IntroductionChild maltreatment is a major public health problem worldwide. As many as one-quarter of the world's population are estimated to have been physically abused as a child1Abbasi MA Saeidi M Khademi G Hoseini BL Moghadam ZE Child maltreatment worldwide: a review article. and more than one-third emotionally abused.2Stoltenborgh M Bakermans-Kranenburg MJ Alink LRA van IJzendoorn MH The universality of childhood emotional abuse: a meta-analysis of worldwide prevalence. Although there are appreciable differences in the reported prevalence of child maltreatment across countries and studies, children are universally acknowledged to be a uniquely vulnerable population who need protection from abuse or neglect, whether by parents or other adult perpetrators.3Child protection: a universal concern and a permanent challenge in the field of child and adolescent mental health., 4UNICEF
Convention on the rights of the child.The many adverse social, behavioural, mental, and physical health consequences of child maltreatment5Australian Institute of Family Studies
Effects of child abuse and neglect for children and adolescents. can persist into adulthood, including unplanned pregnancies, unemployment and welfare dependency, violence and criminal activity, suicide, and homelessness.5Australian Institute of Family Studies
Effects of child abuse and neglect for children and adolescents., 6Chamberlain C Gee G Harfield S et al.Parenting after a history of childhood maltreatment: a scoping review and map of evidence in the perinatal period., 7Gilbert R Widom CS Browne K Fergusson D Webb E Janson S Burden and consequences of child maltreatment in high-income countries., 8Hughes K Bellis MA Hardcastle KA et al.The effect of multiple adverse childhood experiences on health: a systematic review and meta-analysis. Child maltreatment can also have long-term relational effects, which can hinder an individual's subsequent capacity to adequately care for their own children.6Chamberlain C Gee G Harfield S et al.Parenting after a history of childhood maltreatment: a scoping review and map of evidence in the perinatal period., 9Disrupting intergenerational maternal maltreatment in middle childhood: therapeutic objectives and clinical translation. A history of maltreatment is one of the most prominent risk factors for perpetration of maltreatment;10Intergenerational transmission of child maltreatment. this is known as intergenerational transmission of child maltreatment.Many studies have found that people who were maltreated as children are more likely to abuse or neglect their own children; however, rigorous quantitative evidence for intergenerational transmission of child maltreatment is insufficient to allow for conclusions regarding the strength of the association.11Ben-David V Jonson-Reid M Drake B Kohl PL The association between childhood maltreatment experiences and the onset of maltreatment perpetration in young adulthood controlling for proximal and distal risk factors., 12Leve LD Khurana A Reich EB Intergenerational transmission of maltreatment: a multilevel examination., 13Widom CS Czaja SJ DuMont KA Intergenerational transmission of child abuse and neglect: real or detection bias?. A widely cited systematic review concluded that most studies that support the intergenerational hypothesis have weak study designs and cannot provide a definitive confirmation of this association.14Thornberry TP Knight KE Lovegrove PJ Does maltreatment beget maltreatment? A systematic review of the intergenerational literature. A systematic review published in 2020 by Greene and colleagues15Greene CA Haisley L Wallace C Ford JD Intergenerational effects of childhood maltreatment: a systematic review of the parenting practices of adult survivors of childhood abuse, neglect, and violence. investigated associations between a parent's reported childhood maltreatment and their subsequent parenting behaviours. This study showed intergenerational transmission of child maltreatment, but whether the maltreatment reached serious concern thresholds is difficult to establish for many of the included studies and the heavy reliance on self-reporting in adulthood is problematic.Research in context

Evidence before this study

We did a systematic review of the relevant studies using PubMed for scientific articles published in English between Jan 1, 2013, and July 21, 2020, using the terms “intergenerational AND (maltreatment OR abuse OR neglect) AND child*”, which identified 343 articles. All titles and abstracts were reviewed and we identified 13 reviews and 35 primary studies that assessed the extent of intergenerational transmission of child maltreatment. A general search of articles published before 2013 was based on a keyword search of PubMed and references from articles obtained via our systematic search.

All relevant reviews and meta-analyses found statistically significant associations between a parent's history of maltreatment in childhood and maltreatment of their own children, except for a single review that assessed childhood sexual abuse. A synthesis of meta-analyses found a parent's history of maltreatment to be the single largest predictor of maltreatment for their children. Most reviews concluded that the evidence was modest or moderate, and some reviews reported smaller effects among high-quality studies.

22 of the 35 primary studies used cross-sectional or matched control designs. Parent maltreatment during childhood was predominantly ascertained using recall-based questionnaires or parental report. 13 studies used a longitudinal or cohort design. Only four studies used population samples, but these studies had methodological flaws and exhibited potential biases that preclude rigorous estimates (eg, misclassifying time spent in out-of-home care, the use of self-reports of both parental and child abuse, the inclusion of non-substantiated maltreatment allegations with substantiated maltreatment, and the inability to capture the majority of the mother's child protection history).

Added value of this study

Our study found very high risk for familial maltreatment among children whose mothers had been maltreated as children. To our knowledge, this is the first cohort study to analyse the population risk of intergenerational transmission of child maltreatment using definitions of maltreatment based on those used by child protection authorities. This study adds details of the types of child protection system involvement among mothers and finds a greater risk of intergenerational transmission of child maltreatment than in previous studies.

Implications of all the evidence

Given the high prevalence of child maltreatment worldwide and its serious social, health, psychological, and economic burden, knowing the substantial risk associated with a parent's history of maltreatment underlines the importance of identifying and helping people who have a history of abuse in childhood and later become parents. Our findings show that society must strive to mitigate the long-term detrimental effects of child abuse and neglect. Interventions that address the past trauma of child maltreatment, maladaptive parent–child relationships, and social and family risk factors are likely to be crucial in preventing intergenerational transmission of child maltreatment. These interventions will require effective cross-agency collaborations, improved funding, and concerted social and political will.

To date, only four studies11Ben-David V Jonson-Reid M Drake B Kohl PL The association between childhood maltreatment experiences and the onset of maltreatment perpetration in young adulthood controlling for proximal and distal risk factors., 16Font S Cancian M Berger LM DiGiovanni A Patterns of intergenerational child protective services involvement., 17Tracy M Salo M Appleton AA The mitigating effects of maternal social support and paternal involvement on the intergenerational transmission of violence., 18Putnam-Hornstein E Cederbaum JA King B Eastman AL Trickett PK A population-level and longitudinal study of adolescent mothers and intergenerational maltreatment. have used population-based cohort designs to rigorously study the extent of intergenerational transmission of child maltreatment. However, perpetration of maltreatment by the parent was not confirmed in two of these studies11Ben-David V Jonson-Reid M Drake B Kohl PL The association between childhood maltreatment experiences and the onset of maltreatment perpetration in young adulthood controlling for proximal and distal risk factors., 16Font S Cancian M Berger LM DiGiovanni A Patterns of intergenerational child protective services involvement. and one study17Tracy M Salo M Appleton AA The mitigating effects of maternal social support and paternal involvement on the intergenerational transmission of violence. used parental self-reporting of their own and their child's history of maltreatment, which is subject to recall and reporting bias. Only one study18Putnam-Hornstein E Cederbaum JA King B Eastman AL Trickett PK A population-level and longitudinal study of adolescent mothers and intergenerational maltreatment. used administrative records of substantiated maltreatment from a population sample, finding a two-times risk of child maltreatment for individuals whose mothers had been the subject of unsubstantiated maltreatment reports in childhood, and a three-times risk for individuals whose mothers had experienced substantiated child maltreatment. However, this study18Putnam-Hornstein E Cederbaum JA King B Eastman AL Trickett PK A population-level and longitudinal study of adolescent mothers and intergenerational maltreatment. measured second-generation maltreatment only up to age 5 years and did not report on mothers older than 19 years, which restricts the generalisability of the findings. Additionally, maternal history of maltreatment was based only on reports occurring after the mother's tenth birthday, further restricting generalisability through censoring of maltreatment exposure.

Robust and methodologically adequate assessments of the intergenerational hypothesis are needed to inform prevention and intervention efforts. Although administrative records might not capture all instances of child maltreatment in the population, they ensure a consistent threshold of risk or harm and overcome some of the limitations of survey data, including social acceptability bias, the inability of participants to recall events that might have occurred in early childhood, sample bias through low representation of people with the most serious child maltreatment exposure, and inappropriate thresholds for defining maltreatment.

In this study, we aimed to determine the extent of intergenerational transmission of child maltreatment by analysing associations between mother and child involvement with the child protection system (CPS) in a large population-based cohort in South Australia. We hypothesised that a mother's history of child maltreatment would be associated with increased risk of CPS outcomes for their child.

Methods Study design and participants

In this retrospective cohort study, mother–child dyads were identified from a record of all births in South Australia, Australia, between July 1, 1986, and June 30, 2017 (the date of last available CPS data). To obtain information on both mothers and their children, we selected dyads that included mothers who were a part of this birth cohort and who also had live-born children during the study period. Both mothers and their children were followed up from birth.

These analyses form part of the Impacts of Child Abuse and Neglect (iCAN) study. Ethical approval for the iCAN study was obtained from the South Australian Health Human Research Ethics Committee (reference HREC14SAH28), and University of South Australia Human Research Ethics Committee (reference 000032801).

 Data collectionWe used linked administrative data from the South Australian Births Registry, Perinatal Statistics Collection, Department for Child Protection, and hospital inpatient and emergency department datasets (appendix p 1). Data linkage of mother–child dyads was done by SA-NT DataLink.19Schneider M Radbone CG Vasquez SA Palfy M Stanley AK Population Data Centre profile: SA NT DataLink (South Australia and Northern Territory). Data were linked using a combination of probabilistic algorithms and clerical review, incorporating information from more than 50 sources to identify unique individuals. Privacy was maintained using the well established separation principle, in which approved researchers are only provided access to de-identified data.20Kelman CW Bass AJ Holman CDJ Research use of linked health data—a best practice protocol. Outcomes and variablesThere were three main outcome variables for assessing second-generation child maltreatment: any CPS involvement (yes or no), any substantiated maltreatment (yes or no), and any placement in out-of-home care (yes or no). Second-generation child CPS involvement was based on data from the Department of Child Protection and contained records of all children receiving notifications, investigations, and substantiations of maltreatment, in addition to out-of-home care placements. Further details regarding CPS in South Australia, including the legislative definition of child maltreatment, are available in the appendix (pp 2–3). Information on new placements into out-of-home care was available from Jan 1, 1990, to June 30, 2017, and information was also available on the start dates of placements that were active as of Jan 1, 1990.Mother's CPS involvement was defined by categories used by the Department for Child Protection, detailed elsewhere,21Government of South Australia
Child protection systems royal commission., 22Segal L Nguyen H Mansor MM et al.Lifetime risk of child protection system involvement in South Australia for Aboriginal and non-Aboriginal children, 1986-2017 using linked administrative data. and covered the period from birth to age 18 years (or up to June 30, 2017, for mothers not yet aged 18 years). The eight categories of CPS involvement correspond to increasing likelihood of risk or harm: no CPS involvement; notifier-only concern (a notification deemed insufficient or vague, lacking credibility, or failing to meet definitional thresholds); other notification (a notification that meets the definitional threshold for child maltreatment but is not investigated because the event is historical, is being dealt with elsewhere, relates to an adolescent at risk, or the alleged perpetrator is not the child's parent or carer); child protection matter (a notification that meets the definition for familial maltreatment but is not investigated by the CPS); investigation only (a notification that meets definitional thresholds and is investigated by the Department for Child Protection but is not substantiated); substantiated maltreatment but no time in out-of-home care; out-of-home care but no substantiation of maltreatment; and substantiated maltreatment with time in out-of-home care (appendix p 4).

Predictor variables related to the children included: sex, whether the child was still in hospital more than 28 days following birth, and whether the child was born with a congenital condition.

Area-based socioeconomic status of the mother at birth, a potential confounder, was determined using the Index of Relative Socio-economic Disadvantage, which is based on a weighted combination of variables (eg, income, educational attainment, and employment) from the Australian Census, and categorised by quintiles, with quintile 5 being the least disadvantaged and quintile 1 the most disadvantaged.

For each dyad, the mother's age at the time of her child's birth was taken from the birth registry and categorised as age younger than 20 years, age 20–24 years, or age 25 years or older. The number of children per mother was calculated from a count of all second-generation children matched to unique mothers, measured up to June 30, 2017. Counts were recoded into four categories: one child, two to three children, four to five children, and six or more children.

Data on disability of the mother were obtained from Department for Education yearly Census records, as completed by parents. A disability was recorded if there was any Census record in any school year showing that the person had autism spectrum disorder, a language or communication disorder, global development delay, hearing disability, intellectual disability, physical disability, speech or language disability, or vision disability.

Self-reported smoking status was obtained from the Perinatal Statistics Collection. Mothers who reported being a smoker at first visit or who reported smoking any number of cigarettes during the second half of pregnancy were categorised as smoking (any amount), whereas all other mothers were categorised as non-smokers or unknown.

Employment status of the mother at the time of the child's birth was taken from the birth registry, which is based on occupation information reported by the mother. Any occupation with paid employment was categorised as employed and any reported occupation that was not paid (eg, unemployed, home duties, pensioner, student) was categorised as not employed. The employment status of individuals with missing data on employment were categorised as unknown.

Marital status of the mother at the time of the child's birth was based on perinatal data and recoded as married or de facto relationship, not partnered, or unknown or missing. When no information was available, a marriage date recorded on the child's birth registry record was used to infer that the mother was married.

Admission of the mother to hospital for a mental health condition or substance use was identified from an emergency department presentation grouped under a major diagnostic category relating to alcohol or drug misuse, alcohol-induced or drug-induced mental disorders, or psychiatric illness, or an inpatient hospital visit relating to mental diseases or disorders, alcohol or drug use, or alcohol-induced or drug-induced organic mental disorders.

 Statistical analysis

Cumulative incidence curves were plotted to describe the association between a mother's CPS involvement and that of her children. The cumulative probability of children having a CPS event at time t was calculated as the percentage of children with the event of interest by time t (measured from birth), multiplied by the probabilities of having the event at all time intervals preceding that time, reported separately for the eight mother CPS categories. Because no second-generation children reached age 18 years during the study period, measurement was censored at the end of data coverage (June 30, 2017) or, if applicable, the date of the child's death.

Because of the varying exposure (age) of second-generation children, univariable and multivariable Cox regression modelled time to the three child CPS outcomes: first CPS notification, the notification attached to the first substantiation, and first entry into out-of-home-care. The proportional hazards assumption was tested using a Wald test and Schoenfeld residual plots.

We used two multivariable model specifications. In both, adjusted hazard ratios (HRs) with 95% CIs were estimated between child CPS outcomes and their mother's CPS exposure and other explanatory variables. Model 1 included mother's CPS exposure, decade of birth, area-based socioeconomic status at birth, disability status, and child's sex and congenital birth conditions. We considered this model to provide the best estimate of the effect of a mother's CPS exposure on the maltreatment of their child, adjusting for confounders. Model 2 included the variables used in model 1 and several attributes that we hypothesised to lie on the causal pathway between mother and child CPS involvement, mediating associations with child CPS outcomes. These mediating variables are potential consequences of the mother's maltreatment history (mother's mental health, number of children, age at child's birth, marital status, smoking status, employment status, child in hospital >28 days following birth). Model 2 can, therefore, be viewed as an explanatory model.

Because we were interested in how the timing of CPS involvement might affect intergenerational transmission of child maltreatment we did an additional subset Cox regression analysis of the association between the mother's age at first CPS notification and the risk of the child having CPS outcomes, and the association between the mother having a CPS notification during the teenage years (age 13–17 years) on the child's CPS outcomes.

To adjust for the clustering of multiple children with the same mother, SEs were adjusted using the Taylor series linearisation approach, using SPSS Complex Samples procedures in SPSS (version 26.0).

Missing data were addressed by pairwise deletion or by including a category of missing or unknown for some variables where this was considered potentially meaningful (not at random).

 Role of the funding source

The funders of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report.

Results

621 496 children were born in South Australia between July 1, 1986, and June 30, 2017, of whom 299 325 were female. 22 789 (7·6%) of these girls went on to have at least one live-born child in South Australia by the study end-date, with 12 166 (53·4%) of them having had more than one child. This resulted in 38 556 unique mother–child dyads for analysis. As of June 30, 2017, children had a mean age of 4·2 years (SD 3·2 years), and a median age of 3·5 years (IQR 4·6 years).

8358 (36·7%) of the 22 789 mothers had some CPS involvement during childhood, 2293 (10·1%) had substantiated maltreatment, and 1070 (4·7%) had spent some time in out-of-home care. The percentage of mothers in mother–child dyads with CPS contact was slightly higher than this (table 1), as mothers with CPS contact had on average more children.

Table 1Distribution of risk factors for child maltreatment according to mothers' CPS involvement

Data are n (%). CPS=child protection system. OOHC=out-of-home care.

19 478 (83·1%) of 23 437 children whose mothers had no CPS involvement also had no CPS involvement, 3488 (14·9%) had a CPS notification that was not substantiated, 458 (2·0%) had a substantiated report, and 127 (0·5%) spent time in out-of-home care (appendix p 5). 1632 (37·2%) of 4382 children whose mothers experienced substantiated maltreatment had no CPS involvement, 1748 (39·9%) had a CPS notification that was not substantiated, 970 (22·1%) had substantiated maltreatment, and 469 (10·7%) spent time in out-of-home care (appendix p 6). 638 (30·4%) of 2096 children whose mothers spent time in out-of-home care had no CPS involvement, 859 (41·0%) had a CPS notification that was not substantiated, 574 (27·4%) had a substantiated report, and 326 (15·6%) spent time in out-of-home care (appendix p 6).The cumulative incidence of child CPS outcomes stratified by mother's CPS involvement is presented in the appendix (pp 7–9). CPS system involvement for children of mothers with substantiated maltreatment was highly prevalent, with some CPS contact in 63% of these children by study end and approximately 84% modelled to age 13 years. Substantial differences between subgroups defined by mother's CPS involvement were seen for all child CPS outcomes by the time children were aged 1 year. These differences increased across early and middle childhood for children with substantiated maltreatment or who had spent time in out-of-home care.Mothers with CPS involvement indicative of high risk of serious harm tended to be younger (table 1).In univariate analyses, significant associations were found between mother's CPS exposure and child CPS outcomes. HRs for child CPS outcomes increased across categories of mother's CPS exposure, with the strongest association seen for children of mothers with a substantiated report of maltreatment and time spent in out-of-home care (table 2).

Table 2Percentage of children with CPS outcomes and univariable HRs by mothers' CPS involvement and other explanatory variables

In all adjusted models, consistent and strong associations were found between mother's CPS involvement and child CPS outcomes. Using no maternal CPS involvement as the reference category, the strongest association in model 1 was found for children of mothers who had substantiated maltreatment and time in out-of-home care (HR 6·25 [95% CI 5·59–6·98] for the child having any CPS involvement, 13·69 [10·08–16·92] for substantiated maltreatment, and 25·78 [18·23–36·45] for spending any time in out-of-home care; table 3). Effect sizes were attenuated but remained significant in model 2, after adjusting for potential mediating variables (table 3).

Table 3Adjusted HRs from multivariable Cox regression models for child CPS outcomes by mothers' CPS involvement and other explanatory variables

Data are adjusted HR (95% CI). Model 1 includes maternal CPS involvement and covariates considered to be not on the causal pathway. Model 2 includes all explanatory variables including those hypothesised to be on the causal pathway between maternal CPS involvement and that of their child. HR=hazard ratio. CPS=child protection system. OOHC=out-of-home care.

Differences in risk were seen across mother CPS exposure categories (table 3). After controlling for a wide range of child and mother-related variables in model 2, children of mothers with notifications classed as notifier-only concern, who did not receive further investigation, were at increased risk of any CPS involvement (adjusted HR 2·47, 95% CI 2·23–2·72), substantiated maltreatment (3·24, 2·50–4·20), and time in out-of-home care (5·23, 3·38–8·09).The risk of CPS outcomes was substantially increased for the children of mothers who had their first CPS notification before age 1 year compared with the children of mothers who had CPS notifications later in childhood (figure 1). Additionally, children had an approximately 2–3 times increased risk of CPS outcomes if their mothers had a CPS notification aged 13–17 years compared with the children of mothers with a CPS notification earlier in childhood (figure 2).

Figure 1Association between mother's age at first CPS notification and child CPS outcomes (for mothers with CPS involvement)

Show full captionHazard ratios are shown on the figure. Error bars represent the 95% CI; values for 95% CIs are listed in the appendix (p 4). CPS=child protection system. OOHC=out-of-home care.

Figure 2Association between mother's CPS notification during teenage years (age 13–17 years) and child CPS outcomes (for mothers with CPS involvement)

Show full captionHazard ratios are shown on the figure. Error bars represent the 95% CI; values for 95% CIs are listed in the appendix (p 5). CPS=child protection system. HR=hazard ratio. OOHC=out-of-home care.DiscussionThis study found a strong association between the CPS involvement of mothers and their children, supporting the hypothesis that child maltreatment has a substantial intergenerational component. This finding is consistent with the well described pathways between child maltreatment and disrupted brain development23Shonkoff JP Garner AS Siegel BS et al.The lifelong effects of early childhood adversity and toxic stress. and disturbed relational patterning,9Disrupting intergenerational maternal maltreatment in middle childhood: therapeutic objectives and clinical translation. which can result in deeply embedded behaviours and emotional responses that undermine nurturing parenting, even with a strong desire to do better.6Chamberlain C Gee G Harfield S et al.Parenting after a history of childhood maltreatment: a scoping review and map of evidence in the perinatal period., 9Disrupting intergenerational maternal maltreatment in middle childhood: therapeutic objectives and clinical translation.Children of mothers with a history of both substantiated maltreatment and out-of-home care had 6·3 times the risk of any CPS involvement, 13·7 times the risk of substantiated maltreatment, and 25·8 times the risk of having time in out-of-home care, compared with children of mothers with no CPS involvement. Adjusted associations are appreciably larger than those reported in other large cohort studies11Ben-David V Jonson-Reid M Drake B Kohl PL The association between childhood maltreatment experiences and the onset of maltreatment perpetration in young adulthood controlling for proximal and distal risk factors., 15Greene CA Haisley L Wallace C Ford JD Intergenerational effects of childhood maltreatment: a systematic review of the parenting practices of adult survivors of childhood abuse, neglect, and violence., 18Putnam-Hornstein E Cederbaum JA King B Eastman AL Trickett PK A population-level and longitudinal study of adolescent mothers and intergenerational maltreatment., 24Wall-Wieler E Almquist Y Liu C Vinnerljung B Hjern A Intergenerational transmission of out-of-home care in Sweden: a population-based cohort study. and systematic reviews.14Thornberry TP Knight KE Lovegrove PJ Does maltreatment beget maltreatment? A systematic review of the intergenerational literature., 25van IJzendoorn MH Bakermans-Kranenburg MJ Coughlan B Reijman S Annual research review: umbrella synthesis of meta-analyses on child maltreatment antecedents and interventions: differential susceptibility perspective on risk and resilience.HRs were attenuated after adjusting for child and mother-related factors, but remained large and statistically significant. The fully adjusted analysis (model 2) and results of other published studies should be interpreted with caution. Many commonly used covariates, such as maternal age at child's birth, maternal education, history of mental illness, substance use, and employment status are also consequences of child maltreatment.