Estimating morbidity rates from electronic medical records in general practice. Evaluation of a grouping system

Abstract

Objectives: In this study, we evaluated the internal validity of EPICON, an application for grouping ICPC-coded diagnoses from electronic medical records into episodes of care. These episodes are used to estimate morbidity rates in general practice.

Methods: Morbidity rates based on EPICON were compared to a gold standard; i.e. the rates from the second Dutch National Survey of General Practice. We calculated the deviation from the gold standard for 677 prevalence and 681 incidence rates, based on the full dataset. Additionally, we examined the effect of case-based reasoning within EPICON using a comparison to a simple, not case-based method (EPI-0). Finally, we used a split sample procedure to evaluate the performance of EPICON.

Results: Morbidity rates that are based on EPICON deviate only slightly from the gold standard and show no systematic bias. The effect of case-based reasoning within EPICON is evident. The addition of case-based reasoning to the grouping system reduced both systematic and random error. Although the morbidity rates that are based on the split sample procedure show no systematic bias, they do deviate more from the gold standard than morbidity rates for the full dataset.

Conclusions: Results from this study indicate that the internal validity of EPICON is adequate. Assuming that the standard is gold, EPICON provides valid outcomes for this study population. EPICON seems useful for registries in general practice for the purpose of estimating morbidity rates.