Parental Mental Illness and the Likelihood of Child Out-of-Home Care: A Cohort Study

This cohort study used various Swedish national registers linked through the unique personal identification number.¹⁸ We identified all children born in Sweden 2000 to 2011 (N = 1 251 574) using the Total Population Register¹⁹ and their birth parents using the Multi-generation Register.²⁰ Children without known birth parents and adopted children were censored before study start (N = 477). We followed children from birth until first OHC, emigration, or death (parent or child), or December 31, 2016, whichever was earliest. The final cohort included 1 249 463 children linked to 778 170 mothers and 765 843 fathers (Fig 1).

Parental mental illness was identified through inpatient and specialized outpatient visits recorded in the National Patient Register²¹ for nonaffective and affective psychotic disorders, alcohol and drug misuse, depressive disorders, anxiety and stress-related disorders, eating disorders, personality disorders, attention deficit hyperactivity disorder, autism spectrum disorder, and intellectual disability (Supplemental Table 3). The exposure window for parental mental illness spanned 3 years before childbirth until the end of follow-up; exposure was established as a time-dependent variable, such that children were exposed from the date of their mother or father’s first mental illness until end of follow-up.

The outcome was the date of first OHC placement, identified from the National Child Welfare Register.²² OHC placements were categorized as voluntary or involuntary. Where information was available, reasons behind the placement²³ were also identified, categorized into “family-related circumstances” (eg, because of domestic violence), “own behaviors” (eg, because of substance misuse), or “both.”

Demographic variables were identified using the Total Population Register¹⁹ and included child sex, birth year, number of siblings, parental county of birth (Sweden or other), and parental age at birth. Socioeconomic variables at time of child’s birth were identified using the Longitudinal Integration Database for Health Insurance and Labor Market Studies.²⁴ Parental marital status was defined as being married or in registered partnership, or not. We defined parental education as the highest attained education of the mother or father, categorized into compulsory (≤9 years), secondary (10–12 years), and university (≥13 years). Parental employment status was defined as at least 1 parent having gainful employment. Household receipt of welfare benefits was defined as at least 1 parent receiving needs-based financial assistance from the municipality. Household disposable income was defined as yearly income and public benefits earned by all family members after taxation, categorized into quintiles for each calendar year. Parental OHC history was defined as at least 1 parent ever being placed in OHC at any time up to childbirth.

Childhood psychopathology (Supplemental Table 4) was identified during follow-up from the National Patient Register.²¹ Children were considered as having childhood psychopathology from the date of first recorded diagnosis until end of follow-up. Parental history of alcohol and drug misuse was defined as diagnosis of alcohol or drug misuse (Supplemental Table 3) in the National Patient Register any time before childbirth.

We estimated crude incidence rates for OHC comparing children with and without parental mental illness. Cox proportional hazard models, with time from birth as the time scale, were used to estimate hazard ratios (HRs). Model 1 was the crude model, without adjustment. Model 2 adjusted for demographics (child’s sex, birth year, number of siblings, parental country of birth, parental age, and parental marital status). Model 3 adjusted further for socioeconomic variables possibly on the causal pathway between parental mental illness and children’s outcomes (parental education, parental employment status, household receipt of social welfare benefits, and household disposable income). Model 4 adjusted further for parental OHC history. Missing observations in demographic and socioeconomic variables, ranging from 0.0% for parental country of birth to 1.8% in parental education, were included as a separate category in the analyses. Since OHC risk might vary throughout childhood and potentially violate the proportional hazards assumption, we also estimated the HRs separately for each child’s developmental period, including infancy (0–1 years), preschool (2–5 years), school age (6–12 years), and adolescence (13–16 years), by including time-by-exposure interactions in the Cox model.

Effect modification was tested using interaction terms in model 4 between parental mental illness status and potential effect modifiers. Interactions with P < .01 were considered as modifying the relationship between parental mental illness and OHC risk. Variables considered for potential effect modification were: child’s sex, birth year, number of siblings, parental country of birth, maternal and paternal age, marital status, education, employment status, household receipt of social welfare benefits, household disposable income, parental history of OHC placement, and childhood psychopathology.

As a supplementary analysis, we additionally adjusted for parental history of alcohol and drug misuse to account for potential co-occurrence with other parental mental illnesses. We conducted a further sensitivity analysis, including only children known to be living with their parents before the outcome.

Data management and analyses were performed using SAS 9.4, Stata 16.1, and R 4.2.1.

The Regional Ethics Review Board in Stockholm, Sweden (DNR: 2010/1185-31/5, 2013/1118-32, 2016/987-32) approved the study.

Children with parental mental illness were more likely than other children to experience socioeconomic disadvantage (Table 1), to be diagnosed with childhood psychopathology, and have parents with a previous history of OHC placement (Table 1). The majority of children in OHC was placed voluntarily (Supplemental Table 5). However, compared to other children, a slightly higher proportion of children with parental mental illness were placed involuntarily (Supplemental Table 7). Mental illness was diagnosed more commonly among mothers (14.0%) than fathers (9.5%); the most common diagnoses were anxiety or stress-related and depressive disorders (Supplemental Table 6). Overall, children with parental mental illness had markedly higher incidence of OHC compared with other children (6.3 vs 0.8 per 1000 person-years, Table 2). They were 4 times more likely to be placed in OHC than other children with the same distribution of demographic and socioeconomic variables (95% CI 4.24–4.61, Table 2). Furthermore, the association varied by age of the child (P < .0001): the highest incidence (Supplemental Table 8) and strongest association observed in the youngest children, aged 0 to 1 (HR 5.77, 95% CI 5.42–6.14, Supplemental Table 9). The association also differed by diagnoses: highest for parental intellectual disability (HR 4.73, 95% CI 4.09–5.46), parental alcohol and drug misuse (HR 3.91, 95% CI 3.70–4.13), and parental nonaffective psychosis (HR 3.68, 95% CI 3.36–4.03, Table 2). The HR was considerably higher for children with maternal (HR 4.56, 95% CI 4.37–4.76) compared with paternal mental illness (HR 2.68, 95% CI 2.56–2.81, Table 2). Additional adjustments for parental history of alcohol and drug misuse slightly attenuated estimates, but overall associations were similar (Supplemental Table 10). Results did not change materially when restricted to children known to live with parents (Supplemental Table 11).

There was evidence of effect modification (P < .01) for the relationship between parental mental illness OHC risk by all examined demographic, socioeconomic, and other health factors, except for child’s sex (Fig 2, Supplemental Table 12). Compared with children without parental mental illness, or those living in better socioeconomic circumstances, children with parental mental illness in more deprived circumstances had around 8 to 15 times greater likelihood of OHC placement. The increase in OHC likelihood was particularly high when parents had compulsory education (≤9 years, up to lower secondary school, HR 15.08, 95% CI 13.75–16.54), history of OHC themselves (HR 10.52, 95% CI 9.89–11.19), and if the household received benefits (HR 11.96, 95% CI 11.21–12.75).

Children with parental mental illness were up to 8 times more likely than other children to be placed in care outside the home before the age of 18. Socioeconomic confounders influenced this significantly, but the OHC risk remained fourfold greater after adjustment. Greatest risk was associated with parental intellectual disability, nonaffective psychosis, or alcohol and drug misuse. OHC was significantly more likely in children whose mother (rather than father) suffered from a mental illness. Children with the highest OHC risk had mothers with nonaffective psychosis (eg, schizophrenia) or maternal mental illness present continuously 3 years before birth and the first year of the child’s life. Family characteristics including poverty, having parents with only compulsory education (ie, leaving school at 16), or who themselves experienced OHC as a child further increased likelihood of OHC.

Our findings are consistent with 2 studies reporting greatest risk of children with parental mental illness being placed in OHC during the first years of life.^3,9 Previous European studies^3,8,10 reported that children of mothers with schizophrenia are at the highest risk of OHC among all children with maternal mental illness. By contrast, 2 North American studies reported that the greatest increase in risk occurs for children exposed to maternal substance misuse⁵ or anxiety disorder.⁶ Differences in study populations may explain discrepancies between studies or could reflect differences in resources available to support mothers by country.⁶ Kohl et al⁶ suggested that mothers with schizophrenia might have already received other caregiving assistance, which made their children less likely to be placed in OHC compared with mothers with anxiety disorders. Other evidence suggests that staff safeguarding assessments may consider risk of harm to infants greater if a new mother has schizophrenia compared with other diagnoses.¹⁰ 

Two studies^3,8 reported that children of fathers with schizophrenia had the highest likelihood of being placed in OHC compared with other children. Unlike previous studies, we examined paternal neurodevelopmental disorders (intellectual disability, autism spectrum disorders), as well as alcohol and drug misuse, both associated with the highest increased risk. This finding is novel and has not previously been reported.^3,8 

Socioeconomic factors partly explain the associations between parental mental illness and OHC risks which reduced by half after adjustment for socioeconomic variables. They also modified effects, consistent with Ranning et al³ who reported that lower family socioeconomic position exacerbated OHC risk in children with parental mental illness. These authors also suggested higher maternal education or employment might mitigate effects of maternal depression.³ 

Closer monitoring of families with serious parental mental illness might in part be responsible for their higher rates of offspring OHC. In these vulnerable families, the absence of sufficient financial or social support may exacerbate difficulties for parents already struggling to provide care to their children, particularly during early infancy. Current approaches may focus on perceived risks to safeguarding, rather than keeping children with parents. Lack of resources and training for the painstaking, specialist work required to support ill parents to achieve their parenting goals may mean social care systems become less likely solutions compared with measures such as OHC. We did not have access to evidence of what parenting difficulties were experienced or reasons for OHC. Clearly, this would be important if future research is to gain a deeper understanding of the problem. It is no coincidence that this most demanding early postnatal period corresponds with clinical pathways providing closer monitoring of women and babies after birth,^3,25 expressly to ensure healthy infant progression and safeguarding.²⁶ 

Greater OHC placement among children with maternal versus paternal mental illness might be tied to differences in parenting expectations between women and men: women overwhelmingly act as primary caregivers, especially in the first postnatal months²⁷; and, when fathers are ill, mothers maybe expected to fulfill the parenting role, but not vice versa. All parental mental illnesses were associated with higher likelihood of children going into OHC. However, symptoms associated with diagnoses conferring highest risk, eg, nonaffective psychosis or schizophrenia, might have a greater effect on parenting. For example, mothers with schizophrenia can experience negative symptoms and cognitive difficulties that may reduce their ability to “pick up on,” and respond to infant cues²⁸; or, they may be rated by professionals as having more difficulties caring for their infant¹⁰ compared with mothers with affective disorders. Not only is risk of harm to infants potentially greater in mothers with affective disorder than schizophrenia, parenting problems are often reversible or amenable to treatment if recognized.²⁹ Greater awareness about potential clinical bias in the risk assessment of these vulnerable women could mean mothers with schizophrenia (and intellectual disabilities) can be given the right help and support to care for their infants.

Our effect modification analyses suggested that inequalities in family socioeconomic environment contribute significantly to a child’s risk of OHC: socioeconomic adversity is consistently associated with higher risk of removal of children.^3,7 It is notable that parents were also placed in care as children and have a higher risk of their children being placed away from them. Intergenerational transmission of socioeconomic, psychosocial, and parenting risks is a well-recognized phenomenon.³⁰ In the context of parental mental illness, a realistic prospect of improving living conditions for children and families must rely on additional measures aimed at reducing inequalities.

It is important that our work allows for early identification of parental needs – as well as risk. Specific support to mothers who are younger, more deprived, less well educated and alone – with any diagnosis – is crucial. Rather than a clinical focus on maternal diagnosis and its treatment, support should be provided specifically for parenting – tailored to the needs of ill mothers. Greater understanding about the factors determining risk in infants and mitigating risk of OHC should drive future research and be central to its questioning. Supporting parents to keep their infants with them is increasingly recognized as important,³¹ meaning there is an added urgent need to examine risks against potential benefits of OHC in longer studies.

OHC may be a necessary and important option to help children thrive and keep them safe. Nevertheless, the potential adverse child and family effects of OHC^13,–16 suggest we need rigorous decision-making about where the threshold lies for this option of last resort, using thorough and, where possible, standardized, objective assessment. At every stage, consideration should be given to alternative interventions, including temporary arrangements with the intention of replacing children back with families and measures that support the child within its family environment and improve the child’s living conditions. If we were to prioritize systems aimed at ensuring families are afforded every opportunity to continue in-home care, this might have profound effects on the children’s lives. The success of the US and UK’s Family Drug and Alcohol Courts diversion scheme is one such model.³² In Sweden, contact between social services and healthcare professionals is currently limited to child’s healthcare professionals.³³ Strengthening and expanding collaboration between social services, welfare, and healthcare systems, including treating mental health professionals, will ensure integrated assessment and care planning.

Our findings appear consistent across a range of similar settings, eg, United States and Europe. However, we are aware that decisions about OHC largely depend on the organization of welfare and healthcare in each country. Cultures surrounding childbirth are very different in some countries where families offer significantly more input after the birth of a child. In such circumstances, OHC may become more informal and intermittent or less permanent than in Western settings. Infants and childcare may be shared among extended family members when parents are seen to be struggling.³⁴ 

This is the largest, most comprehensive study detailing OHC placement among children with a wide range of parental mental illnesses and considering factors that might modify this relationship. Using high-quality national registers with few missing data provided us with sufficient power to detect associations and interactions. Nevertheless, significant limitations remain. Available information was limited to first OHC, so we were unable to capture more complex placement, eg, whether children were placed long-term or had multiple different placements. Neither did we have detailed information about reasons for placement, including child neglect or abuse or parenting capacities. Finally, we lack data on factors that might enhance our understanding, such as severity of parental mental illness or signs of abuse or neglect.

Children with parental mental illness are more likely to be placed in OHC compared with other children, especially during the first year of life. The risk is highest if the mother is diagnosed with schizophrenia, intellectual disability, or substance misuse. Family socioeconomic deprivation significantly worsens the likelihood of OHC, and we lack information about longer-term risks and benefits of this intervention. Continuous efforts are needed to support families with parental mental illness and prevent OHC. These efforts might have a better chance of success if a core element becomes methods to improve family socioeconomic conditions and support parenting in identifiably vulnerable new parents.:

CI

confidence interval

HR

hazard ratio

OHC

out-of-home care