Optimizing the design of web-based questionnaires--experience from a population-based study among 50,000 women.

Web-questionnaires are an important tool for future epidemiological research because these allow for rapid and cost-efficient assembly of self-reported information on risk factors and health outcomes. However, to achieve high response rates it is essential to accommodate factors that prevent drop out and so insure validity of future studies. We aim to study how socio-demographic variables as well as design issues such as the ordering and level of difficulty (Easy-to-hard vs. Hard-to-easy) of questions in a web-questionnaire affects the probability of drop out and non-response. In 2003 we invited 47,859 women participating in an ongoing prospective study to a follow-up using a web-based mode. Two versions of the questionnaire existed, varying in level of difficulty (Easy-to-hard vs. Hard-to-easy). We report drop out (proportion non-completers) between groups defined by level of difficulty and estimated adjusted risk differences. The drop out differs significantly depending on the order of the questions in the web-questionnaire. The socio-demographic pattern among lurkers (participants that enter, start responding to, but do not complete a web-questionnaire) differs from that among completers of web-questionnaires. An additional 6% units of completers--persons initiating and completing the questionnaire--can be obtained by considering the ordering of questions. A group uniquely identified in web-surveys, as lurkers are potentially easier to persuade to complete an already started web-questionnaire compared to a non-responder. Lurkers thus constitute a unique opportunity of decreasing the drop out rate and therefore merit future research.