During the COVID-19 pandemic, people were often exposed to harmful social media misinformation. Prior studies have devoted their efforts to detecting misinformation and understanding the psychological features related to misinformation. This paper contributes to the literature of handling crisis misinformation by connecting psychological characteristics to people’s actual actions. Anchoring on social media user engagement reflected in the numbers of retweets, we examine the effects of expressed uncertainty and emotions as well as various platform-specific aspects (hashtags and URLs) by extracting features from captured conversations on Twitter social media platform. Subsequently, we quantify expected harms from the chosen COVID-19 misinformation scenarios from the judgements of several healthcare experts, which were then utilized to classify scenarios into different categories for further analyses. With much of the hypotheses supported in both main effects and interaction effects, the study has theoretical contributions in establishing a mechanism to measure expressed uncertainty and emotions from captured Twitter conversations, measuring misinformation harms from professional experts and examining causal relationships between social media behaviour and uncertainty, emotions, harms and several platform specific features. It also has practical contributions of deriving insights to help involved stakeholders in crisis communications understand the role of misinformation harms, and to reduce misinformation diffusion and minimize possible harms.