Long-term health insurance contracts have the potential to efficiently insure against reclassification risk, but at the expense of other limitations like provider lock-in. This paper empirically investigates the workings of long-term contracts which are subject to this trade-off. Individuals are shielded against premium increases and coverage denial as long as they stay with their initial contract, but those that become higher risk are subject to premium increases or coverage denials upon switching, potentially leaving them locked-in with their original network of providers. I provide the first empirical evidence on the importance of this phenomenon using administrative panel data from the universe of the private health insurance market in Chile, where competing insurers o˙er long term contracts. I fit a structural model to yearly plan choices, and am able to jointly estimate evolving preferences for different insurance companies and supply-side underwriting in the form of premium risk-rating and coverage denial. To quantify the welfare effects of lock-in, I compare simulated choices under the current rules to those in a counterfactual scenario with no underwriting. The results show that consumers would be willing to pay around 13 percent more in yearly premiums to avoid lock-in. Finally, I study a counterfactual scenario where long-term contracts are replaced with community-rated spot contracts, and I find only minor general-equilibrium effects on premiums and on the allocation of individuals across insurers. I argue that these small effects are the result of large levels of preference heterogeneity uncorrelated to risk.