To address the limitation of existing bridge maintenance optimization methods that fail to consider interactions between sequential decisions across the entire service life, this study proposes a multi-stage, two-level optimization framework grounded in sequential decision-making principles. The upper level model determines performance improvement goals for the maintenance sequence, while considering the influence of preceding decisions on subsequent maintenance policies. The lower-level model then identifies the optimal maintenance actions for each component at each stage, subject to the upper-level constraints. Case analysis shows that, while maintaining superior structural condition over the full life cycle, the proposed method reduces cumulative maintenance cost by 28.6% compared with the traditional strategies. Moreover, when the average deterioration rate of the performance condition index is below 1.425 per year, total life-cycle maintenance and rehabilitation cost can be further reduced by reducing the number of decision-making stages.