Kypros Kypri1, Shaun Stephenson, John Langley. 1. Department of Preventive and Social Medicine, Injury Prevention Research Unit, University of Otago, Dunedin, New Zealand. kypros.kypri@hunter.health.nsw.gov.au
Abstract
BACKGROUND: Decreasing survey response rates are a growing concern in epidemiological research, principally because prevalence estimates may be biased by selective nonresponse. Internet-based methods have the potential to yield higher-quality data with lower nonresponse rates and at a lower cost than traditional methods. Little research exists on nonresponse bias in Internet surveys of alcohol use. This investigation draws on a study of the implementation of an Internet-based alcohol survey involving a random sample of 1910 university students with a response rate of 82% (n = 1564). Our aim was to identify nonresponse bias and to quantify its effects on estimates of alcohol consumption, the incidence of alcohol-related problems, and the prevalence of hazardous drinking. METHODS: Survey nonresponse has been characterized in terms of a continuum of resistance model, in which the propensity of individuals to respond is inferred from the level of effort required to elicit a response. Two methods were used to test this model: comparison of the demographic characteristics of the target sample with those of the respondents and comparison of alcohol variables for those who responded late with those who responded early. RESULTS: The results attained with method 1 showed that bias varied as a function of gender, age, ethnicity, and living arrangement. The results attained with method 2 showed that the incidence of alcohol-related problems and hazardous drinking prevalence varied as a function of response latency. If only the early and intermediate respondents had participated, the incidence of alcohol-related problems and the prevalence of hazardous drinking would each have been underestimated by 3%. CONCLUSIONS: The findings reported here are consistent with the continuum of resistance model but show that the bias resulting from nonresponse is arguably too small to be of concern with respect to estimating consumption levels, the incidence of alcohol-related problems, and the prevalence of hazardous drinking.
BACKGROUND: Decreasing survey response rates are a growing concern in epidemiological research, principally because prevalence estimates may be biased by selective nonresponse. Internet-based methods have the potential to yield higher-quality data with lower nonresponse rates and at a lower cost than traditional methods. Little research exists on nonresponse bias in Internet surveys of alcohol use. This investigation draws on a study of the implementation of an Internet-based alcohol survey involving a random sample of 1910 university students with a response rate of 82% (n = 1564). Our aim was to identify nonresponse bias and to quantify its effects on estimates of alcohol consumption, the incidence of alcohol-related problems, and the prevalence of hazardous drinking. METHODS: Survey nonresponse has been characterized in terms of a continuum of resistance model, in which the propensity of individuals to respond is inferred from the level of effort required to elicit a response. Two methods were used to test this model: comparison of the demographic characteristics of the target sample with those of the respondents and comparison of alcohol variables for those who responded late with those who responded early. RESULTS: The results attained with method 1 showed that bias varied as a function of gender, age, ethnicity, and living arrangement. The results attained with method 2 showed that the incidence of alcohol-related problems and hazardous drinking prevalence varied as a function of response latency. If only the early and intermediate respondents had participated, the incidence of alcohol-related problems and the prevalence of hazardous drinking would each have been underestimated by 3%. CONCLUSIONS: The findings reported here are consistent with the continuum of resistance model but show that the bias resulting from nonresponse is arguably too small to be of concern with respect to estimating consumption levels, the incidence of alcohol-related problems, and the prevalence of hazardous drinking.
Authors: James A Cranford; Sean Esteban McCabe; Carol J Boyd; Janie Slayden; Mark B Reed; Julie M Ketchie; James E Lange; Marcia S Scott Journal: Addict Behav Date: 2007-08-03 Impact factor: 3.913
Authors: Mary E Larimer; Clayton Neighbors; Ty W Lostutter; Ursula Whiteside; Jessica M Cronce; Debra Kaysen; Denise D Walker Journal: Addiction Date: 2012-02-28 Impact factor: 6.526
Authors: Mary E Larimer; Christine M Lee; Jason R Kilmer; Patricia M Fabiano; Christopher B Stark; Irene M Geisner; Kimberly A Mallett; Ty W Lostutter; Jessica M Cronce; Maggie Feeney; Clayton Neighbors Journal: J Consult Clin Psychol Date: 2007-04
Authors: Jonathan Hallett; Peter M Howat; Bruce R Maycock; Alexandra McManus; Kypros Kypri; Satvinder S Dhaliwal Journal: BMC Public Health Date: 2012-01-16 Impact factor: 3.295