Infrastructure maintenance is a vital aspect for any country to ensure safety and reliability of its infrastructure and the population which use these assets. To ensure that the highest degree of maintenance is performed and recorded for infrastructure, Bridge Management Systems (BMSs) have been developed to allow bridge agencies to have an effective means to determine and understand the best decisions to make for infrastructure maintenance. Various models have been developed for the BMS with the most typical approach being the stochastic Markovian-based method, using currently retrieved bridge data as inputs for predicting the bridges’ future deterioration levels. However, a drawback to this method is the disregard for historical data as references to future predictions. This situation has led to the advancement of BMSs to incorporate Artificial Neural Network (ANN) processes as a means of predicting future bridge deterioration levels. This advancement in ANN-based BMSs is an improvement over the typical model due to the incorporation of historical data curves. However, a drawback to this is the fact that biannual bridge inspection data has only started to be collected within the past 10-20 years, limiting the inputs for ANN methods. Further research into ANN models has developed a means of deriving the missing historical data through the use of current bridge inspection data and non-bridge data collected from various sources. This method is referred to as the Backwards-Prediction Model (BPM) and is an effective method for determining this missing historical data for subsequent use as inputs to further ANN methods for future prediction.