On continuity correction for RSS-structured cluster randomized designs with binary outcomes