Factors affecting adolescents’ participation in randomized controlled trials evaluating the effectiveness of healthcare interventions: the case of the STEPSTONES project

Background Recruitment of adolescents to intervention studies is a known challenge. For randomized controlled trials (RCT) to be generalizable, reach must be assessed, which means ascertaining how many of the intended population actually participated in the trial. The aim of this study was to evaluate the reach and representativeness of an RCT evaluating the effectiveness of a complex intervention for adolescents with chronic conditions. Methods A mixed methods sequential explanatory design was employed. Firstly, quantitative cross-sectional data from the RCT, patient registries and medical records were collected and analysed regarding baseline differences between participants and non-participants in the trial. Secondly, qualitative data on their reasons for participating or not were collected and analysed with content analysis to explain the quantitative findings. Results Participants showed larger differences in effect sizes and a significantly more complex chronic condition than non-participants. No other statistically significant differences were reported, and effect sizes were negligible. Reasons for declining or accepting participation were categorized into three main categories: altruistic reasons, personal reasons and external reasons and factors. Conclusions Integration of quantitative and qualitative findings showed that participation in the RCT was affected by disease complexity, the perceived need to give back to healthcare and research and the adolescents’ willingness to engage in their illness. To empower adolescents with chronic conditions and motivate them to participate in research, future intervention studies should consider developing tailored recruitment strategies and communications with sub-groups that are harder to reach.


Conclusions
Integration of quantitative and qualitative ndings showed that participation in the RCT was affected by disease complexity, the perceived need to give back to healthcare and research and the adolescents' willingness to engage in their illness. To empower adolescents with chronic conditions and motivate them to participate in research, future intervention studies should consider developing tailored recruitment strategies and communications with sub-groups that are harder to reach.

Background
Recruitment of adolescents and young persons with chronic conditions (CC) to research studies is a known challenge (1). For intervention studies, recruitment rates in this population have ranged from 10-50% (2)(3)(4)(5)(6). Factors considered to affect adolescents' participation in intervention studies may be clinical (i.e. disease severity) (7), demographic (i.e. household income and geographic location), and individual (i.e. personal incentives and school performance) (8)(9)(10)(11). Moreover, an additional barrier to recruitment is that participation in intervention studies means the adolescent has to engage in their illness (12). For many adolescents this is challenging, as young persons want to live their lives as a healthy normal teens, and not spend excessive time thinking about their illness (13).
In most settings, adolescents living with a CC have to transfer their care from a pediatric to an adult setting at age 18. During this transitional phase, these individuals are subject to several risks, such as care gaps in medical follow-up (14), increased healthcare utilization (15,16), and di culties in achieving educational and vocational milestones (17). To prepare adolescents for the transfer to adult care and transition to adulthood, transition programs are advocated (18)(19)(20).
Randomized controlled trials (RCT) have been conducted to evaluate the effectiveness of transition programs, with positive effects in disease-related knowledge, self-e cacy, self-management (21) and reduced delay in transfer to adult care (22). However, RCTs evaluating these interventions have lacked reporting of components to assess external validity such as representativeness of their samples and response rates (21,23), so generalizability of current transition programs is limited.
In addition to generalizability, the 'reach' of an intervention study is important to consider when translating ndings to clinical practice. Indeed, transition programs are complex interventions because of their numerous interacting components, challenging behaviours required by those delivering and receiving the intervention, and because they target different organizational levels (i.e. pediatric and adult care) (24). When evaluating a complex intervention, effectiveness data is not enough to understand if the intervention was successful. This is because implementation of the intervention might differ between individual patients, and sites over time (25). Determining the reach of an intervention (and the full RCT) is an essential element in evaluating of complex interventions, as it provides knowledge of the absolute number, proportion and representativeness of individuals who were willing to participate in the study (26).
When evaluating the effectiveness of new interventions such as transition programs, RCTs are viewed as the gold standard (21,27,28). However, RCTs are vulnerable for selection bias. In trials for patients with CCs, as many as 70% of studies lacked representativeness of their study samples (29), with participants often healthier than patients seen in clinical practice (29,30). The lack of knowledge on reach and representativeness of participants in transition programs (21,23) poses several challenges. Firstly, if samples do not represent patients in real life settings, clinicians cannot assess which ndings that can be translated into clinical practice (31). Secondly, RCTs are usually the basis for health economic evaluations and decisions on resource allocation in healthcare (32,33). If samples of RCTs are biased, decision-makers risk making faulty assessments as to which interventions that are cost-effective and could be implemented into practice. Thirdly, we lack knowledge about which adolescents and sub-groups that are more di cult to recruit for complex interventions. The aim of this study was therefore to evaluate the reach and representativeness of an RCT evaluating the effectiveness of a complex intervention for adolescents with CC in transition to adulthood. Two speci c objectives were formulated to achieve this aim: (i) to compare clinical and demographic characteristics of participants and non-participants in the RCT, and (ii) to describe adolescents' reasons for participating or not participating in the RCT from their own perspective.

Design
A mixed methods sequential explanatory design was utilized, consisting of two phases (34). Phase one consisted of quantitative cross-sectional data from the STEPSTONES (Swedish Transition Effects Projects Supporting Teenagers with chrONic mEdical conditionS) project, where a transition program for adolescents with congenital heart disease (CHD) was evaluated through an RCT (35). Phase two consisted of qualitative data on participants' and non-participants' reasons for accepting or declining participation. A mixed methods approach to answering the research question was used to facilitate a deeper understanding of what factors affect reach in complex interventions from a quantitative and qualitative perspective. The Good Reporting of A Mixed Methods Study (GRAMMS) checklist (36) was followed for this article (supplement 1). The present study is part of a larger process evaluation study of the STEPSTONES project in which the implementation, mechanisms of impact and contextual factors are explored within the RCT (25). A study protocol describing the full extent of the process evaluation has been previously published (37).

Setting
In Sweden, care for individuals living with CHD is mainly organized around seven university hospitals (CHD-centres). However, follow-up is also provided at regional hospitals connected to the university hospitals. These seven centres provide specialized care for both pediatric and adult patients with CHD, with the transfer to adult care occurring at age 18. Medical follow up is based on current treatment guidelines following the complexity of the CHD and varies between frequent visits on a yearly basis to every 3-5 years (38).

The intervention study
The STEPSTONES project was performed at the seven CHD centres in Sweden, with the hypothesis that adolescents (age 16-18.5 years) participating in the transition program in addition to usual care will have a higher patient empowerment score than adolescents receiving only usual care (primary outcome) (39).
The transition program is a complex intervention comprising multiple components delivered by a transition coordinator who works at the outpatient pediatric cardiology clinic. In short, the transition program entails three outpatient visits with 1-year intervals and one information day for adolescents and their parents (35). The effects of the transition program are evaluated using a hybrid RCT where a longitudinal observational study is embedded within a conventional RCT (see Fig. 1). The two largest centres are designated as RCT centres, where participants are individually randomized to either the intervention group or the comparison group. The remaining ve centres are control groups in which no contamination from the intervention is possible. The total sample size of the trial is 210 patients with 70 patients in each arm, based on power calculation of the primary outcome (35). For the present study, only the RCT centres were included in the analysis. Indeed, the incentive to participate in the RCT differs from that of the observational study, as patients in the latter group only had to ll out questionnaires and did not have to come to the hospital for additional visits if randomized to the intervention group.

Sample
The inclusion criteria for the present study followed those of the RCT: 1) age 16 at recruitment, 2) literate, 3) Swedish-speaking, and 4) having a CHD as de ned by Mitchell et al. (40) with follow-up planned in an Adult Congenital Heart Disease (ACHD) clinic. Exclusion criteria were: 1) previous heart transplantation, and 2) illiterate or non-Swedish speaking. The cohort for the quantitative study (phase 1) was categorized into two groups: participants (n = 134) and non-participants (n = 223). Participants were patients who had consented to participate in the study and had been randomized to either intervention or control group. Non-participants were patients who either actively declined participation, were unreachable, or did not return the informed consent to participate in the study despite several reminders. For the qualitative study (phase 2), we used nested sampling whereby a sub-sample of patients from the group of participants (n = 10) and non-participants (n = 20) were sampled for participation (34).

Data collection
Quantitative data was retrieved from the national register SWEDCON (SWEDish register of CONgential heart disease) (http://www.ucr.uu.se/swedcon.se) (41) and medical records. Patients' medical records were screened for residential address and Google Maps™ was used to calculate the geographical distance between the patients' home and the hospital where the transition program (i.e. intervention) was delivered.
Qualitative data were collected from two sources. All patients declining participation in the STEPSTONES project at the time of recruitment to the RCT were asked their reasons for doing so by the data collection o cer. If the patient provided a reason, this was documented in the RCT enrolment forms. For patients accepting participation in the RCT, data were collected from patients participating in the intervention group after participation in the RCT through semi-structured interviews. In total, 10 interviews were carried out with adolescents aged 18-19 by the main author (MS). In the sampling of participants, maximum variation was utilized (42) in terms of disease complexity, sex and intervention centre. The interviews were carried out face-to-face (n = 1), over the telephone (n = 5) or email (n = 4) as we wanted the adolescents to choose the communication medium that they were most comfortable with (43,44). The interviews comprised three questions: "Can you tell me why you agreed to to participate in the study?", "What do you think are the bene ts of participating in this kind of study?" and "What do you think are reasons for not wanting to participate in this kind of study?", followed by probing questions. The interviews performed face-to-face or over the telephone were audio-recorded and transcribed verbatim.

Data analysis
Quantitative data Statistical analyses were performed in SPSS (Statistical Package for Social Sciences) v.24. As the continuous variables were not normally distributed, they were represented as medians and interquartile ranges. Categorical variables were expressed as absolute numbers and proportions. Variables concerning clinical, demographic and health service use were compared between participants and non-participants using signi cance testing (signi cance level < 0.05). The chi-square test and Fishers' exact test was used for categorical variables and the Mann-Whitney U test for continuous variables. In order to calculate the magnitude of difference between participants and non-participants, we calculated effect sizes. For categorical variables, we used Cohens' w. For continuous variables, we used the Wilcoxon's test by calculating r = Z/ √n. Z represents the Wilcoxon test statistic, which is the Mann-Whitney U Z statistic. Cut-offs for Cohens w and r were graded as follows: 0.1-0.3 = small difference, 0.3-0.5 = moderate difference, and > 0.5 = large difference (45).

Qualitative data
Inductive content analysis of the manifest content of reasons for participation and non-participation was performed according to Graneheim and Lundman (46). This analysis method is suitable for analysing texts and was therefore appropriate due to our using data from various sources. The analysis was performed with the computer-assisted qualitative analysis software, NVivo version 12. We chose to analyse and present the data from the participants and non-participants as a whole. This was because the texts from the RCT recruitment protocols (non-participants' reasons) where short summaries, and the interviews (participants) provided more depth and explanations that were not present in the summaries from the recruitment protocols. The text from the recruitment protocols and transcribed interviews were read through several times to get a grasp of the whole. Meaning units were then identi ed as constellations of words and sentences that answered the aim of the study and contained a common meaning. These meaning units were then condensed, which in some cases meant a reducing the number of words but keeping the core essence of the meaning unit. In this process, codes were created and compared to each other, with those that were similar in content sorted into the same group. These groups formed the basis for subcategories. This procedure was repeated for the subcategories, resulting in the main categories (46). Two of the authors (MS & ELB) with experience in qualitative methods and content analysis were responsible for carrying out the analysis and continuously discussed the ndings with the other co-authors.

Ethical considerations
The study was approved by the Regional Ethical Review Board in Gothenburg, Sweden (No.931 − 15) and the board of directors of SWEDCON. The study was performed in accordance with the Declaration of Helsinki (47). For the quantitative study, data was retrieved from a national register, so informed consent could not be retrieved from each individual participant for this particular study. However, each patient enrolled in the register is informed that their data could be used for research purposes. In addition, the patient is informed about their right to refuse permission to use their data. For the qualitative study, informed consent was retrieved from participants after they had been given oral and written information about the study. Non-participants who provided a reason for declining were informed that their answers would be used for research purposes and consented to this. All data were stored coded and passwordprotected. During the interview period, the interviews performed over email were copied to a passwordprotected database and stored in a designated folder in the main authors email program. When the analysis was completed, the emails were deleted from the email program to protect the participants' con dentiality.

Phase 1 -Quantitative study
From the total eligible population identi ed in SWEDCON (n = 357), 37.5% (n = 134) participated in the STEPSTONES RCT. Statistically signi cant differences were observed between participants and nonparticipants for primary diagnosis and disease complexity (Table 1). A larger proportion of participants had a more complex CHD than non-participants. The statistically signi cant results also presented large effect sizes for primary diagnosis w = 1.46 and disease complexity w = 0.54. There were not statistical differences between participants and non-participants for sex, geographic distance to hospital, proportion receiving cardiac pharmacotherapy, proportion receiving special aid in school, and the number of cardiac interventions and surgeries, and the effect sizes were negligible.

Helping other adolescents in the same situation
This subcategory included the wish to help other adolescents in the same situation, even if their participation did not help themselves. Believed reasons were that they could be supportive by showing other adolescents in the same situation that they were not alone and that other adolescents were going through the same process of transitioning into adulthood.
Yes, for example, I think mentally they know they are not alone; there are other people who have similar problems and who behave similarly in life. And maybe ... many have different ways of acting ... but I think it helps people to show that they are not alone… 19-year-old male.
Contributing to research and improved care Here, the participants described a wish to contribute to research by sharing their experiences, but also through their participation in the study.
Why I did was because I felt kind of like I might as well help, I think it's good that you carry out studies and research like this and then I thought in that case, I might as well because I'm experiencing this -I can share my experiences. And how I felt about this whole process and all that.
Participating was the right thing to do For some participants, this subcategory was shaped on the assumption that participation in a research study was a good deed and that deciding to contribute came naturally.

Personal reasons
This category highlighted personal reasons for participating or not participating in the study. Four subcategories were identi ed that were considered factors affecting the decision to participate.

Meeting others in the same situation
Being given the opportunity to meet other adolescents in the same situation could lead to participants and others in the same situation feeling less alone. Another aspect was the need to show others that there are young people with CHD who are doing well and living normal lives.
Yes exactly! Seeing that you are not alone is something that very often satis es the brain.
…that there are people who are also doing well with this heart disease… 19-year-old male.
However, meeting other young adolescents with CHD could be a reason for not wanting to participate, as some adolescents felt it was awkward to talk about their condition in front of others.

Educational objectives
Wanting to learn more about CHD and how the condition would affect the adolescents in their daily lives were reasons that facilitated participation. The adolescents considered participating in the study an advantage to participate in the study because they saw it as an opportunity to learn more about their illness, the transfer to adult care and the transition to adulthood with a CC.
It's also clear that there were bene ts from this study for me as well; it gave back to me a bit because the study meant a lot that I would learn about my illness -then I understood 19-year-old male.
Well... maybe I wanted more understanding of what it was because I didn't even know ... I'd just heard that I had something with my heart and I kind of went to the doctor sometimes. I didn't know what it was at all. But then maybe I want more information about it.
Feeling healthy and therefore not needing the study This subcategory described why adolescents did not participate in the study due to not feeling ill or affected by their CHD. For this reason, they also believed they would be a bad t for the study. Moreover, there as a perception among the adolescents that young people participating in studies like this were affected more severely by CHD than they were.
And I also think that there are probably a lot of people who aren't really affected by their CHD, just like me, and then maybe they feel that there's even less reason for them to participate. Like that you don't t in there in any way, that you have nothing to do with the study because you are not characterized by your heart failure.

19-year-old male.
Fearing what the study would evoke and demand of them Practical reasons were also given for not participating, such as not having enough time to contribute to the study, that the hospital was too far away from the patient's home, and that the additional hospital visits that randomization to the intervention group would entail would be emotionally and practically stressful for the adolescents.

External reasons and factors
The nal category included reasons and factors that were external to the adolescents' own perceptions and decisions to participate in the research study. Here, parental in uence on the decision to participate was illustrated in two subcategories.

Parents as facilitators
Parents could be active partners in the decision to participate in the study.
It was kind of a collective decision by both my parents; if you are 16, you are usually a little indecisive, very hard to think that you shouldn't do it. But we talked it over brie y and thought like, why not? So nothing wrong with that.
It was a bit of a joint decision, but it was a decision that I would also like to make myself, you could say.

Parents as barriers
In this subcategory, there were reasons as to why parents could potentially be barriers to participation.
Domestic issues, such as con icts within the family, were also brought up as a barrier to participation.
Maybe that parents are more involved and tell the kids that they should not join, that it is unnecessary or so 19-year-old male.

Discussion
Having integrated ndings from the quantitative and qualitative phases of this study, we can conclude that participants differed from non-participants in regard to disease complexity. A potential reason for this may be that adolescents with a more complex conditions felt grateful towards research and the healthcare system, and were thus motivated to give back. In contrast, non-participants (who as a group generally had a less complex condition) mentioned not wanting to engage in their illness more than necessary and not recognizing the illness as a part of their social identity as reasons for declining participation. These ndings are in line with results from previous studies in this patient population (48). It is also in keeping with observations in other patient populations, for instance, in RCTs concerning young persons with asthma, where non-participants have been shown to be healthier and less likely to require medication (2). Differences in quality of life and health status could also be a reason why patients with more severe conditions want to participate in a transition intervention. Several studies have shown that patients with moderate and severe complexity of CHD have a lower health related quality of life than patients with milder CHD (49)(50)(51). Moreover, adolescents with complex CHD experience lower selfreported health and physical functioning (52,53). Both health status and quality of life might be incentives that affect participation in an intervention aiming to improve the adolescents' ability to manage their health and care in adulthood.
An important aspect to consider regarding representativeness of the sample of this study is that patients with complex CHD generally have a higher frequency of outpatient visits (38). As patients in the STEPSTONES RCT had a more complex CHD than those who declined participation, this will have an impact on upcoming economic evaluations, as patients seen in the RCT are the primary data source for cost-effectiveness ratios. It is therefore important for upcoming studies evaluating similar interventions to consider performing budget impact analyses adjacent to cost effectiveness studies in order to estimate the true healthcare budget impact of complex healthcare interventions (54). In the present study, we did not include a variable on health service use as SWEDCON was established in 2009 and data for this variable has not been added retrospectively for all centres. As young persons with CHD consume more health care resources during the early years of life (55), including this variable would have introduced bias into our results.
Although our ndings suggest that disease complexity may be one factor affecting participation in a complex healthcare intervention such as a transition program, our results should be interpreted bearing in mind some methodological caveats. Firstly, we did not have access to individual level data on socioeconomic status for the complete sample. This has been an important factor in previous studies comparing participants to non-participants (7-10). Indeed, domestic issues were also raised in our qualitative ndings as a potential reason for declining participation. Secondly, we did not have any measure of self-reported health. The health status measure registered in SWEDCON, PedsQL 4.0 (56), has been recently implemented into standard practice in pediatric cardiology and coverage in the register is therefore limited. Thirdly, we only received reasons for declining participation from 8% of all nonparticipants, which makes it di cult to generalize the ndings. Furthermore, as ethical constrictions prevented us from interviewing these non-participants, we had to rely on the short statements they had provided on the RCT enrolment forms.
In the light of these limitations, this study has several strengths. To our knowledge, this is the rst published study assessing the reach of an RCT evaluating a transition program for adolescents with CC.
As evidence on transition programs is scarce (57), this is an important step in providing evidence on effectiveness. The recruitment for the RCT was based on a register which has high concordance with the medical records, therefore increasing validity (58). Moreover, by reporting data on the reach and representativeness of this RCT before the outcomes of the effectiveness evaluation, we can avoid making biased interpretation of outcomes (25). Finally, the use of mixed methods, where qualitative methods were used to explain quantitative ndings, gave a deeper understanding of factors that impact on the decision to participate or not in intervention studies for young persons living with CC. One insight especially gained from mixing methods was that adolescents with a milder CC need tailored messages while being recruited to intervention studies. From our ndings, these patients are more likely to decline participation due to, for example, not wanting to engage in their illness and fear of standing out for the wrong reasons. In future trials with this patient population, researchers and clinicians should carefully consider to this issue. Indeed, it would be useful to involve young persons in the development phase of the intervention, and then not only focus on developing the clinical intervention but also focus on a recruitment strategy with tailored messages that can empower adolescents with milder conditions to participate in research trials (59).

Conclusion
Factors positively affecting adolescents' participation in intervention studies are higher disease complexity and a will to give back to research and healthcare, along with an interest to learn more about their health and care. Factors that negatively affect participation negatively are milder disease complexity, along with the feeling of being healthy and not wanting to engage in the illness. Future intervention studies within this area should investigate the impact of socio-demographic variables on study participation, along with self-reported health. The creation of tailored recruitment messages with relevant stakeholders to reach hard-to-recruit sub-groups is an area which needs increased attention and a topic for future research initiatives.

Declarations
Ethics approval and consent to participate The study has received ethical approval from the Regional Ethical Review Board in Gothenburg, Sweden.
Informed consent for participation was retrieved from the study participants. As participants in this study were 16 years and older at the time of participation, written informed consent was not retrieved from parents or guardian.

Consent for publication
Not applicable.

Availability of data and materials
All data and materials are available from the corresponding author upon request.

Competing interests
The authors declare they have no competing interests. Authors' contributions MS, LW, PM, HG and ELB conceived the study and were responsible for the design and draft of the manuscript. MS was responsible for data collection, analysis, interpretation and rst draft of the manuscript. MS and ELB conducted the analyses. All authors critically revised and approved the manuscript.