There is definitely some clustering of child-level outcomes by the mother and/or household. At DHS we do not adjust for that, but you certainly may want to do so. A complication is that the number of children per woman is low, especially if you are focusing on outcomes within a recent time period and on children in a narrow age range such as 0-4. Other users may be able to suggest how to specify svyset for such outcomes.