We study the optimal subsidy on prevention against premature death in an economy composed of two-person households, where the survival of the spouse matters, either because of self-oriented coexistence concerns, or because of altruism. Under a non-cooperative household model, the laissez-faire prevention levels are shown to be lower than the first-best levels, to an extent that is increasing in self-oriented coexistence concerns and decreasing in spousal altruism. The decentralization of the social optimum requires thus a subsidy on prevention depending on the precise type of coexistence concerns. Our results are shown to be globally robust to the introduction of imperfect observability of preferences, life insurance, imperfect marriage mat...