Identification of the adverse effects of sports supplements through social media text mining

Abstract

Abstract Introduction: Sports supplements, despite their beneficial effects, are sometimes associated with side effects. Sports supplements mostly are not prescribed by medical doctors and athletes generally share their knowledge and experience regarding sports supplements through the social media. Identification of the side effects of medications is far easier since they are prescribed by physicians, and their effectiveness and adverse effects are recorded in medical record systems. Yet, identification of the adverse effects of sports supplements is more difficult due to arbitrary and non-medical use. Thus, the experience of Twitter users may be a good source to identify these adverse effects. The present study aims to identify the adverse effects of the sports supplements through Twitter using text mining techniques. Materials and Methods: This study is an applied developmental study in which 40466 tweets of Twitter users from 2014 to 2019 have been collected for 10 frequently used sports supplements. in the present study, K-Means standard algorithm is applied for the clustering of the adverse effects. Silhouette index is used to determine the optimum K value for clustering of the sports supplements. Results: Comparing the extracted list of adverse effects with the known adverse effects revealed 93 new adverse effects that were not reported and could be considered as side effects for investigated sport supplements. Moreover, sports supplements were divided into three groups according to the K-Means standard algorithm. Also, analysis showed that the highest and lowest discrepancy rates belonged to Dinitrophenol (0.041) and Danazol (0), respectively. Conclusion: Presented research method based on text mining could be used to extract the adverse effects of sports supplements from the texts of tweets. Results of the presents study may be used to update the adverse effects database of sport supplements.