Who do you ask? Predicting injury rates from survey responses

Few studies have addressed whether using manager or worker perceptions of safety policies and practices alone predict reported injury rates less accurately than using both.

This study provides an example and describes a method that can be used to address this issue with survey instruments designed to measure safety climate, policies, or practices.

Using multilevel logistic regression, we estimated the relationship between worker and manager perceptions of a given exposure and the odds of worker injury during the post-survey year for three safety scales. We tested whether surveying both workers and managers provides additional predictive value compared with surveying just one group.

Injury in the year following the survey was significantly associated with worker scores on two of the three scales. Manager responses were not significantly associated with injury and did not significantly improve injury rate prediction when added to a model with only worker survey responses.

The capacity of manager-only or worker-only perceptions of safety policies and practices to predict worker injuries should be established before choosing to survey just one or the other. The approach and findings in this paper can be applied to other survey instruments and in other settings to help make this choice.