Iterated defect correction for differential equations part I: theoretical results