Integrating area-based and national samples in birth cohort studies: the case of Life Study


  • Harvey Goldstein Centre for Multi-level Modelling, University of Bristol, UK
  • Francesco Sera London School of Hygiene and Tropical medicine
  • Peter Elias University of Warwick
  • Carol Dezateux University College London



Birth cohort study, longitudinal, Study design, weighting, pregnancy study, attrition, non-response


The most recent UK birth cohort study, known as ‘Life Study’ was a longitudinal study planned to involve some 80,000 babies and comprised two components. The largest, the ‘Pregnancy Component’ was to consist of around 60,000 pregnant women who were to be recruited when attending for a routine antenatal ultrasound at selected maternity units in England. The other component, the ‘Birth Component’ was to be a random sample of intended size 20,000 live births across the UK. Recruitment to the cohort was to take place over a period of four years starting in 2015. Innovative sampling procedures had been designed and tested and a synthetic dataset produced with similar characteristics to the anticipated survey data was produced to study the performance of the sampling procedures and explore analysis strategies.

This research note describes the proposed sample design, and discusses how the two components were to be integrated to provide a consistent dataset for users.  Approaches to the provision of suitable sampling weights and modelling approaches are also presented. Lessons are drawn for designs of future cohort studies.

Author Biography

Harvey Goldstein, Centre for Multi-level Modelling, University of Bristol, UK

Professor of Social Statistics, University of Bristol.


Multilevel models

missing data

record linkage

educational assessment