Automated data extraction model for the USIDNET registry: Bigger, faster, and better data collection.

Area of Science:

• Immunology
• Genetics
• Medical Informatics

Background:

• The United States Immunodeficiency Network (USIDNET) is a research consortium funded by the National Institutes of Health (NIH) focused on inborn errors of immunity (IEIs).
• Historically, USIDNET utilized an opt-in system with patient informed consent for data collection.
• Transitioned to a semiautomated, de-identified data extraction from the EPIC electronic health record system, employing a consent waiver.

Purpose of the Study:

• To evaluate the effectiveness of the new USIDNET registry methodology in improving patient enrollment and data richness.
• To compare enrollment demographics and data completeness between the previous and current USIDNET registry systems at a single research site.

Main Methods:

• Comparative analysis of registry data from the new (n=2,796) and old (n=551) USIDNET systems.
• Assessment of patient demographics (diagnoses, sex, age), recorded clinical features, and laboratory data.
• Evaluation of data query response times.

Main Results:

• The new registry enrolled a substantially larger number of subjects compared to the old registry.
• Significantly greater data capture per patient in the new registry: six times more clinical features and 22 times more laboratory data.
• Markedly improved query response times, with database queries executed within a day.
• Observed demographic differences in enrollment based on specific underlying diagnoses.

Conclusions:

• The revised USIDNET registry design, utilizing de-identified data extraction with a consent waiver, demonstrates superior patient enrollment and data acquisition capabilities.
• This enhanced data capture is crucial for advancing scientific investigation into inborn errors of immunity.
• The new methodology offers a more comprehensive and efficient approach to studying IEIs, potentially capturing a broader and more representative patient population.