Validation of information recorded on general practitioner based computerised data resource in the United Kingdom

Prior research has shown that electronic health records are increasingly used in primary care settings. However, the accuracy of these records compared to traditional paper-based documentation remains unclear. Existing studies have focused on data completeness and accessibility, but few have evaluated the alignment between electronic and manual clinical records. This gap motivated an investigation into the reliability of computerized data in general practice. No prior work had resolved the extent to which digital records reflect actual clinical encounters. Understanding this discrepancy is essential for studies relying on electronic data. Manual records, such as hospital letters, serve as a benchmark for comparison. This paper's contribution lies in directly comparing two data sources within a real-world setting.

The aim of this study was to assess the agreement between clinical information stored in general practitioners' computer systems and the same information documented in manual records. Specifically, the focus was on hospital consultant letters and their corresponding diagnoses. The motivation stemmed from the need to validate the accuracy of electronic records for use in clinical research. Hospital consultants' letters were selected as a gold standard for comparison. The study targeted patients receiving non-steroidal anti-inflammatory drugs to ensure a consistent clinical context. The goal was to determine if digital records could be reliably used for research purposes. This approach allows for a direct comparison of data sources. The findings may inform the use of electronic records in future studies.

The study involved 2491 patients across 58 general practices in the United Kingdom. These practices used VAMP Health computers for recording clinical data. Hospital consultant letters were manually retrieved and photocopied for analysis. Diagnoses from these letters were compared with diagnoses recorded on the computer systems. A preliminary review ensured the quality of the digital records was acceptable. The sample included patients receiving non-steroidal anti-inflammatory drugs. This selection criterion provided a focused clinical context. The comparison was based on the presence of matching diagnoses in both record types.

Among the 1191 patients with available consultant letters, 1038 had matching clinical diagnoses in their computer records. This represents an agreement rate of 87%. The majority of discrepancies involved missing or incomplete data in the digital records. No significant patterns were identified in the types of diagnoses affected. The study found that electronic records were generally reliable for research use. The high agreement rate suggests that digital data can be used with confidence in many studies. However, the 13% discrepancy rate indicates room for improvement in data entry practices. These findings suggest that electronic records are largely consistent with manual documentation.

The authors propose that the clinical information stored in general practitioner computer systems is sufficiently accurate for use in many clinical studies. The 87% agreement rate supports the reliability of these records for research purposes. However, the study does not claim that all data is error-free. The findings suggest that electronic records can serve as a valid data source. The authors do not assign necessity to any specific data entry practice. The study does not generalize to all electronic health record systems. The results may inform future research on data quality in primary care. The authors do not propose future directions or drug targets.