Multivariate conditional logistic regression model† of risk factors for HEV disease.

†Only the models constructed using the liberal case definition are shown as the conservative case definition was extremely restricted by a low sample size. Variables also significant using the conservative model are those indicated above by an (‡).*Model reflects explanatory variables of interest significant at the 25% level (p**Model selected by forward and reverse stepwise conditional logistic regression, including variables significant at the 25% level (p§Gender was left out of the stepwise model as female gender was strongly inversely correlated with work outside the home or with performing outdoor work. Stratified analysis was not possible or appropriate due to extremely small numbers in each cell.