This paper investigates the impact of usage patterns of water and sanitation on the health of children in Ethiopia, once supply side factors are controlled for. A comprehensive set of controls is employed to account for individual heterogeneity driven by time-invariant unobservables, time-invariant and time-variant observables. The results from the pooled cross-section estimated by OLS suggest that there is a strong relationship between water and sanitation choices of a household and a child's weight-for-age z-scores. This correlation disappears when unobserved time-invariant fixed effects are removed through a fixed effects specification. However, the fixed effects model reveals that children from the poorest families who improved usage patterns of water and sanitation exhibited a significantly larger growth in weight-for-age z-scores between the two periods. The effect remains robust across a number of different specifications, including a dynamic model to allow for catch-up effects and an instrumented version of a dynamic model. Consequently, the effectiveness of investments in water and sanitation infrastructure can be leveraged by influencing the choices made by households regarding the type of water and sanitation.