Objectives. Ensuring safety in high-risk occupations is challenging when human performance is influenced by organizational, technical and environmental factors. This study aims to analyze the causal mechanisms of accidents in high-risk work systems, using China's general aviation (GA) as a case study. Methods. A novel data-driven tree-augmented naive Bayesian network (TAN-BN) model was developed, integrating variables across human, organizational, regulatory, technical and environmental domains. The model was applied to 186 GA accident reports from China (1996-2024), with sensitivity analysis and validation assessing its robustness. Results. Human, technical and environmental factors significantly influence pilot behavior, while preconditions for pilot unsafe acts are shaped by organizational and regulatory oversight. Sensitivity analysis further identified distinct risk patterns: regulatory deficiencies dominate serious accidents, whereas technical failures are more critical in minor and major accidents. The TAN-BN model demonstrated strong predictive accuracy. Conclusions. This study quantitatively maps how context-specific regulatory and organizational conditions shape human performance under pressure and reveals universal risk propagation mechanisms in high-risk work systems. The findings provide a data-informed basis for human-centered safety interventions, highlighting the need to strengthen organizational management, regulatory oversight and environmental risk management in GA.
Ma et al. (Fri,) studied this question.