Traffic conflict-based crash estimation models. Overview and application across various traffic environment

Traditional road safety analysis has predominantly relied on crash data, a method that faces significant limitations, including the need for long observation periods, data unavailability, lack of precision, misclassification, underreporting, and a reactive nature that raises ethical concerns by waiting for accidents to occur. A proactive and innovative alternative involves analysing road safety through the study of traffic conflicts, which are non-crash events described using surrogate safety measures. While numerous applications of traffic conflict methods have shown promise for general safety assessment, their acceptance as a robust alternative to crash-based analysis depends on the product validity of crash-conflict models, defined as the ability of traffic conflicts to reliably estimate crashes. Decades of research on the crash-conflict relationship have led to the evolution of methodologies, transitioning from early crash-based approaches like linear regression to more sophisticated non-crash-based models like extreme value theory models. Despite these advancements, a critical overview is needed to assess their applicability, particularly in the context of low- and middle-income countries where data limitations and diverse traffic environments pose unique challenges. This paper provides a comprehensive review of studies on the crash-conflict relationship, offering a detailed examination of key methodological considerations, including conflict indicators, road user types, road facility types, data sources, model validity, and traffic environments. The review highlights general trends, challenges, and limitations while uncovering gaps in existing research and outlining potential directions for future development, with particular attention to the conditions of low- and middle-income countries.