Optimal cut-off point of the short version of effort-reward imbalance questionnaire

Introduction
Based on the measurement by the recently developed short version of Effort-Reward Imbalance (ERI) questionnaire, considerably high ERI prevalence has been observed in several Japanese settings. Therefore we conducted a validation study to seek the optimal cut-off point of the short version.

Methods
We analyzed randomly selected workers from the database of a commercial investigation company (N = 1,548) in a web-based survey in January 2011. We applied the original and short versions of questionnaire using a cross-over design. We measured K6 as a criterion variable with several potential confounders. A standardized measure of the ERI model was defined a threshold of 1.0 discriminates for a high-risk group (effort/reward ratio >1). We examined the differences between high-risk groups defined by the two versions by using Kappa analysis and comparing the associations with a criterion measure, and then we examined the optimal cut-off point of the short version by ROC curve analysis.

Results
Cronbach’s α coefficients were 0.88-0.91 for the effort and 0.77-0.76 for the reward scales of the short version. Prevalence of the defined high-risk workers was 18.9% for the original version and 41.4% for the short version. Although the two high risk groups were significantly associated with each other (?2=254.1,p<0.001), they were not consistent very much (Kappa=0.357). Multiple logistic regression analysis showed a stronger association between the original version and K6 cases (OR=7.3, 4.9-10.8) than between the short version and K6 cases (OR=4.1, 3.3-5.2). We found the largest AUC values (0.838) when ROC was drawn using the cut-off point 1.8. 

Discussion
In the short version of ERI questionnaire, the ordinal threshold defined by the original ERI notion (effort/reward ratio >1) appears inappropriate to detect psychological distress. Our finding suggests the cut-off point to be reconsidered for the short version.