Dr. McPherson is affiliated with the Institute for Research on Women and Gender, University of Michigan, G120A Lane Hall, Ann Arbor, MI 48109-1290 (e-mail: melneem@umich.edu). Dr. Delva is with the School of Social Work and Dr. Cranford is with the Substance Abuse Research Center, University of Michigan, Ann Arbor.

Severe mental illness, substance misuse, and intimate partner violence have emerged as major intersecting social and public health problems that adversely and disproportionately impact the lives of women in the United States (1,2,3,4), particularly black women, in urban areas of the nation (5,6). Moreover, many women who have a mental illness or a co-occurring substance use disorder (dual diagnosis) are also raising children (7), thus creating another potentially problematic area along the articulated continuum of social problems. For example, women with severe mental illness and their children are at greater risk of emotional, behavioral, and developmental difficulties, particularly in the absence of programs targeting intervention and prevention activities (8).

It is well established that the population of individuals with severe mental illness is at increased risk of having a substance use disorder (dual diagnosis) (9,10,11,12,13,14,15,16,17,18,19,20). An increased risk of sexual and physical violence is associated with dual diagnosis among women. Women from all social and racial backgrounds are victims of partner violence. However, available literature suggests that women who are black, young, separated, residents of urban areas, and living in poverty are most frequently the victims of partner abuse (21), particularly women with substance abuse problems (1,22) and those diagnosed as having mental disorders (2,23)

Until recently dual diagnosis has generally been characterized as a "male" phenomenon (24,25); therefore, very little of the empirical literature discusses the ramifications of this diagnosis for women and the roles, responsibilities, and stressors in their lives, particularly the experience of interpersonal violence while parenting. In addition, few studies have included African-American women with severe mental illness (26,27,28). To our knowledge, no longitudinal studies of intimate partner violence have been conducted in this population. The following research question drives our study: What are the sociodemographic and clinical correlates of intimate partner violence in this sample of women with mental illness?

Participants

Participants were from a longitudinal, community-based study of mothers with severe mental illness (7) who were interviewed in three waves, each approximately 20 months apart over a period of six years (1995—2000). Participants were recruited from 12 community agencies and three psychiatric units in a large metropolitan city in southeastern Michigan. Criteria for study inclusion were being female, having a DSM-IV diagnosis of severe mental illness, being 18 to 55 years old, caring for at least one child between ages four to 16 years, and having an active case at study initiation (7). In this study we focused on the interviews conducted at the first wave (baseline) and the second wave (follow-up). A total of 379 women participated at baseline, and 324 (85%) participated at follow-up. This study focused on the 324 women with data from both waves.

Procedure

Interviewers were all women with at least an undergraduate degree in the human services. They were supervised by an interview coordinator, who ensured interview quality by reviewing audiotapes and checking completed forms. All interviews took place in the participants' homes, and the women were compensated $15 for each interview at baseline (at baseline, a two-part interview session was conducted one or two weeks apart between 1995 and 1996) and $30 for the follow-up session (interviews were conducted between 1997 and 1998). More detailed information on the sample recruitment procedures are described elsewhere (7,28). All participating women provided voluntary informed consent. The study was approved by the University of Michigan Institutional Review Board.

Names of eligible women were obtained from inpatient and outpatient public mental health agencies through case-manager referrals and record searches. To enroll clients in the study, eligible women were contacted either by a community mental health employee funded through the project or by their community mental health case managers. Women in inpatient units were recruited on the wards and interviewed 30 days later. A follow-up interview was conducted 20 months after baseline. At follow-up, 324 women completed interviews, 25 refused, 23 could not be located, and seven had died, leaving a completion rate of 85%.

Sample characteristics

As shown in Table 1, of the 379 participants at baseline, 337 met DSM-IV criteria for having one of six lifetime diagnoses, including schizophrenia (9%), schizoaffective disorder (9%), major depression (35%), major depression with psychotic features (11%), bipolar disorder (11%), and bipolar disorder with psychotic features (14%). The 42 women (11%) with no research diagnosis included 14 women who did not complete the diagnostic interview and 28 women for whom the interview yielded insufficient information for diagnostic determination. The median age reported for onset of mental illness was 27 years. Age of onset was measured by first psychiatric hospitalization, or alternately, those without psychiatric hospitalizations were asked for the age at which they first saw a psychiatrist or other mental health professional.

Sixty percent of the participants were African American, 29% were Caucasian, 8% were Hispanic, and 3% were of other racial or ethnic backgrounds. For purposes of this analysis, race or ethnicity was coded as 1 for African Americans and 0 for Caucasians, Hispanics, and others collapsed into one group. A total of 153 participants (40%) had some college education or more, 93 (25%) had a high school diploma or GED, and 133 (35%) had less than a high school education. Participants' median age was 36.5 years.

All participants had at least one child between the ages of four and 16 for whom they had child care responsibilities, defined as having overnight contact at least once a week. Respondents reported having a median number of three children with a median age of 11 years (ranging from infancy to adulthood). Seventy-five participants (20%) had only one child, 197 (52%) had two or three children, and 106 (28%) had more than three children. A total of 127 participants (34%) were married or living with a partner; 129 (34%) were separated, divorced, or widowed; and the remaining 123 (32%) were single, having never married. The median family income was $929 a month. After the analysis adjusted for individual women's household composition, 261 participants (69%) fell below the federal poverty threshold—that is, a mother with one child receiving $10,815 annually.

Measures

At baseline two interview sessions were conducted with each of the 379 participants (one or two weeks apart, on average), lasting about 60 to 90 minutes each. The material was divided between two interview sessions to ensure that all critical information was gathered in the first interview, should the participant not be interviewed a second time. Information gathered in the first session covered personal history, parenting, social context, and community functioning. The second part (N=324) contained modules of the Diagnostic Interview Schedule (29) and a slightly modified version of the Drug Abuse Screening Test (DAST) (30).

The study's dependent variable, intimate partner violence, was originally constructed by project staff to reflect the content of the physical violence items from the Conflict Tactics Scales (31). At follow-up, the participants were asked four questions: Over the past 12 months did someone you were romantically involved with ever push, grab, or slap you? Did they ever hit you with a fist or an object, kick you, or beat you up? Did they ever choke you, tie you up, or physically restrain you? Did they ever force sexual activity that you didn't want to happen? Intimate partner violence for this study was measured by whether the participant reported the experience of any violence by a romantic partner over the past 12 months (coded as 1, yes, or 0, no).

The study's main independent variable, substance use disorder, was measured with a modified version of the DAST. Unlike the original 20-item DAST, our version included 19 questions that asked participants to report on their lifetime history of substance use, including alcohol.

The study of mothers with severe mental illness used the 20-item version of the DAST and adapted it as described below to be appropriate for use with mothers with mental illness. In general, the word alcohol was added to all the questions. One original question was split into two: Question 11 from the DAST—"Have you ever neglected your family or missed work because of your use of drugs?"—was split into "Have you ever neglected your family obligations because you have been using drugs and alcohol?" and "Have you ever missed work because you have been using drugs or alcohol?" This was done so that specific information about family neglect could be collected and separated from work issues. Questions 12 and 13 of the DAST were combined into one question: "Have you ever been in trouble at work or lost a job because of drug and alcohol use?" This was done because few women in the sample were working or had worked. Questions 19 and 20 were combined into one question: "Have you ever gone to anyone for help or been involved in a treatment program for a drug or alcohol problem?"

Mothers were asked whether they had used ever used alcohol or any drugs, such as illicit drugs, had ever used nonmedical drugs, had ever used prescribed or over-the-counter drugs in excess of the directions, and had experienced drug-related problems in the past 12 months. A cutoff score of 5, rather than 6 as customary, was utilized to indicate whether there was a history of any past problems with alcohol or drugs. This lower cutoff point was used because we had reduced by 1 the total DAST score that could be obtained. Also, as suggested in the literature, lower cutoff scores in psychiatric populations are appropriate (32). The reliability estimate of the 19-item DAST in our sample was α=.94 (mean±SD correct items=5.01±4.76).

Control variables included race or ethnicity, age, marital status, and level of education. Marital status was a dummy variable, coded as 1, for single, separated, or divorced, and 0, for married or widowed. Clinical variables included number of lifetime hospitalizations (from date of first psychiatric hospitalization or first contact with a psychiatrist or other mental health professional), and duration of mental illness was calculated by using the difference between the date respondents reported first having seen a psychiatrist or other mental health professional and the date of the initial interview (coded for this analysis as 0, less than five years, or 1, five or more years). The diagnostic portion of the interview consisted of the depression, mania, and psychosis sections of the Diagnostic Interview Schedule, version III-R, modified for DSM-IV criteria. For the analysis presented here, diagnostic categories used in this study were dichotomized into affective disorders (major depression and bipolar disorder without psychotic features) and psychotic disorders (schizophrenia, schizoaffective disorder, major depression with psychotic features, and bipolar disorder with psychotic features) (coded as 0, psychotic disorders, or 1, affective disorders). Psychiatric symptomatology was measured by using the Colorado Symptom Index (33).

Analysis plan

As described in the measures section, we first calculated the overall prevalence rate of the dependent variable, intimate partner violence at follow-up. Bivariate analyses were conducted to examine relationships between variables in the study. Multiple logistic regression analysis was used to examine the sociodemographic and clinical characteristics (independent variables) at baseline that were correlated with intimate partner violence at follow-up.

Bivariate analyses

At follow-up, the prevalence rate for intimate partner violence in this sample was 19% (N=72). Intimate partner violence at follow-up was significantly associated with elevated rates of alcohol and drug use at baseline (42 persons, or 58%; χ²=7.5, df=1, p=.006). The rate of intimate partner violence for African Americans was higher (50 persons, or 26%) than that of non-African Americans (18 persons, or 19%), but this difference was not significant (p=.053). There were no significant relationships between intimate partner violence and psychiatric diagnoses, marital status, or duration of mental illness.

Independent-samples t tests were used to compare the group that experienced intimate partner violence with the group that did not. Results showed that the women who reported intimate partner violence were younger (mean of 34.6±6.4 years) than those who had not reported it (mean of 36.8±6.7 years). Those who had experienced intimate partner violence also reported more psychiatric symptoms (mean of 3.05±.82) than the group that did not report intimate partner violence (mean of 2.74±.82). There was no association between intimate partner violence and the length of time a woman had been hospitalized over her lifetime.

Multiple logistic regression analysis

Findings from the multiple logistic regression analysis indicated that age, number of hospitalizations, number of symptoms, and elevated levels of alcohol and drug use at baseline were significantly associated with the report of intimate partner violence at follow-up (Table 2). Specifically, older women were at less risk of experiencing intimate partner violence, compared with younger women (odds ratio [OR]=.95, 95% confidence interval [CI]=.90—.99). The odds of experiencing intimate partner violence at follow-up were slightly higher among women who had experienced a greater number of lifetime psychiatric hospitalizations (OR=1.05, CI=1.00—1.09). Perhaps more significantly, mothers had significantly greater risk of experiencing intimate partner violence at follow-up if at baseline they had a greater number of psychiatric symptoms (OR=1.48, CI=1.03—2.11) and elevated levels of alcohol and drug use (OR=1.97, CI=1.10—3.52).

Sociodemographics variables

In our sample we found that 19% of the participants reported intimate partner violence. However, other studies have found prevalence rates between 14% and 28% (6,34,35, Elliott DM, Briere J, unpublished manuscript, 2003). We examined the distribution of intimate partner violence by race or ethnicity, age, marital status, and educational attainment. It would appear that research regarding the relationship between race or ethnicity and partner violence is equivocal. Some studies have found a significant association—that is, they found that African-American women were more likely than Caucasian women to experience intimate partner violence (6,21,36,37)—whereas other researchers have not found a significant association with racial or ethnic background (38).

Our study found age to be the only sociodemographic variable that was significantly associated with intimate partner violence. That is, younger women were more likely than older women to report experiencing violence by an intimate partner. This finding is consistent with the hypothesis that younger women are more likely to be materially dependent on their abusive partners, particularly when they have young children to raise. It also supports previous research showing that women between the ages of 19 and 29 years and women in families with incomes below $10,000 were more likely than other women to be victims of violence by an intimate partner (34).

Extant research also suggests that being single or separated is positively associated with the experience of intimate partner violence (34,39), whereas education generally has an inverse relationship with intimate partner violence (40,41,42). In our study, partner violence was not significantly associated with marital status or with education. To our knowledge, this is the only longitudinal study that examined mothers with severe mental illness to determine whether alcohol and drug use is associated with intimate partner violence. Contrary to previous studies of samples of women without severe mental illness, it is significant that our longitudinal study found that race or ethnicity, education, and marital status were not associated with the risk of intimate partner violence. These findings would suggest that sociodemographic variables may play less of a role than clinical variables or even the mental illness itself in the experience of intimate partner violence. Further research is needed to replicate this finding.

Clinical variables

The severity of mental illness, as reflected by the number of hospitalizations and the level of symptomatology, and elevated levels of substance use were positively correlated with the subsequent experience of partner violence. These findings indicate that women who are dually diagnosed as having a severe mental illness and a substance use disorder are more likely than women with only the diagnosis of mental illness to be involved in a relationship with an abusive partner. It is emphasized here that the positive correlation between dual diagnosis and intimate partner violence is not meant to imply a causal relationship between women's substance use and male-on-female violence. Studies of partner violence have been inconsistent in showing a direct relationship between intimate partner violence and substance use (1). Most research argues that the relationship between intimate partner violence and substance use is a complex phenomenon, whereby one probably increases the risk of the other (1,2). Women with severe mental illness, particularly schizophrenia, are at increased risk of victimization and HIV infection, because deficits in social competence may interfere with, among other things, the ability to negotiate or avoid dangerous situations (21,22,43).

Interestingly, two clinical variables—duration of mental illness and having a diagnosis of psychotic disorder—did not predict partner violence, whereas having a diagnosis of an affective disorder was predictive. It would appear, however, that regardless of the length of time a woman has had a mental illness diagnosis, the number of hospitalizations is a better predictor of whether she reports experiencing intimate partner violence.

Study limitations

Although an important characterization of experiences of intimate partner violence was obtained with this sample, the results may not generalize to women who are not in the care of the public mental health system or who live in nonurban areas. It is possible that including women with severe mental illness from other settings (for example, substance abuse treatment programs, hospitals, jails, or prisons) would have produced different findings. In addition, it is possible that the use of a 20-month interval between interviews was not sensitive enough to capture the full experience of intimate partner violence. Finally, because the study was not originally designed to focus on intimate partner violence, the measure of intimate partner violence may be limited. A more comprehensive battery of questions on intimate partner violence is needed to more accurately capture these experiences. Nonetheless, despite these limitations we were able to longitudinally assess experiences of intimate partner violence in a relatively large sample of women, including African Americans, using a comprehensive assessment of mental illness and alcohol and drug use. Even though the data are nearly ten years old, they represent the most current analysis of substance abuse and intimate partner violence within a population of mothers with severe mental illness. Also, the age of the data suggest that more research should be undertaken to explore this population.

The profile of risk of intimate partner violence that emerges from this study of women with severe mental illness—being young, having a co-occurring substance use disorder, and being diagnosed as having a psychiatric disorder—suggests that more prevention efforts need to target mothers with a dual diagnosis. Violence against women is acknowledged as a major public health problem worldwide (39,44). According to estimates from the National Crime Victimization Survey, in 2001 a total of 588,490 victimizations by intimate partners (85%) were against women (45). Taking into account the likelihood that 25% to 35% of average adult women have been sexually abused as children (46), violence against females takes on a pressing social and political urgency.

Historically, different treatment philosophies, different policies, and different funding streams (12,47) have separated mental health and substance abuse services within the public sphere. However, the need for the integration of comprehensive services has never been greater. We must have professionals and treatment facilities that are qualified to serve the growing group of women who are coping with severe mental illness, substance misuse, and victimization. It was noted above that there is a lack of empirical studies of women that examine the intersection of dual diagnosis and intimate partner violence. It may be the case that in a population of mothers with severe mental illness, clinical variables such as alcohol and drug use are more important than demographic characteristics in predicting the reported experience of intimate partner violence. Further research should continue to examine this phenomenon and to identify effective ways of preventing intimate partner violence in this vulnerable population.

This research was supported by grant R01-MH-54321-0 from the National Institute of Mental Health, by the University of Michigan Substance Abuse Research Center, and by a National Research Service Award (T32-DA-007267) from the National Institute on Drug Abuse. The authors appreciate the helpful comments of Deborah Bybee, Ph.D., concerning the construction of the intimate partner violence measure.

The authors report no competing interests.