Students in the early years of medical school should learn the skills of clinical site risk assessment. However, the effect of this training on clinically inexperienced students is not clear, and it is difficult for them to predict risks from a wide range of perspectives. Therefore, in this study, based on Kiken-Yochi Training (KYT) for risk prediction using what-if analysis, we examined how to expand risk prediction among clinically inexperienced medical students. We divided 120 students in the first year of medical school into small groups of seven to eight students. First, each group predicted risks in the standard KYT (S-KY) method, stating what risks exist in the illustrations. Next, they conducted a What-If KYT (W-KY) analysis, brainstorming situations that differed from the illustrations, and again conducted risk prediction. Three kinds of illustrations depicting medical scenes were used. Last, each student proposed solutions to prevent risks. In this study, we clarify differences in risk assessment tendencies for students between W-KY and S-KY. We found that students could predict a wide variety of risks about illustrations using W-KY, particularly risks about patient and medical personnel. However, for risks regarding management, clinical rules, and stakeholders, prediction in both S-KY and W-KY was difficult due students' lack of knowledge, but solutions proposed by students covered these elements. Improving the format of discussion in W-KY might allow students to predict risk from a wider range of perspectives.