Introduction

Participant recruitment and retention for longitudinal studies remain a major hurdle in cohort research and must be addressed to increase data collection efficiency and scientific robustness [1, 2]. Although longitudinal cohort research designs offer unique strengths mainly due to their capacity for assessing temporal changes and causal effects, participant retention and engagement over time is a challenge that intensifies with increasing length of follow-up. Indeed, prospective longitudinal cohort studies are time-consuming and require substantial financial investments to initiate and collect data over time [3]. For instance, one of the longest longitudinal studies, the Nurses’ Health Study established in 1976, led to an expenditure of over one hundred million US dollars [4]. Furthermore, attrition is commonly reported in prospective studies where participants are hard to trace for follow-up research [5, 6]. Attrition compromises the integrity of research by threatening both the internal and external validity of empirical results, weakening the accuracy of statistical inferences and the generalizability of findings [7–13].

A possible option that has the potential to facilitate longitudinal research is to trace original participants from historical cohorts, and ultimately conduct follow-up research. The term “tracing” or “tracking” is used to outline the process of locating, targeting, and re-engaging with participants during specific periods to measure changes and trends in variables of interest [14, 15]. Traditionally, common methods such as advertisement posters, business cards, mailshots, newspapers, and radio advertisements have been used to recruit participants, but tend to be time-consuming, costly, and produce low response rates [16]. The past couple of decades have seen remarkable advancements in information systems including new platforms that have contributed to an evolution in participant recruitment and tracing methods. For example, while web search phone directories were used in research to trace cohort participants [11], such methods have now become less useful or obsolete. Recently, social networking sites (SNS) have become an avenue for participant recruitment [5, 17, 18] and may be promising tools for participant tracing and retention. Accordingly, with the increasing use of SNS such as Facebook, Twitter (X), LinkedIn, Instagram, etc., there is an opportunity to leverage these platforms for recruitment and cohort study follow-up, potentially improving efficiency and facilitating the collection of more accurate and timely data. Indeed, with over 4.5 billion global users of SNS [19], researchers can now reach a large and diverse pool of potential participants in a timely manner.

Given the size and reach of SNS, it represents a potential avenue to mitigate the challenges related to participant tracing and attrition in longitudinal cohort studies, including offsetting financial burdens and increasing efficiencies. The use of Facebook in particular has been implemented by researchers to re-engage participants in follow up studies, leading to improved response rates and reduced attrition [5, 6, 18, 20]. For example, a study by Bolanos et al. [20] aimed to maximize follow-up response rates in a longitudinal study of adults using methamphetamine and successfully located 8.7% of participants through Facebook, with 2% completing the follow-up interview. In another study, the use of Facebook led to a 16% reduction in loss to follow-up in a longitudinal research project with two follow-up points at seven and ten years [5]. Moreover, using SNS may also maximize the recruitment and retention of certain cohorts that engage in social media daily, such as adolescents [18]. For example, a study found that targeting adolescent girls through Facebook minimized loss of follow-up, enabling them to find 45% of the lost cohort (n = 175) [18]. SNS searching may be particularly useful for addressing stigmatized health-related concerns and in situations where in-person methods may be restricted, such as during the COVID-19 pandemic [21]. In addition to the use of SNS, studies suggest using a multi-pronged strategy to increase participant recruitment rates by incorporating branding, outreach/networking, on-site presence, and incentives [22, 23].

While previous literature has demonstrated the potential benefits of using SNS to re-engage participants in longitudinal research [5, 22–24], the effectiveness of SNS for participant tracing and re-engagement in more recent years remains underexplored [25]. To our knowledge, no research has explored the effectiveness of using SNS to trace participants after decades of no contact. Considering the exponential growth and changing landscape of SNS, it is essential to re-evaluate the effectiveness of these various platforms in not only recruiting and retaining participants in longitudinal research but also tracing original participants from historical cohorts.

Materials and methods

Ethics consent to conduct this study and secondary analysis of the Hamilton Child Cohort Study (HCC) dataset, was received from Ontario Tech University (REB File #17547). Original participants of the HCC consented to be contacted for follow up research. Participants were traced using publicly available information on SNS.

Results

Using Facebook, a total of 479 individuals were identified who were considered to be participants of the HCC study. Using LinkedIn, 185 participants were identified. Instagram was used to identify nine potential participants. Only one participant was identified on X. Of the 3,166 participants traced, 21% were linked to one social media account and thus identified as having SMP, while 10% were not found or traced to any social media accounts (NSMP). Overall, 68% of participants of the original cohort were not able to be conclusively identified for reasons such as being linked to multiple accounts with the same name, accounts with not enough information, or if ages did not match the estimated cohort age. Of the people identified to have SMP, 15% (n = 479) and 6% (n = 185) of the sample were found through Facebook and LinkedIn, respectively (Figure 1).

Display full size

Figure 1. 

Findings of the total Hamilton Child Cohort Study (HCC) sample that were traced through social networking sites (n = 3,166). DNC: participants who requested in 2006 to not be contacted again. SMP: social media presence; NSMP: no SMP

Demographic and health-related characteristics of the SMP and NSMP cohort are summarized in Table 1. There was a significant difference in relation to the proportion of males (59%) compared to females (41%) traced through social media (χ² = 12.83, p < 0.001). In the current study, more individuals of the SMP cohort resided in the WU neighbourhood during childhood (31%), whereas the highest proportion of the NSMP cohort resided in the EL neighbourhood (34%) (χ² = 8.23, p = 0.042). A higher proportion of participants whose family income in childhood was above $20,000 were located through social media, compared to their counterparts with NSMP (56% and 53%, respectively). In relation to health characteristics, a higher proportion of individuals with NSMP reported ever having asthma (9%), compared to their counterparts traced to a social media account (7%). In addition, about 7% of the NSMP cohort were found to report chest illness during childhood, compared to 6% of the SMP cohort. A higher proportion of individuals in the NSMP cohort were hospitalized at least once compared to those of the SMP cohort (6.9% vs 5.6%, respectively). Compared to their counterparts, a higher proportion of individuals of the NSMP cohort were reported to have smoked during childhood (24% vs 22%). These differences were not statistically significant.

Current demographic characteristics of the SMP cohort

The study was able to deduce the current location of 80% of the SMP cohort (Table 1). Of those individuals, 409 (61.5%) reported that they still live in Hamilton, followed by 20 individuals currently living in other Halton Region cities (e.g., Burlington, Oakville), and the rest dispersed across provinces in Canada, or residing in different countries. In relation to educational status, 64% of individuals had data available on the secondary school (n = 159) or post-secondary institution (n = 266) attended. The proportion of participants of the SMP cohort that had data available on the high school they attended, and who still reside in Hamilton was 61%. The most attended high schools included Hill Park Secondary School (14%), Scott Park Secondary (13%), and Westmount Secondary School (9%). Furthermore, of the SMP cohort that had available data on their post-secondary institution, 70% still resided in Hamilton (Figure 2). The most attended post-secondary institution was Mohawk College (54%), followed by McMaster University (24%), Sheridan College (3%), Humber College (2%), and others (18%). This information suggests that a majority of the original cohort stayed in Hamilton for post-secondary education and indicates that many stayed in this area throughout their lifespan. In relation to marital status, 141 individuals had their marital status available on their social media page, and 60% reported being married, 11% were in a relationship, and 22% were single. The remainder indicated that they were either divorced or in a complicated relationship.

Display full size

Figure 2. 

Flow diagram of the most prevalent post-secondary schools that participants of the SMP cohort attended and the proportion of individuals that are still located in Hamilton, Ontario. SMP: social media presence

Discussion

We traced participants from a previous cohort study, the HCC study, which was conducted between 1978 and 1986. We utilized only publicly available information through social media platforms to determine the effectiveness of using SNS to trace and recruit a reconstructed cohort for follow up research. Through SNS such as Facebook and LinkedIn, we were able to trace 21% of participants to one social media account. Characteristics between participants identified and linked to one social media account and determined as having SMP and those with NSMP differed in terms of sex and neighborhood of residence. In addition, the NSMP cohort had higher rates of childhood illness.

The initial study involved children as participants and aimed to evaluate the long-term effects of childhood exposure to air pollution. As the years have passed and the children from the original study have transitioned into adulthood, locating them for follow-up research has become increasingly challenging due to various factors such as changes in contact information, relocation, and potential name changes. Even for participants that were traced in 2006, the likelihood of maintaining accurate contact information decreased over time. Participants may have changed their names, no longer have contact with parents, passed away, or moved away, further complicating the tracing process. To address these challenges, our study leveraged the power of social media as a tool for locating participants. By utilizing SNS, we were able to conduct searches by names, which facilitated the process of finding participants who may have been otherwise difficult to locate. However, tracing participants with common names through social media sources proved to be a challenge due to the higher frequency of possible matches.

In a previous related study [28] that analyzed predictors of locating participants of the original cohort study, males were more likely to be located when compared to females (59% vs 41%). This is consistent with the current research as more males were found through social media when compared to females. Various other studies also convey the same findings, proposing sex or gender as a factor that influences the success rate of tracking cohorts [29]. This may be due to the fact that most women after marriage are more likely to adopt their partner’s name, making it more difficult for researchers to track them after years have passed with no contact. Moreover, we found that the SMP cohort was more likely linked to the WU region, which indicated that they came from a relatively higher socioeconomic background in childhood. In addition, childhood health history of the SMP cohort was relatively better in comparison to the NSMP cohort. A previous study of the HCC by Barakat-Haddad et al. [27], found that those residing in the EL and WL were more likely to have a family income below the income cut off, higher levels in utilizing gas cooking, parental smoking, and sharing a room with at least two family members. Thus, these demographic characteristics linked to health-related outcomes seem to influence tracing success. Indeed, Barakat-Haddad et al. [28] previously found that participants who were above the low-income cut off in childhood and those who were not exposed to smoking were more likely to be traced after 20 years of no contact.

Other factors that may impact the success rate of tracing participants include the periods between follow up. For instance, findings of a study that focused on the effectiveness of cancer screening trials employed two and five years follow up periods throughout the study and found better results in the shorter periods when contrasted to the longer follow up [30]. When a two year follow up was employed, 2,372 participants out of the original 154,901 were traced. However, after a five year follow up period, this decreased to 909 individuals being located [30]. Another study that implemented a two year follow up survey located 86% of a cohort of 2,631 participants [31]. In this study, between 17 years and 37 years had passed with no contact with participants, which hindered the success rate of tracing cohorts [14, 15]. Our results suggest that participants who included information on their location, age, academic enrollment, and given name were more likely to be successfully traced on Facebook and LinkedIn. Overall, 21% of the cohort was traced using this method. In comparison, a study conducted in 2019 examining the success of relocating 237 participants to complete a postpartum survey after five years had passed—using Facebook search engines, was able to find 47% of the cohort [7].

The use of social media has proven to be a more efficient method of tracing participants, as compared to traditional methods of data collection [32]. Our study found that a notable number of participants were identifiable through their social media profiles, which allowed for effective and efficient re-engagement with the cohort. This aligns with previous literature that also found the use of recruitment through social media to be more efficient when compared to traditional methods of recruitment [33]. In addition, the use of social media platforms such as LinkedIn and Facebook enabled us to reach participants who had moved away from the study area, which would have been difficult to achieve using traditional methods such as phone calls or mailing letters. Facebook’s extensive user base and LinkedIn appeal to professionals and older demographics were also a strength in tracing our cohort. Furthermore, information related to where they attended school may be used to locate and re-engage with participants by contacting that school and organizing alumni events in order to invite participants back as an incentive for future follow-up research. In addition, marital status can be used to reach participants/partners. The cost-effectiveness of using social media is also notable, as it eliminates the need for more expensive traditional methods. This not only saves time and resources but also reduces the potential for errors that may occur during the manual collection of data. Furthermore, the use of social media platforms allowed for easy dissemination of study updates and reminders to participants which may contribute to higher participation rates in subsequent follow-up studies.

Our study has demonstrated the efficacy of SNS in tracing research participants while also revealing demographic limitations that could affect the representativeness of certain groups within the general population. Specifically, our research focused on childhood illness, and we found no significant differences in the prevalence of these conditions between participants identified through social media and those who were not. However, given that this was a pilot study and that the NSMP cohort had relatively worse health outcomes based on the retrieved childhood health history in comparison to the SMP cohort, there may be differences during future research as we trace and re-engage with more participants.

Furthermore, unlike previous studies conducted [18] which focused solely on Facebook, we expanded our approach to include multiple platforms. Notably, our study included LinkedIn, which proved particularly effective for tracing participants due to the rich professional and educational information available on profiles. This addition was useful for our research as it provided deeper insights into the backgrounds of participants, which can be valuable in public health studies.

It is important to note the limitations imposed by using social media for tracing participants. For example, the tracing team encountered problems being labeled as spam on Facebook due to a high volume of messages being sent in a short period of time to potential participants. This resulted in the suspension of the account which delayed the tracing process. The tracing team had to request a re-evaluation of the suspension for Facebook to unsuspend the account. Another barrier when tracing participants on Facebook was that the messages were being sent to their spam box instead of message requests. We addressed this barrier by shortening the informative message of the study to the participants. Facebook also consistently asked for the account to be authorized, putting a halt to advertisements being viewed by users and reaching out to potential participants. To avoid barriers in future projects, researchers should consider requesting verification of a Facebook account before starting tracing and include paid advertisements as this increases the viewership of the Facebook page created for the respective research study.

Likewise, LinkedIn posed limitations as well, including the restriction on messaging non-connected users. In order to overcome this limitation, our research team opted to acquire a LinkedIn business premium account. This provided us with access to the InMail feature, enabling us to send messages to potential participants without being connected to them. However, it is worth noting that the cost of such premium subscriptions can be significant.

Another limitation we encountered was that a number of individuals were not active on social media platforms, which made it difficult to track them. Specifically, the lower success rate on Instagram and X can be attributed to multiple potential factors. These platforms predominantly attract younger users, which conflicts with the study demographic as our participants are middle aged (50–65). For example, a study found that around 71% of Instagram users ranged within the emerging adult population (18 to 29) [34]. Individuals ranging within the older age group are often less represented, having limited to no presence on X and Instagram. This generational gap as well as the platforms’ designs being less focused on sharing personal identifying details—crucial for the accurate tracing of individuals, are potential reasons and possible limitations of using specific social media sites for tracing the sample demographic of this study.

The potential of biases arising from samples for the follow-up study must be considered to ensure accuracy. Sample bias may occur if individuals being traced in the study are more likely to be selected than participants who are lost to follow up due to differences in certain variables that are being examined [35]. To combat possible bias and loss of accuracy in the results of cohort follow up, studies have suggested an ideal percentage of loss to follow up. According to various studies, 5–60% is the agreed upon ideal loss to follow up that will result in an unbiased estimate of influence [36]. In general, researchers trust that an 80% capture rate is sufficient [37].

Moreover, when using SNS to trace participants, many individuals had common names with multiple links, which made it challenging to confirm if the individual found was the original participant. Another limitation was the risk of selecting the wrong individual due to the lack of demographic data available on social media platforms. This limitation made it challenging to differentiate between individuals with the same name, as social media platforms do not always provide enough information to confirm their identity. These limitations highlight the need for caution when using social media as a tracing tool and the importance of utilizing multiple methods of contact to increase the chances of locating all participants.

As the use of social media emerges, opportunities present to enhance the recruitment process for longitudinal cohort studies. Given the growing trend of paid advertisements, it is imperative to assess their efficacy and weigh their advantages against their drawbacks. Moreover, an integrated approach that merges conventional tracing methods (e.g., referrals, phone calls, flyers/posters, and post mail) with modern digital strategies such as SNS, appears promising and warrants further investigation. For future projects involving tracing, it is recommended to allocate a dedicated budget for the purposes of connecting with participants on LinkedIn—particularly given that the sales navigator core subscription—which allows for up to 50 InMails per month, comes at a monthly cost of over $85 CAD. In addition, it is recommended that to increase viewership of recruitment material through Facebook advertisements, a one-time fee ranging between $5–10 CAD should be considered, which can potentially reach around 1,000 individuals. Advertisements can also be adjusted to the specific audience of interest such as age, interests, email address, and location. In upcoming research, our intention is to not only recruit but also re-engage participants who may have been lost to follow-up. As we move forward with subsequent phases of our study, leveraging the power of digital platforms in participant recruitment and retention will be of utmost importance.

Overall, this study has demonstrated the potential of using social media to trace participants from a historical cohort study. Despite the challenges faced in relation to significant time gaps between studies, and the limitations inherent to using social media platforms, research and results have shown that this method of participant tracing can be efficient and cost effective. However, while social media is a valuable tool to track participants, it is key to acknowledge its limitations such as account suspensions, spam bots, difficulties differentiating between individuals with similar names, and sampling bias to name a few. It is also important to recognize that due to a range of structural factors, and from an equity perspective, those most marginalized are the most difficult to find. To reduce these challenges and minimize bias, future research should implement a multi-faceted approach to participant re-engagement, using both traditional and modern methods. This study adds to the growing body of literature that supports the use of social media as a participant tracing tool in longitudinal studies. This may help improve processes of retaining participants for longitudinal research and the validity of findings from the long-term effects of various exposures and interventions on human health and wellbeing.