SEAT-BELT LAWS: TRAPS TO AVOID IN EVALUATION STUDIES

This article discusses several statistical issues that need to be considered in analyzing benefits of traffic safety interventions such as mandatory seat belt laws. The article is intended for persons without a strong statistical background, so it uses a minimum of technical terminology. Nevertheless, the issues addressed are important, and persons responsible for evaluating the effects of a belt use law must be careful to use statistical techniques properly. Matters that must be considered are discussed under four headings: (1) Determining whether a relationship (between the law and the outcome) really exists; (2) Determining whether that relationship is causal; (3) Choosing a valid measure of the outcome; and (4) Determining whether findings are applicable to other settings and time periods.