Quality assessment of automatically extracted data from GPs' EPR

Area of Science:

• Health Informatics
• Primary Care Research
• Data Quality Assessment

Background:

• Routine primary care data offers significant secondary benefits for research.
• Understanding the properties of this data is crucial before its widespread use.
• Electronic health records (EHRs) are increasingly used in primary care settings.

Purpose of the Study:

• To describe a method for assessing the Positive Predictive Value (PPV) and sensitivity of data extracted from Belgian General Practitioners' (GPs) Electronic Patient Records (EPR).
• To evaluate the accuracy of diagnoses, drug prescriptions, referrals, and specific parameters within GP EPR systems.
• To establish a reliable method for validating primary care data for research applications.

Main Methods:

• Utilized an electronic questionnaire as the gold standard for data validation.
• Assessed data extracted from Belgian GPs' EPR systems, including diagnoses, drug prescriptions, referrals, and parameters.
• Conducted the ResoPrim phase 2 project involving 4 software systems and 43 practices, encompassing 10,307 patients.

Main Results:

• Detailed the methodology for calculating PPV and sensitivity of extracted EPR data.
• Presented the results from the ResoPrim phase 2 project, demonstrating the application of the assessment method.
• The study involved a large patient cohort across multiple software systems and general practices.

Conclusions:

• The described method provides a robust approach to assessing the quality of routinely collected primary care data.
• This validation technique enhances the reliability of data extracted from Electronic Patient Records for research.
• The assessment methodology is adaptable and can be applied to other research networks to ensure data integrity.