Quantitative risk assessment: is more complex always better? Simple is not stupid and complex is not always more correct

Abstract

In quantitative risk assessments a large variety of complexities can be found, from simple and deterministic to very extensive and stochastic. This publication advocates that both simple and complex approaches have their value and should be done in parallel. The simple analysis gives much insight and can help to detect main factors and potential errors in the complex analysis. Extensive analysis with increased complexity suggests better precision but might not increase the accuracy, due to the uncertainty in the additional parameters. However, complex analysis supplies more confidence in certain phenomena and might also increase insight. This is shown with two examples. The first is the effectiveness of sampling plans for powdered infant formula, for factories operating at various levels of contamination. The results of a simple determination, an analysis including a within batch variability and an analysis including both within batch and between batch variability will be compared. The last approach has as advantage that apart from determining the probability of rejection of a batch, it can determine also the reduction of the health risk in the population following a certain sampling plan; it is more complex but it also does bring additional information. However the conclusions still contain large uncertainty, due to the difficulty of obtaining realistic values of the within batch and between batch variability. The second example is dose-response relations comparing the exponential model (one parameter), the beta-Poisson model (two parameters) and the Weibull-gamma model (three parameters). The conclusion is not that simple is best, but that simple is not stupid, and provides valuable information. Complex, on the other hand, is not always by definition more correct, but also does have its merits.