Quasi-experimental study which aimed to verify the influence of nursing activities redesign in the reduction of medication errors in three pediatrics wards of a university hospital. Types and frequencies of medication errors identified in a study carried through the wards guided the redesign and exerted the function of dependent variable in the assessment of the intervention. To errs identification 556 documents on 77 children's medical charts were analyzed. In 8550 medication doses analyzed, in 1498 (17,5%) errors were evidenced, an inferior ratio (21,1%) of the control study. Globally the intervention generated small changes in medication errors ratio and type, being effective to reduce dose omission (p< 0.0001), medication suspended by physicians and not registered as suspended for the nurses (p<0.0001) and wrong hour (p= 0,0002).