The looked-after child in time: Creating and analysing longitudinal data on placement history and educational outcomes

Gillian Raab & Cecilia MacIntyre - The International Journal of Population Data Science (IJPDS)

Background with rationale

The Scottish government collects data on Looked-after-children (LAC) from the 32 local authorities (LAs) in Scotland. Since 2008, the LAs have provided individual data on a yearly basis that covers every child who was looked after at any time during the year up to the Census date. These data have been linked to data on educational outcomes for these children. Cross-sectional analysis by the Scottish Government show that the educational outcomes for these children are much poorer than for other children in Scotland. This presentation will discuss methods to create a longitudinal data set from these data and thus infer how a child’s lifetime history of care relates to their educational outcomes.

Main aim

To relate the children’s educational outcomes (school attendance and exclusion) to their history of being in care. Features of their care history include, age a start of care, number and type of care episodes (e.g. at home, with relatives, in foster care or residential care) as well as their legal reasons for being in care.


These data present a number of challenges to achieving the aim. The process of creating longitudinal records from nine cross-sectional samples revealed many data problems. A child’s history in care can potentially last from birth to aged 16, or even older, but the data available was for a nine-year window (2008 to 2017) along with some details of episodes that started before 2008 for those in care later. Individual histories were either right-censored or left-truncated. The latter posed a bigger problem for our analyses since most of the early history is missing for a large proportion of the children. The methods used involved using multiple imputation methods to infer the age at start of care and the number of previous episodes for these children.


We will compare the results from the longitudinal analyses with those obtained from cross-sectional results.


Longitudinal data provides the opportunity to understand which patterns of care are associated with the poorest educational outcomes