Project Title:

Linkage validation study

Project Number:

2007_013

Researchers:

Ganka Mueller
Lin Hattersley
Susan Wallace
Mary Macdonald
Neil Bowie

Start Date:

Approved on 22-01-2008

Summary:

The Ethnicity and Health Study, led by the University of Edinburgh, involves the creation of a database combining information from the 2001 Census and administrative health records in order to study differences by ethnic background in 4 areas of health outcomes: coronary heart disease, cancer, maternity and child health, and mental health. The study uses probability record linkage methods to link Census records to the Community Health Index (CHI), which is a database of everyone registered with a GP or in receipt of screening services in Scotland.

The SLS database uses the NHS Central Register (NHSCR) as a tracing tool to build longitudinal records for its sample members. This involves a record linkage between Census records and the NHSCR by means of automated deterministic matching and clerical methods which may in addition use information about other household members or vital events registrations. The NHSCR serves as a reference dataset for the CHI and a one-to-one link between the two databases exists.

The proposed study seeks to make use of the existence of two independent linkages of 2001 Census records to administrative health registers in order to gain insights into the quality aspects of the linkage. The SLS seeks to minimise the level of false positives and puts greater effort and resource into the linkage. The Ethnicity and Health study can tolerate some statistical error and does not prioritise between false positive and false negative linkages. High levels of agreement between the two sets of outcomes, across the whole of the population and/or for individual socio-demographic groups, would be encouraging and indicate good reliability. Levels and patterns of disagreement would point to areas where the outcomes are sensitive to the type of methodology used.

The proposed comparison assumes high quality of the underlying NHSCR-CHI linkage. Where this fails, disagreement between the SLS and the Ethnicity and Health Study linkage results will not necessarily signify poor quality of the Census linkage. To evaluate the validity of the basic assumption, therefore, additional smaller studies will be considered within the data requested for this project.

Appendix 1 sets out a the main steps and data flows. The proposal has been designed to ensure that no additional sharing of personally identifiable data is required for this research.

The objective of the proposed research is to compare the linkage between the 2001 Census and the NHS Central Register (NHSCR) achieved in the SLS with that obtained in a similar study carried out by the University of Edinburgh (Ethnicity and Health), which links 2001 Census records to the Community Health Index (CHI).

Neither the SLS nor the Ethnicity and Health study has the benfits of an independent validation of the linkage which is at the core of the database they create. It is hoped that the results from the proposed comparison will throw light on the quality of the linkage between the 2001 Census and GP regsitrations in Scotland, which is central to both of these studies and has an important potential application in the design of the 2011 Census.

Related Outputs (viewable on CALLS Hub):

• Sorry, no related items were available.