Analysis of Diffusion Rates of Drugs Incorporated to the Brazilian Public Health Care System

Abstract

and support vector machine prediction models. The remaining data was equally divided into a validation dataset for parameter tuning and a test dataset for evaluation. Results: The Random Forest model achieved an area-under-ROC of 0.88 with precision= 0.82 and recall= 0.37. The SVM model achieved an area-under-ROC of 0.72 with precision= 0.56 and recall= 0.44. In particular, we found Tramadol, a pain killer, might be effective to treat the obesity. ConClusions: The discussions on drug usage are dramatically expanding on social media sites with some drugrepositioning mentions not sufficiently evaluated by researchers. Machine-learning methods are capable of capturing the drug-repositioning contents. Integrating social media data with other data sources, such as biological, chemical, literature and EHR would boost the identification of drug-repositioning opportunities in a real-time and systematic fashion.