A research routine to assess bias introduced by low response rates in postal surveys.

Low response rates to postal surveys potentially bias study results. We used three approaches to determine why 46 per cent of a sample were non-responders, and to analyse any potential bias. Labour force data, telephone interviews with a number of non-responders and trend examination showed that our study sample was no different to the known nursing population, that there were few differences between responders and non-responders and that there were no trends in differences between early and late responders respectively. Results suggest 'intenders', or potential responders who do not complete and return the survey, are a key factor in non-response in surveys of nurses. Analysis for response bias increases confidence in the interpretations and conclusions of any study and should therefore become standard survey practice.