Quantifying unrecognised replication present in reports of HIV diagnoses
Nikolaos Sfikas1, David Greenhalgh, Wenwen Huo
1Novartis Pharma AG, Basel, Switzerland.
Statistics in Medicine
|February 28, 2014
Summary
This study analyzed HIV diagnoses data, finding a significant decrease in duplicate reports from 3.37% in 1991 to 0.58% in 1994. This improvement reflects enhanced data accuracy in national HIV surveillance.
Area of Science:
- Epidemiology
- Public Health Surveillance
- Biostatistics
Background:
- New Human Immunodeficiency Virus (HIV) diagnoses were confidentially reported to the Public Health Laboratory Service AIDS Centre via a national voluntary surveillance scheme.
- Two distinct national datasets from 1991 and 1994 were utilized for analysis, with the latter showing reduced report replication.
Purpose of the Study:
- To quantify the extent of duplicate report entries within national HIV surveillance data.
- To assess changes in data replication over time using statistical methods.
Main Methods:
- Analysis of reported full dates of birth to identify replicated individual records.
- Application of a nonparametric maximum likelihood estimation method to quantify data replication.
- Comparison of replication rates between 1991 and 1994 datasets.
Main Results:
- The estimated replication rate in the 1991 dataset was 3.37% (95% CI: 0.98%, 11.83%).
- The estimated replication rate significantly decreased to 0.58% (95% CI: 0%, 2.64%) in the 1994 dataset.
- This indicates a substantial reduction in data duplication over the study period.
Conclusions:
- The national HIV surveillance system demonstrated improved data accuracy and reduced report replication between 1991 and 1994.
- The findings suggest successful enhancements in data management or reporting processes within the surveillance scheme.
- Continued monitoring of data quality is essential for reliable HIV epidemiology.
