Correctional Officer Fatalities in Line of Duty During 2005 to 2015: A Survival Analysis

Abstract

Using correctional officer data from the National Law Enforcement Officers Memorial (NLEOMF) database, we describe the characteristics of U.S. correctional officers killed in line of duty in the past decade, and explore how the hazard probability changed over the life span and factors related to these changes. Survival analysis results show that several demographic characteristics were significantly related to changes in hazard probabilities. Study results provided the most recent statistics on profiles of U.S. correctional officers killed in line of duty. Findings have implications for prevention programs aimed at increasing correctional officer safety.

Keywords

correctional officer fatality survival analysis correctional officer safety hazard probability

Introduction

While research has been conducted on police officer fatalities (Blair, Fowler, Betz, & Baumgardner, 2016; Brandl & Stroshine, 2012; Fridell, Faggiani, Taylor, Brito, & Kubu, 2009; LaTourrette, 2010; Liu & Taylor, 2017) and specialized databases, such as the Law Enforcement Officers Killed and Assaulted (LEOKA) Program database, that are available to researchers on police officer fatalities and injuries, very little systematic research has been done on correctional officer fatalities. This is surprising given that there are close to a half a million correctional officers in the United States (Bureau of Justice Assistance, 2016) responsible for supervising over 2 million inmates in the unique controlled environment of a prison (Konda, Tiesman, Reichard, & Hartley, 2013). Correctional officers are an important group to study, as they have one of the highest rates of nonfatal, work-related injury rates, among all U.S. workers (Bureau of Labor Statistics [BLS], 2015). In 2014, correctional officers had 491 work-related injuries or illnesses per 10,000 full-time employees (FTEs), which were serious enough to require that they missed a day of work (BLS, 2015). This was almost 3 times greater than the rate for all public sector workers who missed a day of work (167 cases per 10,000; BLS, 2015). Correctional officers experienced 153 work-related injuries per 10,000 full-time equivalent employees (FTEs) due to assaults and violent acts in 2014 which is much higher than the rate for all workers (seven per 10,000 FTEs; BLS, 2015).

One important reason for the limited research on correctional officer fatalities is the lack of data. Most of what is known about correctional officer fatalities comes from broader studies on occupational fatalities conducted by the BLS (2016) through their Census of Fatal Occupational Injuries (CFOI) database. Unfortunately, the CFOI data are a restricted access database not easily available to researchers, with access to most of the data fields limited to BLS or National Institute for Occupational Safety and Health (NIOSH) researchers (Tiesman, Swedler, Konda, & Pollack, 2013). In addition, the CFOI database lacks details on the circumstances of the fatalities (other than a short narrative text field), for example, it does not include the use protective equipment or the type of weapons used in an assault (Konda et al., 2013).

Another dataset that includes correctional officer fatalities is the National Violent Death Reporting System (NVDRS), a state-based surveillance system that collects information on violent deaths in participating states in the United States (Blair et. al., 2016). However, the NVDRS dataset has a number of limitations. For example, only 17 states currently participate in NVDRS; thus, the data are not nationally representative. While NVDRS collects fatality information on all law enforcement officers (LEOs), it focuses heavily on police officers and includes very few correctional officer fatality cases, making a separate analysis of correctional officer fatalities difficult. In addition, NVDRS does not collect information on accidental deaths in-the-line-of-duty. Last, NVDRS does not collect any information on the use of body armor or any protective method used by the victims.

While generally known for its police officer fatality data, the National Law Enforcement Officers Memorial (NLEOMF) database (www.nleomf.org/facts/) also contains correctional officer fatality data, providing another valuable source to study correctional officer fatalities. However, the correctional officer portion of the dataset is less known and has not been used for research purposes. The NLEOMF is a private, non-profit foundation founded in 1984 to honor federal, state, and local LEOs who have died in the line of duty (www.nleomf.org/about/). Headquartered in Washington, D.C., the Memorial Fund built and continues to maintain the NLEOMF—the nation’s monument to LEOs killed in the line of duty. The NLEOMF also maintains a database of all officer fatalities that meet their criteria to be entered onto the memorial wall. That is, the NLEOMF database includes data on all forms of “law enforcement officer” line of duty fatalities including correctional and police officers if they have law enforcement status by their employing jurisdiction (www.nleomf.org/memorial/names/nomination-and-approval-process/). Also, fatalities are included in the NLEOMF if the officer “died as a direct and proximate result of a personal injury sustained in the line of duty” (www.nleomf.org/memorial/names/nomination-and-approval-process/). Excluded are fatalities due to natural causes except for those arising from work-related physical exertion including training programs, fitness tests, lifting heavy objects, or a stressful response to an emergency situation including physical struggles, search and rescue missions, performing emergency medical treatment, or high speed pursuits (www.nleomf.org/memorial/names/nomination-and-approval-process/). Also excluded are fatalities caused by intentional misconduct, suicides, and those attributed to an officer performing his or her duty in a negligent manner (www.nleomf.org/memorial/names/nomination-and-approval-process/).

The NLEOMF has a number of strengths that support our use of this database. First, the NLEOMF is a census of all fatalities for officers from all 50 states and territories (and D.C.) meeting the NLEOMF definition. Second, the NLEOMF is rigorous in assessing these fatalities based on the use of a fallen officer notification form, death certificates, state and local police memorials, newspaper accounts, officers’ relatives, agency incident reports, medical records, and autopsies. Eligibility is determined after a review of all available data and issues such as circumstances of death and job descriptions (www.nleomf.org/memorial/names/nomination-and-approval-process/). For an officer to be considered for inclusion on the memorial wall, officer data forms must be completed by the head of the agency for which the officer worked at the time of his or her death (www.nleomf.org/memorial/names/nomination-and-approval-process/). NLEOMF research staff review the forms to determine if the line of duty death meets the organization’s criteria, and the Names Committee, a standing committee of the NLEOMF’s Board of Directors, thoroughly reviews each case to confirm that the fatality meets the criteria for inclusion (www.nleomf.org/memorial/names/nomination-and-approval-process/).

To our knowledge, this is the first research study using the NLEOMF correctional officer data in the published literature. We had three research goals for this study. First, we describe the characteristics of all correctional officers killed from 2005 to 2015 by key background factors (e.g., in terms of marital status, having children, age, years of service, region, etc.). Second, given the evidence that fatalities increase drastically as correctional officers get older (e.g., according to earlier CFOI analyses, most of the fatalities occurred to officers aged 40 and older; Konda et al., 2013), we explore how the hazard probability changes over the life span and factors related to these changes (e.g., marital status, gender, and race). Third, we explore how incident characteristics (e.g., accidental vs. felonious, weapon used) are related to the age of the victims. We also explore the proportion of officers killed who were using safety equipment (i.e., body armor or seat belts) and variations in such use by age. Study results will provide the most recent statistics on profiles of correctional officers killed in line of duty, and findings will have implications for prevention programs aimed to increase correctional officer safety and reduce job-related injury and fatalities.

Method

Sample

Data for the analysis are from the NLEOMF database (http://www.nleomf.org/facts/), including correctional officers with sworn law enforcement status killed in the line of duty in the past 11 years (2005-2015) in all 50 U.S. states, territories, and Washington, D.C. (N = 80). The analysis sample was selected using the criteria of no missing data on variables used in the final multivariate model. Using this criteria, five officers were excluded from the analytical sample. An attrition analysis revealed no significant difference between the 75 selected cases and the five excluded cases; thus, we concluded that this exclusion did not bias the study results.

Variables

Victim’s age (ranged from 22 to 76 years old) was grouped into five age intervals (i.e., 20-29, 30-39, 40-49, 50-59, 60+) used as the time scale in discrete time survival analysis. In addition to age, distribution of fatality is also described by length of service (⩽1 year, 1-5 years, 6-10 years, 11-20 years, 21-30 years, >30 years). Gender (1 = female and 0 = male), marital status (1 = married and 0 = single or divorced), officer rank (1 = manager/commander and 0 =line officer, that is, non-manager/commander), race (1 = non-White and 0 = White), number of children (1 = 1+ children; 0 = no children), agency type (1 = state/federal prison and 0 = local correctional facilities), and region (northeast, southwest, west, U.S. territory, south = 0) were used as exogenous variables to predict hazard probabilities at each age interval. Incident characteristics, including circumstance (1 = accidental and 0 = felonious), whether other officers were injured (1 = yes or 0 = no), reasons of death (1 = traffic-related death, 2 = attacks, 3 = other reasons such as illness), weapon used in felonious incidents (1 = knife, 2 = handgun, 3 = other weapons), whether the victim was wearing body armor in attacks (1 = yes or 0 = no), and whether the victim was wearing seat belt in traffic-related death (1 = yes or 0 = no), were also described and compared across age intervals.

Analytical Plan

Univariate analyses were used to describe the profiles of correctional officer killed, and bivariate analyses were applied to explore the variation of incident characteristics and the use of safety equipment at the time of the incident across victims’ age, using SPSS 20. Discrete time survival analysis was used to analyze the data. In Mplus 7.3 (L. K. Muthén & Muthén, 1998-2015), discrete time survival analysis is modeled using a latent class analysis framework with a single class (Figure 1, see B. O. Muthén & Masyn, 2005, for details). Survival analysis, also known as event history modeling, is a statistical method that allows the study of not only if but also when an event of interest occurred (Singer & Willett, 1993). In this study, the event was defined as a correctional officer being killed. While the time-to-event process in reality tends to be continuous, that is, officers could be killed in the line of duty at any time and day, the timing of the event was recorded in years of age; thus, the process need to be modeled using a discrete time scale. The time scale of officer death was recorded into discrete age intervals by grouping officer’s age at death into 10-year intervals (i.e., age 20-29, age 30-39, age 40-49, age 50-59, and age 60+), to accommodate low frequencies.

Figure 1.

Discrete time survival analysis.

Five binary event indicators corresponding to the five age intervals were coded in a way that an event indicator had the value of 1 if the officer was killed in that interval and 0 if the officer had not been killed by the end of that interval. Once an officer was killed, he was she was counted as “lost to follow up”; that is, the remaining event indicators were coded missing (also known as “right centering” in survival analysis). The probability of an event occurring in a certain age interval is defined as the “hazard probability” for that period (B. O. Muthén & Masyn, 2005); that is, the probability of a correctional officer being killed in that age interval provided that he or she was not killed in the previous age interval. The hazard probability for each age interval is freely estimated rather than being constrained by a linear or curvilinear function of time.

The hazard probability was then related to covariates via logistic regression, so that the effect of a covariate on the timing of an officer being killed is estimated as effect on the log hazard odds of an officer being killed in a given age interval. The covariate effect on the likelihood of being killed in each age interval is expressed as the hazard odds ratio (hOR) that has a similar interpretation as an odds ratio (OR). Covariate effects can be time-invariant (covariate effects are constrained to be the same across each age interval, or hOR is constant over time, also known as the “proportional hazard odds assumption”) or time-variant (covariate effects are allowed to be different across age intervals). A time-variant effect of a covariate would indicate the risk of being killed associated with that covariate which may change as officers get older and may be overwritten by other covariates.

We first conducted bivariate analyses relating each covariate to predict the hazard probability in each age interval, and test the proportional-hazard-odds assumption. After evaluating the results, we then added all covariates with significant (or borderline significant) effects into a multivariate model. Results from the multivariate analysis are presented as adjusted hORs. In addition, we present the predicted-hazard probability for each time interval for several selected groups of interest.

Results

Characteristics of Correctional Officers Killed: 2005 to 2015

Table 1 describes the distribution of correctional officers killed in the line of duty during 2005 to 2015. The officers were killed at a mean age of 46 with an average length of service of 13 years. The majority were male (82.7%), White (61.3%), married (62.7%), and with children (84.0%). Close to three quarters (72.0%) worked in federal or state prisons. Over half were from agencies in the South (56.0%) and the rest distributed across other regions. The distribution over the 11 years shows a slight overall increasing trend, with the number of fatalities in 2013 being the highest and that in 2014 being the lowest. About half (53.3%) of the fatalities were due to accidents and the other half were felonious in nature. The majority of the incidents (80%) did not involve other officers injured or killed. Four in 10 officers were killed in attacks, one third were killed in traffic-related incidents, and the rest were killed for other reasons such as illness directly related to their job. Among the felonious incidents, 80% were killed by fire arm or knife/other blunt instrument. Among traffic-related incidents, only about one third (30%) wore seat belt at the time of the incident. Among attacks, only one in five (20.7%) wore body armor at the time of the incident.

Table 1.

Descriptive Analysis (N = 75).

Variable	M (SD)/%
Officer characteristics
Age	45.6 (11.8)
20-29	9.3%
30-39	22.7%
40-49	26.7%
50-59	30.7%
60+	10.7%
Gender
Male	82.7%
Female	17.3%
Officer’s rank
Correctional line officer	69.3%
Supervisor/commander/manager	30.7%
Officer service length	12.84 (9.36)
⩽1 year	10.7%
1-5 years	18.7%
6-10 years	16.0%
11-20 years	30.7%
21-30 years	21.3%
>30 years	2.7%
Officer’s race
White	61.3%
Non-White	38.7%
Officer’s marital status
Married	62.7%
Other (divorced and single)	37.3%
Having children
Yes	84.0%
No	16.0%
Agency type
Federal/State prison	72.0%
Local correctional facility	28.0%
Region
Northeast	9.3%
Midwest	17.3%
South	56.0%
West	14.7%
U.S. territory	2.7%
Incident characteristics
Year
2005	5.3%
2006	5.3%
2007	9.3%
2008	6.7%
2009	8.0%
2010	8.0%
2011	13.3%
2012	13.3%
2013	14.7%
2014	4.0%
2015	12.0%
Circumstance
Accidental	53.3%
Felonious	45.7%
Other officers involved (Yes)	20.0%
Reasons of death
Traffic-related death	32.0%
Attacks	38.7%
Other causes	29.3%
Weapon use^a
Blunt	48.4%
Firearm	29.0%
Other	22.6%
Seat belt^b	30%
Body armor^c	20.7%

Among the 35 correctional officers feloniously killed.

Among the 20 correctional officers killed in automobile accidents (excluding those struck by a car).

Among the 29 correctional officers killed by attacks.

Survival Analysis

Unconditional model

We first estimated an unconditional model, leaving the baseline hazard probabilities (i.e., the hazard probabilities for each age interval without adding covariates) unstructured so that we estimated five separate baseline hazard probabilities, one for each age interval (see Figure 2). On average, the hazard probability increased as correctional officers get older with a twofold increase every 10 years.

Figure 2.

Hazard probability of being killed in the line of duty for correctional officers killed during 2005 to 2015.

Bivariate analysis

We then examined the effect of each individual covariate in separate bivariate models (Table 2). As a first step, these effects were constrained to be time-invariant. We then built the model by relaxing the proportionality assumption, allowing the effects of the given covariate to be different across age intervals. As the hazard probability for the last age interval, age 60+, was set at 1 (everyone was killed in this sample by this age interval, by definition of the sample), the effect on the last age interval was not estimated. We used the likelihood ratio chi-square test to compare the fit of the two models. Gender, χ²(3, N = 75) = 23.03, p < .001; marital status, χ²(3, N = 75) = 8.4, p = .038; and having children, χ²(3, N = 75) = 10.88, p = .012, had time-variant effects; that is, the significance level and magnitude of the effects differed across age intervals. In the multivariate model, these covariates were included with nonproportional effects. In addition, officer rank had a borderline significant effect on the hazard probability of being killed at each age interval; thus, it was also included in the multivariate model. None of the other covariates had a significant effect.

Table 2.

Bivariate Analysis Results of Covariate Effects on Correctional Officer Hazard of Fatality.

	LL^a	No. of parameters	Adjusted^b χ²	Δdf	p value	Effect in hOR^c
Gender (reference category: male)
Invariant	−113.216	5				Age 20-29: 8.741*
Invariant	−113.216	5				Age 30-39: 0.336
Variant	−105.462	8	23.03	3	<.001	Age 40-49: 0.180
Variant	−105.462	8	23.03	3	<.001	Age 50-59: NA**
Officer rank (reference category: correctional line officer)
Invariant	−111.738	5				Age 20-29: 0.523^†
Invariant	−111.738	5				Age 30-39: 0.523^†
Variant	−111.425	8	0.63	3	ns	Age 40-49: 0.523^†
Variant	−111.425	8	0.63	3	ns	Age 50-59: 0.523^†
Marital status (reference category: not married)
Invariant	−112.94	5				Age 20-29: 0.080*
Invariant	−112.94	5				Age 30-39: 1.773
Variant	−108.794	8	8.4	3	.038	Age 40-49: 1.021
Variant	−108.794	8	8.4	3	.038	Age 50-59: 0.519
Agency type (reference category: local correctional facilities)
Invariant	−113.033	5				Age 20-29: 0.740
Invariant	−113.033	5				Age 30-39: 0.740
Variant	−111.674	8	2.89	3	ns	Age 40-49: 0.740
Variant	−111.674	8	2.89	3	ns	Age 50-59: 0.740
Region (reference category: South)
Invariant	−112.261	8				Northeast	0.436
Invariant	−112.261	8				Southwest	1.025
Variant	−108.943	20	10.06	12	ns	West	0.991
Variant	−108.943	20	10.06	12	ns	U.S. territory	1.110
Race (reference category: White)
Invariant	−112.62	5				Age 20-29: 1.512
Invariant	−112.62	5				Age 30-39: 1.512
Variant	−112.433	8	0.37	3	ns	Age 40-49: 1.512
Variant	−112.433	8	0.37	3	ns	Age 50-59: 1.512
Having children
Invariant	−112.633	5				Age 20-29: 0.203^†
Invariant	−112.633	5				Age 30-39: 0.622
Variant	−109.099	8	10.88	3	.012	Age 40-49: 3.654
Variant	−109.099	8	10.88	3	.012	Age 50-59: NA**

Note. NA = hOR could not be estimated due to a zero cell, that is, all females and those without children were killed by age 59. LL = loglikelihood; hOR = hazard odds ratio.

Log likelihood ratio for model fit.

Chi-square test adjusted for scaling correction factor, for details see https://www.statmodel.com/chidiff.shtml.

When chi-square p value is significant, hORs from the variant model were presented; when chi-square p value is not significant, hORs from the invariant model were presented.

†

p < .10. *p < .05. **p < .01.

Multivariate analysis

Results from the final multivariate model are presented in Table 3. Holding the effects of marital status, having children, and officer rank constant, female officers were 6 times as likely to be killed on the job than male officers between age 20 and 29. They were also more likely to be killed on the job than their male counterparts between age 50 and 59 (however, the hOR could not be computed as all female officers in the sample were killed before age 59). However, females were not more likely to be killed than males between age 30 and 49. Between age 20 and 29, married officers were 86% less likely (p < .10) to be killed than their unmarried counterparts. However, marriage did not have a significant effect on the odds of being killed in other age intervals. Having one or more children significantly reduced the likelihood of being killed for the age group of 50 to 59 (hOR cannot be calculated as all officers without children in this sample were all killed before age 59), but did not have an impact on the odds of being killed in any other age interval. Finally, officers’ rank had a time-invariant effect, with mangers/supervisors/commanders 50% less likely to be killed in each age interval compared with their correctional line officer counterparts (but with only boarder line significance).

Table 3.

Effects of Covariates on the Hazard of Being Killed at Job for Correctional Officers Killed During 2005 to 2015 in hOR in a Multivariate Model (N = 75).

	Age 20-29	Age 30-39	Age 40-49	Age 50-59
Gender (reference category: male)	6.67*	0.44	0.22	NA**
Marital status (reference category: not married)	0.14^†	1.92	0.95	1.16
Having children	0.36	0.57	3.75	NA**
Officer rank (reference category: correctional line officer)	0.49^†	0.49^†	0.49^†	0.49^†

Note. NA = hOR could not be estimated due to a zero cell, that is, all females and those without children were killed by age 59. hOR = hazard odds ratio.

†

p < .10. *p < .05. **p < .01.

Predicted hazard probabilities

Predicted hazard probabilities for a number of selected groups are also computed to assist with the interpretation of the results (see Figure 3; predicted probabilities for other groups are available upon request). For males, the hazard of being killed showed a general increasing trend with age. For married male officers with children and with a manger/commander rank, the trend displays a close to linear change with age; for example, the odds of being killed during age 20 and 29 is close to 0, but steadily increased in each age interval after that, with a sharper increase from age 50-59 to age 60+. This trend is slightly different for single male officers without children and with a line officer rank: the odds of being killed stayed stable between 0.2 and 0.3 up till age 49, with a very sharp increase after age 50 (the probability of being killed is close to 1 for the age 50-59 interval).

Figure 3.

Predicted hazard probabilities of being killed for selected subgroups of correctional officer killed during 2005 to 2015 (N = 75).

A slightly different picture is shown for female correctional officers. For both groups (unmarried, no children, and line officers vs. married with children and manager and commander), the probability of being killed stays relatively low and stable at less than 0.2 up till age 49, with the exception of unmarried line officers with no children, who experienced a probability of being killed between age 20 and 29 as high as 0.7. The probability of being killed increased very sharply after age 50 for both groups of female officers to close to 1.

When comparing male and female officers who were unmarried, with no children, and line officers, males were at a higher rate of being killed than females before age 50 (with the exception of age 20 to 29), but the rates for males and females converge after age 50. When comparing male and female officers who were married, with children, and managers/commanders, they both start at a very low level of risk (less than 0.1). While the probability of female officers being killed stayed low (at less than 0.1) and stable until age 50, that of male officers steadily increased to close to 0.4 by age 50. However, female officers in this group experienced a sharp increase in their likelihood of being killed after age 50 and well surpassed their male counterparts (i.e., between age 50 and 59, males had a .5 probability of being killed, while females had a close to 1 probability of being killed).

Incident Characteristics by Age

Table 4 presents the distribution of incident characteristics by age. Overall, although there are some differences across the age groups, incident characteristics do not differ significantly by age. This could be due to the small sample size; thus, we interpret the results only in an explorative fashion and results need to be interpreted with caution. For example, there is a potential curvilinear relationship between age and reasons for being killed; that is, certain age periods face more danger of being killed from felonious attacks (e.g., age 20-29, age 40-49, age 60+).

Table 4.

Incident Characteristics by Age Among Correctional Officers Killed in Line of Duty During 2005 to 2015 (N = 75).

	Age 20-29 (%)	Age 30-39 (%)	Age 40-49 (%)	Age 50-59 (%)	Age 60+ (%)	p value
Circumstance
Accidental	42.9	64.7	40.0	65.2	37.5	.316
Felonious	57.1	35.3	60.0	34.8	62.5	.316
Other officers involved
Yes	42.9	5.9	20.0	26.1	12.5	.266
No	57.1	94.1	80.0	73.9	87.5	.266
Reasons of death
Traffic-related death	28.6	35.3	25.0	39.1	25.0	.768
Attacks	57.1	29.4	50.0	26.1	50.0
Other causes (e.g., illness)	14.3	35.3	25.0	34.8	25.0
Weapon use^a
Blunt	50.0	20.0	63.6	42.9	50.0	.505
Firearm	0	40.0	27.3	28.6	50.0
Other (e.g., vehicle, fists)	50.0	40.0	9.1	28.6	0
Seat belt^b
Yes	0	20	20	57.1	0	.377
No	100	80	80	42.9	100	.377
Body armor^c
Yes	25	0	30	16.7	25	.739
No	75	100	70	83.3	75	.739

Among the 35 correctional officers feloniously killed.

Among the 20 correctional officers killed in automobile accidents (excluding those struck by a car).

Among the 29 correctional officers killed by attacks.

Discussion

While there is a whole field of research on incarcerated individuals, there is little research on the safety of those who are responsible for keeping those individuals incarcerated. This is an important omission for correctional officers who have one of the highest rates of nonfatal, work-related injury rates, including those due to assaults and violent acts and those injuries requiring treatment in emergency rooms (BLS, 2015). Most of what is known about correctional officer fatalities comes from broader studies on occupational fatalities using BLS’s (2016) CFOI database. Given that this database has restricted access, we made use of another important database involving correctional officer fatalities within the NLEOMF database. The NLEOMF has a number of strengths that support our use of this database such as serving as a census of all fatalities for officers in the United States and the NLEOMF is rigorous in assessing these fatalities. To our knowledge, this is the first analysis focusing on the correctional officer component of the NLEOMF data in the published literature.

Despite that the CFOI and NLEOMF have different criteria for determining their eligible population of correctional officers (CFOI allows for all correctional officer fatalities to be counted but the NLEOMF counts fatalities only if the officer had law enforcement status in their employing jurisdiction) and we are not looking at the same exact years, the overall number of correctional officer fatalities per year recorded in each database are relatively close. The CFOI data covers the period of 1999 to 2008 and the NLEOMF data we examined covers the period 2005-2015. The CFOI from 1999 to 2008 shows an average of 11 (113/10) correctional officer work–related fatalities per year (Konda et al., 2013)and NLEOMF from 2005 to 2015 shows an average of seven (80/11) correctional officer work–related fatalities per year. In essence, both databases show that correctional officer fatalities are rare events. Both databases show that male correctional officers experienced most of the fatalities (89% for CFOI and 83% for NLEOMF) and most of the fatalities (58% for CFOI and 68% for NLEOMF) occurred to officers aged 40 and older. While violence against correctional officers is certainly a key factor in the fatalities in both databases, a large number of fatalities are for other non-felonious reasons. In the CFOI database, assaults and violent acts accounted for 40% of all fatalities and transportation-related fatalities accounted for 40% of all fatalities (falls and other events accounted for the other 20%). The proportions in the NLEOMF database are similar; that is, 39% of the deaths were accounted for by assault and attacks, traffic-related fatalities accounted for 32%, and the other 29% were accounted for by other reasons, for example, falls and illness.

On the other variables in the NLEOMF database, we found that most of the correctional fatalities involved White, married with children correctional line officers. Most of the fatalities occurred in federal or state prisons and in the south of the United States. Also, most of the fatalities involved cases where safety equipment was not used (among traffic-related incidents, only one third wore seat belt at the time of the incident and only one in five wore body armor at the time of the incident).

Next, given the importance of age from the earlier CFOI analyses of correctional officer deaths (most of the fatalities occurred to officers aged 40 and older), we next explored the hazard of fatalities by age. More specifically, we examined the change in hazard of fatalities over time for correctional officers and if the observed change in hazard probability relates to demographic variables, such as marital status, gender, and race.

We found that overall, the hazard of a fatality increased as correctional officers got older with a twofold increase every 10 years. Gender, marital status, and having children were found to have time-variant effects, and officer rank had a borderline significant effect on the hazard probability of being killed at each age interval. Female officers were more likely to be killed on the job than male officers between age 20 and 29 and between age 50 and 59, even after controlling for marital status, having children, and officer rank. However, females were not more likely to be killed than males between age 30 and 49. Married officers were somewhat less likely to be killed than their unmarried counterparts, but only for the young age between 20 and 29. Having one or more children is also a protective factor, but only between age 50 and 59. The finding of time-variant effects of gender, marital status, and having children indicates that risk factors and indicators change over time, with important implications that prevention programs aiming to increase correctional officer safety should take age into consideration. Finally, supervisors and commanders were less likely to be killed in every age interval compared with their correctional line officer counterparts, perhaps due to the different nature of work between line officers and supervisors/commanders. Agency type or region did not predict fatality risk, indicating correctional officer risk of being killed in the line of duty is mostly driven by individual factors rather than agency-level factors. This further highlights that prevention efforts should be tailored to individual officer needs and that threats to officers are invariant across the country.

We also examined the predicted probabilities for a number of selected groups. For males, the hazard of being killed showed a general increasing trend with age. Line officers who are single males without children are the most likely to be killed at almost every age interval (with the exception of age 40-49), suggesting this group might be taking more risks than the other male group, that is, married, with children, of manager/commander rank. It is particularly alarming that such differences grow larger over time. The fatality rate of male line officers who are single and without children skyrocketed when they entered age 50 when compared with married male commanders/supervisors with children (nearly 1 compared to 0.5). It is possible that single line officers without children take even more risk at an older age, as the perceived risk may be less for this group. It is also possible that they are more likely to suffer from other age-related problems that place them in a more vulnerable position. Further research is needed to understand the alarmingly high fatality rate for this group, and prevention efforts to increase correctional officer safety should also focus on this group.

In contrast, the probability of being killed for female correctional officers stay relatively low and stable up until age 49. This group needs to be studied more to understand potential protective factors in place for this group. One exception to this trend is that young unmarried female line officers (age 20-29) with no children had a much higher than average fatality risk (0.7). It is possible that they are particularly vulnerable to assault by inmates as a more attractive target. This subgroup of female officers needs to be studied in greater detail to understand why they are at a greater fatality risk.

It is also interesting to find that being married, having children, and being a supervisor/commander has a protective effect only for men but not for women, especially as correctional officers get older. After age 50, male supervisors/commanders who were married with children have a much lower fatality risk than their counter parts, that is, non-married line officers without children (0.5 versus close to 1). This difference is not presented for females, as the probability of being killed increases very sharply after age 50 for both groups of female officers (whether married or not, with children or not, and irrespective of rank). This could reflect the decreasing resilience to attacks or other traumatic experiences with age is more prominent for females than for male correctional officers.

When examining the relationship between incident characteristics and age, we found that the danger for being killed from felonious attacks was not evenly distributed across the life span; for example, officers face greater risk of being killed from attacks during age 20 to 29, age 40 to 49, and age 60+. It is possible that younger correctional officers are more likely to take risk, and older correctional officers are less likely to survive an injury. It is also possible that older officers may be more likely to “let their guard down” as they get older. The overall lack of a significant relationship between incident characteristics and age may be due to the lack of statistical power with our modestly sized sample. To explore in greater details these and other subgroups, including whether injuries follow a similar patterns, we recognize the need for an expanded database.

We acknowledge that there are several limitations associated with our study. First, the NLEOMF was not designed to serve as a public health surveillance system. The NLEOMF was designed to commemorate the service and sacrifice of LEOs and memorialize as many members of the law enforcement community as possible (Tiesman et al., 2013). Therefore, we do not have a comparison group of non-fatally injured officers within a database dedicated to fatalities only, making it challenging to study whether certain safety equipment such as seat belt and body armor is effective in protecting correctional officers. Furthermore, the NLEOMF does not count all correctional officer fatalities, only those where the correctional officer has “law enforcement status” in their employing jurisdiction. Outside of the NLEOMF, no other group is collecting similar fatality data on correctional officers. Finally, the high correlation between age and tenure of service (bivariate correlation coefficient > 0.6; for example, all officers who died before age 29 had less than 5 years of service, and those who had over 30 years of service were at least 50 years of age when they were killed) and the small number of cases in the dataset prevented us from further disentangling the effect of age versus service tenure on the risk of correctional being killed over time. As risk of being killed depends not only on service length but also on other factors that are closely related with age (e.g., health status and ability to survive injury; Evans, 1999), we chose to focus on age–fatality relationship in the survival analysis, with descriptive analysis on the bivariate relationships separately by age and length of service.

Despite the above limitations, our study is the first published research analyzing the correctional officer fatality data in the NLEOMF database. The danger related to working as a correctional officer does not draw nearly as much attention as that related to working as a patrol police officer. Yet, close to half of the correctional officer fatalities in the line-of-duty were accounted for by felonious assault, which can be prevented. Our results provided a much-needed up-to-date profile of correctional officers killed in the line of duty in the past decade, and painted a picture of how risk of being killed on the job changes across the life span for groups of correctional officers with different risks. We found that the risk of being killed in the line of duty is not equally distributed across the life span, nor across different demographic groups. Specifically, a few groups of correctional officers were identified to have particularly high fatality risk, for example, older unmarried male line officers without children and young unmarried female line officers without children. While we were able to review potential reasons, for example, risk-taking behavior and vulnerability, the limited data did not allow us to test these hypotheses. In addition, findings in this study also raised interesting future research questions. For example, are there different risk factors for each particular type of incident, and how do these risk factors change over the life course of a correctional officer? Together, such research would have implications for targeted prevention programs aimed to increase correctional officer safety, and provide guidance for correctional agencies to change their policies and practices (e.g., alert staff to the greater risks faced by single young female correctional officers) to reduce job-related injury and fatalities.

As researchers and practitioners become more aware of the dangers that correctional officers face when doing their daily work, the field is in critical need for a national database of all serious attacks (e.g., aggravated assaults) against correctional officers, equivalent to the LEOKA dataset. While this expanded database would be a major undertaking and expensive, if properly leveraged and analyzed could lead to better prevention efforts (based on empirical risk factors) that could save correctional officer lives. Such a database would also allow the study of the effectiveness of safety equipment, such as body armor, in saving correctional officers’ lives. In addition, prospective survey data following a group of nationally representative correctional officers would also be fruitful in further investigating their risk-taking behavior and have important implications for correctional policies and practices.

Footnotes

Declaration of Conflicting Interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship and/or publication of this article: This study was funded by the National Institute of Justice (Grant # 2011-IJ-CX-K056 and 2012-R2-CX-K008).

Author Biographies

Weiwei Liu, PhD, is a criminologist and prevention scientist at NORC at the University of Chicago, focusing on the developmental vulnerability of risky behaviors over the life course. Prior to joining NORC, she completed an NIH-funded postdoctoral fellowship in Prevention Science at Johns Hopkins University.

Bruce G. Taylor, PhD, is a senior fellow at NORC at the University of Chicago. A criminologist by training, his work has focused on identifying demographic and contextual explanations for a variety of forms of violent and related risky behaviors. His major research interest is violence prevention related to intimate partner and youth violence.

References

Blair

J. M.

Fowler

K. A.

Betz

C. J.

Baumgardner

J. L

. (2016). Occupational homicides of law enforcement officers, 2003–2013: Data from the National Violent Death Reporting System. American Journal of Preventive Medicine, 51(5, Suppl. 3), S188-S196. doi:10.1016/j.amepre.2016.08.019

Brandl

S. G.

Stroshine

M. S

. (2012). The physical hazards of police work revisited. Police Quarterly, 15, 262-282.

Bureau of Justice Assistance. (2016). Occupational employment and wages, May 2015 (33-3012 Correctional Officers and Jailers). Retrieved from www.bls.gov/oes/current/oes333012.htm.

Bureau of Labor Statistics. (2015). Nonfatal occupational injuries and illnesses requiring days away from work in 2014. Retrieved from www.bls.gov/opub/ted/2015/nonfatal-occupational-injuries-and-illnesses-requiring-days-away-from-work-in-2014.htm

Bureau of Labor Statistics. (2016). Census of fatal occupational injuries Summary, 2014. Retrieved from www.bls.gov/news.release/cfoi.nr0.htm

Evans

. (1999). Age dependence of female to male fatality risk in the same crash: An independent reexamination. Annual Proceedings/Association for the Advancement of Automotive Medicine, 43, 225-238.

Fridell

Faggiani

Taylor

Brito

C. S.

Kubu

. (2009). The impact of agency context, policies, and practices on violence against police. Journal of Criminal Justice, 37, 542-552. doi:10.1016/j.jcrimjus.2009.09.003

Konda

Tiesman

Reichard

Hartley

. (2013). U.S. correctional officers killed or injured on the job. Corrections Today, 75, 122-123.

LaTourrette

. (2010). The life-saving effectiveness of body armor for police officers. Journal of Occupational and Environmental Hygiene, 7, 557-562.

10.

Liu

Taylor

. (2017). The effect of body armor on saving officers’ lives: An analysis using LEOKA data. Journal of Occupational and Environmental Hygiene, 14, 73-80. doi:10.1080/15459624.2016.1214272

11.

Muthén

B. O.

Masyn

. (2005). Discrete-time survival mixture analysis. Journal of Educational and Behavioral Statistics, 30(1), 27-58.

12.

Muthén

L. K.

Muthén

B. O

. (1998-2015). Mplus user’s guide (7th ed.). Los Angeles, CA: Author.

13.

Singer

J. D.

Willett

J. B

. (1993). It’s about time: Using discrete-time survival analysis to study duration and the timing of events. Journal of Educational and Behavioral Statistics, 18, 155-195.

14.

Tiesman

H. M.

Swedler

D. I.

Konda

Pollack

K. M

. (2013). Fatal occupational injuries among U.S. law enforcement officers: A comparison of national surveillance systems. American Journal of Industrial Medicine, 56, 693-700. doi:10.1002/ajim.22182