Innovations in applied decision theory for health and safety

There are established methods for occupational epidemiological cohort analysis, such as proportional hazards regression, that are well suited to aetiological research and yield parameter estimates that allow for succinct communication among academics. However, these methods are not necessarily well suited for evaluation of health impacts of policy choices and communication to decision makers. An informed decision about a policy that impacts health and safety requires a valid estimate of the policy's potential impact.

We propose methods for data summarisation that may facilitate communication with managers, workers and their advocates. We calculate measures of effect in a framework for competing events, graphically display potential impacts on cause-specific mortality under policy alternatives and contrast these results to estimates obtained using standard Poisson regression methods. Methods are illustrated using a cohort mortality study of 28 546 Ontario uranium miners hired between 1950 and 1996 and followed through 2007.

A standard regression analysis yields a positive association between cumulative radon progeny exposure and all-cause mortality (log(RR per 100 WLM)=0.09; SE=0.02). The proposed method yields an estimate of the expected gain in life expectancy (approximately 6 months per worker) and reduction of 261 lung cancer deaths under a policy that eliminated occupational radon progeny exposure.

The proposed method shifts attention from covariate-adjusted risk ratios or rate ratios to estimates of deaths that are avoided or delayed under a potential policy. The approach may help inform decision-making and strengthen the connection of epidemiological approaches to data analysis with developments in decision theory and systems engineering to improve health and safety.