Machine learning model for predicting the risk of back pains among porters in Singer and Kurmi Market, Kano State, Nigeria

Porters in large Nigerian markets are routinely exposed to repetitive lifting and load carriage. This study aims to build and deploy a machine learning model to predict the risk of back pain from anthropometric parameters. A descriptive cross-sectional study was conducted among 300 active porters. Back pain was measured using an interviewer-administered questionnaire adapted from the Standardized Nordic Musculoskeletal Questionnaire. Body weight, height, and circumferences of neck, calf, and waist were measured using standardized protocols. The Tidymodels framework of the R software was used for all stages of the model development. The model was deployed as an R Shiny web application and hosted on shinyapp.io. The mean age, height, weight, body mass index, waist circumference, and waist-to-height ratio of porters with back pains were respectively 26.4 ± 7.2 years, 1.70 ± 0.07 meters, 64.8 ± 11.1 kg, 22.4 ± 4.1 kg/m2, 82.9 ± 7.6 cm, and 0.49 ± 0.05. The respective values for porters without back pains were 24.8 ± 6.1 years, versus 1.74 ± 0.08 meters, 62.4 ± 9.3 kg, 21.0 ± 3.5 kg/m2, 81.5± 8.2cm, and 0.47 ± 0.05. The prevalence of back pain among the subjects was 87.3%. The sensitivity, specificity, overall accuracy, and area under the ROC curve of the model for the test data were found to be 85%, 42%, 79% and 66%, respectively. In conclusion, predictive analytics using simple anthropometric measurements has the potential to be used in screening the risk of back pain among porters in the large markets of Kano State, Nigeria.

Key Words: porters, back pain, machine learning, Kano, shiny application