Self-selection into healthcare options on the basis of severity likely biases estimates of the effects of healthcare choice on health outcomes. Using an instrumental variables strategy which exploits exogenous variation in the cost of formal-sector care, we show that using such care to treat acute sickness decreases the incidence of fever and malaria in young children in Tanzania. Compared to the instrumental variables estimates, ordinary least squares estimates significantly understate the effects of formal-sector healthcare use on health outcomes. Improved information and more timely treatment, rather than greater access to medicines, seem to be the primary mechanisms for this effect.