We found a starkly higher risk of premature death among mothers who experienced care proceedings compared to those who did not. Once age-standardised, and accounting for immortal time bias, women died at a rate 21 times higher in periods following care proceedings than in periods without. Most deaths among the care proceedings group were potentially preventable, being related to suicide, homicide, drugs/alcohol, or accidental injury in 73% of cases, regardless of age. By contrast, among women who had not experienced care proceedings, the percentage of deaths related to these causes was 21% for mothers aged 25 years or older at their first live birth and 44% for mothers aged under 25. The substantially higher percentage of deaths related to these adversity causes and the lack of an association with age further underscore the extreme vulnerability of mothers who experience care proceedings. Our findings suggest that action could be taken to prevent most of these deaths.

As noted in the introduction, mechanisms explaining a higher risk of death are complex, including pre-existing poorer mental and physical health, a theory supported by our data. It may also be that child removal increases maternal mortality risk. ^5–7 The profound impact of child removal has been described, including long-lasting grief, guilt and stigma; loss of identity as a parent; a sense of failure; and adverse health consequences—all without access to post-removal support. ¹⁰ Quantitative evidence further indicates outcomes for mothers experiencing custody loss may be worse than those who experience child death. ^6,8,20 We found that in care proceedings, where orders were made related to child removal, the ten-year mortality risk was between 2·4% and 2·6%, compared to 0·8% to 0·9% where orders were made where the child likely remained at home. The data are therefore consistent with an increased risk of maternal death due to child removal, though such interpretation must be extremely cautious as other contributing factors are likely driving “selection” into these orders, such as more severe previous abuse and health problems. In other words, it is impossible using these data to estimate the causal effect of court decision on maternal mortality. Additionally, no data were available on actual placements (for example, a care order can include a placement at home and between 6% and 9% of children looked after are placed with parents depending on calendar year^21,22). This could be remedied with linkage to children’s social care data, ²³ provided these be improved by inclusion of natural identifiers enabling linkage of younger children’s data.

A summary of previous research on mortality risk among similarly deprived mothers can be found in Table 3. ^4–7,24 Our crude mortality rate (2·3/1,000 person-years) among mothers experiencing care proceedings was slightly lower than the rates of 3·7/1,000 person-years and 3·8/1,000 person-years observed in Manitoba⁶ and Sweden, ⁷ respectively, for mothers with children removed to care (including via non-judicial routes). Our estimated ten-year mortality risk (2·1%) was lower than those of Pearson et al⁴ (among mothers accessing south London mental health services with care proceedings, 3·1%) and Guttmann et al²⁴ (among mothers of infants with neonatal abstinence syndrome, 5·1% in England and 4·6% in Ontario). However, the relative difference (age-standardised mortality ratio of 21·0) was higher in our study than in others (hazard ratios of 2·2 to 12·1). In these other studies, comparison groups were similarly disadvantaged, underestimating relative mortality differences vis-à-vis the general population. For example, Pearson et al⁴ compared with other mothers accessing secondary mental health services and Wall-Weiler et al’s^5,6 comparison groups in Manitoba consisted of biological sisters who did not have children removed or mothers receiving other forms of social care intervention. Additionally, some studies employed matching and adjusted for covariates which likely attenuated differences. We provide unadjusted results, meaning our findings reflect actual real-world risks. ¹²

Strengths of this study include the use of whole population data, providing a cohort of 2·7 million first time mothers all with at least 3 years’ follow-up available for analysis across healthcare, family court and mortality data. Mothers with deliveries in hospital data are representative of all women delivering in England, but the methodology excludes mothers who delivered their first child outside of the NHS, either within private hospitals (<1% of all births), at home (∼2% of all births) or outside of England. Long-term physical or mental health problems are not fully ascertained in hospital data and we likely underestimated the extent to which mothers experienced drug/alcohol use and mental health conditions. To improve case-ascertainment for these conditions, linkages to other datasets are required, including primary care and mental health service data.

Preventive interventions might reduce the need for care proceedings. General practitioners and health visitors are the first point of contact for mothers at their first pregnancy and provide the bedrock of universal healthcare for children and families. These universal healthcare services underpin targeted programmes such as health visiting and the Family Nurse Partnership targeted for first-time teenaged mothers. ²⁵ Family support centres, such as Sure Start centres or family hubs, ²⁶ may also offer mental health, domestic abuse and substance abuse support, parenting training, or advice about housing and financial problems. However, both universal and targeted services suffer from rising demands, understaffing, and patchy commissioning. For example, a 2022 study reported Family Nurse Partnership commissioning in only two-thirds of local authorities in England, with capacity to enrol only one-quarter of all first-time teenaged mothers intended for support. ²⁵ Further, intensive support primarily focuses in the first 1,000 days whereas our findings indicate intensive support may be needed beyond this window and providing support may be difficult where mothers lack trust in public services and/or are stigmatised. ¹⁰ On-going monitoring and evaluation of mortality and other health outcomes could be supported by routine anonymised linkage between Cafcass and hospital data.

Findings also have implications for family law and policy, including the need to adopt an holistic approach to the family and their problems. The Family Drug and Alcohol Court, which provides a “problem-solving” approach to adjudication in contrast to the family court’s traditional adversarial nature, have shown promise in addressing the underlying problems that led to proceedings in the first place with improved drug use outcomes among parents and higher family reunification rates. ²⁷ However, there are only 13 such courts operating in 35 local authorities, which require specialist training and provision. An holistic view should also consider private proceedings (usually between parents following relationship breakdown), which our study did not examine. Those involved in private law proceedings also have poorer mental and physical health and experience deterioration in health as a result of proceedings. ²⁸ This is especially so in cases involving domestic abuse, estimated to be the majority of private proceedings. ²⁸ Research is needed on mothers involved in both public and/or private proceedings. Such research could also make use of mother-baby links in hospital data²⁹ to examine outcomes for children, other datasets including mental health data in England and substance misuse and primary care data in Wales, and data on fathers, who have been understudied to date. Ultimately, a public health approach to family justice is needed that examines why families come to court in the first place, their journeys through the courts and their long-term outcomes following proceedings. ³⁰