regression deficit