[Evaluation of electronic medical records from general practices using the example of the diagnosis of community-acquired pneumonia: A qualitative feasibility study].

Background: Data collected by general practitioners (GPs) may provide potential for health services research. In this study, we investigated if clinical questions can be answered with GPs' electronic medical records (EMRs) by means of diagnosing community-acquired pneumonia (CAP).

Method: Patients diagnosed with CAP, defined as ICD code J18.9, were identified in the fourth quarter of 2021. The data were derived from the EMR system (Medical Office®) of a central German association of 30 general practices, using three different approaches: 1. The integrated statistic tool was used to record whether patients were referred for radiological diagnostic confirmation. 2. Retrospectively, EMRs were evaluated manually by a doctor familiar with the EMR. 3. The raw data of the EMR system were extracted by automated export. The information obtained through the three types of access was compared. For each patient case, detailed comments on problems and specifics were documented and evaluated by qualitative content analysis (QCA) according to Mayring.

Results: In total, 164 patients diagnosed with CAP were identified. The numbers of documented radiological diagnostic confirmations varied between data approaches: While the manual evaluation of the EMRs revealed 60 referred patients, the statistics tool identified 38 of these cases. The export of the raw data identified 58 referrals to radiography after adjustment. According to QCA, there was a high variation in applied diagnostics and time of diagnosis. Referrals for radiography were made both before and after coding of the diagnosis. In case of hospitalization, X-rays were usually performed during the inpatient stay. Laboratory tests were performed as an alternative to radiography. There was also a high variation in the documentation of risk factors and diagnostic certainty.

Discussion: The statistics tool integrated into the EMR system is a quick way to perform simple queries but proved to be impracticable for complex questions. The EMRs provide detailed information but need to be evaluated manually. An automated data export from the raw data offers both detailed information and access to large volumes of data but requires complex preparation and appropriate IT expertise.

Conclusion: Based on the example of diagnosed CAP in a GP setting, the use of data extracted from an EMR system seems to be feasible to answer simple clinical questions. However, it is necessary to adapt the data export, and a comparison with a small number of manually evaluated cases is useful to achieve valid results.