This study was registered with the Australian New Zealand Clinical Trials Registry (ACTRN) on August 14, 2018 (ACTRN12618001364268). The study protocol was published on April 23, 2020 (international registered report identifier: DERR1-10.2196/16108) [12].

This pilot longitudinal cohort study was conducted at Macquarie University in Sydney, Australia. It involved collecting information via web-based surveys (ie, at baseline and 12 months post baseline) and health data record linkage.

All students graduating from Macquarie University in 2018 (N=8532) were eligible to participate in this study. Macquarie University is a large public university located in a suburban area of Sydney, Australia. At the time of participant recruitment, the university comprised five faculties (ie, Faculty of Arts, Faculty of Business and Economics, Faculty of Human Sciences, Faculty of Medicine and Health Sciences, and Faculty of Science and Engineering), which collectively hosted approximately 45,000 students, including 33,000 undergraduate students, 9000 postgraduate students, and 1500 higher-degree research students.

The graduates were invited to participate via email during the autumn (ie, April) and spring (ie, September) graduation periods. Email invitations included a unique link to a purpose-made website where the graduates were informed about the study before consenting to participate. The initial invitation was followed by 3 reminder emails over a 6-week period. After completing the web-based registration form, participants received an email with an individualized link to the baseline survey.

An incentive to participate was introduced for the second (ie, September) graduation period. The incentive to participate was an entry into a random draw to win 1 of 3 prizes, namely an iPad mini (Apple Inc, first prize) or movie tickets (second and third prizes).

Surveys were administered via the web-based Qualtrics XM platform (Qualtrics International Inc) at baseline and 12 months post baseline. The baseline and 12-month follow-up surveys comprised the same battery of validated questionnaires and instruments designed to capture data regarding sociodemographic factors, education, employment, job satisfaction, mentoring, self-perceived physical and mental health status, work-life balance, connectedness, resilience, injury, risk behaviors, and life events, as well as social media and technology use. For a detailed overview of the domains and specific questionnaires included in the web-based surveys, see the published study protocol [12].

Survey responses were linked to personal health information from administrative data collections (ie, ambulance dispatches, emergency department presentations, hospital admissions, cancer registry, and mortality records) in New South Wales (NSW) from April 1, 2018, to 12 months after the completion of the baseline survey. Participants had an opportunity to opt out of having their survey responses linked to their health records during the web-based registration process. The secure health data linkage was conducted by the Centre for Health Record Linkage (CHeReL).

The specific feasibility outcome measures for this pilot study were as follows:

All data were analyzed using SAS, version 9.4 ( SAS Institute). The recruitment rate, response rate, retention rate, data linkage opt-out rate, data linkage rate, and linked record availability rate were calculated as described above and presented as proportions. Descriptive statistics were used to evaluate the representativeness of the sample and completeness of the baseline and follow-up surveys.

Of the 2.8% (238/8532) of invited graduates who agreed to participate in this pilot cohort study, 59.7% (142/238) responded to the baseline survey. This is considerably lower than the response rate for, for instance, the pilot SUN study (11%) [13] and the Australian arm of the World Health Organization’s World Mental Health Surveys International College Student initiative (7%) [14]. Although the overall recruitment and response rates are disappointing, they are perhaps unsurprising given the general decline in survey participation observed in recent decades [15-17]. Superimposed on this general decline, there is a myriad of factors that may have contributed to the relatively poor recruitment and response rates observed in this pilot cohort study. Dillman’s extension of social exchange theory, the tailored design method, is a theoretical framework that seeks to explain why individuals are motivated to engage in certain social behaviors such as survey participation [18]. This framework suggests that survey response rates depend on reward, cost, and trust. For instance, survey participation is typically more rewarding when participants have a vested interest in the topic [19]. The time spent completing a survey is an important cost consideration for survey participants [19], with longer stated survey length resulting in fewer respondents [20]. In addition to survey length, poor survey structure and design can increase the perceived cost of responding to surveys [19]. In regard to trust, perceived trustworthiness of the organization or institution responsible for administering the survey, confidential use of data, and adequate privacy protections are key elements for reassuring survey participants and improving response rates [19].

It is difficult to determine to which extent each of the abovementioned factors have influenced the recruitment and response rates in this pilot cohort study. Although one might expect recent university graduates to have a vested interest in the topic of health and well-being in emerging adults, perhaps the relative absence of health problems in this age group resulted in a lower interest in the topic and thus lower perceived reward and motivation for participation. For instance, there is evidence suggesting that emerging adults are less motivated by long-term health concerns and lifestyle interventions than older counterparts [21-24]. Additionally, perhaps graduating from university is accompanied by a sense of separation and decreased interest in participating in university-based research surveys. There is also the potential issue of different surveys competing for graduates’ attention and motivation. For instance, Australian university graduates are regularly invited to complete the nationwide Graduate Outcomes Survey, which is one component of the Quality Indicators for Learning and Teaching suite of surveys conducted for the Australian Government Department of Education, Skills and Employment [25]. Competition for attention and motivation becomes a particularly important consideration in the context of emerging adults’ perceived scarcity of time [26]. The graduates invited to participate in this pilot study were informed that it would take approximately 40 minutes to complete each survey (ie, baseline and follow-up). It is conceivable that many potential participants considered the perceived costs of participation in this study to be too high.

Although there is conflicting evidence, some studies have demonstrated that incentives such monetary rewards or lotteries can positively impact response rates [27,28]. In an attempt to improve the recruitment rate in this pilot study, an incentive to participate was introduced for the second (ie, September) graduation period. Consequently, the recruitment rate increased from 2% for the first (ie, April) graduation cohort to 3.8% for the second graduation cohort. The improvement after introducing incentives notwithstanding, the recruitment rate remained disappointingly low. This suggests that the incentives to participate in this pilot study were insufficient to counteract the perceived costs of participation for the vast majority of invited graduates. Perhaps the proliferation and ubiquity of mobile devices in modern society means that the main incentive used in this pilot study (ie, a chance to win an iPad mini) is not perceived as an attractive reward for participating in research.

Previous research has demonstrated that multiple reminders are an effective way to increase response rates [29-31]. However, studies have also shown that the recruitment yield typically declines rapidly with each subsequent reminder [32-34]. This rapidly diminishing marginal return suggests that increasing the number of reminders beyond a small number may not be a cost-effective measure [35]. Hence, Saleh and Bista [19] recommended sending at least 1 reminder, but not more than 3. Furthermore, the desire for increased survey response rates also needs to be balanced with the concern among human research ethics committees that multiple reminders may result in potential survey participants feeling harassed or coerced into participating in the research [36,37].

Survey completeness can be used to refer to 3 different concepts: (1) completeness of the achieved sample with respect to the original one, (2) participation of the respondents throughout all the phases predicted by a research design, and (3) respondents’ propensity to answer all the questions within the questionnaire [29]. The latter 2 concepts were considered in this pilot study. The first of which corresponds to the retention rate between the baseline and follow-up surveys, which was found to be 69.7% (99/142). In regard to the propensity to answer all the questions within a survey, the completeness was excellent, with the proportion of questions answered in each survey domain ranging from 87% to 100% in both the baseline and follow-up surveys. This suggests that the participants did not consider the survey length to be excessive in the pilot study. This is important because previous research has shown that longer survey length can result in both poorer completeness and quality of responses [26].

Approximately two-thirds (161/238, 67.6%) of the participants in this pilot study did not opt out of having their survey responses linked to their health records. This suggests that the acceptability of data linkage in the present study was very similar to that in previous Australian studies. For instance, one study reported that 66% of older Australians found it acceptable to have their health data accessed and linked in a registry [38]. Another study of young Australian parents reported, unsurprisingly, that privacy protection was an important consideration for most participants [39]. However, it also was noted that protection measures adopted in best practice health data linkage studies were viewed by most participants as adequate protection for data linkage to proceed without specific individual consent.

This pilot study offers insights into the feasibility of recruiting recent university graduates to establish a longitudinal cohort study to investigate health and well-being in emerging adults. It is unrealistic to expect a reversal of the general decline in survey response rates. However, it has been suggested that reducing nonresponse rates is less important than minimizing bias in estimates. That is, despite preconceived notions of a good response rate, neither a 5% response rate nor even a 75% response rate necessarily provides unbiased estimates [40,41]. Nonresponse bias occurs when subgroups respond at different rates. In this pilot study, female graduates were more likely to respond to the baseline survey. Overrepresentation of females is common in health-related survey research. Although postsurvey adjustment techniques can be implemented to help reduce nonresponse biases, it is often preferable to prevent nonresponse bias by designing surveys that are more acceptable to the target population in the first instance [42].

In the context of this pilot study, future cohort studies of emerging adults should target populations that are more likely to hold interest in the research. For instance, it might be more useful to target populations at an earlier stage, such as at first enrollment at university or before leaving high school. There are potentially more opportunities for targeted engagement with a student cohort who are commencing postsecondary education. Only a subset of emerging adults undertakes postsecondary education. In Australia, 39.7% of people aged 25 to 34 years had a bachelor’s degree or higher in 2018 [43]. Thus, a cohort of high school graduates offers a less biased sample.

Apart from the choice of specific target populations, future studies should consider the need for personalizing invitations, reducing survey length, crafting surveys that are simple to complete, administering surveys via smartphone apps, and incentivizing participation (eg, through gamification). Lastly, future studies are strongly encouraged to use research codesign to optimize survey parameters. Incorporating the lived experiences of emerging adults into the development and implementation of future research will increase its likelihood of success and impact [44,45].

The overall recruitment rate was poor, while the completeness of survey responses among respondents was good to excellent. There was reasonable acceptability for conducting data linkage of health records from administrative data collections and survey responses. Future research aiming to establish a longitudinal cohort study to investigate health and well-being in emerging adults should carefully consider the target population as well as how best to obtain an unbiased sample and craft surveys to maximize participation.