Automatic infection detection based on electronic medical records

Making the most relevant patient care decision, as early as possible, is a constant challenge, especially for physicians in the emergency department. The increasing volumes of electronic medical records (EMR) open new horizons for automatic diagnosis. In this paper, we propose to use machine learning approaches for automatic infection detection using EMR. Three categories of features are extracted, including text-based features, vital signs and clinic tests. Experimental results on a newly constructed EMR dataset from emergency department showed that our model can achieve a decent performance for infection detection with F1 score of 0.88 using text features only. Further analysis reveals that our model can identify indicative symptom expressions about infection.