Reliability of proxy-reported and self-reported household appliance use

Exposure assessment presents a major challenge for studies evaluating the a ssociation between household exposure to electric and magnetic fields and a dverse health outcomes, especially the reliance on proxy respondents when s tudy subjects themselves have died. We evaluated the reliability of proxy- and self-reported household appliance exposure. We recruited 92 healthy cou ples through either random digit dialing or newspaper advertisements. Train ed interviewers administered questionnaires to each member of a couple inde pendently to assess the reliability of proxy reported household appliance u se. Eighty five couples completed a second interview 2 months later to asse ss the reliability of self-reported appliance use. Reliability uf proxy-rep orted appliance exposure was good when we inquired about having any exposur e to each of ther eight indicator appliances during the past year (range of kappa coefficients = 0.63-0.85; median = 0.76) hut was lower with increase d time to recall or increased detail. Reliability of self respondents repor ting 2 months apart was excellent (range of kappa coefficients = 0.75-0.94; median = 0.87) for having any exposure to the eight indicator appliances d uring the past year, but reliability was again lower with increased detail. When we used self reports at the first interview as the standard, little s ystematic over or underreporting occurred for proxy respondents or fur self respondents reporting 2 months later. Because this study did not include c ases of specific disease, these findings of no systematic differences in re porting do not refer to case or control status. In summary, reliability of self respondents' reports of appliance use is very good for recent time per iods and good for broad aspects of exposure in distant time periods. Proxy respondents can provide information regarding broad aspects of appliance ex posure in the past year, but detailed aspects of exposure or exposure in mo n distant time periods is not reliable.