In this paper, we discuss the theoretical framework upon which observational studies of occupational injuries are based. Following a general description of how causal effects are estimated, the challenges faced by researchers working in this area are outlined, with an emphasis on case-control studies. These challenges include defining the at-risk period for workers whose tasks change over time and whose hazard period may be very brief, evaluating the underreporting of both exposures and injuries, and considering the effects of multiple injuries per individual on study design and data analysis. We review both the theoretical and practical considerations in the design and conduct of traditional case-control studies, based on the collection of individual level data, as well as other approaches, such as using information culled from administrative and descriptive databases, and case-control studies in which the plant or work site is the unit of analysis. The case-crossover design is also reviewed and its utility for reducing confounding due to differences between individuals by self-matching is highlighted. While this design has not yet been applied to the work setting, its potential for increasing our understanding of the causes of acute-onset occupational injuries seems promising. Finally, a variety of hybrid designs are discussed, including combinations of case-control, case-crossover, and cohort designs.