Use of an HR survey to construct indicators of plant-level risk factors for injury

Background/Purpose

Administrative data can be a valuable source of information for the study of occupational outcomes. Human resource departments of large firms often commission surveys of their workforces focusing on job satisfaction and engagement, however these data are typically not used in injury/health research. These surveys may capture dimensions known to be associated with injury, such as decision authority. The de-identified nature of these data prompts their use as aggregate variables in a multilevel analysis of risk factors for work-related injury.

Methods

A survey of an aluminum manufacturing workforce administered by an HR firm was obtained. Principal components analysis was used to construct four dimensions of responses: 'Job satisfaction', 'Perception of supervisor', 'Work environment 'and 'Stress'. Anonymous individual responses were aggregated into plant-level indicators. A cohort dataset was created from personnel records. Multilevel poisson regression models were constructed to model individual risk factors and plant level factors, while controlling for hours worked. The outcome was all acute, traumatic work-related injuries entered into an incident management system over the 24-month period following survey administration.

Results

Sex, race, tenure on the job, and job grade are significant individual-level risk factors for work-related injury in this cohort. Plant-level variables are strong predictors but do not always achieve significance. Positive perception of the work environment is a strong and significant predictor of injury.

Conclusions

HR survey data may be an untapped source of data that can be explored for occupational health investigations. Moreover, multilevel models allow for the exploration of plant level risk factors.