Premium
Building the foundation for veterinary register‐based epidemiology: A systematic approach to data quality assessment and validation
Author(s) -
Birkegård Anna Camilla,
Fertner Mette Ely,
Jensen Vibeke Frøkjær,
Boklund Anette,
Toft Nils,
Halasa Tariq,
Lopes Antunes Ana Carolina
Publication year - 2018
Publication title -
zoonoses and public health
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.87
H-Index - 65
eISSN - 1863-2378
pISSN - 1863-1959
DOI - 10.1111/zph.12513
Subject(s) - quality (philosophy) , register (sociolinguistics) , data quality , workflow , danish , quality assessment , medicine , animal husbandry , environmental health , epidemiology , veterinary medicine , database , external quality assessment , geography , computer science , operations management , pathology , engineering , metric (unit) , philosophy , linguistics , archaeology , epistemology , agriculture
Epidemiological studies often use data from registers. Data quality is of vital importance for the quality of the research. The aim of this study was to suggest a structured workflow to assess the quality of veterinary national registers. As an example of how to use the workflow, the quality of the following three registers was assessed: the Central Husbandry Register ( CHR ), the database for movement of pigs ( DMP ) and the national Danish register of drugs for veterinary use (VetStat). A systematic quantitative assessment was performed, with calculation the proportion of farms and observations with “poor quality” of data. “Poor” quality was defined for each measure (variable) either as a mismatch between and/or within registers, registrations of numbers outside the expected range, or unbalanced in‐ and outgoing movements. Interviews were conducted to make a complementary qualitative assessment. The proportion of farms and observations within each quality measure varied. This study highlights the importance of systematic quality assessment of register data and suggests a systematic approach for such assessments and validations without the use of primary data.