Heterogeneity-Based Survival Analysis of NBI Condition Ratings for Concrete Highway Bridge Decks in Oregon by Condition Group

Abstract

This study applied a heterogeneity-based survival analysis to determine significant contributing factors to a bridge deck’s time-in-condition rating (TICR) by condition rating (CR) group: high CR, medium CR, and low CR. Although previous work has shown the effects of different CR groups on time in condition, CR-group-specific effects of NBI-related variables on time in condition by CR group are less documented. Additionally, heterogeneity issues with NBI data have focused on between-group heterogeneity, and not on bridge-deck-specific variation due to unobserved heterogeneity. In the context of statistical and econometric models, unobserved heterogeneity refers to unobserved characteristics that may influence the process of interest—in this case, the likelihood of being assigned a lower CR. This work uniquely addressed both these gaps. First, three separate survival models by CR group were fit and a parameter transferability test was conducted. Results from the parameter transferability test indicated that separate models by CR group should be fit (i.e., parameters and their effects are not transferable among CR groups). Secondly, to account for bridge-deck-specific variation due to unobserved heterogeneity and to address potential bias in parameter estimates, random parameters were estimated. Model specifications suggested that the NBI data are susceptible to large amounts of heterogeneity, as indicated by the number of random parameters and the heterogeneity-based model being preferred over the fixed parameters model. For CR-group-specific variables and effects, only two variables were found to be significant in each CR group: mountainous climate, and design period after 1970. For both variables, the effects differed by CR group.