In recent years, governments of more than 200 countries and regions have enacted measures to control the spread of COVID-19. A precise and comprehensive evaluation of policy effect provides important grounds for policy-making. Since the whole world has entered the post-epidemic era, prevention policies are inclined to strike a trade-off between controlling confirmed/death cases and the economic rebound. Furthermore, with the increasing vaccination rate, vaccination has become a considerable factor in determining policy stringency. However, the existing approaches are still limited in efficiency due to the following reasons: (1) They are still confined to policies’ containment effect on COVID-19, neglecting the impact of vaccination on policy effect and the impact of policies on economy; (2) While evaluating policy effect in different regions, most existing models lack robustness. To address these problems, we propose a multi-dimensional evaluation model for more effective assessment of epidemic prevention policies in post-epidemic era. The proposed model consists of two modules: (1) A multi-dimensional objective-programming module is raised to evaluate the policy effect comprehensively, where vaccination, policy stringency, economy indicators, confirmed cases, and reproductive rate are taken into account; (2) A vaccine-dependent parameter learning (VDPL) module based on Bayesian deep learning (BDL) models a vaccine-dependent parameter which indicates the relationship between vaccination and policy stringency. The module also strengthens the robustness of the proposed model with the help of BDL since BDL can adapt the data of different regions better through resampling the probability distribution of network weights. Finally, We evaluate our model on the data of the US. The results demonstrate that the proposed approach performs better in depicting the spread of COVID-19 under the influence of policy.