A public health approach to family justice and the use of administrative data

Matthew Jay, Senior Research Fellow in Epidemiology, UCL Legal Epidemiology Group, Great Ormond Street Institute of Child Health

Keywords: family law, public health, administrative data, quantitative methods

Family proceedings resolve disputes about children. In England, they are private proceedings between parents (~60,000 children/year), or public, where a local authority seeks to remove a child into care due to abuse/neglect (~25,000 children/year). Those involved in family proceedings are at the sharp end of disputes about children. Private proceedings, marked by high rates of socioeconomic deprivation, involve allegations of domestic abuse in 40%-60% of cases. There is also significant overlap between public and private proceedings, either because of pre-existing welfare concerns or because concerns arise during proceedings.

Evidence shows the health of those coming to family court is worse than those who do not. We also know that undergoing family proceedings is itself harmful to health, both private and public. Understanding the pre-existing health of those who come to family court and the effects of family court on health means that we can both help people earlier on (with dual benefits of improving health and wellbeing and reducing burden on the justice system) and reform court processes so that they do not cause harm and result in better, safer decisions. This is a public health (or social policy) approach: one that seeks upstream causes and downstream effects with the intention of improving outcomes across the population.

However, as the former President of the Family Division, Munby P, put it, “there is still too much – far too much – wrong with the system [yet there is a] striking lack of any rigorous, independent, research.” Addressing evidence gaps requires interdisciplinary, mixed-methods research. Here I focus on one of these tools, administrative data, but that is not to say that such data are superior to other forms: they are simply different and offer one piece of the jigsaw.

Administrative data

Administrative data are collected routinely by public services. Their purpose is operational, for example to monitor trends, plan services or administer remuneration. Because administrative datasets typically collect information on all services users, they offer population-level insights that are not possible with traditional, sampling-based methodology. Administrative data offer scale (whole population) and comprehensive coverage (daily contacts) at a level that is unfeasible with primary data collection.

Crucially, not only is it possible to examine data from single services (such as courts, hospitals, schools or children’s social care), but datasets can be linked together to enhance analyses. This is because datasets typically contain the same identifiable information (name, address, date of birth) that can be used to cross-reference records in each dataset. Linkages have the potential to transform research on the justice system, especially on topics pertinent to populations who are less likely to take part in traditional research designs. This is particularly the case, for example, with family justice, where recruitment into studies can be difficult and attrition (drop-out) can be very high. The scale of the datasets also means that it is possible to examine rare outcomes such as mortality and disaggregate results at a much finer level (e.g., minor ethnic groups) than would otherwise be possible.

Linking administrative datasets

Here I show an example from our on-going research that links all-of-England NHS hospital data with data from the Children and Family Court Advisory Support Service (CAFCASS). CAFCASS are a court-based social work organisation that holds a comprehensive database covering almost all family proceedings involving children and including information such as:

• Person demographics,

• Relationship status (person X is mother to person Y),

• Orders applied for, granted and their dates.

The data are very much of the “process” type. While CAFCASS does of course hold detailed case files generated by their social workers, these are not contained in the administrative data.

We are linking these data to data from all NHS hospitals, known as the Hospital Episode Statistics (HES). HES includes data on all NHS hospital inpatient stays and contains coded diagnoses and procedures, meaning it is possible to characterise groups of people with reference to health conditions. We showed, for example, that:

• 1.3% of mothers with a first live birth in England had care proceedings within 10 years.

• This was higher for those with chronic health conditions before birth (3.8%) and especially for the sub-groups with mental health conditions (5.7%), admissions related to drugs/alcohol, violence or self-harm (12.8%) or intellectual disability (30.1%).

• Most mothers involved in care proceedings had two or fewer births during the 10-year follow-up, challenging stereotypes about repeat births.

• One in 62 mothers (1.56%) involved in care proceedings died within 10 years of their first birth, compared to one in 357 (0.27%) of mothers not involved, showing the risk of early death is much higher for mothers with care proceedings.

More detail can be found in our end-of-study report and up-coming publications. By linking CAFCASS to HES, we characterised the health of mothers in care proceedings across England for the first time and answered questions, about differences in mortality risk, for example, that simply would not have been possible without linking these data.

Cautionary notes

I finish with some cautionary notes. No dataset is perfect, but administrative data suffer the disadvantage of not being collected for research purposes. Significant work is required to understand the data and what is and is not possible. Most time spent working with administrative data involves data processing and wrangling to ensure they are analysis-ready.

Ethical and legal considerations abound and there is a social licence that could be broken without care. I have focused on using anonymised data for researching groups of people. This is very different to using data on known individuals to make predictions or decisions about them. I have also not touched on the use of administrative data for routine monitoring of the system. Such use has the potential to enhance transparency overall, although how variation is examined, how different units are compared, and the consequences of this need careful consideration.

Despite this, administrative data certainly have a role to play in enhancing our understanding of the family justice system. By knowing it, we stand a much better of chance of improving it.