Approval of the project was gained through the British Columbia Cancer Agency Research Ethics Board (UBC BCCA REB Certificate #H09-00681). The study data was collected as part of a previous project, where a breast cancer risk assessment was conducted for the women participating in the SMPBC [20]. The focus of this manuscript is a subpopulation of women from the assessment who were attending screening mammography at Victoria General Hospital in Victoria, British Columbia (BC) from August 2009 to January 2010. All participants gave voluntary informed consent prior to study involvement, and were 40 to 79 years of age at the time of the study. The study involved the development and administration of a questionnaire to collect personal information related to common risk factors for breast cancer of women attending the SMPBC [20]. Within the consent form, participants were also asked to state their preferred method of response as either mail, telephone, or the study website platform, should they agree to participate [21]. The questionnaire was distributed by both clerks and on-site volunteers, with each participant invited to complete it at her own convenience using her preferred method of response. All participant information and responses were stored in a study database. Once the questionnaire was completed, an unique study identification was assigned to each participant to ensure confidentiality and anonymity.

In order to better understand patient preferences and factors influencing questionnaire response type, both univariate and multivariate analyses were used to explore demographic factors associated with certain response types. All analyses were performed using STATA 13.1.

A total of 1192 women successfully completed the study questionnaire at Victoria General Hospital. Of the three survey modes, distribution by mail was the most preferred (509/1192, 42.70%), followed by website platform (422/1192, 35.40%), and telephone (147/1192, 12.33%; Table 1). Approximately 10% of participants (114/1192, 9.56%) did not specify a preferred method of response in their consent form.

Survey mode preference differed between younger (<60 years) and older (>60 years) patients, as preference for mail and telephone questionnaires strengthened with age (Figure 1). In contrast, the Web-based option became less preferred with each subsequent age grouping. Higher levels of post-secondary education led to a stronger preference for Web-based questionnaires, while the opposite trend was observed for mail. With regards to ethnicity, nearly all groups expressed a preference for the mail mode.

In all, 80.12% (955/1192) of patients completed the questionnaire in the mode that was previously specified as their most preferred method (Tables 2, 3, Figure 2). For those who responded in a method other than their original preference, mail was the most common response type. In addition, for those who did not specify a preferred response method, mail was also the most common response type (Table 3).

The study uptake at the Victoria General Hospital for which this analysis is performed was found to be 47% [20]. This is comparable with the on-site study uptake of 31% to 50% in a similar risk assessment study conducted by Evans and colleagues [22]. Of the questionnaires that were completed, 57.72% (688/1192) were submitted via mail, followed by the 31.12% (371/1192) completed using the website platform, and the 11.16% (133/1192) by telephone (Tables 2, 3). The overall average age for participation was 55 years (95% confidence interval [CI] 54.4-55.5). The average age of participants was significantly lower for those who responded via the website platform (52.9, 95% CI 52.1-53.7) compared with those who used telephone or mail (55.9, 95% CI 55.2-56.5; P<.001).

Each additional level of education showed an increased number of respondents, with university graduates being the largest group to complete the questionnaire (483/1192, 40.52%). Web-based participation was also more likely in higher education groups (P<.001). With regards to ethnicity, the study population was dominated by White participants (1108/1192, 92.95%); additional ethnic groups were then consolidated into a “Nonwhite” group for subsequent analysis.

Logistic regression demonstrated that, after adjusting for age and education level, ethnicity (White vs nonwhite) was not predictive of survey mode usage (P=.96) (Table 4). Both age and education were significantly associated with Web-based participation versus other modes (P<.001 for both variables).

Each decade of increased age was associated with a 0.97-fold decrease in the odds of a patient participating on the website platform when compared to other available methods (odds ratio [OR] 0.97, P<.001). Each interval increase in education level saw a 1.83-fold increase in the odds of website platform usage (OR 1.83, P<.001). However a low R ² value (0.06) denotes that age and education alone are poor predictors of response type, even when considering them together. Therefore, these results suggest that there are factors other than age and level of education that are likely to influence which mode people use to participate.

Consistent with other single-mode survey studies, response rates were higher in the mode that was originally stated as the participant’s preferred mode [16,23]. Mail was the most preferred type of questionnaire response, and was also the most commonly used method when participants responded in a mode other than their original preference. Previous studies conducted on men and women have found that both participant attitudes toward, and familiarity with, a certain type of survey technology play an important role in their usage of that particular mode [16,24]. Therefore, it is not surprising that the mail method was commonly reverted to, especially due to the massive use of paper (mail) in our current culture. However, as Internet usage among adults continues to grow in our society, from a rate of 14% in 1995 to 87% in 2014, for example [25], we may begin to see a shift away from more traditional methods, such as mail toward Internet technologies such as Web-based questionnaires (as offered through this study), mobile applications, email, and text message. Moreover, familiarity with specific survey modes may vary across many demographic groups. For example, studies have shown that both younger participants and those with higher education are more likely to use the Internet and Web-based technologies [26-31]. This is consistent with our findings. In turn, our results did not show ethnicity to be a predictor for (or against) Web-based questionnaire usage in this study. This is consistent with a recent systematic literature review that reported research based in five multiethnic developed countries (Canada, United States, United Kingdom, New Zealand, and Australia); results found that Nonwhite populations (eg, African American, Asian, and Hispanic) who participate in surveys are as likely to participate in research as Whites [32].

For all respondents included in the Weisstock et al. study [20], a mean age of 56.6 years (SD 9.6 years) was observed, which was comparable to our subpopulation mean of 55 years (95% CI 54.4-55.5). These respondents displayed an age distribution comparable to that of women in the SMPBC and British Columbian residents at the time of the study. Our subpopulation was dominated by White participants (1108/1192, 92.95%), which included those of British / French / Irish / Scottish / Welsh / (Northern / Southern / Eastern / Western) European descent (Tables 1, 2). This is consistent with the overall study population for Weisstock et al. [20], where the most prevalent ethnicity was British/Irish/Scottish/Welsh (54% of respondents); furthermore, 49% of British Columbians self-reported ethnic backgrounds of British Isles / European / English / Scottish / Canadian / Irish / Welsh descent during the 2006 census. As with our results, age was found to influence questionnaire response mode, as younger women responded more frequently using the Web-based questionnaire, while older participants tended to respond via mail (Table 2).

As Internet technology continues to be integrated into society at various social, educational, and institutional levels, these demographic groups may not be so predictive of Web-based survey usage. While we expect the Internet to continue to become more accepted as a common means of communication, there are still many individuals, particularly those from older generations and lower education levels, who prefer mail and may be intimidated by their lack of familiarity with mobile platforms or anticipated difficulties in navigating Web-based systems [24,27]. This is especially significant in Canada, where approximately 1 in 6 Canadians are at least 65 years of age [33]. In turn, the current estimated growth rate of this population is 3.5%, approximately four times that of the total population growth rate. While many individuals within this age group may become more comfortable with technology use as they age, we must still prepare for a large proportion of this population to still prefer mail for the duration of their lifetime. Therefore, studies similar to this one will be important in future work, in order to assess how people over the age of 40, the typical age demographic for cancer diagnoses, respond to various forms of survey methods.

With regards to breast cancer, the use of alternative approaches to screening has been proposed [34]. These alternative approaches would involve making more individualized decisions based on a woman’s breast cancer risk factors and beliefs about the risks and benefits of screening mammography. This will require the collection of detailed breast cancer risk information from the eligible women, in order to estimate their risk of developing breast cancer. Web-based collection of the breast cancer risk factors may be an effective way to collect this information and, therefore, the results of this study will be of importance as personalized breast screening regimens become more main stream.

Moving forward, there is a need to protect already declining response rates [5,15,16]. The inclusion of a paper option to a Web-based survey (or vice versa) may be needed so as to prevent a premature dependency on mobile survey methods. Furthermore, it will be beneficial to develop platforms that are user-friendly and intuitive for various age groups, especially those of older generations. This is especially important for research and health service work that aims to obtain a representative sample of the population of interest. It will also be helpful to provide support through study personnel or educational materials, in order to instruct patients on how to properly use a Web-based survey system and increase their confidence with its usage [24]. Integration of such strategies will help minimize the sample bias that may typically occur should survey participation factors not be considered.

The predominant limitation of this study is that all participants involved were registered with the SMPBC, and were consequently adult females in the age range eligible for screening mammography (40-79 years). With the median age of Canadians being 39.9 years, we acknowledge that women within our study are generally older than the general population in Canada [35]. In turn, our population does not allow for any determination of the effect of gender on response rate and mode. Therefore, our results regarding the effects of participant preference and demographic effect on survey mode response may not be generalizable to populations that include males and females of other ages.

Our findings suggest that it is important to consider participant’s survey mode preference when designing and implementing PRO surveys. In turn, younger participants and those with higher levels of education were more likely to use the mobile platform; Web-based participation failed to vary across ethnic groups. Overall, this information will be very valuable during the planning stages of future studies. PROs are becoming increasingly used in breast cancer screening regimens and health care systems of the developed world, making it imperative that approaches to obtaining patient feedback are representative of the population being served. For this reason, clinicians and researchers must be diligent in implementing survey modes that capture data of their desired patient cohorts.