A dynamic population-based model for the development of work-related respiratory health effects among bakery workers

This paper presents a dynamic population-based model for the development of sensitisation and respiratory symptoms in bakery workers. The model simulates a population of individual workers longitudinally and tracks the development of work-related sensitisation and respiratory symptoms in each worker.

The model has three components: a multi-stage disease model describing the development of sensitisation and respiratory symptoms in each worker over time; an exposure model describing occupational exposure to flour dust and allergens; and a basic population model describing the length of a worker's career in the bakery sector and the influx of new workers. Each worker's disease state is modelled independently using a discrete time Markov Chain, updated yearly using each individual's simulated exposure. A Bayesian analysis of data from a recent epidemiological study provided estimates of the yearly transition probabilities between disease states.

For non-atopic/non-sensitised workers the estimated probabilities of developing moderate (upper respiratory) symptoms and progression to severe (lower respiratory) symptoms are 0.4% (95% CI 0.3 to 0.5%) and 1.1% (95% CI 0.6 to 1.9%) per mg/m(3)/year of flour dust, respectively, and approximately twice these for atopic workers. The model predicts that 36% (95% CI 26 to 46%) of workers with severe symptoms are sensitised to wheat and 22% (95% CI 12 to 37%) to alpha-amylase. The predicted mean latency period for respiratory symptoms was 10.3 years (95% CI 8.3 to 12.3).

While the model provides a valuable population-level representation of the mechanisms contributing to respiratory diseases in bakers, it was primarily developed for use in quantitative health impact assessment. Future research will use the model to evaluate a range of workplace interventions, including achievable reductions in exposure and health surveillance. The general methodology is applicable to other diseases such as chronic obstructive pulmonary disease, silicosis and musculoskeletal disorders and could be particularly valuable for forecasting changes in long latency diseases.