A longitudinal analysis of data quality in a large pediatric data research network

Area of Science:

• Pediatric Health Research
• Clinical Data Research Networks
• Health Informatics

Background:

• PEDSnet aggregates electronic health record data from children's hospitals for large-scale research.
• Ensuring data quality is crucial for the validity of research conducted within PEDSnet.
• This study evaluates PEDSnet's research capacity by analyzing identified data quality issues.

Purpose of the Study:

• To present and interpret data quality issues identified over 18 months within PEDSnet.
• To assess the research capacity of the PEDSnet clinical data research network.
• To provide insights into data quality assessment for other clinical data research networks.

Main Methods:

• A semiautomated data quality assessment workflow was employed.
• Two investigators reviewed programmatic data quality issues.
• Discussions with data partners' extract-transform-load analysts identified issue causes.

Main Results:

• A longitudinal summary identified 2182 data quality issues across 9 submission cycles.
• The most frequent issues included missing data and outliers.
• Medications and lab measurements were the most complex domains; source data characteristics were the primary cause.

Conclusions:

• Longitudinal findings show PEDSnet's evolution in data alignment and understanding pediatric clinical norms.
• Data quality assessment findings are critical but often unpublished.
• This study offers a real-world account of data quality interpretation in a pediatric CDRN, providing lessons for others.