The impact of data quality and source data verification on epidemiologic inference: a practical application using HIV...

Area of Science:

• Epidemiology
• Biostatistics
• Health Informatics

Background:

• Data audits are typically assessed immediately after completion.
• The long-term impact of data quality improvements from audits on subsequent analyses is not well understood.
• This study investigates the effect of the entire data audit process on statistical analyses.

Purpose of the Study:

• To assess the impact of the complete data audit process on subsequent statistical analyses.
• To evaluate how data quality improvements influence epidemiological inferences.
• To quantify changes in mortality and AIDS-defining event estimates post-audit.

Main Methods:

• On-site data audits were conducted at nine international HIV care sites.
• Error rates for demographic and clinical variables were quantified.
• Time-to-event analyses (mortality, AIDS-defining events) were performed using pre-audit, audit, and post-audit data.

Main Results:

• An overall discrepancy rate of 17.1% was found between pre-audit and audit data.
• Audited data showed higher estimated probabilities of mortality and AIDS-defining events compared to pre-audit data.
• Post-audit data also yielded higher AIDS and mortality estimates.

Conclusions:

• Improved data quality following audits may independently influence epidemiological inferences.
• The data audit process can impact the estimation of health outcomes.
• Findings highlight the need to consider data quality changes in epidemiological studies.