Treated Mental Illness and the Risk of Child Abuse Perpetration

The MacArthur database included 1,136 individuals who were followed after a psychiatric hospital discharge from three acute civil inpatient facilities (in cities in Pennsylvania, Missouri, and Massachusetts). By definition, these 1,136 participants included individuals who had been at recent elevated risk from mental illness or substance misuse and who had been treated. English-speaking patients ages 18 to 40 were included. Mean length of hospitalization was nine days. In addition to the recently hospitalized sample, a comparison group of 519 persons who were matched for neighborhood (census tract) was included.

In this secondary analysis, we included all those with follow-up data at the first ten-week interview who reported that they had children. We categorized parents in the participant group into four diagnostic groups: serious mental illness only, a substance use disorder only, both a serious mental illness and a substance use disorder, and another issue that required psychiatric hospitalization. [A flowchart in an online supplement to this article illustrates inclusion in the participant and comparison groups.] We used the diagnosis at hospital discharge; these were considered most accurate because they were based on the participant’s entire hospitalization. Serious mental illness was defined as including schizophrenia or a psychotic disorder, bipolar disorder, or recurrent or severe major depressive disorder: schizophrenic disorders (ICD-10-CM codes F20.X), schizotypal disorders (F21), schizoaffective disorders (F25), other nonorganic psychotic disorders (F28 and F29), bipolar disorder (F31.X), serious depressive episode with psychotic symptoms (F32.3), and recurrent serious depressive disorders (F33). Substance use disorders were then categorized separately from serious mental illness. Examples of diagnoses that were not included as serious mental illness but included as “another issue” were personality disorder, marital problem, other interpersonal problem, phase-of-life problem, malingering, no axis I diagnosis, dementia, anxiety disorders, eating disorders, adjustment disorder, dysthymia, cyclothymia, posttraumatic stress disorder, intermittent explosive disorder, and attention-deficit disorder.

Collateral informants were chosen by participants and comparison group members at each follow-up as the person most familiar with the participant’s behavior during the time period. The collateral informants for the participant group were most often family members (47%) or friends (24%) but also included mental health professionals (6%). Participants and their collateral informants were interviewed at baseline and every ten weeks for the next year. Comparison group members and their collateral informants were interviewed about the past 10 weeks. Interviews included a detailed description of violence. Data from arrest records and rehospitalization were also obtained for both groups.

The outcome measure used in the study reported here was violence by a parent toward a child. The MacArthur study modified the Conflicts Tactics Scale for the interviews. Participants, comparison group members, and their collateral informants were asked about violence by the participant or comparison group member in the past 10 weeks. The recorded violence included both whether or not a violent incident had occurred, and if such an incident had occurred, the number of violent acts occurring in that incident. We included all acts of violence, which was defined to include assaultive acts, whether or not the act resulted in physical injury or involved use of a weapon; sexual assaults; and threats made with a weapon in hand. This was a more inclusive definition of violence than in the original study because each of the listed acts may be conceptualized as child abuse. Participants, comparison group members, and collateral informants were asked for detailed information about the target of the violence. We considered cases of either self-report or collateral report of violence toward a child. The MacArthur study used two independent coders when considering acts of violence, in order to reconcile reports.

To describe rates of violence against children by parents in both groups and related factors, analysis was conducted in SPSS, version 21. Chi-square tests were used as appropriate to analyze comparisons. Our analysis of the publicly available database was exempt from review by the university’s ethics committee.

Of the 1,136 participants, 549 (48%) reported that they had children (total of 1,228 children). Data on psychiatric diagnosis, gender, and parenthood for the participants are presented in Table 1. Depending on the serious mental illness requiring hospitalization, 46%−63% of women were mothers and 28% to 37% of men were fathers. Of the 525 participants who qualified for a diagnosis of serious mental illness at hospital discharge, 221 (42%) reported that they had children. A total of 846 had follow-up data, and of those, 416 had children (total of 929 children) [see online supplement]. Of the 519 individuals in the comparison group, 299 (58%) reported that they had children. Table 2 presents data on demographic and clinical characteristics of the participants and the comparison group.

Among the 846 participants with follow-up interview data at 10 weeks, 22 committed violence toward a child (3%) in the first 10 weeks. This included two violent incidents committed by participants who were not parents. Of the 416 parents in the participant group with follow-up data, 20 (5%) were violent toward a child in the ten weeks after hospital discharge (Table 3). One incident involved two child victims. Among those with a diagnosis of serious mental illness only, 8% committed violence toward a child, as did 3% of those with substance use disorder only, 4% of those with serious mental illness and substance use disorder, and 7% of those with another issue. Of these violent parents, ten had serious mental illness (either alone or with a co-occurring substance use disorder); five of the ten had recurrent major depressive disorder, and five had bipolar disorder; four had a co-occurring substance use disorder. The remainder of parents in the participant group who committed violence toward a child either had diagnoses of a substance use disorder or had another issue.

Of the 20 parent participants who committed violence toward a child in the ten weeks after discharge, all but four (20%) reported the violent incident to interviewers (that is, in only four reconciled cases did a collateral informant assert that there was violence toward a child but the participant did not report it). All 20 parents reported that they were not experiencing the psychotic symptoms of hallucinations or delusions around the time of the violent incident. However, the presence of delusions at the time of the interview was correlated with violence in the preceding ten weeks (p=.012). One of the violent parents in the participant group was rehospitalized during the follow-up period, and one participant was arrested after the violent incident. Of the 20 parents who were violent toward children, 13 were prescribed psychiatric medications. Eleven of the 20 were receiving additional mental health or substance abuse treatment during the ten-week follow-up period; however, treatment had ceased prior to the interview for five. In the ten weeks after hospital discharge, parents perpetrated 35 incidents of violence toward children, including 14 incidents described as slapping; 12 as pushing, grabbing, or shoving; six as hitting with a fist or object; two as throwing something; and one as kicking, biting, or choking the victim. In a single incident, the participant described slapping the child more than 20 times. None of the participants with serious mental illness caused injury to the child that required medical attention. Two participants with a substance use disorder and one with another issue caused injury to the child requiring medical attention (seven incidents committed by these three participants resulted in bruises or cuts).

Of the 299 parents in the comparison group, 41 (14%) committed violence toward a child in the prior ten weeks. This included both self-report and collateral report. In the comparison group, many parents who committed violence toward a child were detected by collateral informants rather than by self-report, in contrast to the patient group in which most parents who committed violence toward a child were detected by self-report. In the comparison group, 27 parents admitted violence toward a child in the prior ten weeks, and four of the 27 said that violence was severe enough to leave a bruise. In the comparison group, 18 collateral informants reported violence toward a child, and in only four of the 18 cases did the parent self-report the violence. Of the 213 acts of violence toward children committed by the comparison group members, 130 were described as slapping; 81 as pushing, grabbing, or shoving; and two as throwing something. This included four parents who said that they slapped, pushed, or grabbed their child between 20 and 70 times in one incident. Some parents in the comparison group had a mental disorder or a substance use disorder (Table 2), but none had required recent psychiatric hospitalization. Of the 41 comparison group parents who were violent toward a child, seven (17%) were taking psychiatric medication and five (12%) had delusions present at time of the ten-week interview.

The prevalence of violent acts was 5.81 violent acts per 100 children for the patient group and 31.14 violent acts per 100 children for the comparison group. There were no child sexual assaults nor threats with weapon in hand reported in either group.

Violence was assessed only for one ten-week period among the community comparison group, whereas it was considered over the course of a year for the participants. Over the 50-week period, 37 parents in the participant group committed an act of violence toward a child that was corroborated by a collateral informant. In sum, participants were significantly less likely than members of the comparison group to have committed violence toward a child during the 10-week period, a period that immediately followed hospital discharge for the participants (5% versus 14%; χ²=16.55, df=1, p<.001) (Table 3). Among mothers, those in the participant group were significantly less likely than those in the comparison group to have committed violence toward a child (9% versus 18%, p=.008).

Individuals with mental illness are often parents. Our findings suggest that parents whose serious mental illness is treated are not at increased risk of violence toward their children and that, in fact, those who have been admitted to an acute psychiatric facility, treated, and recently discharged appear to be at lower risk of violence toward their children than other parents in their communities. Furthermore, use of collateral informants for both groups allowed us to conclude that parents with mental illness were more likely than their community counterparts to report violence toward their child when it occurred.

Most psychiatrists report that they routinely ask their female patients whether they are mothers. However, in one survey, 10% admitted that they ask less than one-third of the time or never (16). In the study reported here, of those released from a psychiatric hospital, 46% to 63% of the women with serious mental illness and 65% of the women with substance use disorders were mothers. Similarly, 28% to 37% of the men with serious mental illness and 48% of the men with substance use disorders were fathers. Parenting should be routinely considered in mental health evaluations of both women and men. Among parents, maintaining good mental health may be particularly important. Seeman (17) recommended that parents, in collaboration with mental health services, learn to monitor themselves for warning signs of relapse, take advantage of available parenting resources, and develop a crisis plan.

Mental illness is stigmatized, and parents are often dissuaded from seeking treatment because of concerns about being reported to CPS and custody loss (1,7). Instead, seeking psychiatric care should be seen as a protective factor against violence by parents with mental illness. Seeman (17) noted, “As symptoms of psychosis decline, parenting stress is reduced and the quality of parenting inevitably improves.” As we hypothesized, parents who were recently treated and released from psychiatric hospitals were no more likely than parents in the comparison group to be violent toward children. In fact, they were significantly less likely to be violent, which may be attributable to treatment of their symptoms of mental illness and substance use disorders. The symptomatic presence of delusions at the time of the interview correlated with violence against children in the previous ten weeks. Studies about child abuse often collapse mental illness, substance abuse, and personality disorders into a single category. Substance use disorders and personality disorders should be considered as risk factors separate from serious mental illness in assessment of children’s risk, as they are for other types of violence, and should be managed accordingly.

Of note, the rates of violence against children in both the participant and the comparison groups were greater than commonly reported rates (18). This was likely because of the use of both self- and collateral informant reports, an approach that is likely to yield more accurate findings. The rate of parental violence toward a child in the participant group was lower over the study year than the rate in the community group in only 10 weeks. On the basis of collateral informant reports, we found that parents recently discharged from a psychiatric hospital were more likely than their community counterparts to self-report their own violence. If replicated, this finding may suggest that persons without mental illness underreport their own violence toward children, compared with those with mental illness. Thus surveys may appear to overrepresent persons with mental illness among perpetrators.

Along with its strengths, the study had some methodological limitations. Using the MacArthur database allowed us to avoid most of the biases in studies of parental mental illness and child abuse. The MacArthur data were collected contemporaneously rather than, as in most studies, retrospectively. Comparison of parents with treated mental illness with parents in the same community was novel. Furthermore, careful definitions of serious mental illness and data from discharge records were used, rather than vague self-reports. We were able to consider substance use disorders separately from mental illness, another frequent methodological limitation. We used a specific definition of violence rather than vague maltreatment definitions, which do not include verification but merely suspicion. Collateral reporting also increased the likelihood that information about violence was accurate. However, this secondary analysis of data was unable to examine neglect or emotional abuse.

This study considered serious mental illness, including individuals who were so unwell as to require psychiatric hospitalization. Compliance with appointment attendance was known, but compliance with medications was not known. Some persons in the comparison group also experienced symptoms of mental illness (although not severe enough to warrant psychiatric hospitalization) and substance use disorders. However, there was not a significantly increased proportion of persons with mental illness among those who committed violence against children.

Even in this well-designed study with a large sample of persons released from psychiatric hospitals, the subsamples with specific mental illnesses were too small for analytical comparisons. Therefore, we placed those with serious mental illnesses in a single category.

This study did not address filicide (child murder by a parent). More filicides occur because of fatal child maltreatment than because of purposeful homicide by a parent with mental illness. The literature suggests that when a mother with mental illness commits purposeful homicide (rather than killing “accidentally” due to abuse), she is often acutely psychotic, depressed, or suicidal (19,20). However, with treatment and symptomatic improvement, the risk of homicide is expected to decrease.

Finally, the rates of parental violence toward children were higher than published rates in other studies. Because of the downward socioeconomic drift of persons with serious mental illness, the participants may have been living in relatively impoverished neighborhoods and were compared to those living in these neighborhoods. Thus it is possible that child abuse is higher in the community group than usual for this reason.

This study found that parents with serious mental illness who had been recently discharged from psychiatric hospitals were not at higher risk than other parents in their community of perpetrating violence toward children. Studies which purport to demonstrate an elevated risk of child abuse among parents with mental illness often have serious methodological issues, including conflating substance misuse and mental illness, and do not consider whether the mental illness is treated. Furthermore, inappropriately labeling individuals with mental illness as violent is likely to lead to lower rates of help seeking and thus increased symptoms, resulting in higher rates of CPS intervention and loss of child custody. Instead, parents with mental illness should be identified, treated, and encouraged to continue treatment.

The authors appreciate the insights and suggestions of Joshua Friedman, M.D., Ph.D.