Research utility of the National Violent Death Reporting System: a scoping review

Background To better understand and prevent suicide and homicide, the National Center for Injury Prevention and Control of the US Centers for Disease Control and Prevention launched the National Violent Death Reporting System (NVDRS) in six states in 2002. As of 2018, the NVDRS has been expanded to include all 50 states, the District of Columbia and Puerto Rico. The purpose of this review was to assess the research utility of the NVDRS based on studies indexed in major bibliographical databases. Methods We performed a scoping review of published studies that were based on data from the NVDRS, identified by searching six electronic databases: PubMed, EMBASE, Google Scholar, OVID, Scopus, and Web of Science. We examined the time trend of annual NVDRS-based research output, generated a word cloud using the keywords listed in the publications, and mapped the knowledge domains covered by NVDRS-based studies. Results Our review included a total of 150 studies published between 2005 and 2018. There was a marked increase in the annual number of NVDRS-based publications, with 120 (80.0%) of the 150 studies published between 2011 and 2018. Overall, 104 (69.3%) studies focused on suicide and 39 (26.0%) on homicide. Of the included studies, 100 (66.7%) were descriptive epidemiology, 31 (20.7%) were risk factor analyses, 9 (6.0%) were evaluations, 7 (4.7%) were trend analyses, and 4 (2.7%) were data quality assessments. Knowledge domain mapping identified two major clusters of studies, one on suicide and the other on homicide. The cluster on suicide was commonly linked to “circumstance,” “alcohol” and “substance abuse” and the cluster on homicide was commonly linked to “firearm,” “injury,” and “gang.” The two clusters were interlinked to overlapping networks of keywords, such as “firearm” and “mental health problem.” Conclusions Research utility of the NVDRS has increased considerably in recent years. Studies based on data from the NVDRS are clustered in two knowledge domains – suicide and homicide. The vast potential of the NVDRS for violence research and prevention remains to be fully exploited.


Background
In 2016, approximately 65,000 deaths were attributed to violence-related injuries in the United States (Fatal Injury Reports, 1981. Of these violent deaths, about 70% resulted from suicide and 30% from homicide. Violent deaths account for about 30% of the total injury mortality in the United States and violence prevention presents a major public health challenge. Information about the circumstances surrounding violent deaths is crucial for developing effective intervention programs. In an effort to understand and prevent suicide and homicide, the National Center for Injury Prevention and Control of the Centers for Disease Control and Prevention (CDC) started the National Violent Death Reporting System (NVDRS) in 2002 to monitor violent deaths and their circumstances .
Prior to the establishment of the NVDRS, data sources such as the National Vital Statistics System and the Uniform Crime Reporting Statistics provided only limited information about the circumstances of violent deaths (Crosby et al. 2016). As such, there were calls for the creation of a national data system on violent deaths similar to the Fatality Analysis Reporting System developed by the National Highway Traffic Safety Administration (Mercy and Houk 1988;Teret et al. 1992;Barber et al. 2000;Hemenway et al. 2009). After several failed attempts at developing and enhancing violent injury surveillance systems, the CDC halted all activities related to violent injuries in 1996 Rovner 1996). However, in 1998, a group of private foundations provided temporary funding to revive the initiative and in 1999, in response to an Institute of Medicine report calling for a national fatal injury system, a pilot program called the National Violent Injury Statistics System (NVISS) was established .
With growing support for the NVISS, in 2002 the Congress allocated $1.5 million to the CDC to establish the NVDRS (Paulozzi et al. 2004). It became the first multistate system to provide detailed information on the circumstances precipitating violent deaths by culling data from multiple sources (Crosby et al. 2016). In 2003, the surveillance system began data collection in six states: Maryland, Massachusetts, New Jersey, Oregon, South Carolina, and Virginia McNally et al. 2016). Since then, the US government has provided an additional $20 million in annual funding to expand the NVDRS to collect data from all states (Barber et al. 2013). As of 2018, the NVDRS has been expanded to all 50 states, the District of Columbia, and Puerto Rico, pooling data from coroner and medical examiner reports, death certificates, law enforcement reports, and toxicology reports using standardized protocols.
Since the launch of the NVDRS, data generated from the system have been used to identify research needs of at-risk populations and develop targeted intervention programs to prevent violent deaths. For instance, the NVDRS data have been used to study homicide and suicide incidents involving older adults and young children, veterans, and law enforcement officers and to document the prevalence of putative risk factors such as alcohol abuse, substance abuse, and financial distress. One study found a significant increase in suicide related to the US housing crisis resulting in foreclosures ). Another study showed that veterans had a higher risk of suicide in comparison to non-veterans . The NVDRS has helped build alliances among key stakeholders, inform intervention programs, raise awareness, and catalyze new project developments (Horan and Mallonee 2003;Cambell et al. 2006). For instance, the Oregon Violent Death Reporting System staff collaborated with the Oregon Veteran Health Administration to establish prevention programs for veterans (Shen and Millet 2014).
Violent death is defined as a death resulting from either the intentional use of physical force or power against oneself or others . The NVDRS categorizes violent deaths into the following groups: suicide, homicide, unintentional firearm death, deaths of undetermined intent, deaths due to legal intervention (excluding executions) and deaths due to terrorism (excluding acts of war). Reportable cases are identified based on the codes of the International Classification of Diseases, Tenth Revision, or the manner of death assigned by the coroner, medical examiner or law enforcement ). The NVDRS system is incident-based and records data on both the perpetrator and the victim. In the event of multiple-death homicide, it links all the deaths into one incident record. The NVDRS has been used extensively by researchers. The purpose of this scoping review is to provide a summary assessment of the research utility of the NVDRS through bibliographical analysis, including knowledge domain mapping and network visualization.

Methods
This scoping review was an analysis of bibliographical data for NVDRS-based studies published in peerreviewed journals. We followed the framework outlined by Arksey and O'Malley (2005).

Eligibility criteria
In the first phase, only the titles and abstracts of potentially relevant studies were screened to match the minimum inclusion criteria. In case of titles for which abstracts were not available, full articles were obtained and reviewed. Publications that were letters, editorials, commentaries, or abstracts only were excluded. Also excluded were articles describing the NVDRS without analyzing the data. Studies eligible for inclusion in this review were those that were based solely or partially on the NVDRS data.

Data sources and search criteria
The search query consisted of "National Violent Death Reporting System" and "NVDRS" and was conducted in six databases: PubMed, EMBASE, Google Scholar, OVID, Scopus, and Web of Science. Search criteria included all the articles published from 2002 to 2018 in English.

Citation management
All citations were imported into the EndNote-PC software. Duplicate articles were manually excluded by sorting and reviewing individual article titles.

Analytical tool
Bibliographical data from studies meeting the inclusion criteria were analyzed with the software VOSviewer (ver. 1.6.10) (Van Eck and Watman 2016) and the online word cloud generator (Word Clouds, 2019). VOSviewer is a literature knowledge visualization software based on the Visualization of Similarities (VOS) technology. It displays knowledge domain maps and clusters based on network data. VOSviewer is well suited for analyzing large-scale bibliographical data and for constructing complex networks and interactions. Using bibliographic and text data, co-occurrence network, citation network and coupling network can be constructed (Van Eck and Waltman 2010). In this review, we used VOSviewer to map knowledge domains and display the development process and structural relationships of the scientific knowledge covered in the studies. Keywords in the same cluster show the greater similarity with respect to the highest node keyword and positioned within the same cluster. The size of the text within each circle of the  Year of Publication a network visualization map corresponds to the weight/ frequency of that keyword within and between the clusters. The software also provides a quantifiable framework through "Total Links Strength" and "Occurrences" analysis. Total links strength is calculated as: P Links to other keywords Total links . The higher the total links strength, the higher the co-occurrence of given keywords. "Occurrences" are the total frequency of a keyword within the data (Van Eck and Waltman, 2016).

Results
The comprehensive search returned a total of 1339 potentially relevant articles. Of the 1339 articles, 151 were indexed in PubMed, 164 in Embase, 500 in Google Scholar, 230 in OVID, 138 in Scopus, and 156 in Web of Science. After removing duplicates, 271 articles were included for further review. Of these, 69 did not meet the eligibility criteria due to non-related topics or articles that were thesis, books, commentaries, dissertations, and opinions, leaving 202 articles that were reviewed in full-text.
After excluding 52 full-text articles that contained retracted articles, did not use NVDRS or were abstracts only, 150 met the inclusion criteria and were included in the scoping review ( Fig. 1 and Table 1). Of the 150 articles, 132 were indexed in PubMed and 147 in Google Scholar.
There were no articles published between 2002 and 2004. The annual number of articles based on data from NVDRS increased from 1 in 2005 to 28 in 2018 (Fig. 2). Based on size and location of the words in the word cloud, the most frequently used words in the articles were homicide, suicide, NVRDS, and deaths (Fig. 3). Frequently used words and their interrelations are displayed in the density map (Fig. 4). Suicide was the most researched topic and was commonly linked to "circumstance," "alcohol" and "substance abuse." Knowledge domain mapping and network visualization revealed two major clusters and their interconnections (Fig. 5). The larger cluster was on suicide, with total links strength of 7902 and 529 occurrences (Table 2). Alcohol and substance abuse were most frequently examined risk factors associated with suicide. The second cluster was on homicide with total links strength of 3585 and 175 occurrences and was commonly linked to "firearm," "injury," and "gang" (Fig. 5 and Table 2).

Discussion
This scoping review provides a general assessment of the research utility of the NVDRS. The results indicate that research output as measured by the number of peer-reviewed publications has increased markedly in recent years. However, the majority of the authors and their publishing outlets are within the field of public health and medicine. The NVDRS is still underutilized by researchers in other disciplines, such as criminology, sociology, economics and policy. NVDRS-based studies are clustered into two knowledge domains; suicide and homicide. Knowledge domain mapping reveals that these studies covered a variety of interrelated topics. Given that suicide comprises the majority of violent deaths in the United States, it is not surprising that suicide is the most frequently researched topic covered in NVDRS-based studies. The term most commonly linked to studies on suicide is "circumstance," followed by "alcohol," and "substance abuse." Homicide is the second most frequently covered topic in NVDRS-based studies. The term most commonly linked to studies on homicide is "firearm", followed by "injury" and "gang." Our analysis also shows that studies clustered in the two knowledge domains (i.e., suicide and homicide) are densely interlinked to overlapping networks of keywords, such as "firearm," "mental health problem," "substance abuse," and "men." Although knowledge domain mapping sheds light on the development progress and structural relationship among NVDRS-based studies, this scoping review has several limitations. First, our search was limited to six major bibliographical databases for studies published in peer-reviewed journals. NVDRS-based studies published in the "gray" literature, such as theses, dissertations, and conference proceedings, are not included. Second, given the nature of the scoping review that focuses on general exploration of the research topic and its narrative presentation (Peterson et al. 2017), we did not evaluate the quality and findings of the individual studies. Finally, our review was restricted to assessing the research utility of the NVDRS. Use of the NVDRS for policy briefing, advocacy and other functions (Sundararaman et al. 2008) is not included in this review but could be as important as for knowledge creation.

Conclusion
Bibliographical analysis from this scoping review indicates that research output using the NVDRS has increased substantially since 2005. With the expansion of the NVDRS, this upward trend is likely to continue in the coming years. Results from knowledge domain mapping and manual review suggest that although NVDRS-based studies have covered a wide array of topics related to suicide and homicide, they are often focused on a few putative risk factors, such as alcohol and substance abuse for suicide and firearms and gang involvement for homicide. As more data are collected and become available, the vast potential of the NVDRS for violence research and prevention will be explored by investigators from a multitude of scientific disciplines.