Assessing the Severity of Health States based on Social Media Posts

The unprecedented growth of Internet users has resulted in an abundance of unstructured information on social media including health forums, where patients request health-related information or opinions from other users. Previous studies show that online peer support has limited potency without expert intervention. Therefore, a system capable of assessing the severity of patients from their social media posts can help health professionals (HP) in making a timely intervention. In this study, we inspect the efficacy of different aspects of Natural Language Understanding (NLU) to identify the severity of users’ health state over two perspectives (a) Medical Condition (i.e., Recover, Exist, Deteriorate, Other) and (b) Medication (i.e., Effective, Ineffective, Serious Adverse Effect, Other) in online health communities. We propose a deep learning framework that models both the textual content as well as contextual-information to assess a users’ health state. Specifically, our model utilizes the NLU features such as sentiment, emotions, personality, and use of figurative language to extract the contextual information. These multifaceted NLU features help in understanding an individual’s feelings, mental state, and behavior and thereby assist the model in capturing the health states more accurately along with the content feature extracted from social media medical blog-posts. We compare the performance of our framework on a publicly available dataset against the state-of-the-art baselines which are based on deep learning (CNN, LSTM, and Adversarial Learning) algorithms. The experimental results show that our proposed model significantly outperforms the baseline methods.