|
Abstract
|
Background
Studies regarding predictors of underground continuous miner technology-related injuries have been scarce in number, especially from the human error perspective. This study deals with risk factors (personal, including behavioral and work organizational) that influence human error and evaluates those risk factors as predictors of injury occurrences at continuous miner worksites. This study also deals with small sample size and collinearity between predictors.
Material and methods
A match case-control study design was adopted. 135 cases were matched with 270 controls using matching criteria age (±5 years) and job occupation maintaining matching ratio of 1:2. A standard protocol was followed, which included: (1) a participation request to the mine management; and (2) administration of the standardized questionnaire, known as a worker's response device questionnaire, to the selected workers for data collection. Four models (statistical and machine learning) were explored through the empirical training-testing approach.
Results
Bootstrap integrated conditional logistic regression was found to be the best model based on predictability and interpretability.
RESULTS revealed that injury occurrences were associated with multiple factors. Personal factors included big family size (ORa = 3.37, p < 0.001), no formal education (ORa = 2.80, p < 0.01), regular alcohol consumption (ORa = 2.43, p < 0.05), and presence of disease (ORa = 3.87, p < 0.01), along with high risk-taking behavior (ORa = 2.33, p < 0.05) and poor work organization (ORa = 4.75, p < 0.001).
Conclusion
Bootstrap integrated conditional logistic model produced efficient results for the study dataset characterized by small sample size and collinearity. Injury occurrences are multifactorial; addressing these issues through appropriate prevention programs would provide a safer workplace.
Language: en |