The paper by Swedler et al1 provides us with new insights into the relationship between workplace factors, safety performance (in this case, the probability of employees wearing slip-resistant shoes), and risk of slip injuries over a 12-week follow-up period. While one could get caught up with potential sample selection issues (eg, the 38% response rate) or measurement issue (if wearing slip-resistant shoes, which required an out-of-pocket purchase by the employee, accurately represents safety performance), the positive aspects of this paper are more worthy of comment.

In particular, the use of structural equation modelling (SEM) allows insights into the relationship between the main independent variables and the outcome, which would likely have been missed using a more traditional regression approach.2 In addition to practical issues of model identification, the SEM process requires the one to distinguish between variables that confound the relationship between independent variable(s) and the outcome, and variables that mediate the relationship, with these choices based on a good theoretical …