennings et al., 2014). The Colorado YVPC utilized Denver’s violent crime data to select intervention and control neighborhoods for their study. Neighborhoods that ranked in the upper one-third of violent crime for Denver were deemed eligible to participate. Neighborhood level violent crime data are also being used to evaluate outcomes.

Although incident data can sometimes be difficult to obtain and have not always been released in a timely fashion, these data are increasingly accessible through government-sponsored websites that make crime maps and the raw incident data routinely available. For example, the City of Chicago hosts the Chicago Data Portal (www.​data.​cityofchicago.​org), a data repository that contains a wide range of data related to the city (City of Chicago, 2014). Included on this site are all crime incidents from 2000 to the present. These data are updated daily and can be accessed and downloaded without restriction. These data were central to the process used by the Chicago YVPC to select their target and comparison communities. The Chicago YVPC continues to use crime incident data to monitor crime in the target community and to evaluate the impact of the YVPC over time.

In addition, researchers from the Chicago YVPC have explored the use of crime incident data as a form of sentinel surveillance (Henry et al., 2014). Henry and colleagues used data on incidents of disorderly conduct, vandalism, and weapons violations to predict future increases in neighborhood-level violent crimes. The study suggested the potential use of crime incident data to anticipate neighborhoods at risk for future increases in violence and to target prevention strategies accordingly.

Despite the strengths of crime incident data, there are some limitations that must be considered when using these data as part of a surveillance system. As with any single component of surveillance, there are biases inherent in incident data. The primary issue to consider is that for most types of crime, not all cases are reported to the police and recorded in the database (e.g., Skogan, 1984). Typically, more severe crimes (e.g., homicide, violent crimes that involve major injuries, major theft) are more likely to be reported and recorded than less severe crimes; however, underreporting is a particular problem for sexual assault. Data from the 2013 National Crime Victimization Survey rates of reporting victimizations to police were 36.1 % for property crimes, 45.6 % for violent crimes, and 61.0 % for serious violent crimes (Truman & Langton, 2014). Of note, the reporting rate for rape and sexual assault was only 34.8 %.

Within challenged communities that are the focus of the YVPCs, there may be several competing biases in the rates for crime incident and arrest data. Within inner city, low income neighborhoods or other communities that are frequently patrolled by law enforcement officers, individuals may be more likely to be cited or reported for minor crimes than incidents occurring in other areas. Further, racial stereotyping in predominantly minority neighborhoods could also lead to disproportionate arrest rates. While policing patterns may differentially raise incident and arrest rates, distrust of law enforcement, perceptions of ineffective policing, and social norms against reporting to law enforcement can serve to lower rates in these areas. In the context of using incident data to evaluate community-level interventions, changing reporting practices, whether intended or not, may alter measured incidence rates when actual crime rates have not changed. For example, efforts to enhance block club participation such as neighborhood walking or gardening clubs and neighborhood watch may empower residents to report crimes that previously went unreported.

All of the YVPCs use disciplinary incident data to measure outcomes and monitor youth violence in schools. The YVPCs collaborate with schools and state and county departments of education to obtain data on disciplinary incidents, remedial actions, referrals and school characteristics. The type of data available varies by YVPC but typically includes school and student characteristics.

Data from schools provide useful information to understand school-based violence. For instance, the YVPC at VCU monitors trends of youth violence rates including weapon possession, fights, bullying, truancy, and drop outs (Masho & Bishop, 2014a). Furthermore, school disciplinary data are linked to student survey data to measure individual level outcomes. UNC uses school disciplinary data to monitor acts of youth violence committed during the school year. Data related to weapon possession, possession of controlled substances and alcoholic beverages, assault, and sexual assault are examined. The Chicago YVPC is incorporating school disciplinary data into the evaluation of school-based interventions that are included in their comprehensive violence prevention strategy. Disciplinary data at the school and grade (within school) level allow the evaluation team to assess changes in disciplinary incidents and to compare schools in the intervention community to those in the comparison community as well as to all Chicago Public Schools. By linking schools to neighborhoods, covariates in these models include community-level as well as school-level factors.

Despite these benefits, disciplinary incident data have some limitations. First, the data are limited to incidents that occurred on school grounds and do not provide accurate estimates of violence outside school. Second, the data are limited to students who attend school. Drop out and chronically tardy students who are at a higher risk of violence may not be adequately represented. Third, data are limited to certain demographic and school characteristics and do not allow for the examination of risk and protective factors. Fourth, the uniformity of the data may vary by school culture and school-specific initiatives. For instance, a school with a bullying prevention program may be more diligent in collecting violence related data which may indicate changes in bullying. Therefore, a clear understanding of contextual issues affecting data collection is important.

Few of the YVPCs use juvenile justice services data to monitor the impact of their program. The type of data available varies by YVPC site; however, data typically include individual-level data on youth who were referred to the juvenile justice system or have been formally processed or sentenced by juvenile court. Although the data collected may differ by locality, demographic information, and type of offense, sentence and referral services provided are consistently documented across sites. The Office of Juvenile Justice and Delinquency Prevention (OJJDP) has a website that offers a collection of national juvenile justice datasets (OJJDP, 2014).

The juvenile justice services data allow a close examination of youth who are at increased risk of behavioral problems. UNC uses juvenile justice services data to examine juvenile arrests, delinquent acts and complaints against juveniles. VCU has tracked juvenile justice services data since 2003 to monitor violence-related offenses including homicide, assault, sexual assault, weapons and bombs, robbery and kidnapping. The data are also being used to monitor trends and changes in the community. For instance, a recent analysis of the data in Richmond, VA found a gradual increase in the rate of females referred to juvenile justice services (Masho & Bishop, 2014b). This information provided the impetus to monitor and understand gender differences in youth violence. Furthermore, the data are being used to assess the impact of targeted interventions among these high risk youths at VCU.

Although the juvenile justice services data offer several advantages, they have some limitations. Juvenile justice system data are administrative data collected on youth who have been identified and referred to the juvenile justice system. The introduction of initiatives targeting youth may lead to over- or under-reporting that may inflate or deflate data. For example, in Richmond, VA, a truancy initiative resulted in increased reporting of truancy making other forms of violence seem lower (unpublished data).

ED data are being used by all of the YVPCs. The type of data collected may vary by site. However, they typically include patients’ demographic characteristics, diagnosis and procedure codes. To determine injury groupings, the recommended framework for presenting injury mortality data, the International Classification of Diseases, Ninth Revision (ICD-9) provided by CDC is being used to summarize the data. External Cause of Injury Codes (E-codes) are used to determine injury groupings. These groupings include presence of injury, external injury, motor vehicle crash, assault, child abuse, suicide and firearm injury (CDC, 2014o). The YVPC at VCU has partnered with its tertiary trauma center and collects data to assess the burden of youth violence locally (Masho & Bishop, 2014c). The data are being used to monitor trends in youth violence in Richmond City and identify populations that are disproportionately affected by the problem. The data are also being used to estimate rates of ED visits attributed to violence and examine the impact of local interventions.

ED data are also available nationally from the National Electronic Injury Surveillance System—All Injury Program (NEISS-AIP). The system is operated by the U.S. Consumer Product Safety Commission with CDC’s National Center for Injury Prevention and Control (NCIPC). These data incorporate reports with national estimates of nonfatal injuries treated in U.S. hospital emergency departments broken down by intent (physical assault, sexual assault, legal intervention, self-harm, unintentional) and cause and mechanism (e.g., firearm, cut or pierce). Using these same data, the CDC’s NCIPC generates the leading cause of nonfatal injury rankings for the top 10 causes of nonfatal injuries treated in EDs by the age and sex of the injured patient, intent of injury, and disposition when released from the ED. The YVPCs use these data to compare local rates with national rates. Although the NEISS-AIP offers similar information, the data usually lag by a few years. Local ED data offer more timely and robust information that can provide immediate local benefit.

ED data are one of the most commonly used data sources to measure violence, and include demographic and medical information. However, ED data capture only victimized youth who seek medical attention—typically more severe violence. These data exclude the majority of youth involved with violent episodes who are not injured as well as those who are injured but do not require medical attention.

VCU uses ambulance data to assess the trends of more severe violence in the community. Similar to ED data, ambulance data are used to estimate rates of violence in the community. These data include information on the demographic characteristics of the victim and call disposition. The call disposition information is used to determine violent incidents. Events where an assault occurred, such as rape, a fight or brawl, shooting, or stabbing are categorized as violent injury-related events.

The data are used to evaluate community-wide efforts and guide policies. For instance, a report by Masho et al. (2014) utilized ambulance data to examine the impact of restricted alcoholic beverage licenses on rates of ambulance pickup in areas where violence rates were high. For example, a report of the surveillance data depicting higher rates of violence surrounding grocery stores selling less expensive alcoholic beverages informed the effort made by a local civic organization to place restrictions on the sale of single-serve alcoholic beverages.

Despite their strengths, ambulance data have a number of limitations. Similar to ED data, ambulance data only represent injuries that require immediate medical attention. Furthermore, dispositions are usually based on patient report and initial examination and may be inaccurate. It is also important to note that most cities have more than one ambulance provider, and data collected by multiple companies may lack uniformity.

The YRBSS is a national survey administered by the CDC biennially (CDC, 2014f). All of the YVPCs use YRBSS data to monitor youth violence risk factors. The YRBSS began in 1991 to systematically collect data on a wide range of adolescent behaviors, including violence related activities. The school-based survey, which covers grades 9 through 12, is anonymous and cross-sectional. A middle school YRBSS is also established, but is currently used by only 18 states. Five states where the YVPCs are located (Colorado, Maryland, Michigan, North Carolina, and Virginia) collect data on middle schools. However, the data may not be representative of specific localities to evaluate programs.

YVPCs rely on the YRBSS to measure current behavior and monitor trends over time at the state level. YVPC researchers often use YRBSS measures to ensure that their measures of youth violence and risks are comparable to data available in many cities and states. Although 47 states implement the survey, participation rates by school may differ by states. Decisions to participate in the YRBSS may be mandated by government officials, or left to the discretion of school administrators. Thus, a significant proportion of schools may opt not to participate, limiting its generalizability. For example, in Colorado, state level data are available dating back to 1995, but Denver public schools did not participate in the YRBSS until 2007—therefore, arguably the highest risk population in Colorado was not represented until then. Additionally, YRBSS data are based on students who were present or available on the day of the survey; thus, the highest risk students may be truant and not represented. The YRBSS website has an extensive explanation of study design, and multiple reports and manuscripts have been published on the most recent trends in adolescent behavior (CDC, 2014m).

The YRBSS data are publicly available and can be downloaded for further analysis. Additionally, summary statistics from the YRBSS data can be accessed from the Youth Online portal (CDC, 2014n). The portal provides statistics and trends at the national, state, and local levels, depending on the availability of weighted data for a particular geographic location. For instance, weighted data are available for Colorado for years 2005, 2009, and 2011. The data revealed that there has been a decrease in physical fighting for the first time in a decade. Comparing data from 2009 to 2011, physical fighting decreased from 32 to 24.9 % p < 0.001. Further analysis showed that the differences are accounted for by changes in fighting among males: 42–30.3 %, p < 0.001 and are not due to change in female fighting, which also declined from 21.8 to 18.2 %, (p = 0.16) (CDC, 2014l).

The Colorado YVPC incorporates several YRBSS violence related questions into their community survey. This allows for the comparison of state level changes for physical fighting (as an example), over time to intervention community changes to help determine whether their intensive community level efforts to decrease youth violence can be attributed to the intervention.