To mitigate safety risks inherent in highway bridge construction, this research establishes a practical framework for assessing workers’ fitness for work. Using grounded theory, we analyzed interview records and documented accident cases through systematic coding, identifying critical indicators spanning physiological states, safety training effectiveness, and atypical behavioral markers. Rather than relying on single-method approaches, we combined Delphi expert consultation with entropy weighting to capture both professional judgment and data-driven variance, thereby reducing bias while preserving information richness. The resulting assessment protocol enables quantifiable classification of workers into distinct risk tiers. Implementation at the Zhangjinggao Yangtze River Bridge demonstrated the system's discriminatory power through field data collection and direct behavioral monitoring, successfully segmenting the workforce into low-, medium-, and high-risk categories. Results suggest the tool functions effectively as a pre-employment screening mechanism, allowing project managers to intercept potentially unfit workers before they enter hazardous work zones, consequently lowering the incidence of human-factor accidents.