Quality of response in different population groups in mail and telephone surveys.

Mail and telephone survey methods, with follow-up by other methods, can provide high response rates. However, it is not clear whether different population groups provide responses of different quality, thus creating risk of biased comparisons. A closely related problem is whether proxy response adequately substitutes for self-response. This study addressed these issues in the context of parallel mail and telephone health surveys carried out in Montreal. In the telephone survey, proxy respondents provided lower estimates of morbidity and health care utilization than self-respondents; in the mail survey, there was no difference between proxy and self-response. Response validity was assessed by comparing reported physician visits with those recorded by the government-run universal health insurance plan. In general, mail responses were more valid than telephone responses. In both methods, there were suggestive but not persuasive differences in validity among sociodemographic subgroups. In both methods, those reporting illness or medication use had less underreporting of physician visits than those not reporting such things.