Effects of survey mode on self-reports of adult alcohol consumption: a comparison of mail, web and telephone approaches.

OBJECTIVE: Mail and Web surveys are two possible alternatives for reducing potential nonresponse bias in telephone-based studies of alcohol consumption. Mail surveys have been a staple of epidemiological research, but there is a dearth of similar research on Web surveys. Most studies using Web surveys have been conducted among college students or other specialized populations where Internet penetration is relatively high. The present study examines differences in self-reports of alcohol consumption among adult members of the general public surveyed by mail, Web and telephone.
METHOD: Web and mail versions of the 2003 Behavioral Risk Factor Surveillance System (BRFSS) instrument, traditionally a telephone-only survey, were developed and administered to address-matched households drawn from the random-digit dialed sampling frame used for the BRFSS. Comparisons were made with results from the ongoing telephone-based surveillance.
RESULTS: A total of 4,051 interviews (836 mail, 1,143 Web, and 2,072 telephone) were completed with adults ages 18 or older in four U.S. states. The findings indicate considerable variation in the estimates for heavy drinking (five or more drinks on an occasion during the past 30 days) obtained across these modes and population subgroups, particularly among the Web respondents.
CONCLUSIONS: Mail surveys appear to be a viable alternative to more traditional telephone surveys, whereas use of Web surveys with the general public appears more problematic. Both approaches, however, may complement telephone studies when used in mixed-mode designs. Because alcohol consumption is associated with a number of diseases, caution is required when using different modes of data collection in epidemiological studies.