This article has pretty good data on the patient safety impacts of doctor shift length.
In a secondary analysis that was adjusted for the number of patients per resident physician as a potential confounder, intervention schedules were no longer associated with an increase in errors.