UK surveillance: Provision of quality assured information from combined datasets

G. A. Paiba, S. R. Roberts, C. W. Houston, E. C. Williams, L. H. Smith, J. C. Gibbens, S. Holdship, R. Lysons

Research output: Contribution to journalArticlepeer-review

10 Scopus citations

Abstract

Surveillance information is most useful when provided within a risk framework, which is achieved by presenting results against an appropriate denominator. Often the datasets are captured separately and for different purposes, and will have inherent errors and biases that can be further confounded by the act of merging. The United Kingdom Rapid Analysis and Detection of Animal-related Risks (RADAR) system contains data from several sources and provides both data extracts for research purposes and reports for wider stakeholders. Considerable efforts are made to optimise the data in RADAR during the Extraction, Transformation and Loading (ETL) process. Despite efforts to ensure data quality, the final dataset inevitably contains some data errors and biases, most of which cannot be rectified during subsequent analysis. So, in order for users to establish the 'fitness for purpose' of data merged from more than one data source, Quality Statements are produced as defined within the overarching surveillance Quality Framework. These documents detail identified data errors and biases following ETL and report construction as well as relevant aspects of the datasets from which the data originated. This paper illustrates these issues using RADAR datasets, and describes how they can be minimised.

Original languageEnglish (US)
Pages (from-to)117-134
Number of pages18
JournalPreventive Veterinary Medicine
Volume81
Issue number1-3 SPEC. ISS.
DOIs
StatePublished - Sep 14 2007
Externally publishedYes

Keywords

  • Data presentation
  • Metadata
  • Quality
  • Reporting
  • Surveillance

ASJC Scopus subject areas

  • Food Animals
  • Animal Science and Zoology

Fingerprint

Dive into the research topics of 'UK surveillance: Provision of quality assured information from combined datasets'. Together they form a unique fingerprint.

Cite this