Currently, communications of the COVID-19 vaccine risk and benefit have been confusing and ineffective. This research presents a visual analytical approach to enable decision-driven, multi-perspective risk characterization of COVID-19 vaccination. Using data collected from Vaccine Adverse Event Reporting System (VAERS), we designed multiple-views dash-boards based on identified risk factors to support interactive explorations of anaphylactic risks from policy, clinical, and personal contexts. Based on the hypothetical scenarios, we showed that our visual analytical approach offers multiple benefits for risk characterization tasks, including flexibility in focusing on the subset of risk factors that are specific to user's decision context, exploring and assessing risk in multiple levels of details, and characterizing risk metrics together with uncertainties. Our method and tools have potentials of improving COVID-19 vaccine risk communication to address vaccine hesitancy and to inform public policy.