Rural cancer survivors have worse outcomes than their urban counterparts. To improve outcomes, it is essential that rural survivors participate in research, yet they are underrepresented in cancer research. The aim of this study was to assess urban-rural differences in participation in a cancer survivorship survey and differences in mode of participation (mail, online, or phone) by rurality and age.

We developed a survivorship needs assessment survey and invited cancer survivors to participate by mail, online, or phone. We compared participation between rural and urban invitees and examined differences in mode of participation by rurality and age.

A quarter (25.47%) of invited rural patients and 27.84% of invited urban patients participated in the survivorship study. The probability of participation by urban survivors was approximately 1.09 times higher than for rural survivors (χ2(1) = 4.31, p = 0.038). Rural survivors were more likely to participate by mail (average difference [Rural-Urban] = 9.64%, p < 0.001), while urban survivors were more likely to participate online (average difference [Urban-Rural] = 8.77%, p < 0.001). As participant age increased, the likelihood of survey participation by mail increased (1.16% per year of age, p < 0.001) while the probability of participating online decreased by 1.20% per year of age (p < 0.001).

To ensure equitable access to research for rural and older cancer survivors, researchers should design studies with a range of participation modes. Non-digital methods, such as mailed paper surveys, appear to promote participation among rural and older survivors.

To undertake an assessment of survey participation and non-response error in a population-based study that examined the relationship between socio-economic position and food purchasing behaviour.

The study was conducted in Brisbane City (Australia) in 2000. The sample was selected using a stratified two-stage cluster design. Respondents were recruited using a range of strategies that attempted to maximise the involvement of persons from disadvantaged backgrounds: respondents were contacted by personal visit and data were collected using home-based face-to-face interviews; multiple call-backs on different days and at different times were used; and a financial gratuity was provided.

Non-institutionalised residents of private dwellings (n = 1003), located in 50 small areas that differed in their socio-economic characteristics.

Rates of survey participation – measured by non-contacts, exclusions, dropped cases, response rates and completions – were similar across areas, suggesting that residents of socio-economically advantaged and disadvantaged areas were equally likely to be recruited. Individual-level analysis, however, showed that respondents and non-respondents differed significantly in their sociodemographic and food purchasing characteristics: non-respondents were older, less educated and exhibited different purchasing behaviours. Misclassification bias probably accounted for the inconsistent pattern of association between the area- and individual-level results. Estimates of bias due to non-response indicated that although respondents and non-respondents were qualitatively different, the magnitude of error associated with this differential was minimal.

Socio-economic position measured at the individual level is a strong and consistent predictor of survey non-participation. Future studies that set out to examine the relationship between socio-economic position and diet need to adopt sampling strategies and data collection methods that maximise the likelihood of recruiting participants from all points on the socio-economic spectrum, and particularly persons from disadvantaged backgrounds. Study designs that are not sensitive to the difficulties associated with recruiting a socio-economically representative sample are likely to produce biased estimates (underestimates) of socio-economic differences in the dietary outcome being investigated.