Determining the interviewer effect on CQ Index outcomes: a multilevel approach