We evaluate the effect of the size of deductibles in the basic health insurance in Switzerland on the probability of a doctor visit. We employ nonparametric bounding techniques to minimise statistical assumptions. In order to tighten the bounds we consider three further assumptions: mean independence of an instrument, monotone treatment response, and monotone treatment selection. Under the first two assumptions we are able to bound the treatment effect of high deductibles compared to low deductibles below zero. Adding the third assumption allows to tighten the bounds further. We conclude that there is a negative treatment effect.