The work aims to optimize the interaction mode of smart medicine box based on the data of interaction failure mode of the elderly-oriented smart medicine box, to improve the interactive experience and effectively reduce the interaction error rate. Firstly, an expert group for improving FMEA was established to carry out SHERPA (Systematic Human Error Reduction and Prediction Approach) analysis and build an interaction error prevention model for smart medicine box. Then, the FMEA (Failure Mode and Effects Analysis) was conducted for the interactive design of the elderly-oriented smart medicine box. Finally, the smart medicine box was designed and optimized according to the failure mode required to be improved. According to the experimental results, there were many interaction failure modes in the "empirical object". Optimizing the corresponding interaction failure mode could effectively improve the interactive experience of the elderly in operating the smart medicine box. With the improved FMEA method, the interaction failure mode data of the elderly-oriented smart medicine box can be specifically quantified. The interactive optimization design of the smart medicine box corresponding to the failure mode data is feasible and effective, and can effectively reduce the risk of drug use for the elderly. This method provides a new method for the interactive design research of product.