This dissertation presents the development and implementation of a comprehensive automated software framework for probabilistic bridge deterioration modeling that takes into account the time dependent nature of deterioration as well as the impact of various functional, design, and geographic factors on the deterioration rate. Deterioration models are a critical component of the bridge management systems (BMS) used by transportation departments to optimize the allocation of increasingly constrained resources for maintenance, repair, and rehabilitation (MR&R). Since deterioration models are used to predict the MR&R needs at both the bridge and the network levels, the effectiveness of BMS-driven investment decisions related to the repair and preservation of bridge components and, consequently the economy of bridge management actions and safety assurance of the traveling public, is directly affected by the accuracy of the bridge deterioration models. Although probabilistic approaches have been employed for construction of deterioration models, prior studies have largely been constrained by excessive reliance on practitioner opinion surveys and limited application of statistical analytics. Survival analysis-based approaches implemented to date have been parametric in nature and have neither examined the suitability of the pre-existing bridge classifications nor extended the probabilistic methodology to fully realize the predictive potential of such models. In this study, semi-parametric multivariable proportional hazards modeling of survival functions is combined with application of semi-Markovian theory to develop probabilistic deterioration models that reflect the time dependence as well as effects of explanatory variables on deterioration rates of individual bridge components throughout their life cycle. A user-friendly standalone graphical user interface (GUI) is designed for use by transportation personnel to develop and update these models for obtaining future expected condition rating forecasts over specified planning horizons during network-level multi-objective optimization analyses. The developed framework is implemented on North Carolina's statewide bridge database consisting of over 17,000 bridge records spanning 35 years of historical general condition ratings (GCR) assigned during bridge inspections. As a result, significant factors affecting deterioration rates over different bridge components are identified over the life cycle of component and their time-varying influence is quantified in terms of state-dependent hazard ratios. Comparison of the predictive fidelity of the developed probabilistic models to the currently used deterministic deterioration models is used to characterize the improvement in accuracy afforded by the new technique. A strategy for probabilistically incorporating the effects of maintenance action on deterioration rates in the proposed model is discussed as well as potential secondary applications of the developed framework, including quantifying the value of preventative design measures and preservation actions.