The paper combines the textual uncertainty knowledge (evidence) with structured knowledge graph, and puts forward several methods to calculate the uncertainty level of medical knowledge, including scale, probability, information entropy, evidence comment network and so on. It is proposed that the decision can be made by the machine for the knowledge with higher certainty level. For such conditions where there is only evidence with lower certainty level, it is important to join machines and doctors (scientists) together for shared decision-making, so as to improve the efficiency of knowledge-driven decision support.