In order to analyze the causes of blowout accidents in shallow gas drilling on shallow water drilling platforms at home and abroad, a risk evolution analysis method combining accident tree model and Bayesian network is proposed. Various causes of shallow-water gas blowout accidents were obtained through investigation, logical relationships among various risk sources were sorted out, accident tree models were established, and different risk evolution paths leading to shallow-water gas blowout accidents were obtained. By using GENIE, the accident tree model was transformed into a Bayesian network, and the occurrence probability of the underlying events in the accident cases was counted and sorted. Five key risk factors and their main risk evolution paths were obtained through the inverse computing capability of the Bayesian network. The research results show that the five key risk factors are “piston pulling”, overflow, pumping low-density wash fluid spacer during drilling, delay in filling drilling fluid, too long pumping time, low drilling fluid density and kill equipment failure, which can be used to develop targeted risk identification and control prevention technology in actual drilling operations.