The purpose of this paper is to investigate the factors affecting hospital stay involving drunk driving and non-drunk driving of accident victims in Phuket, Thailand. However, the decisions to drink and drive are made by the road users before the accidents occurred. Self-selection bias may arise when some component of the drink and drive decision is relevant to the length of stay in hospital of accident victims. We discuss a new approach to specifying and estimating zero-inflated negative binomial models with endogenous switching, based on copula function. These models provide a framework of analysis for the effect of self-selection in drunk-driving behaviors on the length of stay in the hospitals. We use the concept of pair-copula constructions for discrete margins to derive the likelihood function of our model system. The results suggest that drunk-driving accident victims are positively selected. © Springer International Publishing Switzerland 2014.