Firearm-Related Deaths in Multnomah County,Oregon,2010-2016: Linking Medical Examiner Data to State Vital Records Data

Abstract

Objectives:

Violence due to firearms is a substantial public health problem. Death data from medical examiner and vital records were linked to evaluate the use of medical examiner data to augment routine surveillance and determine any differences in sex, age, manner of death, or race and ethnicity between the 2 data systems.

Materials and Methods:

Medical examiner data were searched for keywords of interest, and vital records data were obtained and linked for deaths occurring in Multnomah County, Oregon, from January 1, 2010, through December 31, 2016. Both data sets were compared for the number and proportion of firearm-related deaths by sex, age, manner of death, and race/ethnicity. Sensitivity and positive predictive values were calculated for variables that had discordant results.

Results:

A total of 568 firearm-related deaths were identified in the medical examiner data. After matching with manual review, the 2 data systems had 100% case agreement. A reverse match showed that most cases not found in medical examiner data were due to transfer of case jurisdiction. The 2 systems matched nearly perfectly in sex, age, and manner of death but differed in characterization of race and ethnicity. Sensitivity was 62% for Hispanic ethnicity but 93% for white and black race.

Practice Implications:

Using medical examiner data was a useful way to augment routine surveillance of firearm-related deaths in our jurisdiction in close to real time. However, caution is needed when analyzing data by subgroups because of discordant classifications of race between the data systems.

Keywords

firearms surveillance death

Violence due to firearms is a substantial public health problem. In 2017, a total of 39 773 firearm-related deaths occurred in the United States, which represented 16% of all injury deaths that year. From 2016 to 2017, the age-adjusted firearm-related death rate in the total population rose significantly, from 11.8 per 100 000 population in 2016 to 12.0 per 100 000 population in 2017.^1,2 For US children aged 1-17 years, suicide was the second leading cause of death in 2017, with firearms accounting for 41% of those deaths. Furthermore, international studies have shown that more than 90% of firearm-related deaths in children aged ≤14 years occur in the United States.^2,3 In Oregon, the age-adjusted rate of firearm-related deaths per 100 000 population rose from 10.4 in 2009 to 12.1 in 2017, mainly through an increase in suicides.^4,5

One way to address violence, including firearm-related violence, is through a public health approach. Such an approach uses a multifactorial process to identify the magnitude of the problem, identify risks and protective strategies, develop and test the strategies, and ensure widespread adoption of the strategies.⁶ Many of these processes overlap with the 10 essential public health services, including the monitoring of data to inform program development and policy.⁷ Local public health authorities can contribute to this process by analyzing cause-of-death data, such as data provided by state vital statistics programs. These data (provided by 57 vital statistics programs in the United States and its territories) are a stable and comprehensive source of mortality data in the United States.⁸ However, a delay in reporting (about 1 year) can be challenging for local public health purposes. Other potential data sources include the National Violent Death Reporting System (NVDRS); these data cover 40 states (including Oregon) and are available through the Web-based Injury Statistics Query and Reporting System.⁹ This system also includes data on nonfatal injuries, occurrence, and other factors. However, this system has a reporting delay of 1.5 years. More unique and timely data sources are needed to fully implement a public health approach to firearm-related violence.

Medical examiner data are a potential timely source of information on violent deaths. In Oregon, these deaths are automatically investigated by the Office of the State Medical Examiner (Oregon Revised Statutes 146.090).¹⁰ The main purpose of the medical examiner is to conduct a medicolegal investigation of the circumstances of death and to certify the cause and manner of death. Medical examiner data are used to augment the surveillance of opioid overdose deaths in Multnomah County, Oregon,¹¹ and provide as near to real-time updates on overdose deaths as possible.¹² Other jurisdictions have used medical examiner data to augment their surveillance of poisoning/overdose deaths,^13

-20 traffic injury/deaths,²¹ drownings,²² and other conditions of public health importance (eg, trauma). Few studies have addressed the use of medical examiner data specifically (ie, not as part of NVDRS) to augment the surveillance of firearm-related violence, especially at the local level. Oklahoma compared its medical examiner data with data from its state vital records programs and found that each system augmented the other.²³ However, Oklahoma did not compare demographic characteristics between the 2 systems. Connecticut used medical examiner data on firearm type to compare guns used in violent crimes with guns collected at buyback events.²⁴ The dearth of research on gun violence using medical examiner data may reflect a scarcity of research at the national level.^25,26

The objectives of the current study were to evaluate the use of medical examiner data for firearm surveillance by (1) describing firearm-related deaths in the Multnomah County medical examiner data, (2) linking medical examiner data to vital records to assess the completeness of data obtained from the medical examiner, and (3) comparing the matched medical examiner–vital records data sets on sex, age, and racial/ethnic classification by calculating sensitivity and positive predictive value (PPV) for demographic characteristics that differed.

Methods

Data from the medical examiner were extracted by selecting deaths occurring from January 1, 2010, through December 31, 2016, for all jurisdictions in Oregon. Data were downloaded from a secure database and saved as text files (unpublished data). To find deaths with Multnomah jurisdiction, any instance of Multnomah County in the county variable field was retained. Occasionally, cases had multiple jurisdictions associated with them; in these instances, data were manually reviewed to determine the true jurisdiction. Keyword searches for the words firearm and gun were conducted by using text-search functions in 4 fields: cause of death, second cause of death, how the injury occurred, and other significant findings. Cases were manually reviewed to exclude misspelling errors (eg, “exsangunation” instead of “exsanguination”). Two deaths that had the keyword of interest but had more than 1 year between injury and death were excluded.

Death data were obtained from Oregon vital records from a secure county structured query language server, where the data are hosted and updated weekly (unpublished data). The data include deaths of residents and nonresidents of Multnomah County, because the medical examiner investigates all deaths occurring in Multnomah County, regardless of the decedent’s residence. Variables needed for matching (full name, date of birth, date of death) were retained, along with cause-of-death information coded in the International Classification of Diseases, 10th Revision (ICD-10) codes.²⁷

To ensure completeness and validity of data, the medical examiner data were matched to the Oregon vital records data, which ensured that all deaths investigated by the medical examiner would also be captured in the state data after the reporting lag. A matched data set was created by matching on first name, middle name or initial, last name, date of birth, and date of death. A manual review of all nonmatches was conducted to examine discrepancies or name/date transpositions or inversions. Matching was performed by using the SAS-callable program Link King, which is a combination deterministic/probabilistic matching program.²⁸ The recommended mapping protocol and default probabilistic weight settings were used, but the blocking level was set too high to ensure maximum linkages were found. In the medical examiner data, race and ethnicity were populated as 1 variable, whereas the vital record had multiple racial and ethnic variables. In the analysis for the third objective (comparing the matched medical examiner–vital records data sets), the vital records “bridged race” variable was used, which assigns multiple-race decedents to a single race category for calculation of race-specific estimates. For Hispanic ethnicity, 4 variables were combined (Mexican, Puerto Rican, Cuban, or other Hispanic). The “other” racial category in both systems can include multiracial persons. To calculate sensitivity and PPV, the vital records coding was used as the gold standard. Sensitivity was defined as the proportion of “true” events identified by the medical examiner data, and PPV was defined as the proportion of events identified by the medical examiner data that were “true” events. Sensitivity and PPV were not calculated for the “other” category because no gold standard was defined.

To further assess the completeness of the medical examiner data, vital records data were linked backwards to medical examiner data. A commonly used combination of ICD-10 codes related to firearm mortality (W32-W34, X72-X74, X93-X95, Y22-Y24, Y350) was applied.⁸ All analyses were conducted by using SAS version 9.4.²⁹ Because the work presented here was an evaluation of surveillance data that are routinely collected and inform our public health practice, institutional review board approval was not sought.

Results

Firearm-Related Deaths in Medical Examiner Data

Overall, firearm-related deaths accounted for 9.2% (568 of 6386) of deaths investigated by the medical examiner (Figure). Most deaths occurred among males (n = 492, 86.6%), among persons aged 25-44 (n = 200, 35.2%), and by suicide (n = 404, 71.1%). Males accounted for 88.0% (357 of 404) of all suicide deaths and 81.3% (126 of 155) of all homicide deaths. White decedents accounted for 440 (77.5%) and black decedents accounted for 67 (11.8%) firearm-related deaths.

Figure.

Process flow diagram to create matched medical examiner and vital records data set, Multnomah County, Oregon, 2010-2016. The medical examiner deaths were deaths that occurred in Multnomah County and were investigated by the Office of the State Medical Examiner as written in Oregon Revised Statutes Chapter 146.¹⁰ Oregon vital records deaths were all persons who lived or died in Multnomah County.

Assessing the Completeness of Medical Examiner Data

The vital records data identified 55 958 deaths occurring from January 1, 2010, through December 31, 2016, where residence or place of death was Multnomah County; 42 870 (76.6%) deaths were among residents. After the first match of medical examiner records and vital records, 542 of 568 (95.4%) medical examiner records matched to vital records (Figure). After manual review, minor corrections, and a second match, all records matched, resulting in a final data set with 568 records. With the exception of race/ethnicity data, the data between the 2 data sets were virtually identical. A total of 435 (76.6%) observations were among Multnomah County residents, and this number constituted the final number of observations for the matched data set.

Application of commonly used ICD-10 codes revealed 19 deaths that were not in the original medical examiner data set. A review of these cases found that 18 deaths were omitted from the medical examiner data set because death occurred in a Multnomah County hospital but another county medical examiner office assumed jurisdiction. The other 1 death was omitted because of an error in the medical examiner database, in which the cause of death was left blank.

Sex, Age, and Racial/Ethnic Classification of Matched Medical Examiner–Vital Records Data Sets

In the matched data set, each source was virtually identical with respect to sex, age, and manner of death (Table 1). As such, sensitivity and PPV were not calculated for these variables. In contrast, as with the separate medical examiner and vital records results, the coding of race and ethnicity in the matched data set varied depending on the data source. The percentage of persons coded as white and Hispanic was higher in the vital records data than in the medical examiner data, whereas the percentage of persons coded as black was higher in the medical examiner data than in the vital records data. Using the vital record as a gold standard, the sensitivity of white and black was 93.2% and of Hispanic was 61.9% (Table 2).

Table 1.

Demographic characteristics of persons who died by firearms, by data source, Multnomah County, Oregon, 2010-2016

Characteristic	Matched Medical Examiner and Vital Records		Limited Matched Medical Examiner and Vital Records^a
Characteristic	Medical Examiner, No. (%)	Vital Records, No. (%)	Medical Examiner, No. (%)	Vital Records, No. (%)
Sex
Male	492 (86.6)	490 (86.3)	372 (85.5)	371 (85.3)
Female	76 (13.4)	78 (13.7)	63 (14.5)	64 (14.7)
Age, y
≤24	95 (16.7)	95 (16.7)	63 (14.5)	63 (14.5)
25-44	200 (35.2)	200 (35.2)	151 (34.7)	151 (34.7)
45-64	189 (33.3)	189 (33.3)	149 (34.3)	149 (34.3)
≥65	84 (14.8)	84 (14.8)	72 (16.6)	72 (16.6)
Manner of death
Suicide	404 (71.1)	403 (71.0)	315 (72.4)	315 (72.4)
Homicide	155 (27.3)	155 (27.3)	117 (26.9)	117 (26.9)
Undetermined/accidental	9 (1.6)	10 (1.8)	3 (0.7)	3 (0.7)
Race/ethnicity^b
White	440 (77.5)	460 (81.0)	334 (76.8)	344 (79.1)
Black	67 (11.8)	63 (11.1)	60 (13.8)	55 (12.6)
Hispanic	20 (3.5)	32 (5.6)	15 (3.4)	21 (4.8)
Other^c	23 (4.0)	45 (7.9)	17 (3.9)	35 (8.0)
Missing	18^d (3.2)	1 (0.2)	9 (2.1)	1 (0.2)
Total	568 (100.0)	568 (100.0)	435 (100.0)	435 (100.0)

^a Matched data set limited to Multnomah County residents.

^b Race and ethnicity are treated as 1 variable in medical examiner data but as separate variables in vital statistics data. As such, numbers may add up to more than the total.

^c Other includes >1 racial category or Asian/Pacific Islander, Native American/Alaska Native, or other.

^d As determined by medical examiner; some records had missing data on race.

Table 2.

Sensitivity and positive predictive value of race and ethnicity^a coding in medical examiner data of residents among matched medical examiner and vital records data (n = 435), Multnomah County, Oregon, 2010-2016^b

Variable	Vital Records (Gold Standard)
	Black		Hispanic		White
	Yes	No	Yes	No	Yes	No
Medical examiner data
Yes	57	3	13	2	329	5
No	4	371	8	412	24	77
Sensitivity, %	93.4		61.9		93.2
Positive predictive value, %	95.0		86.7		98.5

^a Race and ethnicity are treated as 1 variable in medical examiner data but as separate variables in vital statistics data.

^b Matched data limited to Multnomah County residents.

Given the lower sensitivity and PPV for black and Hispanic coding in the medical examiner data using vital records as a gold standard, the rates across systems for these groups varied. For black decedents, the rate per 100 000 population could be as high as 16.3 (vital records) or as low as 14.9 (medical examiner); for Hispanic decedents, the rate could be as low as 2.5 (medical examiner) or as high as 3.5 (vital records; Table 3).

Table 3.

Crude death rates of residents, by data source–specific racial/ethnic classification in matched medical examiner–vital statistics data, Multnomah County, Oregon, 2010-2016

Race/Ethnicity^a	No.		Crude Rate per 100 000 Population
Race/Ethnicity^a	Vital Records	Medical Examiner	Vital Records	Medical Examiner
Black	55	60	14.9	16.3
Hispanic	21	15	3.5	2.5
White	344	334	7.7	7.5

^a Race and ethnicity are treated as 1 variable in medical examiner data but as separate variables in vital statistics data.

Discussion

This study demonstrates the practicality of using medical examiner data to augment other data sources for violent deaths (eg, vital records or NVDRS). Although the study period ended with deaths occurring on December 31, 2016, so that deaths could be matched to the finalized vital records data, data from the medical examiner can be obtained as soon as case reports are entered into the system. Given that the initial match rate between the 2 systems was 95%, it appears the medical examiner data have high usefulness for investigating firearm-related deaths in close to real time. Timely analysis of medical examiner data could supplement state reporting and inform local action, for example, in suicide prevention messaging or health care provider alerts.⁵ It could also be important to identify seasonal trends, given the epidemiologic link between heat and mortality³⁰ as well as heat and crime.³¹

Differences in racial and ethnic classification were found in the 2 sources of death data. This inconsistency is not surprising given the differing purposes of the 2 data sources. The current study found that the sensitivity of the medical examiner data was 69% to classify decedents of Hispanic ethnicity but was >90% to classify white or black decedents. Sensitivity was calculated by using the vital records coding as the gold standard. However, misclassification of race and ethnicity in death certificate data is a known problem, especially for persons of American Indian and Alaska Native ancestry. Death certificate data are thought to provide “reasonably good” reporting for Hispanic decedents and better reporting for black or white decedents.^32,33 Careful consideration must be taken if using the medical examiner data to summarize data or produce rates by race, because this analysis shows a low sensitivity for Hispanic decedents. Overall, the number misclassified (compared with vital records) was low for any particular year but could be magnified when data are combined. Given known disparities in the risk of violent death by race, this misclassification could be an important consideration in the use of medical examiner data.^5,9,34,35

A 2005 study by Comstock et al²³ calculated both a sensitivity and a PPV for their state’s medical examiner and vital records systems. This calculation required the authors to define a gold standard study definition for a violent death. They found that the sensitivity of Oklahoma’s medical examiner system was >99% for all violent deaths. The present investigation did not calculate the sensitivity or PPV for the system in general, because the main purpose of the analysis was to link the data systems, examine the matching rate, and describe differences in demographic characteristics across the 2 systems. Future analyses could calculate these values for both data systems. However, after a manual review, the nearly complete match rate between the Multnomah County medical examiner data and the vital records data indicates that the medical examiner data have good sensitivity in identifying true cases of firearm-related deaths.

Limitations

This evaluation had several limitations. First, text searches were used for keywords from 4 variables in the medical examiner data but not from the narrative for the incident. The narrative is a detailed, rich source of information. It is possible that the keyword searches used in this study excluded cases of firearm-related deaths that might have been apparent in the narrative. However, the reverse match from vital records to medical examiner data showed that few cases would have been excluded by using the keywords of choice; as such, this limitation likely did not affect the study findings. Future use of the narrative data should be explored. Second, not all states have a medical examiner data system. Oregon is 1 of 16 states with a centralized medical examiner system.³⁶ It is possible that the study findings are not generalizable to other states or to the United States as a whole. Third, race and ethnicity variables are separate in the vital records data and combined in the medical examiner data. Because of small numbers in categories other than white, black, or Hispanic, more detailed information about “other” groups could not be calculated. It is possible that a more careful review of the narrative in the medical examiner data would have provided more detailed data on race or ethnicity, especially because data on race and ethnicity were missing during this study period. Finally, use of the vital records as the gold standard could be questionable, given known misclassification in death certificate coding for race and ethnicity. However, the most recent analysis by the National Center for Health Statistics showed a drop from 5% to 3% in the misclassification of Hispanic ethnicity.³²

Conclusion

Using medical examiner data to augment other sources of death data could provide valuable, close-to-real-time information on firearm-related deaths. The medical examiner–vital records linkage analysis demonstrates that true cases of firearm-related violence are not likely missing and that cases not captured in the medical examiner data are more likely caused by jurisdictional issues in the creation of the data set than by other sources of misclassification. Given known disparities in the rates of firearm-related deaths by race, timeliness should be considered in the context of accuracy when reporting from the medical examiner data. Adoption of a nationwide medical examiner coding system could help address differences between jurisdictions, for race and ethnicity and other variables.²³ Data can help describe and understand the epidemiology of firearm-related deaths, but work remains to identify other risk factors.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Jaime Walters, MPH

References

Kochanek

Murphy

Arias

. Deaths: final data for 2017. Natl Vital Stat Rep. 2019;68(9):1–77.

National Center for Injury Prevention and Control, Centers for Disease Control and Prevention. Web-based Injury Statistics Query and Reporting System (WISQARS). https://www.cdc.gov/injury/wisqars/index.html. Published 2003. Accessed April 17, 2019.

Fowler

Dahlberg

Haileyesus

Gutierrez

Bacon

. Childhood firearm injuries in the United States. Pediatrics. 2017;140(1):e20163486. doi:10.1542/ped.2016-3486

Oregon Health Authority. A new approach to preventing firearm deaths. CD Summary. 2017;66(13):1–3. http://www.oregon.gov/oha/PH/DISEASESCONDITIONS/COMMUNICABLEDISEASE/CDSUMMARYNEWSLETTER/Documents/2017/ohd6613.pdf. Accessed January 23, 2018.

Centers for Disease Control and Prevention. Underlying cause of death, 1999-2017. http://wonder.cdc.gov/ucd-icd10.html. Accessed September 11, 2018.

Centers for Disease Control and Prevention. Violence prevention. https://www.cdc.gov/violenceprevention/overview/publichealthapproach.html. Accessed January 5, 2018.

Centers for Disease Control and Prevention. The public health system. https://www.cdc.gov/stltpublichealth/publichealthservices/essentialhealthservices.html. Published 2018. Accessed January 5, 2018.

Murphy

Kochanek

Curtin

Arias

. Deaths: final data for 2015. Natl Vital Stat Rep. 2017;66(6):1–75.

Centers for Disease Control and Prevention. Violence prevention: National Violent Death Report System. https://www.cdc.gov/violenceprevention/datasources/nvdrs/index.html. Accessed February 16, 2018.

10.

Ore Rev Stat, §146.090 (2017).

11.

Multnomah County Health Department. Tri-county region opioid trends: Clackamas, Multnomah, and Washington, Oregon, 2016. https://multco.us/file/59894/download . Published 2017. Accessed September 10, 2018.

12.

Oregon Pain Guidance. Real-time opioid overdose monitoring. https://portlandprofessional.oregonpainguidance.org. Published 2018. Accessed September 10, 2018.

13.

Landen

Castle

Nolte

, et al. Methodological issues in the surveillance of poisoning, illicit drug overdose, and heroin overdose deaths in New Mexico. Am J Epidemiol. 2003;157(3):273–278. doi:10.1093/aje/kwf196

14.

Shah

Lathrop

Reichard

Landen

. Unintentional drug overdose death trends in New Mexico, USA, 1990-2005: combinations of heroin, cocaine, prescription opioids and alcohol. Addiction. 2008;103(1):126–136. doi:10.1111/j.1360-0443.2007.02054.x

15.

Soslow

Woolf

. Reliability of data sources for poisoning deaths in Massachusetts. Am J Emerg Med. 1992;10(2):124–127. doi:10.1016/0735-6757(92)90043-w

16.

Hurstak

Rowe

Turner

, et al. Using medical examiner case narratives to improve opioid overdose surveillance. Int J Drug Policy. 2018;54:35–42. doi:10.1016/j.drugpo.2017.12.017

17.

Hudson

Klekamp

Matthews

. Local public health surveillance of heroin-related morbidity and mortality, Orange County, Florida, 2010-2014. Public Health Rep. 2017;132(1 suppl):80S–87S. doi:10.1177/0033354917709783

18.

Moore

Weber

Cina

Aks

. Syndrome surveillance of fentanyl-laced heroin outbreaks: utilization of EMS, medical examiner and poison center databases. Am J Emerg Med. 2017;35(11):1706–1708. doi:10.1016/j.ajem.2017.05.003

19.

Hargrove

Bunn

Slavova

, et al. Establishment of a comprehensive drug overdose fatality surveillance system in Kentucky to inform drug overdose prevention policies, interventions and best practices. Inj Prev. 2018;24(1):60–67. doi:10.1136/injuryprev-2016-042308

20.

Jiang

McDonald

Wilson

, et al. Rhode Island unintentional drug overdose death trends and ranking—Office of the State Medical Examiners database. R I Med J (2013). 2018;101(1):33–36.

21.

Zonfrillo

Ramsay

Fennell

Andreasen

. Unintentional non-traffic injury and fatal events: threats to children in and around vehicles. Traffic Inj Prev. 2018;19(2):184–188. doi:10.1080/15389588.2017.1369053

22.

Browne

Lewis-Michl

Stark

. Investigation and reporting practices for drownings: implications for injury prevention research in New York State. Am J Forensic Med Pathol. 2002;23(4):398–401. doi:10.1097/00000433-200212000-00021

23.

Comstock

Mallonee

Jordan

. A comparison of two surveillance systems for deaths related to violent injury. Inj Prev. 2005;11(1):58–63. doi:10.1136/ip.2004.007567

24.

Baumann

Clinton

Berntsson

, et al. Suicide, guns, and buyback programs: an epidemiologic analysis of firearm-related deaths in Connecticut. J Trauma Acute Care Surg. 2017;83(6):1195–1199. doi:10.1097/TA.0000000000001575

25.

Kellermann

Rivara

. Silencing the science on gun research. JAMA. 2013;309(6):549–550. doi:10.1001/jama.2012.208207

26.

Rubin

. Tale of 2 agencies: CDC avoids gun violence research but NIH funds it. JAMA. 2016;315(16):1689–1691. doi:10.1001/jama.2016.1707

27.

World Health Organization. International Statistical Classification of Diseases and Related Health Problems. 10th ed, 2nd rev. https://apps.who.int/iris/handle/10665/42980. Accessed August 28, 2019.

28.

Campbell

. Rule your data with the Link King (a SAS/AF application for record linkage and unduplication). https://support.sas.com/resources/papers/proceedings/proceedings/sugi30/020-30.pdf. Published 2005. Accessed August 28, 2019.

29.

SAS [statistical software]. Version 9.4. Cary, NC: SAS Institute Inc; 2013.

30.

Basu

. High ambient temperature and mortality: a review of epidemiologic studies from 2001 to 2008. Environ Health. 2009;8:40. doi:10.1186/1476-069X-8-40

31.

Schinasi

Hamra

. A time series analysis of associations between daily temperature and crime events in Philadelphia, Pennsylvania. J Urban Health. 2017;94(6):892–900. doi:10.1007/s11524-017-081-y

32.

Arias

Schauman

Eschbach

Sorlie

Backlund

. The validity of race and Hispanic origin reporting on death certificates in the United States. Vital Health Stat 2. 2008;(148):1–23.

33.

Arias

Heron

; National Center for Health Statistics, Hakes

; US Census Bureau. The validity of race and Hispanic-origin reporting on death certificates in the United States: an update. Vital Health Stat 2. 2016;(172):1–21.

34.

Kalesan

French

Fagan

Fowler

Galea

. Firearm-related hospitalizations and in-hospital mortality in the United States, 2000-2010. Am J Epidemiol. 2014;179(3):303–312. doi:10.1093/aje/kwt255

35.

Kalesan

Vasan

Mobily

, et al. State-specific, racial and ethnic heterogeneity in trends of firearm-related fatality rates in the USA from 2000 to 2010. BMJ Open. 2014;4(9):e005628. doi:10.1136/bmjopen-2014-005628

36.

Centers for Disease Control and Prevention. Public health professionals gateway: public health law. https://www.cdc.gov/phlp/publications/coroner/death.html. Accessed March 1, 2018.