53. Data quality in ANCA-associated vasculitis: an analysis of the FAIRVASC registries
Creators
- 1. 1Institute of Infection, Immunity and Inflammation, University of Glasgow, Glasgow, United Kingdom
- 2. 2European Institute for Innovation through Health Data, Gent, Belgium
- 3. 3Division of Rheumatology, Department of Clinical Sciences, Lund University, Lund, Sweden
- 4. 4Telemedicine Technologies, Liège, Belgium, Liège, Belgium
- 5. 6Department of Nephrology, First Faculty of Medicine, Charles University and General University Hospital in Prague, Prague, Czech Republic
- 6. 5Centre for Clinical Studies, Freiburg University Medical Centre, Freiburg, Germany
- 7. 7Department of Internal Medicine, Jagiellonian University Medical College, Krakow, Poland,
- 8. 8Trinity Health Kidney Centre, Trinity Translational Medicine Institute, Trinity College Dublin, Dublin, Ireland
- 9. 9ADAPT Centre, Trinity College Dublin, Dublin, Ireland
- 10. 8Trinity Health Kidney Centre, Trinity Translational Medicine Institute, Trinity College Dublin, Dublin, Ireland, 9ADAPT Centre, Trinity College Dublin, Dublin, Ireland
Description
Background: The FAIRVASC project seeks to federate the data of seven ANCA-associated vasculitis registries across Europe using semantic web technology. A high standard of data quality (DQ) is required for the types of data analysis planned for the FAIRVASC architecture. We sought to design and implement a DQ assessment of the FAIRVASC registries.
Methods: A Data Quality Group was established within the consortium. This group consisted of individuals from a variety of specialist backgrounds including clinician scientists, health informaticians, statisticians and computer scientists. DQ domains selected for evaluation were Uniqueness, Consistency, Completeness and Correctness. These dimensions were prioritised by investigator consensus from a pool of nine candidate dimensions drawn from the literature and assessed using statistical methods and tools developed through prior published research. A DQ worksheet was designed using an iterative approach. A representative at each registry used the worksheet to evaluate their local registry DQ.
Results: Registry participants identification numbers were 100% unique across all seven registries. Consistency of data class was 100% across all measured variables. Consistency on logic testing was 99.9% across all registries. Completeness was 94.3% across all registries. Correctness was still under assessment at the time of this report. Where missing data were present due to an assessed variable not being present in a registry dataset, these were removed prior to analysis. Percentages represent the mean of summary percentages reported for each registry as a whole and were not adjusted for registry size.
Conclusions: This analysis demonstrated a high level of DQ across the initial seven FAIRVASC registries. The registry data were therefore deemed highly suited to FAIRVASC objectives including epidemiological analysis of European data and cluster analysis to determine novel disease phenotypes. Future work will include a DQ improvement process with multiple potential objectives such removal of duplicates, selection of highest quality records, imputation of missing values, re-entry of data and increased specificity of registry metadata.
Disclosures: None
Files
Files
(15.6 kB)
| Name | Size | Download all |
|---|---|---|
|
md5:c8283eb89610b0fb31a236b6d03e7401
|
15.6 kB | Download |