Challenges and lessons learned from Scotland and England using linked administrative data to evaluate the Family Nurse Partnership: two administrative data cohort studies

Abstract

BackgroundThe Family Nurse Partnership (FNP) is an early intervention aiming to support adolescent mothers and their children. The FNP has been evaluated in England and Scotland in two separate studies using linked administrative data from health, education, and social care. We aimed to make recommendations for studies using linked administrative data to evaluate public health interventions. MethodsWe constructed two cohorts of all mothers aged 13–19 years giving birth in NHS hospitals between 2010 and 2016–17 using data from Hospital Episode Statistics (England) and Maternity Inpatient and Day Case (Scotland). FNP participation was identified through linkage to FNP programme data. We also linked to health, educational, and social care data for mothers and their babies (data from the National Pupil Database and electronic Data Research and Innovation Service). Ethical approval was obtained but no consent was required because we used secondary data. We used these data to evaluate the effect of the FNP on maternal and child outcomes. FindingsKey challenges included characterising the intervention and usual care, understanding quality of multi-sector data linkage, data access delays, constructing appropriate comparator groups, and interpreting outcomes captured in administrative data. Lessons learned included that evaluations required detailed data on intervention activity (dates and geography) and assessment of usual care, which are rarely readily available and are time-consuming to gather; data linkage quality information was variable or not available, making defining denominators challenging; data access delays impeded on data analysis time; and unmeasured confounders not captured in administrative data possibly prevented the generation of an appropriate comparator group. We recommend that characteristics informing targeting should be explicitly documented, and could be enhanced by using linked primary care data and information on household members (eg, fathers). Process evaluation and qualitative research could help to provide a better understanding of mechanisms of effect. InterpretationLinkage of administrative data presents opportunities for efficient evaluation of large-scale, complex public health interventions. However, sufficient information is needed on programme metadata, targeting, and important confounders to generate meaningful results. Our findings should help to stimulate exploration by practitioners about how such programmes can be improved. FundingNational Institute for Health and Care Research.