Online discussions are widely adopted in higher education to promote student interaction. However, prior research on online discussions falls short to estimate the effect of multiple factors collectively shape student interaction in online discussion activities. In this study, we applied a dynamic network analysis approach named relational event modelling to a data set from an online course where students participated in weekly discussion activities. In the relational event models, we incorporated multiple factors including participant characteristics, network formation mechanisms and immediate participation shifts. Results indicated that the instructor was more likely to initiate interactions but less likely to receive responses. Popularity, activity and familiarity established in prior relational events positively affected future events. Immediate participation shifts such as local popularity, immediate reciprocation and activity bursts also played a positive role. The study highlights the importance of considering multiple factors when examining online discussions, demonstrates the utility of relational event modelling for analysing online interaction and contributes empirical insights into student interaction in online discussions.