Modeling the relation of hand-arm vibration exposure and occupation characteristics using occupational health and safety administrative data

Abstract

Background Hand-arm vibration (HAV) is an occupational hazard which, cumulatively, leads to hand-arm vibration syndrome (HAVS). Detection and reduction of HAV can help prevent the disease or slow down its progress. Unfortunately, assessment of HAV through direct measurement is difficult due to the high cost of measurement equipment, interruption of work performance, and long travel cost to worksites. An alternative assessment method is through development of an exposure prediction model to identify workplace, tool, and task characteristics which significantly predict HAV exposure. Purpose The purpose of this study is twofold: 1) to determine the extent and nature of previously published scientific journal articles on exposure prediction modeling of HAV through performing a systematic review; and 2) to develop a new exposure prediction model using administrative data to find significant HAV predictors. Methods A systematic review of relevant studies involved humans aged 18 or over, applicable to occupational setting, with vibration measured using tri-axial accelerometer and statistical modeling of the effects of occupational characteristics on measure HAV were identified. Five online bibliographic databases (Medline, CINAHL, Web of Science, Scopus, and EMBASE) were searched using a combined word list of terms under three categories: “occupational diseases”; “Hand-arm vibration”; and “Statistical Modelling”. Two multiple linear regression models predicting average hand-arm vibration exposure over 8 hour day, A(8), were built using enforcement data collected by Ministry of Labour Relations and Workplace Safety Saskatchewan using standard model building method. In addition, GEE was used to account for repeated data collection within workers and worksites. Results In the first manuscript, 7 articles found were deemed relevant. Four studies built regression models, while three used ANOVA to find significant HAV predictors. Examples of significant HAV predictors included: tool age, tool weight, participant, and operating power; the proportion of HAV variance explained varied from 46 to 90%. In the second manuscript, the models based on administrative data explained 16% to 27% of A(8) variance. The included variables were tool power source, vibration control, and job type for the parsimonious model; the comprehensive model has the same variables as parsimonious along with accelerometer attachment method and tool brand. Conclusion HAV prediction through exposure prediction modeling is a relatively new method for assessing occupational HAV. It is feasible to find HAV predictors using low-cost administrative data, and variables such as tool power source, job type, and vibration control make promising predictors. However, the variance explained will be lower than using data collected for research purposes. The significant predictors found in the systematic review can be considered when installing protective measures in the future. The significant predictors found and procedures used from the modeling study can be considered for future HAV exposure prediction modeling studies.