Sexual Victimization and Completed Suicide Among Danish Female Adults

Abstract

Although sexual victimization has been associated with suicidal behaviors, its association with completed suicide has not been examined. We investigated this association among Danish women using longitudinal data and a conservative definition of victimization. This population-based case-control study included 476 suicide cases and 12,010 matched controls. Seven cases (1.5%) and 5 controls (0.04%) experienced sexual victimization that was reported to the police and resulted in a conviction. Sexual victimization was associated with a 14-fold increased rate of suicide, controlling for confounders and matching (95% CI: [3.4, 59]). Completed suicide is an important potential outcome of sexual victimization, warranting further examination.

Keywords

crime case-control studies sex offense suicide

Survivors of attempted or completed rape are more likely to contemplate suicide or to attempt suicide than people without these experiences (Kilpatrick et al., 1985). In a cross-sectional community-based study, women reporting a history of sexual assault were about six times more likely to have attempted suicide than those who did not report experiencing a sexual assault, adjusting for demographics, posttraumatic stress disorder, depression, alcohol use, and social support (Davidson, Hughes, George, & Blazer, 1996). Similarly, in a national sample of U.S. women (Ullman & Brecklin, 2002) and in a population-based study of women in Virginia (Masho, Odor, & Adera, 2005), survivors of adult sexual assault had a higher prevalence of suicidal ideation and attempts than women without these experiences.

Various mechanisms by which sexual victimization may be associated with suicidal behavior have been proposed. Sexual victimization is associated with psychiatric morbidity (e.g., posttraumatic stress disorder [PTSD] and depression symptoms; Winfeld, George, Swartz, & Blazer, 1990), which in turn is associated with suicide attempts and completion (Gradus et al., 2010; Kessler, Borges, & Walters, 1999). In fact, the association between intimate partner rape and suicidal ideation was mediated by PTSD and depression symptoms in a sample of women living in a shelter (Weaver et al., 2007). In U.S. college women, sexual victimization was also found to be associated with hopelessness, a widely accepted, consistent predictor of suicidal behavior (Stepakoff, 1998).

Although the association between sexual victimization and suicidal ideation or suicide attempts is well-established, sexual victimization as a risk factor for completed suicide has not been examined. Studies of suicidal ideation and attempt are important in their own right; however, they are imperfect substitutes for studies of suicide completion because people who think about or attempt suicide differ from people who complete suicide with regard to important characteristics (Maris, Berman, & Silverman, 2000a). For example, males are four times more likely to complete suicide than females, but females are three times more likely to attempt suicide than males (Maris, Berman, & Silverman, 2000b), and only 15% of people who attempt suicide ever go on to complete suicide (Maris et al., 2000a). Additionally, while prior suicide attempts are known to be an important predictor of completed suicide, other major factors predicting completed suicide among suicide attempters have been noted (e.g., depression, alcohol and drug abuse, social isolation, older age, White ethnicity, work and/or marital problems, stress, and physical illness; Maris, Berman, & Silverman, 2000c). In addition, the literature summarized above includes only cross-sectional examinations of the association between sexual victimization and suicidal ideation or attempts. For these reasons, a longitudinal study of the association between sexual victimization and completed suicide fills an important gap.We undertook such a study using national Danish healthcare and social registries. According to the World Health Organization (2006), the suicide rate in Denmark was approximately 11.9 per 100,000 as of 2006, and a survey study of a representative sample of female adults in Denmark reported the prevalence of sexual victimization during adulthood to be about 7% (Sundaram, Laursen, & Helweg-Larsen, 2008). Health care and social data are recorded for the entire population of Denmark, and registries containing these data can be readily linked, thereby providing a unique opportunity to evaluate longitudinally the association between sexual victimization and completed suicide. We hypothesized that survivors of sexual victimization would have a higher rate of completed suicide than women who were not similarly victimized.

Method

Study Population

The source population for this case-control study was the female population of Denmark, 15-54 years old from 2001 to 2006. Sexual victimization was only reported by women between the ages of 15 and 54 in the dataset used for the current study; therefore, the analyses in this article are restricted to that age range. Suicide cases included any citizen of Denmark who committed suicide between 2001 and 2006 (n = 476). For each case, we chose up to 30 controls from a 25% representative sample of the Danish population, matched to cases on date of birth and calendar time (n = 12,010). Controls were citizens of Denmark who had not committed suicide prior to 2006. Thirty controls per case were chosen because of low prevalence of sexual victimization exposure in the current sample. Participants had to be residing in Denmark in the calendar year before the year of suicide (or match date for controls) to obtain sufficient social data.

Data Collection

We linked the registries described below using the Central Population Registration number, a unique identification number assigned to all citizens and residents of Denmark and recorded in the Civil Registration System.

We ascertained suicide cases from the Danish Cause-of-Death Registry (Qin, Agerbo, & Mortensen, 2003), coded according to the International Classification of Diseases, tenth revision (ICD-10; codes X60-X84; Janca, Ustun, van Drimmelen, Dittmann, & Isaac, 1994). All unnatural deaths in Denmark result in an inquest conducted by independent forensic medical doctors. This inquest involves collecting information about the deceased and the death itself from the person’s physician, family, friends, and the person(s) who discovered the body. Information is obtained about place of death, method of dying, and presence of a suicide note. If the cause of death is still uncertain following this inquest, an autopsy is conducted. The final determination of death from suicide is made by the independent forensic doctor based on information from all of these sources. The process by which suicide is assigned as the cause of death in Denmark is described in more detail elsewhere (Statens Institute for Folkesundhed, 2005). Control participants matched to suicide cases were selected from the Civil Registration System.

In the Danish registry of crimes reported to the police, from which we obtained data on history of sexual victimization, only crimes that are reported to the police and that result in a perpetrator conviction are coded and available for research. Therefore, we used the following conservative definition of sexual victimization: any sexual touching with force that occurred during the period 2001 to 2006, was reported to the police, and for which the perpetrator was convicted. We will abbreviate this definition as “sexual victimization” and will discuss the advantages and disadvantages of this definition below.

We used the Psychiatric Central Registry (Munk-Jørgensen & Mortensen, 1997) to obtain information on psychiatric diagnoses preceding the crime, which were candidate confounders beyond the matched factors. We did not include psychiatric diseases diagnosed after the crime because these diseases may be on the causal pathway between sexual victimization and suicide. Psychiatric diseases were coded according to the International Classification of Diseases, eighth revision (ICD-8) until 1994 (World Health Organization, 1967) and ICD-10 after 1994 (Janca et al., 1994). Due to sparse data concerns, we combined any history of the following psychiatric diagnoses into one dichotomous variable in the analyses: depressive episodes, recurrent depressive disorder, dysthymia, mental/behavioral disorders due to use of alcohol, mental/behavioral disorders due to use of drugs, phobic anxiety disorders, other anxiety disorders, obsessive-compulsive disorder, acute stress reaction, posttraumatic stress disorder, adjustment disorder, and other reactions to severe stress (ICD-8 codes: 291, 294, 296, 300, 300.2 - 300.4, 303, 304, 307; ICD-10 codes: F10, F11-F19, F32, F33, F34.1, F40 - F43, F43.1, F43.2, F43.8).

We used the Integrated Database for Labor Market Research (Qin et al., 2003) to collect data on social variables, which were also candidate confounders beyond the matched factors. We based annual income on the calendar year before suicide (or match date) and categorized it into quartiles within strata of sex and age groups. We categorized marital status (living in a married or cohabiting status, single status, or unmarried minor) as of November in the year before suicide or the match date.

Data Analysis

We conducted descriptive and stratified analyses to examine the distribution of variables among cases and controls. We then calculated years between sexual victimization date (index date) and suicide/match date by subtracting the index date from the date of suicide/match date. We then categorized the study population into 4 groups: those who were not sexually victimized, those with index date and suicide/match date in the same year, those with index date 1-2 years before the suicide/match date, and those with index date 3 or more years before the suicide/match date. Finally, we used conditional logistic regression to estimate the association between sexual victimization and completed suicide, accounting for the control-to-case matching. As risk-set sampling was used to sample control participants, the odds ratios provide estimates of the corresponding incidence rate ratios (Rothman, Greenland, & Lash, 2008). In addition to adjustment for the matched factors by the conditional analyses, we adjusted for pre-existing psychiatric diagnoses, marital status, and socioeconomic status. Analyses were conducted in SAS version 9.1.3. The study was approved by the Institutional Review Board at Boston University Medical Center, USA, and the Danish Data Protection Agency.

Results

Table 1 displays the characteristics of cases and controls. Cases were more likely to be single (55%) and in the lowest income quartile (54%) than controls (27% and 26%, respectively). There were 7 cases (1.5%) and 5 controls (0.04%) with a history of sexual victimization. Four of the seven cases who were sexually victimized committed suicide in the same year (57%), with the remainder of sexually victimized cases committing suicide 1-2 years after victimization (43%).

Table 1.

Characteristics of Suicide Cases and Matched Controls, Denmark 2001 to 2006

	Suicide cases (n = 476)	Matched controls (n = 12,010)
Age group
15-24 years	40 (8.4%)	796 (6.6%)
25-34 years	82 (17%)	2062 (17%)
35-44 years	161 (34%)	4248 (35%)
45-54 years	193 (41%)	4904 (41%)
Marital status
Married	135 (28%)	6627 (55%)
Cohabitating	69 (15%)	1973 (16%)
Single	263 (55%)	3228 (27%)
Unmarried Minor	9 (1.9%)	182 (1.52%)
Income
Lowest quartile	259 (54%)	3081 (26%)
Second quartile	99 (21%)	2991 (25%)
Third quartile	66 (14%)	3023 (25%)
Highest quartile	52 (11%)	2915 (24%)
Psychiatric diagnoses	345 (72%)	986 (8.2%)
Sexual victimization^a	7 (1.5%)	5 (0.04%)

Sexual victimization is defined as sexual touching with force that occurred during the period 2001 to 2006, was reported to the police, and for which the perpetrator was convicted.

Regression analyses conditioned only on the matched factors revealed that those experiencing sexual victimization had 30 times the rate of suicide as those who did not (95% CI: [9.4, 94]). Pre-existing psychiatric diagnoses were the only identified confounder of this association (OR for association with completed suicide = 20, 95% CI = [17, 23]). After controlling for pre-existing psychiatric diagnoses, the rate of suicide among those who experienced a sexual victimization was 14 times greater than those who did not (95% CI: [3.4, 59]). The results of these analyses are displayed in Table 2.

Table 2.

Results of the Conditional Logistic Regression Analyses for the Association between Sexual Victimization and Completed Suicide

	OR	95% Confidence Interval
Sexual victimization and suicide conditioned on matched factors only	30	[9.4, 94]
Association between identified confounders and suicide
Psychiatric diagnoses	20	[17, 23]
Sexual victimization and suicide conditioned on matched factors and identified confounders	14	[3.4, 59]

Discussion

This study is the first to prospectively examine the association between sexual victimization and completed suicide. Sexual victimization was associated with a substantially increased rate of completed suicide among women, even after adjusting for pre-existing psychiatric diagnoses and the matched factors. Several mechanisms may account for this association. As noted in the introduction, sexual victimization is associated with hopelessness, which is an established predictor of suicidal behavior (World Health Organization, 2006). Posttraumatic stress disorder or depression may develop following sexual victimization and increase suicide risk. The fact that most exposed cases completed suicide soon after the victimization is consistent with the potential action of these mechanisms and previous literature on sexual victimization and suicidal behavior. A study conducted by Stepakoff (1998) found that adult sexual victimization, but not childhood sexual victimization, was associated with current hopelessness and suicidal ideation in a sample of college females.

The current investigation has limitations worth noting. We restricted our sample to women and, though still substantial, the psychological impact of sexual victimization is usually different among men (Sundaram et al., 2008). Second, confounding by preceding psychiatric diagnoses reduced the association more than two-fold. It is possible that adjustment for undiagnosed psychiatric conditions, residual confounding by pre-existing psychiatric disorders, or psychiatric diagnoses recorded in somatic hospitals would further reduce the strength of association. Also, we did not account for the effect of victimization prior to the sexual victimization incident reported in the current study. Given an odds ratio of 14, however, it is unlikely that these sources of residual confounding fully account for the observed association, because their associations with both sexual victimization and suicide risk would have to be at least as strong as this reported odds ratio. In addition, we collapsed across various types of sexual victimization with force to create our sexual victimization variable, and across different psychiatric disorders to create our psychiatric diagnosis variable, due to concerns about sparse data. It is possible that associations observed between specific forms of sexual victimization and suicide would have a different magnitude than the association presented in the current study. It is also likely that adjusting for specific psychiatric disorders would have had varying impacts on the association presented here. Nonetheless, we remain confident that these changes would only marginally affect the strength of association, so that a strong association between sexual victimization—as we have measured it—and completed suicide would remain.

Our conservative definition of sexual victimization is both a strength and limitation of the current study. The advantage of this definition is that the sexual victimization was recorded in advance of the suicide and confirmed, so far as able, by conviction of the perpetrator. To the best of our knowledge, no other registries exist with data quality of this type on reports of sexual victimization. We expect that some sexual victimization events in the study population were not reported to police, or the perpetrator was never apprehended or convicted, and therefore the unexposed group in the current study almost certainly includes some survivors of sexual victimization. As a result, the prevalence of sexual victimization among controls should not be considered a population-based estimate of this prevalence. The inclusion of survivors of sexual victimization in the unexposed group would result in non-differential and independent misclassification of this dichotomous exposure, which biases the observed association toward the null.

A less conservative definition of sexual victimization, including victimizations not reported to the police, would result in a larger number of observed exposed persons, which would improve the precision of the estimate of association. Despite the width of the interval, this study’s result provides “strong” evidence against the null according to one classification scheme (Goodman, 2001), which is especially valuable when the association has never before been studied and even surrogate outcomes have only been studied cross-sectionally. Furthermore, a more inclusive definition of sexual victimization, and the methods that would be used to obtain these data, may introduce biases that do not affect the current study. For example, retrospective history of sexual victimization reported by next-of-kin in a case-control study would be susceptible to recall bias, resulting in differential exposure misclassification that may bias away from the null. History of sexual victimization might be reported by women enrolled in a prospective cohort study, but such a study would have to have a substantial study size and follow-up period before sufficient number of suicide cases occurred to allow an estimate of the association. In addition, a less conservative definition of sexual victimization recorded by either of these designs would also be susceptible to exposure misclassification, since underreporting of these events is pervasive and individual perception of what constitutes sexual victimization varies (Kahn, Jackson, Kully, Badger, & Halvorsen, 2003). These classification errors would reduce the validity of the measured association without necessarily providing further evidence against the null hypothesis (Greenland & Gustafson, 2006).

Despite the limitations of the current study, our results make a unique contribution to the literature as an initial examination of the association between sexual victimization and suicide, using longitudinal data and prospective assessment in a population-based sample. Identifying survivors of sexual victimization for longitudinal research purposes is difficult, and often requires large samples. Going forward, we encourage a longitudinal examination of the impact of sexual victimization on completed suicide in existing large, prospective epidemiologic cohorts. These cohorts could ascertain history of sexual victimization using a much broader definition by relying on self report. The current study highlights completed suicide as an important potential outcome of sexual victimization and provides strong evidence that this association warrants further examination.

Footnotes

Acknowledgements

The authors thank Emily Rothman for her helpful comments on an earlier draft of this article, and also thank Preben Bo Mortensen for enabling access to the data and for providing research resources.

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.

Bios

Jaimie L. Gradus, DSc, MPH, is an epidemiologist at the National Center for PTSD, VA Boston Healthcare System and an assistant professor of psychiatry and epidemiology at Boston University Schools of Medicine and Public Health. Her main research interests are the epidemiology of trauma and PTSD, as well as the associations between trauma, PTSD, and suicidal behaviors. She was the winner of the 2009 Lilienfeld Prize from the Society for Epidemiologic Research for her work examining the association between PTSD and completed suicide.

Ping Qin, MD, PhD, is an associate professor in psychiatric epidemiology at the National Centre for Register-based Research at the University of Aarhus in Denmark. Her main research area is suicide and psychiatric illness. She has been a key figure in epidemiological research on suicide with expertise in utilizing data from various Danish population longitudinal registers. Her outstanding work has generated a number of influential publications in prestigious journals. She was among the founders of the Danish Association for Suicide Research and Prevention.

Alisa K. Lincoln, MPH, PhD, is an associate professor of health sciences and sociology at Northeastern University. Her research focuses on public mental health and mental health care. Her research contributes to our knowledge about mental health disparities focusing on the ways in which social factors and system level factors interrelate to perpetuate these disparities.

Matthew Miller, MD, ScD, is an associate professor of health policy and management at the Harvard School of Public Health and Deputy Director of the Harvard Injury Control Research Center. His research comprises pharmacoepidemiologic and epidemiological studies on suicide, homicide, and unintentional injuries.

Elizabeth Lawler, DSc, MPH, is the deputy director of the Epidemiology Center at the Massachusetts Veterans Epidemiology Research and Information Center (MAVERIC). Her areas of expertise are chronic disease epidemiology, pharmacoepidemiology, and related methodologies. She directs the Pharmacoepidemiology Research Group at MAVERIC. A theme in her research is mental health epidemiology relevant to veterans and identification of biases inherent in using EMR data. She is a coinvestogator on numerous Department of Veterans Affairs Healthcare System (VA) cohort studies such as the Normative Aging Study, and she is a coinvestigator for the establishment of a very large genomic data collection through the VA.

Henrik Toft Sørensen, PhD, DMSc, is a specialist in internal medicine and hepatology. He has conducted research on several epidemiological topics, in particular within clinical epidemiology. His main research interest areas are determinants for prognosis of severe diseases, in particular venous thromboembolism, gastrointestinal diseases and cancer; and observational studies of outcome and safety of medical interventions, in particular drugs and procedures. He is professor of clinical epidemiology and is head and founder of the Department of Clinical Epidemiology at Aarhus University, Denmark. He is an adjunct professor of epidemiology at Boston University and visiting professor at Dartmouth Medical School.

Timothy L. Lash, DSc, MPH, holds appointments as professor of cancer epidemiology in the Department of Clinical Epidemiology at Aarhus University Hospital and adjunct associate professor of epidemiology and medicine at the Boston University Schools of Public Health and Medicine. His primary research interest is in prognostic markers of cancer outcomes, including drug-drug and gene-drug interactions. He also actively researches clinical outcomes of many common chronic diseases as well as methods to improve this research.

References

Davidson

J. R. T.

Hughes

D. C.

George

L. K.

Blazer

D. G.

(1996). The association of sexual assault and attempted suicide within the community. Archives of General Psychiatry, 53, 550-555.

Goodman

S. N.

(2001). Of p-values and Bayes: A modest proposal. Epidemiology, 12, 295-297.

Gradus

J. L.

Qin

Lincoln

A. K.

Miller

Lawler

Sørensen

H. T.

Lash

T. L.

(2010). Posttraumatic stress disorder and completed suicide. American Journal of Epidemiology, 171, 721-727.

Greenland

Gustafson

(2006). Accounting for independent nondifferential misclassification does not increase certainty that an observed association is in the correct direction. American Journal of Epidemiology, 164, 63-68.

Janca

Ustun

T. B.

van Drimmelen

Dittmann

Isaac

(1994). ICD-10 symptom checklist for mental disorders, Version 1.1. Geneva, Switzerland: World Health Organization.

Kahn

A. S.

Jackson

Kully

Badger

Halvorsen

(2003). Calling it rape: Differences in experiences of women who do or do not label their sexual assault as rape. Psychology of Women Quarterly, 27, 233-242.

Kessler

R. C.

Borges

Walters

E. E.

(1999). Prevalence and risk factors for lifetime suicide attempts in the National Comorbidity Survey. Archives of General Psychiatry, 56, 617-626.

Kilpatrick

D. G.

Best

C. L.

Veronen

L. J.

Amick

A. E.

Villeponteaux

L. A.

Ruff

G. A.

(1985). Mental health correlates of criminal victimization: A random community survey. Journal of Consulting and Clinical Psychiatry, 55, 866-873.

Maris

R. W.

Berman

A. L.

Silverman

M. M.

(2000a). Introduction to the study of suicide. In Maris

R. W.

Berman

A. L.

Silverman

M. M.

(Eds), Comprehensive textbook of suicidology (pp. 3-25). New York: Guilford Press.

10.

Maris

R. W.

Berman

A. L.

Silverman

M. M.

(2000b). Suicide, gender and sexuality. In Maris

R. W.

Berman

A. L.

Silverman

M. M.

(Eds.), Comprehensive textbook of suicidology (pp. 145-169). New York: Guilford Press.

11.

Maris

R. W.

Berman

A. L.

Silverman

M. M.

(2000c). Suicide attempts and methods. In Maris

R. W.

Berman

A. L.

Silverman

M. M.

(Eds.), Comprehensive textbook of suicidology (pp. 284-310). New York: Guilford Press.

12.

Masho

S. W.

Odor

R. K.

Adera

(2005). Sexual assault in Virginia: A population-based study. Women’s Health Issues, 15, 157-166.

13.

Munk-Jørgensen

Mortensen

P. B.

(1997). The Danish psychiatric central register. Danish Medical Bulletin, 44(1), 82-84.

14.

Qin

Agerbo

Mortensen

P. B.

(2003). Suicide risk in relation to socioeconomic, demographic, psychiatric, and familial factors: A national register-based study of all suicides in Denmark, 1981-1997. American Journal of Psychiatry, 160, 765-772.

15.

Rothman

K. J.

Greenland

Lash

T. L.

(2008). Case-control studies. In Rothman

K. J.

Greenland

Lash

T. L.

(Eds.), Modern epidemiology. (3rd ed., pp. 111-127). Philadelphia, PA: Lippincott, Williams & Wilkins.

16.

SAS Institute, Inc. (2000-2004). SAS [computer program]. Version 9.1.3. Cary, NC: Author.

17.

Statens Institute for Folkesundhed [National Institute for Public Health]. (2005). Selvmord i Danmark—Markant fald i selvmord, menstigende antal selvmordsfprsøge, Hvorfor? [Suicide in Denmark—A decrease in suicide completion but a increase in suicide attempt, why?] København, Denmark: Kailow Graphic A/S.

18.

Stepakoff

(1998). Effects of sexual victimization on suicidal ideation and behavior in U.S. college women. Suicide and Life-Threatening Behavior, 28, 107-126.

19.

Sundaram

Laursen

Helweg-Larsen

(2008). Is sexual victimization gender specific? The prevalence of forced sexual activity among men and women in Denmark, and self-reported well-being among survivors. Journal of Interpersonal Violence, 23, 1414-1440.

20.

Ullman

S. E.

Brecklin

L. R.

(2002). Sexual assault history and suicidal behavior in a national sample of women. Suicide and Life-Threatening Behavior, 32, 117-130.

21.

Weaver

T. L.

Allen

J. A.

Hopper

Maglione

M. L.

McLaughlin

McCullough

M. A.

(2007). Mediators of suicidal ideation within a sheltered sample of raped and battered women. Health Care for Women International, 28, 478-489.

22.

Winfeld

George

L. K.

Swartz

Blazer

D. G.

(1990). Sexual assault and psychiatric disorders among a community sample of women. American Journal of Psychiatry, 147, 335-341.

23.

World Health Organization. (1967). Manual of the international classification of diseases (ICD-8). Geneva, Switzerland: Author.

24.

World Health Organization. (2006). Suicide prevention: Suicide rates (per 100,000), by gender, Denmark, 1950-2006. Retrieved from http://www.who.int/mental_health/media/denm.pdf