Novel cross-sectoral linkage of routine health and education data at all Scotland level: a national demonstration project.

Wood, R., Clark, D., King, A., Mackay, D. and Pell, J. (2013) Novel cross-sectoral linkage of routine health and education data at all Scotland level: a national demonstration project. Lancet, 382(Sup 3), S10. (doi: 10.1016/S0140-6736(13)62435-6)

Abstract

Background: Analysis of routine data provides an efficient way of exploring health risks and outcomes. We aimed to undertake the first Scotland-wide linkage of children's health and education data to show the feasibility of such cross-sectoral linkage for research.<p></p> Methods: We undertook a data linkage study of children in Scottish schools between 2006–07 and 2011–12. The main datasets were the annual Scottish Government pupil census, which contains Scottish candidate numbers (SCNs; unique identifier used on education records) and personal identifiers for children in publicly funded schools; and the Community Health Index (CHI) database held by NHS National Services Scotland (NSS), which contains CHI numbers (unique identifier used on health records) and identifiers for patients registered with a GP. Restricted pupil identifiers available for linkage in the pupil census (date of birth, sex, home postcode) were matched against patient identifiers held on the CHI database by a bespoke version of NSS's in-house medical record linkage software (previously validated). A best match CHI and additional possible (rival) CHIs were identified for each pupil and assigned probabilistic scores suggesting the amount of agreement between identifiers. Links were then partitioned into categories depending on scores. Links within specified categories (best match CHI with exact agreement on all identifiers or with postcode differing by one character only, and nearest rival CHI with lower score) were regarded as secure, and best match CHIs were accepted. Two methods of linkage were explored: linking identifiers from each year's pupil census separately (A) and combining all the available pupil identifier data from each census into one record per pupil and then linking the combined records (B). Two quality checks were undertaken to assess whether accepted CHIs were correct. Pupil names are available in pupil census records, but the government is unable to share these for linkage purposes. NSS therefore returned SCNs and full names from the best matching CHI records to the government. The government then used name-matching algorithms to compare these with the names in the pupil census. The Scottish Qualifications Agency provided SCNs and full pupil identifiers (including names) for the subset of children registered for Scottish examinations. NSS ran these data through their established probabilistic matching algorithms and the CHI numbers obtained were compared with those from the pupil census linkage. The SCN–CHI key resulting from the linkage was used to construct an anonymised analysis dataset including children's delivery records and their educational attendance, needs, and attainment records. This dataset was used to explore educational outcomes for children with different birth presentations and delivery modes to assess its utility in answering research questions. The results are reported elsewhere.<p></p> Findings: Using linkage method A, an acceptable CHI number could be found for 607 115 (86·3%) of 703 500 children in the 2006–07 census, increasing to 623 396 (92·9%) of 671 264 in 2011–12. Comparable figures using method B were 655 429 of 703 500 (93·2%) in 2006–07 and 638 011 of 671 264 (95·0%) in 2011–12. Linkage method B was therefore regarded as preferable. Using linkage method B, both quality checks suggested that over 99% of the accepted CHIs were indeed correct for children in the 2006–07 and 2011–12 censuses.<p></p> Interpretation: Routine health and education data in Scotland can be linked to an acceptable quality for public health research purposes, despite the absence of names available for linkage within education datasets. This finding opens up the potential for a range of policy-relevant life-course research drawing on routine data from different sectors.<p></p>

• Item Type: Articles
• Status: Published
• Refereed: Yes
• Glasgow Author(s) Enlighten ID: Mackay, Professor Daniel and Pell, Professor Jill
• Authors: Wood, R., Clark, D., King, A., Mackay, D., and Pell, J.
• College/School: College of Medical Veterinary and Life Sciences > School of Health & Wellbeing > Public Health
• Journal Name: Lancet
• Publisher: The Lancet Publishing Group
• ISSN: 0140-6736
• ISSN (Online): 1474-547X