A novel method to assess data quality in large medical registries and databases

Area of Science:

• Medical Informatics
• Health Services Research
• Critical Care Medicine

Background:

• Assessing data quality in large medical registries lacks a gold standard.
• Data protection regulations can hinder traditional data auditing processes.

Purpose of the Study:

• To evaluate the applicability and utility of funnel plots for data quality control in critical care registries.
• To explore funnel plots as a novel tool for ensuring the accuracy and completeness of registry data.

Main Methods:

• The Swiss ICU-Registry (2014-2015) underwent quality assessment for completeness and accuracy.
• Logical rules and cross-checks were developed to analyze data accuracy.
• Error types, coding errors, and implausible data were quantified for each ICU using funnel plots.

Main Results:

• The Swiss ICU-Registry demonstrated excellent data completeness, with minor exceptions in trauma diagnosis data.
• A total of 3121 coding errors and 31,265 implausible data situations were identified.
• The overall error rate was 7.6%, with implausible data often stemming from non-specific information or diagnostic/treatment incoherence.

Conclusions:

• The Swiss ICU-Registry exhibits high data completeness and adequate quality.
• Funnel plots are proposed as a practical and implementable tool for registry quality assurance.
• The method allows for targeted feedback to ICUs, promoting data quality improvement through identification of special-cause variation.