The National Death Index as a Source of Homicide Data: A Methodological Exposition of Promises and Pitfalls for Criminologists

• Author: Richard G. Rogers,David C. Pyrooz,Ryan K. Masters,Jennifer J. Tostlebe
• DOI: 10.1177/1088767920924450
• Date: 01 February 2021
• Published date: 01 February 2021
• Subject Matter: Articles

https://doi.org/10.1177/1088767920924450

Homicide Studies

2021, Vol. 25(1) 5 –36

© 2020 SAGE Publications

Article reuse guidelines:

sagepub.com/journals-permissions

DOI: 10.1177/1088767920924450

journals.sagepub.com/home/hsx

Article

The National Death Index as

a Source of Homicide Data:

A Methodological Exposition

of Promises and Pitfalls for

Criminologists

Jennifer J. Tostlebe1, David C. Pyrooz1,

Richard G. Rogers1, and Ryan K. Masters1

Abstract

Criminologists largely rely on national deidentified data sources to study homicide

in the United States. The National Death Index (NDI), a comprehensive and well-

established database compiled by the National Center for Health Statistics, is an

untapped source of homicide data that offers identifiable linkages to other data

sources while retaining national coverage. This study’s five aims follow. First, we

review the data sources in articles published in Homicide Studies over the past decade.

Second, we describe the NDI, including its origins, procedures, and uses. Third, we

outline the procedures for linking a police gang intelligence database to the NDI.

Fourth, we introduce the St. Louis Gang Member-Linked Mortality Files database,

which is composed of 3,120 police-identified male gang members in the St. Louis area

linked to NDI records. Finally, we report on preliminary cause-of-death findings. We

conclude by outlining the benefits and drawbacks of the NDI as a source of homicide

data for criminologists.

Keywords

homicide, data linkage, National Death Index, gang membership, St. Louis, Missouri

Social scientists of all stripes recognize the value of quality instrumentation and mea-

surement. Criminologists are intimately familiar with these challenges as the key

1University of Colorado Boulder, USA

Corresponding Author:

Jennifer J. Tostlebe, Department of Sociology, University of Colorado Boulder, UCB 327, Boulder, CO

80309-0401, USA.

Email: jennifer.tostlebe@colorado.edu

924450HSXXXX10.1177/1088767920924450Homicide StudiesTostlebe et al.

research-article2020

6 Homicide Studies 25(1)

sources of data—official records and self-report surveys—used to examine the depen-

dent variables in the field are subject to a variety of well-documented limitations

(Eterno et al., 2016; Mosher et al., 2010; Piquero et al., 2014; Pyrooz et al., 2019;

Tcherni-Buzzeo, 2018). However, nearly all criminologists agree that homicide is the

most reliable and valid type of crime measured. Sherman and Langworthy (1979)

notably observed that the “difficulty of disposing of bodies, the generally high level of

agreement between the [UCR] and the [NVSS], and the monitoring function of coro-

ners in recording homicide events all support the view that official statistics provide a

highly accurate measure” (p. 546). Criminologists, in turn, rely primarily on two

sources of data to study homicide victimization in the United States: (a) the Federal

Bureau of Investigation’s (FBI) Supplementary Homicide Reports (SHR), which are

part of the Uniform Crime Reporting (UCR) program; and (b) the National Center for

Health Statistics’ (NCHS) National Vital Statistics System (NVSS; Regoeczi & Banks,

2014). Of course, there are a variety of other national homicide data sources on spe-

cialized populations (e.g., police officers) or regional data sources covering general

populations (e.g., Chicago Homicide Dataset; Los Angeles Police Department/FBI

Homicide Library).

While these data sources have yielded tremendous knowledge about the extent and

nature of homicide, the current inventory of homicide victimization data is limited in

two ways. First, and most critically, leading homicide data sources—particularly

national-level data—are unable to be linked to populations of criminological interest

at the individual-level. The SHR and NVSS, for example, are used primarily by crimi-

nologists to examine national trends and cross-jurisdictional variation in homicide or

to study the features of homicide events or correlates vis-à-vis other causes of death

(Baumer et al., 2018; Cooper & Smith, 2011; Gaston et al., 2019; Kochanek et al.,

2019; McCall et al., 2010; Rosenfeld & Fox, 2019). However, these aggregate and/or

deidentified data sources are unable to be individually linked to other sources of data.

A consequence of unlinked data is that it narrows the scope of research owing to the

limited demographic information it provides on victims. In contrast, linked data could

yield richer information about victim social networks, exposure to violence, criminal

propensities, personal experiences, and physical or mental health. The second problem

applies primarily to regional sources of homicide data—cities, counties, or states—

that may permit linking, thus capable of generating comprehensive linked mortality

files, but suffer in scope because mortality surveillance coverage is limited to the

respective jurisdiction. It is unclear whether regional data sources fully enumerate

homicide for the population subject to the ascertainment of vital status.

In this article, we introduce the National Death Index (NDI) as a source of homicide

victimization data. The NDI is a restricted-access computerized database maintained

by the NCHS. It is an annual compilation of individual-identified death information,

derived from death certificates provided by state vital registrars. The collaboration

motivating this article involved criminologists and social demographers who aimed to

uncover mortality risk among gang members. Prior research on the topic relied primar-

ily on word-of-mouth, online searches, or local law enforcement or coroner records for

the ascertainment of vital status (Decker & Pyrooz, 2010a; Decker & Van Winkle,

Tostlebe et al. 7

1996; Hagedorn & Macon, 1998; Kennedy et al., 1996; Levitt & Venkatesh, 2001;

Loeber et al., 2011; Loeber & Farrington, 2011), raising serious questions about the

complete enumeration of homicide specifically and mortality generally.

We provide a methodological exposition of the procedures used to link NDI records

to individuals listed in a police gang database. Our aims are as follows. First, we

review the sources of homicide victimization data that have appeared in Homicide

Studies over the past decade to illustrate the need for a data source such as the NDI.

Second, we provide an overview of the NDI by detailing its origins, procedures, and

uses. We also distinguish the NDI from the ways in which criminologists have used the

NVSS in the past, as well as other sources of homicide data. Third, we introduce the

project motivating our use of the NDI. Our purpose is to examine all-cause and cause-

specific mortality risk among people identified by law enforcement as gang members

in St. Louis, Missouri. In this study, we sought to ascertain the vital status of 3,164

males who were included in a St. Louis area police gang database between 1993 and

2003. We detail our procedures to create St. Louis Gang Member-Linked Mortality

Files (STL GM-LMFs) across four stages: (a) review of GM data; (b) submission of

the GM data to the NCHS and receipt of NDI records; (c) supplemental verification of

NDI records; and (d) merging the respective data sources to create the STL GM-LMFs.

Finally, we describe the promises and pitfalls of our efforts to generate the STL

GM-LMFs and conclude by discussing the multiple uses of the NDI for criminological

research.

Sources of U.S. Homicide Victimization Data: A Look at

the Last Decade of Homicide Studies

To determine the data sources researchers use to study homicide victimization in the

United States, we gathered information from empirical studies published in Homicide

Studies between February 2010 and November 2019 (40 issues in total).1 Table 1 rank

orders data sources by the frequency with which they are used to study homicide vic-

timization in published works.2 Below, we briefly review the two national-level sys-

tems most often used to provide information about homicides in the United States

(Regoeczi & Banks, 2014).

The Uniform Crime Reporting (UCR) program was established by the FBI in 1930

to track crime statistics in the United States. The UCR is a compilation of individual

law enforcement agency reports that are submitted by non-federal law enforcement

agencies to the FBI either through a state UCR program or directly to the FBI’s UCR

program on a monthly basis. The data are publicly available and provide nationwide

information on crime in nearly 18,000 jurisdictions at the local, state, and national

level. SHR are part of the UCR program and were developed in the early-1960s to

collect and provide detailed information about homicide incidents. The methodology

for the UCR and SHR are well established, and these data are considered valid counts

of homicide (Riedel, 1999). The SHR data are nationally representative and provide

incident-level and demographic information on homicide victims, known homicide