Customize your JAMA Network experience by selecting one or more topics from the list below.
Anderson M, Kaufman J, Simon TR, et al. School-Associated Violent Deaths in the United States, 1994-1999. JAMA. 2001;286(21):2695–2702. doi:10.1001/jama.286.21.2695
Author Affiliations: Division of Violence Prevention (Drs Anderson, Simon, Paulozzi, and Hammond) and Office of Statistics and Programming (Dr Ryan), National Center for Injury Prevention and Control, and Division of Adolescent and School Health, National Center for Chronic Disease Prevention and Health Promotion (Dr Barrios), Centers for Disease Control and Prevention, Atlanta, Ga; Department of Sociology, University of Miami, Coral Gables, Fla (Dr Kaufman); Safe and Drug Free Schools Program, US Department of Education, Washington, DC (Mr Modzeleski); National Institute of Justice, US Department of Justice, Washington, DC (Dr Feucht); and Education Development Center, Inc, Newton, Mass (Dr Potter).
Context Despite the public alarm following a series of high-profile school shootings
that occurred in the United States during the late 1990s, little is known
about the actual incidence and characteristics of school-associated violent
Objective To describe recent trends and features of school-associated violent
deaths in the United States.
Design, Setting, and Subjects Population-based surveillance study of data collected from media databases,
state and local agencies, and police and school officials for July 1, 1994,
through June 30, 1999. A case was defined as a homicide, suicide, legal intervention,
or unintentional firearm-related death of a student or nonstudent in which
the fatal injury occurred (1) on the campus of a public or private elementary
or secondary school, (2) while the victim was on the way to or from such a
school, or (3) while the victim was attending or traveling to or from an official
Main Outcome Measures National estimates of risk of school-associated violent death; national
trends in school-associated violent deaths; common features of these events;
and potential risk factors for perpetration and victimization.
Results Between 1994 and 1999, 220 events resulting in 253 deaths were identified;
202 events involved 1 death and 18 involved multiple deaths (median, 2 deaths
per multiple-victim event). Of the 220 events, 172 were homicides, 30 were
suicides, 11 were homicide-suicides, 5 were legal intervention deaths, and
2 were unintentional firearm-related deaths. Students accounted for 172 (68.0%)
of these deaths, resulting in an estimated average annual incidence of 0.068
per 100 000 students. Between 1992 and 1999, the rate of single-victim
student homicides decreased significantly (P = .03);
however, homicide rates for students killed in multiple-victim events increased
(P = .047). Most events occurred around the start
of the school day, the lunch period, or the end of the school day. For 120
(54.5%) of the incidents, respondents reported that a note, threat, or other
action potentially indicating risk for violence occurred prior to the event.
Homicide offenders were more likely than homicide victims to have expressed
some form of suicidal behavior prior to the event (odds ratio [OR], 6.96;
95% confidence interval [CI], 1.96-24.65) and been bullied by their peers
(OR, 2.57; 95% CI, 1.12-5.92).
Conclusions Although school-associated violent deaths remain rare events, they have
occurred often enough to allow for the detection of patterns and the identification
of potential risk factors. This information may help schools respond to this
In the latter half of the 1990s, the United States experienced a series
of high-profile school shootings that generated considerable media attention
and public alarm. Despite the intense interest, relatively little information
has been collected on violent deaths associated with schools. These deaths
are not routinely reported to state or federal agencies and cannot be identified
using traditional public health or criminal justice data sources.1 In 1996, the Centers for Disease Control and Prevention
(CDC), in conjunction with the US Departments of Education and Justice, published
the first systematic review of school-associated violent deaths, covering
2 academic years (1992-1993 and 1993-1994).1
However, many highly publicized incidents of school violence have occurred
More recent reports have examined violent deaths associated with schools,
including the highly publicized events mentioned, but these reports were restricted
by either the limited data collected on each event9
or small, nonrandom samples.10-15
The current study is an extension and expansion of the prior CDC study.
To our knowledge, it is also the only systematic investigation of all recent
school-associated violent deaths in the United States between 1994 and 1999.
The study was undertaken in collaboration with the US Departments of Education
and Justice with the following objectives in mind: (1) to systematically collect
information on all identified school-associated violent deaths, (2) to provide
a national estimate of risk for school-associated violent death, (3) to assess
national trends in school-associated violent deaths from 1992 through 1999,
(4) to identify common features of these events, and (5) to describe potential
risk factors for perpetration and victimization.
A school-associated violent death was defined as a homicide, suicide,
legal intervention (victim killed by police officer in the line of duty),
or unintentional firearm-related death in which the fatal injury occurred
between July 1, 1994, and June 30, 1999, in one of the following locations:
(1) on the campus of a functioning public or private elementary or secondary
school in the United States, (2) while the victim was on the way to or from
regular sessions at such a school, or (3) while the victim was attending or
traveling to or from an official school-sponsored event.1
Cases included the deaths of both students and nonstudents (faculty, school
staff, family members, and community residents).
We identified cases of school-associated violent death using 2 case-finding
strategies. The first method involved a systematic search of 2 computerized
newspaper and broadcast media databases (Lexis-Nexis and Dialog).16,17 The second method, which has been
used by the US Department of Education and the National School Safety Center9 since 1992, made use of a newspaper clipping service
and voluntary reports from state and local education agencies. The 2 methods
generated more than 18 000 articles, which were then reviewed. The review
identified 424 events with potential cases. After collecting additional media
reports on the 424 events, we were able to exclude 179 events because the
deaths involved were not associated with a functioning elementary or secondary
school in the United States (Figure 1).
We then contacted at least 1 law enforcement or school official familiar
with each of the remaining 245 events. This process disqualified 25 of the
245 events for various reasons (Figure 1).
A total of 220 events involving 253 cases were confirmed.
Once events were confirmed, we obtained data directly from 2 official
sources: the police report and/or a structured telephone interview with a
police officer who investigated the event, and a structured telephone interview
with the school principal or another knowledgeable school official. These
sources provided detailed information about the victims and alleged perpetrators,
the school associated with each death, and the circumstances of the fatal
injuries. For the subset of student victims and perpetrators, these sources
provided additional information on the students' criminal, psychological,
family, victimization, and school histories.
Several steps were taken to ensure the reliability and validity of the
data. First, 2 researchers worked independently and applied preestablished
coding criteria to abstract information from police reports, with 93.6% interrater
agreement. To avoid data entry errors, the same 2 researchers independently
entered coded data from all interviews and reports into 2 separate databases.
Finally, using an algorithm to reconcile discrepant responses, we combined
the school and police datasets to create a final working dataset. Specifically,
for school-related variables, we used responses provided by school officials
over those given by the police. For variables related to criminal or law enforcement
issues, we used the police response over the school response. In addition,
for variables where both the school and police officials could provide valid
information, we used respondents' assessments of their own degree of knowledge
of the event to determine which source was likely to be more accurate.
School-level data (eg, urbanicity, school type, and school size) from
the US Department of Education's Common Core of Data18
and Private School Universe Survey19 were added
to the merged dataset.
To facilitate comparisons between perpetrators and victims, we created
dichotomous variables for relevant characteristics, with responses separated
into "characteristic is present" ("yes" responses) vs "characteristic not
known to be present" ("no" or "unknown" responses).
Rates were calculated to estimate the risk of student school-associated
violent death. Denominators for the rate estimates were obtained from the
US Department of Education, which provided national school enrollment figures
for the 1994-1995 to 1998-1999 academic years broken down by sex, race/ethnicity,
grade level, and type of community.18,19
We also used mortality data compiled by the National Center for Health Statistics
for the period July 1, 1994, through June 30, 1999, to estimate the proportion
of violent deaths among all school-aged children (age 5-18 years) that were
Trends in school-associated violent death event rates and school-associated
student homicide rates were assessed using Poisson regression models with
a systematic component incorporating year as a linear term. Because students
and nonstudents (faculty, staff, and community members) of all ages died in
school-associated violent death events, we calculated event rates using population
data from the postcensal annual estimates compiled by the US Census Bureau.21
Univariable and bivariable analyses were conducted on event characteristics
using the Yates corrected χ2 test; the 2-tailed Fisher exact
test was used when expected values were less than 5. We attempted to identify
potential risk factors for homicide perpetration by comparing characteristics
of homicide perpetrators to homicide victims, a convenient referent population.
These bivariable analyses, based on the Wald χ2 test, were
performed in SUDAAN (SUDAAN Statistical Software Center, Research Triangle
Park, NC) to adjust for clustering by event. Confidence intervals (CIs) obtained
from SUDAAN are reported and used when describing significant differences
The study protocol was approved by the institutional review board of
the CDC, Atlanta, Ga. When necessary, we also applied for and received approval
from local institutional review boards.
Data from at least 1 official source were obtained for all 220 events.
Data from police sources were obtained for 213 (97%) events. We interviewed
a school official for 172 (78%) events.
Of the 220 school-associated violent death events, 202 involved the
death of 1 victim, while 18 events involved the deaths of multiple victims
(median = 2 victims per multiple-victim event). Overall, 131 events involved
1 perpetrator and 76 involved more than 1 perpetrator; for 13 events, the
number of perpetrators was unknown.
In total, 253 victims died in 220 school-associated violent death events.
Of these victims, 172 (68.0%) were students, 18 (7.1%) were faculty/staff,
12 (4.7%) were family members of students, 30 (11.9%) were residents of the
surrounding community, 4 (1.6%) were associated with the school in other ways,
12 (4.7%) were not directly associated with the school or surrounding community,
and 2 (0.8%) were police officers; school association was unknown for 3 (1.2%)
victims. Among the 279 known perpetrators, 103 (36.9%) were students, 2 (0.7%)
were faculty/staff, 7 (2.5%) were family members of a student, 72 (25.8%)
were residents of the surrounding community, 50 (17.9%) were not directly
associated with the school or surrounding community, and 5 (1.8%) were associated
with the school in other ways; school association was unknown or missing for
35 (12.5%) perpetrators. The actions of 5 police officers in the line of duty
led to a school-associated violent death.
The average annual rate of school-associated violent death for students
was 0.068 per 100 000 students (Table
1). The rate of school-associated violent death for male students
was more than twice as high as the rate for female students. The rate for
non-Hispanic, black students was more than 3 times higher than the rate for
non-Hispanic, white students. Students in senior high schools (grades 9-12)
or combined grade schools (schools that combined high school grades with lower
grades) had a school-associated violent death rate that was nearly 14 times
higher than students in elementary schools (preschool-grade 8). Students in
urban school districts experienced a rate of school-associated violent death
(0.085 per 100 000 students) almost twice as high as students in rural
areas (0.045 per 100 000 students). The pattern for homicides was consistent
with this overall pattern; for suicides, the rates were higher, although not
significantly, among whites and in suburban areas.
Between July 1, 1994, and June 30, 1999, 20 541 school-aged children
(5-18 years) died as result of homicide (n = 12 376) or suicide (n =
8165) in the United States.20 During the same
time period, 155 homicides and 28 suicides that were associated with a school
occurred among children 5 to 18 years of age (includes homicides and suicides
associated with a homicide-suicide event). Thus, 0.9% of homicides and suicides
among school-aged children were school associated (1.3% of all homicides and
0.3% of all suicides).
We included data from the previous CDC study1
on school-associated violent deaths to examine trends in these events for
the academic years 1992-1993 through 1998-1999 (Figure 2A). The rate of school-associated violent death events has
decreased significantly since the 1992-1993 school year (P = .03). However, during the same period, the rate of events in which
more than 1 victim was killed due to homicide increased significantly (P = .047).
We also examined trends in student homicide rates by year (Figure 2B). The death rate for single-victim
student homicides declined between 1992 and 1999 (P
= .007). The rate for multiple-victim student homicides has increased since
the 1994-1995 school year (P<.001). Consequently,
total homicide rates for students killed in school-associated violent death
events have increased in recent years.
Of the 220 school-associated violent death events, 172 were homicides,
30 were suicides, 11 were homicide-suicides, 5 were legal intervention deaths,
and 2 were unintentional firearm-related deaths (Table 2a).
Fifty percent of all school-associated death events (n = 110) occurred
while official school activities were in progress, most often during classes
(n = 42 [19.1%]) or after school activities (n = 52 [23.6%]). When examined
by the hour in which the fatal injury occurred (Figure 3), 17.8% (n = 45) of the 253 school-associated violent deaths
occurred near the start of school (7-9AM), 17.8%% (n = 45) during the lunchtime
hours (11AM-1PM), and 21.3% (n = 54) near the end of the school day (2-4PM).
We also sought to understand how often homicide perpetrators' or suicide
victims' actions (eg, threats, notes, journal entries) prior to the incident
may have indicated a potential risk for this type of behavior. We asked police
and school officials to be as inclusive as possible when listing the potential
signs that were made and to include actions that others may not have recognized
as a sign at the time they happened. Over half of the events (n = 120 [54.5%])
were preceded by some action that indicated potential for the coming event.
When we compared the characteristics of suicide and homicide events, we found
that, overall, significantly more suicide events were preceded by a potential
A significantly greater share of homicide events than suicide events
occurred in urban areas (57.8% vs 36.7%) and at off-campus locations (38.4%
vs 10.0%). Suicide events more frequently involved the use of a firearm (90.0%
Within the total population of all perpetrators and all victims, homicide
perpetrators were far more likely than homicide victims to have expressed
suicidal behaviors such as suicidal thoughts, plans, or actual attempts before
the event (odds ratio [OR], 6.96; 95% CI, 1.96-24.65) (Table 3). Overall, homicide perpetrators were also more likely than
homicide victims to have had a history of criminal charges (OR, 6.12; 95%
CI, 3.81-9.82), been a gang member (OR, 4.96; 95% CI, 3.18-7.74), have associated
with high-risk peers or be considered a loner (OR, 6.22; 95% CI, 4.02-9.61),
or used alcohol or drugs on a weekly basis (OR, 3.38; 95% CI, 2.01-5.67).
These patterns were also apparent for students and nonstudents.
Among students, homicide perpetrators were more than twice as likely
as homicide victims to have been bullied by peers (OR, 2.57; 95% CI, 1.12-5.92).
Student homicide perpetrators were also more likely than homicide victims
to be reported to the principal's office for disobeying an authority figure
(OR, 2.50; 95% CI, 1.35-4.62) or fighting peers (OR, 2.87; 95% CI, 1.42-5.78)
and less likely than homicide victims to have participated in extracurricular
activities (OR, 0.26; 95% CI, 0.13-0.50).
School-associated violent deaths represent a small fraction of all homicides
and suicides that occur among school-aged children. However, overall school-associated
student homicide rates appear to have increased in recent years, which can
be attributed to an increase in homicide rates for students killed in multiple-victim
homicide events. The proportion of all school-associated student homicides
that involved multiple victims has risen from 0% in 1992 to 42% in 1999. At
the same time, the rate of single-victim student homicides has declined. Our
findings show that in recent years there were fewer school-associated violent
death events but more deaths per event.
This study includes several important findings that might guide violence
prevention activities. First, most deaths occurred during the transition times
around the start of school, the lunch period, and the end of the school day.
Efforts to reduce crowding, increase supervision, and institute plans for
handling disputes during these intervals may reduce the likelihood that conflicts
will occur and injuries will result when they do.23
In over half of the incidents we examined, some type of potential signal
(note, threat, journal entry, or other action) had been given prior to the
event; in one third of the events, a threat had been made. We do not have
information on how often threats and other potential signs were received by
school officials or what, if any, actions were taken if they were informed
of the threat. These results highlight the importance of investigating the
relationship between threats and school-associated violent deaths.
Homicide perpetrators were nearly 7 times as likely as homicide victims
to have expressed some form of suicidal behavior (thoughts, plans, or attempts)
prior to the event. Homicides followed by suicides and isolated suicides accounted
for nearly 1 in 5 of the violent deaths in this study. These findings, as
well as the results from a nationally representative sample of high school
students indicating that nearly 20% had seriously considered attempting suicide
in the past 12 months, underscore the importance of suicide and suicidal behavior
for schools.24 It is important that we consider
risk factors for suicidal behavior in our efforts to prevent both interpersonal
and self-directed school-associated violence.
Our findings also support recent work demonstrating a link between bullying
victimization and aggressive behavior. In our study, perpetrators were more
likely than victims to have been described as having been bullied by their
peers. These bullied youth may represent the "provocative" or "aggressive"
victims described in recent studies on bullying behavior, who often retaliate
in an aggressive manner in response to being bullied.25-28
This group represents a particularly high-risk population. Data from a nationally
representative sample indicate that 16.9% of students have been bullied on
more than 1 occasion.29 Combined with our findings,
these data demonstrate the importance of programs designed to help teachers
and other school staff recognize and respond to incidents of bullying between
Finally, the results presented in this study emphasize the need for
routine surveillance of school-associated violent death events. Efforts should
be made to make these events reportable to statewide public health, education,
and criminal justice agencies. With complete surveillance information on school-associated
violent deaths, we can address public concerns and develop prevention strategies
Because these data are based on a small number of deaths, some of the
risk estimates presented may be unstable and should be viewed with caution.
In addition, the data described are from secondary sources and are subject
to error and bias. The high visibility and traumatic nature of these events
may differentially affect how respondents recall the characteristics of victims
and perpetrators. For instance, victims may have been viewed in a more positive
light than perpetrators, and this may have influenced the responses.
In summary, this study provides data from a systematic examination of
all known school-associated violent deaths from 1994 to 1999. In thinking
about the prevention of school-associated violent deaths, it is important
to remember that they are rare but complex events. There are no simple solutions;
violence prevention efforts are needed to address risks to young people at
school, at home, and in their communities. By describing the features of these
events and comparing homicide perpetrators to homicide victims, this study
provides some directions for responding to the problem of school-associated
violent deaths now and preventing more deaths in the future.
Create a personal account or sign in to: