An integrated deterioration prediction method has been developed to predict long-term performance of bridge elements for various situations in terms of the quantity and distribution of available condition rating data. The method employs a categorisation and selection process in conjunction with the Backward Prediction Model (BPM) as well as the time-based and state-based models. To check the accuracy of the proposed integrated method, a series of case studies are carried out based on the U.S. National Bridge Inventory (NBI) datasets. A total of 40 bridges with 464 bridge inspection records are selected from the New York State region. Of these, 315 records are used as input for the proposed method to predict the long-term performance of the concerned bridges. The predicted bridge condition ratings are compared with the actual condition ratings - i.e. the remaining 149 inspection records. The accuracy of the prediction is reasonable. To demonstrate the superiority and merits of the proposed method, a detailed comparison is made between the proposed integrated method and the standard Markovian-based procedure.