<p>The blood transfusion process is characterized by stringent risk control requirements, and any minor mistakes can lead to serious medical accidents. To strengthen the control of possible hazards and risks, this paper proposes a novel large-scale group decision-making (LSGDM) failure mode and effect analysis (FMEA) method for risk assessment of the blood transfusion process. First of all, experts are allowed to freely express their evaluation information using the context-free grammar method, which is based on the multi-granularity hesitant fuzzy linguistic term set (MGHFLTS). The comparison method based on MGHFLTSs is put forward to display the expert evaluation preferences. Then, a comparison method and similarity measure based on MGHFLTSs in the LSGDM environment are proposed. Furthermore, a dual vigilance fuzzy adaptive resonance (DVFAR) clustering method based on the expert evaluation similarity is introduced to cluster experts. Next, the clustering similarity matrix is used to calculate the objective weight of each cluster, and then a comprehensive clustering weight method is proposed by combining the objective weight with the relative percentage of experts in each cluster. Additionally, the attribute intensity preference matrix is applied to determine the attribute weights. After that, a novel large-scale group decision-making FMEA method is introduced. Finally, a case study of risk analysis of the blood transfusion process is selected to illustrate the suitability and viability of the present method, and sensitivity and comparative analyses are also conducted to validate the effectiveness of the method.</p>