Using Patient Descriptions of 20 Most Common Diseases in Text Classification for Evidence-based Medicine

Evidence-based Medicine (EBM) reflects a combination of clinical expertise, patient’s values, and best available evidence in the decision-making process related to healthcare. In EBM, the medical professional prescribe medicine based on information from previous medical records (which is available in textual format). This information is often used in clinical practice and recently proved to be very useful in predicting diseases with computational approaches. This paper presents an extensive dataset of 11.8K patient descriptions of the most common 20 diseases, and contribute to their classification through unpretentious supervised machine learning techniques. After rigorous experiments under the Monte Carlo method, we found Random Forest Trees (RFT) outperformed all algorithms by achieving the overall highest accuracy of 83%, followed by Linear-Support Vector Machines (SVM) with 81% accuracy.