Mining in Twitter for adverse events from malaria drugs: the case of doxycycline

During the post-marketing period, when medicines are used by large population contingents and for longer periods, unexpected adverse events (AE) can occur, potentially altering the drug’s risk-benefit ratio enough to demand regulatory action. AE are health problems that can occur during treatment with a pharmaceutical product, which in the drug’s post-marketing period can require a significant increase in health care and result in unnecessary and often fatal harm to patients. Therefore, a key objective for the health system is to identify AE as soon as possible in the post-marketing period. Some countries have pharmacovigilance systems responsible for collecting voluntary reports of post-marketing AE, but studies have shown that social networks can be used to obtain more and faster reports. The current project’s main objective is to build a totally automated system using Twitter as a source to detect both new and previously known AE and conduct the statistical analysis of the resulting data. A system was thus built to collect, process, analyze, and assess tweets in search of AE, comparing them to U.S. Food and Drug Administration (FDA) data and the reference standard. The results allowed detecting new and existing AE related to the drug doxycycline, showing that Twitter can be useful in pharmacovigilance when employed jointly with other data sources.

Keywords:
Drug and Narcotic Control; Biological Ontologie; Natural Language Processing; Social Media; Database