Parental substance use disorder and child abuse: risk factors for child maltreatment?

Although the literature consistently shows an association between parental substance use disorders (SUDs) and child abuse, it is unknown what factors discern non-abusive and abusive parents with SUDs. This study aims to investigate which specific risk factors are associated with child abuse in clinically treated parents with SUDs in the Netherlands. It examines two groups of parents with SUDs in a clinical setting, with and without known instances of child abuse. These groups were compared on SUD-related factors such as the type and severity of the SUDs, and variables addressing psychological disability such as comorbid diagnoses and quality of life. Other than a marginally significant difference in severity of addiction and a lower mean age of the parents in the child abuse group, no significant differences were found. The small sample size and the inherent difficulty of studying SUDs in a clinical sample is likely to have affected the results.

Introduction

Substance use disorders (SUDs) are a widespread and harmful condition. The prevalence of 12-month and lifetime drug use disorder, as diagnosed with the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM-5) (American Psychiatric Association, 2013), is estimated at 3.9% and 9.9%, respectively (Grant et al., 2016). Crucially, children growing up with parents with SUDs suffer serious consequences from such circumstances (Brown, Cohen, Johnson, & Salzinger, 1998). Besides other negative effects, parental SUDs are consistently identified as a serious risk factor for child physical abuse (Dubowitz et al., 2011; Laslett, Room, Dietze, & Ferris, 2012). Children of parents with SUDs, therefore, are at risk for abuse, which could manifest in serious physical, emotional and psychological trauma (e.g. Raitasalo & Holmila, 2016; Schury & Kolassa, 2012). In the Netherlands, an estimated 3% of all youth in the country are subjected to child abuse each year (Alink et al., 2018), and an estimated 3 to 10% of children under the age of 18 have experienced or been threatened with physical abuse (ten Boom & Wittebrood, 2019). A large portion of this population – around 23,000 children in the Netherlands – live with parents who suffer from SUDs (Goossens & van der Zanden, 2012).

Due to the danger that parents with SUDs may pose to their children, it is essential to be able to adequately identify those parents who are at risk of committing child abuse. There is a large body of literature on the risk factors for child abuse and more generally domestic violence, including intimate partner violence. Even though intimate partner violence is not the same as child abuse, there is an overlap; in 40% of the families in which intimate partner violence occurs, the child or children also suffer abuse (Herrenkohl, Sousa, Tajima, Herrenkohl, & Moylan, 2008). Drug use has been consistently related to all forms of aggression and violence, including intimate partner violence (Cafferky, Mendez, Anderson, & Stith, 2018; Foran & O’Leary, 2008) and child abuse (Stith et al., 2009). In a review of 285 studies analysing the relationship between substance use and intimate partner violence, Cafferky et al. (2018) found alcohol and different types of drugs to be equally strongly related to intimate partner violence, whilst an earlier meta-analysis of 96 studies reported similar findings, with an especially strong relationship for cocaine (Moore et al., 2008). With regard to child abuse, Stith et al. (2009) conducted a meta-analysis of 155 studies that had empirically examined the relationship between an identified risk factor and either child physical abuse or neglect, in order to generate composite effect sizes for 39 risk factors of child physical abuse. They found that alcohol abuse is more significantly related to child physical abuse than is drug abuse. As such, a potential factor for discerning violent parents with SUDs from non-violent parents is the type of substance being abused.

In Stith et al.’s (2009) meta-analysis, medium to large effect sizes were also found between child physical abuse and several parental characteristics, such as depression and psychopathology (conceptualized as any measure of mental illness besides depression or anxiety, such as schizophrenia, bipolar disorder, etc.). This suggests that comorbid disorders may play a role in the relationship between SUDs and child abuse. Especially relevant is the comorbidity of Cluster B personality disorders, which are consistently linked with intimate partner violence as well as SUDs (e.g. Kelley & Braitman, 2016) and aggression in general (e.g. Goldstein et al., 2006). Furthermore, Stith et al. found large effect sizes for anger and hyper-reactivity and moderate effect sizes for parental coping and problem-solving skills, suggesting that it may be important to address personality disorders when studying the relationship between SUDs and child abuse.

Lastly, environmental and other psychosocial factors have been identified as risk factors for child abuse. For instance, strong and very strong correlations have been found between neighbourhood poverty and child physical abuse and neglect, respectively (Drake & Pandey, 1996). Similarly, an overview of 18 ecological studies of child maltreatment found that areas with higher levels of poverty have higher rates of child maltreatment. In addition, relationships have been found between drug and alcohol availability in neighbourhoods and child maltreatment (Freisthler, Merritt, & LaScala, 2006) and between unemployment and child physical abuse (Stith et al., 2009). Furthermore, weak family bonds (CASAColumbia, 2012; Stith et al., 2009) and parenting stress (Rodriguez & Green, 1997; Stith et al., 2009) – meaning the amount of stress experienced by individuals in their role as a parent – have been found to exhibit correlations with child abuse.

In a longitudinal study of risk factors for child maltreatment, it was found that the risk of child maltreatment increases with the number of risk factors present, but especially when more than four risk factors are present (Brown et al., 1998). The effects of accumulating risk factors have also been shown to be predictive of child maltreatment in other studies (Ridings, Beasley, & Silovsky, 2017; Solomon, Åsberg, Peer, & Prince, 2016). This underlines that an interplay of factors may be responsible for child abuse.

Many of the risk factors for child abuse and maltreatment discussed above are highly prevalent among patients with SUDs. As such, it may not be possible to use some of these risk factors to distinguish between parents with SUDs who abuse their children and those who do not. Evidently, not all parents who suffer from SUDs commit violence against their children. This raises the question of whether or not there are differentiating factors that could be used to discern non-abusive parents with SUDs from abusive ones. This article aims to identify such factors by comparing two groups of parents with SUDs who were enrolled in a SUD treatment programme, either with or without documented instances of abuse.

It was hypothesized that parents who abuse their children are more likely to have psychological comorbidity, particularly Cluster B personality traits (Kelley & Braitman, 2016). It was also hypothesized that abusive parents with SUDs are more likely to use alcohol or cocaine than non-abusive ones (see also Moore et al., 2008; Stith et al., 2009). Lastly, based on the idea that there may be cumulative dose-response effects of multiple risk factors (see Brown et al., 1998; Ridings et al., 2017; Solomon et al., 2016), it was expected that child abuse would be found to be associated with a higher amount of psychosocial and environmental problems.

Method

Participants

Both the abuser group and the comparison group consist of 6 female patients (32%, 95% CI [18.4, 47.4]) and 13 male patients, a distribution that is similar to the distribution in the general patient population in mental health institutions for addicts in the Netherlands, in which 34% of the patients are women (Wisselink, Kuijpers, & Mol, 2016). All individuals are of Dutch nationality. The mean age is significantly lower in the abuser group (M = 34.1 years, SD = 9.2) than in the comparison group (M = 42.6 years, SD = 10.7), t(36) = −2.6, p < .01, meaning that the average age in the comparison group is comparable to the average age in the general patient population in mental health institutions for addicts in the Netherlands (M = 42.0 years). There is no statistical difference in the average number of children, t(36) = −0.3, p = .79, between the abuser group (M = 2.50 children, SD = 0.96) and the comparison group (M = 2.60 children, SD = 1.43). The distribution of the type of problematic substance across the patients in the comparison group is similar to that of the general Dutch patient population (cf. Wisselink et al., 2016). Alcohol is the most common problematic substance in the comparison group (61%, 95% CI [44.7, 76.3]), which is also the case in the Dutch patient population (45%). Also, the distribution of the other substances in the general population all fall within the confidence intervals of the comparison group. Hence, the comparison group can be considered a representative sample of the general SUD patient population that is currently in mental health care institutions for addicts in the Netherlands (Wisselink et al., 2016).

All participants were enrolled in outpatient treatment plans such as Flexible Assertive Community Treatment (FACT), methadone treatments or family treatment programmes. All patients in the comparison group were self-referred for treatment, whereas those in the abuser group were referred (albeit not court-mandated) into care by a domestic violence agency. Unfortunately, the ages of the patients’ children were not registered in the electronic patient files.

Instruments

The Measurements of Addiction for Triage and Evaluation (MATE; Schippers, Broekman, Buchholz, Koeter, & van den Brink, 2010) instrument was used to obtain information about the severity of the SUDs and the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition, Text Revision (DSM-IV-TR) diagnosis (American Psychiatric Association, 2000). Conceptually, the MATE was constructed according to the International Classification of Diseases (ICD) and the International Classification of Functioning, Disability and Health (ICF) in the World Health Organization’s (2016) classification system. The instrument consists of 10 components that explore, amongst other aspects, current problematic substance abuse and history of drug use, history of healthcare, somatic complaints and social and environmental support networks. It assesses the diagnoses of dependence and abuse according to the DSM-IV-TR. The MATE yields 20 sum scores, which can be summarized into 4 dichotomized dimension scores: severity of addiction, severity of psychiatric comorbidity, severity of social disintegration and history of treatment for SUDs. The severity of addiction is classified as either high or low, based on relatively high and frequent consumption of psychoactive drugs and/or consent with a high number of items in the DSM’s dependency and abuse section and/or a very high score on the craving for the main problem drug (Schippers & Broekman, 2007). As such, this score is an automatic outcome of the MATE. The severity of social disintegration is derived from substantial problems in areas of basic life conditions (housing, feeding, clothing, etc.) and/or substantial negative external factors in the environment that influence the person’s health (Schippers & Broekman, 2007). The MATE is considered to have good inter-rater reliability (Schippers et al., 2010) and adequate validity when compared to the World Health Organization’s Disability Assessment Schedule (Schippers & Broekman, 2007).

The Manchester Short Assessment of Quality of Life questionnaire (MANSA; Priebe, Huxley, Knight, & Evans, 1999) was used as an indication of the self-perceived quality of life. The Dutch MANSA is a self-assessment questionnaire consisting of 12 items that address the quality of a wide variety of environmental, social and occupational areas of one’s life – for instance: ‘How happy are you with your current job, or in case of unemployment, how happy are you with being unemployed?’. The areas of life are scored on a 7-point Likert scale ranging from 1 (could not be worse) to 7 (could not be better). This results in an overall score ranging from 12 to 84, with higher scores indicating a higher perception of quality of life. The psychometric properties of the MANSA are satisfactory, with good reliability and sufficient validity scores (Priebe et al., 1999).

Psychosocial and environmental problems were derived from the registrations in the electronic patient files on Axis IV in the DSM-IV-TR, which is used descriptively to identify the different life domains in which the patient is experiencing difficulties – occupational problems, problems with their primary support group, economic problems and so forth. The number of factors specified in the diagnoses serve as an indicator of the amount of psychosocial and environmental problems. Additionally, the DSM-IV-TR diagnosis was used to establish whether or not there was psychiatric comorbidity by making note of any non-substance-related diagnoses on Axes I and II. A potential comorbid disorder was categorized as a mood disorder, anxiety disorder, personality disorder or developmental disorder to allow for analysis of the type of psychiatric comorbidity. Importantly, this study used patient files from 2011 until 2016, when the files and treatment locations did not yet apply the newer DSM-5 criteria.

Procedure

The study was conducted on the electronic patient files of a large addiction care organization in the north of the Netherlands called Verslavingszorg Noord Nederland (VNN). During the admission intake at VNN, the MATE (Schippers et al., 2010) is administered by MATE-trained professionals to acquire information about the severity of the substance abuse, the severity of social disintegration and the DSM-IV-TR classification. Based on the information derived from the MATE and the admission intake, (clinical) psychologists, masters in addiction medicine or psychiatrists determine the DSM-IV-TR classification of substance abuse or dependence disorder. The MANSA (Priebe et al., 1999) is also administered as part of the standard intake procedure. This study made use of the data regarding the demographic characteristics, MATE, MANSA and the psychiatric diagnoses in the electronic patient files.

This study used a comparative research design wherein two groups of parents with SUD – one abusive to their children and one non-abusive – were assessed for differences and similarities. Parents with a known history of child abuse were selected from the client database of an agency that focuses on safety and protection in cases of domestic violence in Friesland, a province in the north of the Netherlands, called the Regiecentrum Bescherming en Veiligheid (Control Centre for Protection and Safety, hereafter referred to as the Control Centre). Their client database contains different types of reported domestic violence and child abuse in the province. The participants in the abuser group were selected from this database based on three inclusion criteria. Firstly, the individual needed to be reported at the Control Centre for child physical abuse – which is defined as including acts of physical intimate partner violence committed in the presence of a child or children. Secondly, the individual needed to have children who were underage, i.e. below the age of 18 years. Lastly, the individual needed to be referred for treatment, guidance or rehabilitation for SUDs at VNN. Participants were selected by conducting a search query in the Control Centre database with the keyword VNN and a notion of domestic violence. This resulted in 9 cases that matched the inclusion criteria. In addition, cases were retrieved by asking case managers from the Control Centre to bring any cases forward that matched the inclusion criteria. This resulted in another 10 cases, increasing the total to 19 cases that met the preset criteria. Although it was ensured that all children were under the age of 18, there is no information as to the exact ages of the children of the parents included in this study. This group of parents is referred to as the abuser group.

In order to establish a credible comparison group of parents who have SUDs but do not have a known history of committing child physical abuse, the 19 cases in the abuser group were individually matched with patients that were selected from VNN’s electronic patient database. The parents in the comparison group are matched on gender, treatment location at VNN and date of admission to ensure that the comparisons cannot be skewed by potential variances in intake procedures or potential changes in procedures over time. For each individual in the abuser group, all possible matches from within the treatment location were retrieved. From these matches, the individual with children who had the date of admission closest to the individual in the abuser group was chosen. It was possible to match all cases on the same gender and treatment location. The date of enrolment was at most 84 days apart (M = 14.2, SD = 25.3). Potential matches with a more proximate date of enrolment were not selected because of the patients in question either being of a different gender or not being a parent. All participants from the comparison group were registered as having children, although they were not always underage. Moreover, upon verification at the Control Centre, these parents were never reported for possible child abuse. Unfortunately, it was impossible to match the patients on age due to the small number of potentially matching cases. Because of the high variability in treatment location procedures and staff, priority was given to matching cases on treatment location rather than age in order to minimize the variance in the data.

Results

Primary and secondary problematic substances

In the studied cohort, the most common problematic substance was alcohol in both groups. Within the abuser group, 53% primarily used alcohol compared to 68% in the comparison group. In the abuser group, 74% had a diagnosed secondary SUD, which in 79% of cases was tobacco. A similar distribution was found in the comparison group, in which 79% had a diagnosed secondary SUD, which was tobacco in 80% of cases (see also Table 1). These differences were not found to be statistically significant in chi-square analyses.

Table 1.

Primary problematic substance.

	Abuser group (n = 19)	Comparison group (n = 19)
Substance	%	%
Alcohol	52.5	68
Cannabis	31.5	5
Cocaine	16	0
Opioids	0	16
Other	0	11

Psychiatric comorbidity, psychosocial and environmental problems and quality of life

The abuser group was not found to have significantly more cases of concurrent diagnosed psychiatric comorbidity (53%) than the comparison group (37%), χ2(1) = 0.96, ns. In addition, no significant differences were found between the types of psychiatric comorbidity, t(136) = 0.91, ns. The abuser group had 2.8 reported psychosocial and environmental problems (SD = 1.2) while the comparison group averaged 2.4 (SD = 1.3). Furthermore, the mean MANSA scores are not statistically different between the abuser group (M = 48.5, SD = 8.8) and the comparison group (M = 51.6, SD = 12.4) (see also Table 2). Both groups reported just about adequate quality of life, on average.

Table 2.

Psychiatric comorbidity.

	Abuser group (n = 19)	Comparison group (n = 19)
Type	%	%
Mood disorder	16	0
Anxiety disorder	11	5
Personality disorder	5	16
Developmental disorder	21	16
None	47	63

History of treatment, severity of addiction and severity of social disintegration

There are no statistically significant differences between the two groups regarding their prior number of healthcare admissions for SUDs, severity of addiction and social disintegration. In the abuser group and the comparison group the patients had received roughly equal numbers of prior treatments (M = 1.1, SD = 1.0), t(36) = 0.17, ns. There is a marginally significant percentage difference in the severity of addiction between the groups – more patients in the abuser group were considered severely addicted to the primary substance (63%) than in the comparison group (37%), χ2(1) = 2.6, p = .10. Social disintegration does not differ significantly between the groups; two patients in the abuser group were considered highly disintegrated (11%) compared to three in the comparison group (16%) (see also Table 3).

Table 3.

Averages of psychosocial and environmental problems, MANSA score and history of treatment.

	Abuser group (n = 19)		Comparison group (n = 19)
Variable	M	SD	M	SD
Psychosocial and environmental problems	2.8	1.2	2.4	1.3
MANSA score	48.5	8.8	51.6	12.4
History of treatment	1.1	1.0	1.1	1.0

Discussion

The results of the current study show hardly any differences between parents with SUDs who do not have a known history of committing child abuse and those who do. Specifically, a marginally significant difference was observed between the two groups in terms of the severity of the addiction. However, no statistically significant differences emerged in the severity and type of the problematic substance, the amount and type of psychiatric comorbidities, the number of problems across the socio-environmental life domains and the reported quality of life. This was unexpected, as it was the results of previous studies which led to the hypotheses that the abuser group would have more psychological comorbidities (cf. Stith et al., 2009) particularly regarding personality disorders (cf. Kelley & Braitman, 2016), would be more likely to have a SUD involving alcohol or cocaine (cf. Moore et al., 2008; Stith et al., 2009) and would have a higher number of problems across the life domains (Brown et al., 1998; Ridings et al., 2017; Solomon et al., 2016) when contrasted with the comparison group. The lack of differences between the groups may be explained by several factors, of which the most recognizable is the small sample size, which significantly reduces the strength of the results in terms of statistical power; with only 19 cases per group, the statistical power is undermined. When looking at the data, the difference in addiction severity was found to be statistically insignificant but on face value appears interesting enough to warrant further attention. As such, further research with larger groups is needed to address this potential difference.

An important explanation for the lack of differences could also lie in the clinical nature of the current sample. All participants in both groups were being treated, rendering them by definition two clinical samples – but the abuser group was referred for treatment by an external institute (i.e. the Control Centre). In general, only a minority of individuals suffering from SUDs receive healthcare. For instance, in a study of American individuals, only 11% of those who needed specialized treatment actually received it (Substance Abuse and Mental Health Services Administration, 2014), and in the Netherlands only one out of every ten individuals with SUDs eventually reach a healthcare institution (Wisselink et al., 2016), resulting in the biggest treatment gap across all psychiatric disorders (Kohn, Saxena, Levav, & Saraceno, 2004). Herewith, it seems that the few patients with SUDs who seek treatment may possess characteristics that distinguish them from non-clinical patients (Grant et al., 2016; Moos & Moos, 2006; Raimo, Daeppen, Smith, Danko, & Schuckit, 1999). For instance, one study has shown that individuals with alcohol dependency who recover without professional care have fewer socio-environmental difficulties and lower and less severe rates of psychological and substance-related comorbidities than those who have been enrolled in healthcare institutes (Raimo et al., 1999). As such, although they are clinical samples, the abuser and comparison groups may not be similar enough in their clinical nature due to the former being referred for treatment and the latter seeking it voluntarily. Therefore, the differences between parents with SUDs who abuse their children and those who do not could potentially be more visible in two samples that are either fully community based or truly clinical.

Regarding the lack of differences between the two groups, it is also important to realize that a large proportion of child abuse cases remain unknown to authorities. It was found that, in high-income, Western countries including the Netherlands, only 10% of all children facing physical abuse are reported to and identified by the relevant authorities (Alink et al., 2018; Gilbert et al., 2009). Due to such observations, it is possible that some of the parents in the comparison group had committed child physical abuse that had not been identified. In addition, the between-group differences in the ages of the parents – and consequently most likely also in the ages of the children – may have affected the findings, because it is known that younger parents are at higher risk for committing child abuse than older parents (Sidebotham & Golding, 2001). Because it was not possible to match the two groups in terms of age, this variable could not be controlled for.

While retrieving the data and analysing the results, two other observations were made that are worth mentioning. Most striking is the patients who were referred for SUD treatment based on a warning from child protection services but for unknown reasons never received this treatment after their intake meeting. It is remarkable that these individuals did not receive the support that they need, especially considering the unsafe family circumstances which pose a serious hazard to their children’s development; child abuse is a significant risk factor for depressive disorders, drug use, suicide attempts and risky sexual behaviour later in life (Norman et al., 2012; Steenrod & Mirick, 2016). Unfortunately, this problem seems to be widespread – other studies suggest a similar trend wherein only a small percentage of parents complete treatment programmes, despite being required to do so (Oliveros & Kaufman, 2011). Secondly, the involvement of children in their parents’ treatment and recovery phases is often overlooked. Professionals in mental health institutions for addicts are able to indicate the family situation of the patient during intake – and the institutions strongly recommend that they do so. However, when retrieving the Control Centre clients from the VNN database, some clients were registered as childless, even though they were referred to healthcare based on warnings regarding child abuse. Clearly, the children are often not involved in the treatment process, even though they are one of the primary reasons for enrolment in the first place. This suggests that mental health systems and procedures in general may not be adequately equipped to provide appropriate care for the families of patients, especially the children of parents with SUDs, who ought to be part of the treatment plan since they are a particularly vulnerable group.

To conclude, this study is important in addressing matters of child support and protection in a group that is at high risk for committing child abuse. The strength of this study comes from the comparison of two different groups, with attention to diagnostic features that are usually assessed at intake. This provides a practical and naturalistic overview of variables that can easily be retrieved from electronic data files or intake procedures and assessed by healthcare professionals. Future studies addressing this topic may benefit from the present research set-up, but should employ a larger sample size and maybe also the inclusion of community samples or unselected clinical samples. Distinguishing features and improved procedures could result in better prediction, prevention and rehabilitation of perpetrators of child abuse. Moreover, mental health institutions may become better facilitated to incorporate the well-being of abused children into their protocols and systems for parent treatment and recovery.