Mining Electronic Health Records to Guide and Support Clinical Decision Support Systems

Background

In a recent perspective article published in the New England Journal of Medicine (Gandhi, Zuccotti, & Lee, 2011), the authors summarized an analysis of outpatient records for 1.7 million patients in an integrated healthcare system in the USA. The analysis highlighted that 71% of the patients who had undergone splenectomy did not have it medically coded under the ‘problems list’ in their EHR. As a result of this missing information, only a third of the patients whose records did not mention splenectomy were given pneumococcal vaccination. More interestingly, the authors reported that an alert-and-reminders-based clinical decision support system was already in place and was not of much assistance. In this example, text mining could be employed to medically code splenectomy under the ‘problems list’ in an EHR. Furthermore, data mining techniques like predictive models can be built to predict other associated problems related to splenectomy.