| Keywords |
Vehicle inspection station; Personal protective equipment (PPE); Safety characteristics; Comfort characteristics; Integrated safety importance index |
| Abstract |
This study investigated the working environment of vehicle inspection stations in Korea and developed an integrated safety importance index for vehicle inspectors. Face-to-face interviews were conducted with 77 inspectors regarding safety and comfort characteristics and the importance of personal protective equipment (PPE). Descriptive statistics, Pearson correlation analysis, and stepwise multiple regression were performed, Regression diagnostics, 5,000-sample bootstrap analysis, and 100-repeated 5-fold cross-validation were subsequently conducted to assess model robustness and internal predictive performance. Summer thermal comfort (X ), summer safety-clothing comfort (X ), and helmet wearing (X ) were significant predictors of the 1 2 3 integrated safety importance index (YP). The final model was YP = 1.949 + 0.823X1 ? 0.447X + 0.331X (R = 2 2 3 .421). In the repeated cross-validation, candidate-variable screening and stepwise variable selection were repeated within each training fold before predictions were generated for the corresponding test fold. Based on the resulting out-of-fold predictions, the mean Pearson correlation coefficient was r = .484 (SD = .055), and the mean cross-validated coefficient of determination was CV R = .210 (SD = .066). The selection frequencies of 2 12 3 X , X , and X across 500 training folds were 81.0%, 59.8%, and 64.0%, respectively, indicating that the model retained some internal predictive information while variable selection showed variability within the limited sample. The findings suggest that thermal conditions and PPE wearing and comfort characteristics should be considered together when assessing safety in vehicle inspection work. The proposed index provides a preliminary quantitative basis for integrated assessment of working-environment and PPE-related safety. Further studies using larger, independent, multicentre samples encompassing both public and private vehicle inspection stations are required to evaluate the reproducibility, variable-selection stability, and external predictive performance of the proposed model. |