Development of A Predictive Model for Work-Relatedness of MSDs Among Semiconductor Back-End Workers

Purpose: There are limited models available to predict work-relatedness of musculoskeletal disorders (MSDs) among semiconductor back-end workers. The study attempts to address the gap through development of a predictive model for this specific population. Method: Potential MSDs risk factors were extracted from 277 ergonomics investigation reports for work compensation claim cases conducted between 2011 - 2019. A binary logistic regression approach was used to determine predictors from extracted data. Results: Significant predictors (p<0.05) include poor posture (OR = 1.822;95%CI [1.261,2.632]), forceful exertion (OR = 1.741;95%CI [1.281,2.367]), static posture (OR = 1.796;95%CI [1.367,2.378]), lifting-lowering (OR = 1.438;95%CI [0.966,1.880]), transferring (OR = 1.533;95%CI [1.101,2.136]), pushing-pulling (OR = 0.990;95%CI [0.744,1.317]), repairing (OR = 0.845;95%CI [0.616,1.159]), preventive maintenance (OR = 1.061;95%CI [0.765,1.471]) and quality inspection (OR = 0.982;95%CI [0.729,1.322]). Confounding factors and employment duration also played crucial roles in this predictive model. The model accuracy conducted through cross-validation with 30 sets of new data was 86.2%. Face validation activity among 30 experts shows a promising result, with mean score of agreements on predictors inclusion to be rated at 7.9/10 (SD = 1.9). Conclusion: The model allows practitioners to predict potential MSD cases among semiconductor back-end workers, and proactively plan appropriate mitigation measures.