Comparison between Generalized Linear Modelling and Additive Bayesian Network

Identification of Factors associated with the Incidence of Antibodies against Leptospira interrogans sv Pomona in Meat Workers in New Zealand

Additive Bayesian Network (ABN) is a graphical model which extends Generalized Linear Modelling (GLM) to multiple dependent variables. The present study compares results from GLM with those from ABN analysis used to identify factors associated with Leptospira interrogans sv Pomona (Pomona) infection by exploring the advantages and disadvantages of these two methodologies, to corroborate inferences informing health and safety measures at abattoirs in New Zealand (NZ).

In a cohort study in four sheep slaughtering abattoirs in NZ, sera were collected twice a year from 384 meat workers and tested by Microscopic Agglutination with a 91% sensitivity and 94% specificity for Pomona. The study primarily addressed the effect of work position, personal protective equipment (PPE) and non-work related exposures such as hunting on a new infection with Pomona. Directly, significantly associated with Pomona were “Work position” and “Abattoir” (GLM), and “Work position” (ABN).

Facemasks and safety glasses did not show any indication of being protective in GLM, on the contrary, such PPE increased the odds of infection in ABN, but this requires verification using other research methodology. In ABN, all relationships between variables are modelled; hence it has an advantage over GLM due to its capacity to capture the natural complexity of data more effectively.