Periodic assessments of the bearing capacity of operational existing bridge structures are deemed necessary. Probability-based reliability assessment is regarded as an ideal approach for this purpose, although it requires substantial data for thorough evaluation. Commencing from parameter estimation in probability statistics, derived confidence intervals for mean and variance when the mean was either known or unknown. By employing interval analysis methods, confidence intervals of probability reliability indicators were computed, thereby estimating calculation errors in probability reliability indicators due to sampling errors. Through the integration of target reliability indicators of bridge structures and classification criteria for existing bridges, numerical methods were employed to analyze confidence intervals of probability reliability indicators under conventional conditions. The results show that with a limited dataset, the confidence interval range for reliability indicators becomes excessively large, potentially leading to misjudgments or omissions in bridge assessments. A method was proposed for classifying existing structural bridges based on possibility, which assessed the reliability of existing bridge structures with more comprehensive information and produces more objective assessment results, thereby facilitating rational decision-making for bridge maintainers.