Research efforts in the field of occupational injuries have been hampered by difficulties in ascertaining true prevalence rates. Lack of accuracy in reporting and documentation of work injuries leads to loss of information and lack of comparability of work injury statistics. The paper explores some of the factors related to loss of information about work injuries and proposes a descriptive model of the information processing system. The Filter Model describes the way in which information about a given injury proceeds through a series of levels, from true prevalence of injuries in the workplace, to prevalence as indicated by national aggregate statistics. Between each level is a "filter" or barrier that is selectively permeable, ensuring that information about some, but not all, injuries will pass through to the next level. Consequently, the type of work injury data collected at each level will differ. The Filter Model can aid in choice of appropriate level of data for research purposes, and provides a framework for investigation into the nature and extent of filtering processes within work injury reporting systems.