Posture Assessment and Musculoskeletal Disorder Risk Prediction in Gas Company Office Workers Using Machine Learning Techniques Based on ROSA Scoring

Authors

  • Ali Dormohammadi Department of Occupational Health Engineering, School of Public Health, North Khorasan University of Medical Sciences, Bojnurd, Iran
  • Dariush Shahparast Industrial Hygiene, North Khorasan Gas Company, Bojnourd, Iran
  • Azam Orooji Department of Advanced Technologies, Faculty of Medicine, North Khorasan University of Medical Sciences (NKUMS), Bojnurd, Iran
  • Zahra Esmaeil Nezhad Department of Statistics and Performance Analysis, Health Deputy, North Khorasan University of Medical Sciences, Bojnurd, Iran
  • Rajabali Hokmabadi Department of Occupational Health Engineering, School of Public Health, North Khorasan University of Medical Sciences, Bojnurd, Iran

DOI:

https://doi.org/10.18502/jhsw.v16i2.22912

Keywords:

Assessment, Posture, Risk, Musculoskeletal disorders, Machine learning, ROSA method

Abstract

Introduction: Office workers, who comprise a substantial portion of the country’s workforce and use computers extensively, are at risk of developing occupational musculoskeletal disorders. Therefore, the aim of this study is to assess posture and predict the risk of developing musculoskeletal disorders using ML techniques based on ROSA scoring.

Material and Methods: This descriptive-analytical, cross-sectional study was conducted in North Khorasan Gas Company in 1402. This study was conducted in three steps, which included collecting data using the ROSA method and the Nordic questionnaire, creating a prediction model based on machine learning techniques, and evaluating the results based on the criteria of accuracy, precision, sensitivity, specificity, area under the ROC curve, and F criterion. Also, the 10-fold cross validation method was used to evaluate the performance of the models.

Results: The results showed that 32.5% of employees were aged 41 to 45 years and 35% had a work experience of 16 to 20 years. The most pain and discomfort was observed in the back and neck areas with a prevalence of 52.1% and 41.5% of employees, respectively. The results of the ROSA showed that 44.8% of people had a score of 5 and above. The Bayesian Network and RF models achieved accuracy of 80.87 and 80 and sensitivity of 80.9 and 80, respectively. However, the low value of specificity for RF showed that its performance in identifying high-risk cases was low. Considering all performance criteria, the best model was the Bayesian Network, which in addition to high accuracy and sensitivity, also had a specificity of 60.6%.

Conclusion: The results showed that the proposed model can be used to identify and predict musculoskeletal disorders in organizational employees. This tool can also be used as a predictive tool for diagnosing musculoskeletal disorders and interpreting the results effectively and reliably

Published

2026-10-07

Issue

Volume (16) issue (2)

Section

Articles