This paper presents two approaches of modeling the deterioration of highway bridges that can lead to an effective maintenance management strategy. Probabilistic state-based/time-based models are used to predict the global or macro-response of bridge components for network level analysis, while reliability-based mechanistic models are used to predict the detailed or micro-response of bridge components for project level analysis. Probabilistic state-based/time-based models are developed using qualitative performance indicators (condition ratings) that are determined through visual inspections to identify the overall condition of damaged components in a bridge network. Reliability-based mechanistic models are developed using quantitative performance indicators (physical parameters) that are determined through detailed condition surveys, analytical modeling, and empirical investigations to identify the extent and severity of specific deterioration mechanisms for safety critical structures and/or highly damaged components. The condition rating data obtained from the Ministére des Transports du Quebec database and the condition assessment of the Dickson Bridge in Montreal, Canada were used to demonstrate the development of the two approaches in modeling the deterioration of concrete bridge decks.