Construction sites represent complex environments with particular characteristics and high baseline levels of risk in which accidents are a main point of analysis in much of the studies in the literature. Despite the inertia behind this accident-based focus, though, there is criticism that such a reactive approach often involves unreliable information about the special characteristics of the construction site and that the use of lagging indicators is not appropriate. Current trends have taken proactive approaches and made use of analyses based on precursors or leading indicators, which aim to foresee safety issues before they turn into actual incidents. But these two focuses are not independent, and new proactive measures must be developed and rigorous empirical validation. The objective of the current work is to present a critical review that considers these two approaches based on a meta-classification of studies on health and safety on construction sites.