Hot topics, as a common phenomenon on social media platform, play a major role in public opinion. This paper aims to discuss the issues about the duration of hot topics: which factors influence the duration of a hot topic on a social media platform? To answer this question, Cox regression model of survival analysis was introduced to make empirical analysis. The survey data containing 60 hot topics from 2011 to 2013 was collected from a popular social media platform of China. Besides, model verification is implemented to validate the adopted model and the calibrated parameters. As a result, this paper finds that the number of participants and opinion leaders have a significant positive influence on an incident’s duration, and official response times have a significant negative impact; whereas, the number of attending media has no significant impact. Furthermore, the prediction accuracy in the validation is up to 85 %, which implies that the obtained results would be robust. The established model and the designed empirical analysis are demonstrated to be adaptable to analyze the duration of hot topics, and three factors are found to effect the duration of a hot topic based on our collected dataset. Our analytic method could also be adopted in numerous problems related to the duration issues in the field of information management.