Effective and resource‐efficient strategies for recruiting families in physical activity, sedentary behavior, nutrition, and obesity prevention research: A systematic review with expert opinion

We systematically identified effective and resource‐efficient strategies for recruiting families into health promoting intervention research. Four databases were searched for reviews. Interventions were extracted from included reviews. Additionally, a Delphi study was conducted with 35 experts in family‐based research. We assessed extracted data from our review and Delphi participants' opinions by collating responses into overarching themes based on recruitment setting then recruitment strategies to identify effective and resource‐efficient strategies for recruiting families into intervention research. A total of 64 articles (n = 49 studies) were included. Data regarding recruitment duration (33%), target sample size (32%), reach (18%), expressions of interest (33%), and enrollment rate (22%) were scarcely reported. Recruitment settings (84%) and strategies (73%) used were available for most studies. However, the details were vague, particularly regarding who was responsible for recruitment or how recruitment strategies were implemented. The Delphi showed recruitment settings, and strategies fell under six themes: school‐based, print/electronic media, community settings‐based, primary care‐based, employer‐based, and referral‐based strategies. Underrecruitment in family‐based trials is a major issue. Reporting on recruitment can be improved by better adherence to existing guidelines. Our findings suggest a multifaceted recruitment approach targeting adults and children with multiple exposures to study information.


Summary
We systematically identified effective and resource-efficient strategies for recruiting families into health promoting intervention research. Four databases were searched for reviews. Interventions were extracted from included reviews. Additionally, a Delphi study was conducted with 35 experts in family-based research. We assessed extracted data from our review and Delphi participants' opinions by collating responses into overarching themes based on recruitment setting then recruitment strategies to identify effective and resource-efficient strategies for recruiting families into intervention research. A total of 64 articles (n = 49 studies) were included. Data regarding recruitment duration (33%), target sample size (32%), reach (18%), expressions of interest (33%), and enrollment rate (22%) were scarcely reported. Recruitment settings (84%) and strategies (73%) used were available for most studies.
However, the details were vague, particularly regarding who was responsible for recruitment or how recruitment strategies were implemented. The Delphi showed recruitment settings, and strategies fell under six themes: school-based, print/ electronic media, community settings-based, primary care-based, employer-based, and referral-based strategies. Underrecruitment in family-based trials is a major issue.
Reporting on recruitment can be improved by better adherence to existing guidelines. Our findings suggest a multifaceted recruitment approach targeting adults and children with multiple exposures to study information.

K E Y W O R D S
Delphi study, healthy eating, physical activity, screen time

| INTRODUCTION
Childhood overweight and obesity remains to be an omnipresent global public health issue as the prevalence has risen steadily worldwide over the past few decades. [1][2][3] Contributing to the increasing waistlines of young people is the proliferation of poor lifestyle behaviors, with few children meeting physical activity, sedentary behavior, and fruit and vegetable consumption recommendations internationally. [4][5][6][7] Parents can influence their children's health behaviors through a variety of mechanisms, including their general parenting style, parenting practices (e.g., rule setting), and their control over the home environment. 8,9 Therefore, health promotion interventions targeting families may therefore be a valuable way to improve lifestyle behaviors physical activity among children. 10,11 A vital first step towards this goal is the development of strategies to overcome barriers to recruitment.
The recruitment of participants into intervention research has been notoriously difficult for research teams around the world. 12,13 Two reviews of publicly funded trials in the United Kingdom (through the National Institute for Health Research) found that only about half of the included trials recruited 100% of their target sample size within their pre-agreed timescale. 14,15 The overall start to recruitment was delayed in 41% of trials, early recruitment problems occurred in 63% of trials, 15 and just over one third received an extension of some kind. 14,15 There is little evidence that recruitment into intervention research is improving over time. 12,15 Recruitment of families to research projects is particularly challenging. 9,16-18 Elsewhere, we have described specific recruitment challenges we have encountered in previous work, [19][20][21] but there has not been a comprehensive assessment of how to recruit families to family-based health promotion research.
The aim of this study was, therefore, to systematically identify effective and resource-efficient strategies for recruiting families into intervention research aimed at improving physical activity or nutrition or reducing levels of sedentary behavior (including screen time) and overweight/obesity. Our objectives were to (1) describe procedures used and outcomes related to recruitment (e.g., recruitment duration, strategies used, recruitment settings, reach, expressions of interest, and enrollment rates) and (2) determine the most optimal family-based recruitment strategies.  Reporting of the systematic review was guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses. 22 In short, we identified relevant intervention studies through a systematic search of published reviews on the relevant topic. Intervention studies were then extracted from those included reviews. Subsequently, a forward search of the included intervention studies identified more recently published studies not captured in the included reviews.

Systematic reviews
All types of reviews describing the results of family-based experimental studies with outcomes related to physical activity, sedentary behavior, nutrition, or obesity prevention were eligible for inclusion.

Intervention studies
Intervention studies were eligible for inclusion if they met the following inclusion criteria: • Participants. Generally healthy school aged children and youth and at least one adult primarily responsible for their care. Studies focused on preschool or postsecondary aged youth samples were excluded, as were those with clinical populations (e.g., populations affected by any illness, disorder, or disability) or exclusively targeting children and youth affected by overweight/obesity.
• Interventions. Interventions that deliberately attempted to implement a change in physical activity, sedentary behavior, screen time use, nutrition, or prevent overweight/obesity were included. No restriction was placed on the type of comparison. Treatment interventions (e.g., weight management interventions) were excluded.
• Study type. All experimental (e.g., randomized controlled trials [RCT] and cross-over designs) and quasi-experimental designs were included. Cross-sectional and cohort studies were excluded.
No limitations were set regarding the duration of the intervention or the follow-up period.
• Types of outcome measures. Included studies could have employed any outcome measure related to physical activity, sedentary behavior, screen time use, nutrition, or overweight/obesity prevention. However, outcomes must have been measured on at least one child and at least one adult primarily responsible for their care.
For both reviews and intervention studies, we set no limits on the earliest publication date. We included English language, peerreviewed full-text articles that reported primary data or protocols and had been published by February 2019. Forward searching was conducted in August 2019.

| Search strategy
We conducted a systematic search for review articles in Cochrane Library, PubMed, PsycINFO, and Scopus. The search included keywords to the population ("children/young people" and "parents"), interventions ("physical activity," "nutrition," etc.), and study type (e.g., "review"); Table S1 shows an example of the full search strategy used in our Scopus database search. Identified references were imported into EndNote reference manager and duplicates were removed.
Titles and abstracts were screened by a single reviewer, with a second reviewer double screening a random 10%. Two reviewers independently screened full-text papers, with any discrepancies resolved by discussion. Reasons for exclusion were identified at this full-text screening stage. References of included reviews were reviewed in duplicate and references of potentially relevant intervention studies or reviews extracted into EndNote. Following de-duplication, two reviewers independently screened titles/abstracts and then full-text versions of interventions studies identified; reasons for exclusion were identified at this stage. Any disagreements were discussed by the two reviewers until consensus was reached.

| Data extraction
The following data were extracted from each intervention study: characteristics of study design and sampling, recruitment duration and strategies used, recruitment settings, and information about reach, expressions of interest, and enrollment (see Table 1). We sent the extracted data to first and last authors of studies published within the last 5 years (i.e., since 2014), inviting them to check the extracted data for accuracy and to add any missing information, if possible. We only contacted authors of articles published within the last 5 years as we believed this was a reasonable timeframe for records to be available and researchers to have adequate recall of the study.

| Risk of bias in individual studies and across studies
We were only interested in examining the strategies used for recruiting families into family-based intervention research, which does not inherently affect the internal validity (risk of bias) of a study. Therefore, we have decided not to include a risk of bias (quality) assessment.

| Summary measures and synthesis of results
As indicated above, this study focused on family-based recruitment strategies, rather than study findings, which were therefore only presented descriptively.

| Study design
The Delphi procedure or technique is a group process involving the interaction between the researcher and a group of identified experts on a specified topic. 23 This procedure is appropriate for research questions that cannot be answered with complete certainty but rather by the subjective opinion of a collective group of informed experts. 24 Here, we used a Delphi procedure to determine, through the consensus of experts, the most effective and resource-efficient strategies for recruiting families into intervention studies. Ethical approval for the study was obtained in July 2019 through MRC Epidemiology Unit departmental ethical review.

| Study procedures
Two groups of experts were selected to participate in this study and received an email invitation: (a) all first and senior authors of the intervention studies identified in Phase 1 and (b) known experts in the field as identified by the study team. Delphi participants were also permitted to suggest other experts for invitation. All participants were asked to complete an informed consent online prior to the start of the study.
The Delphi study included three rounds using an online questionnaire created in Qualtrics. To start each round, participants were sent an email containing a direct link to the online questionnaire and given 1-2 weeks to complete. One reminder was sent 3 days before the deadline. After each round, a summary of the findings was fed back to the participants.
Our protocol for each round of the Delphi study was based on a similar published study. 25 In Round 1, participants responded to questions related to the most recent family-based study they had conducted (e.g., recruitment strategies, recruitment duration, and sample size), and to provide their top two strategies for recruiting families in intervention studies (see Table 2 for questions). Following the deadline, the study team then reviewed the panel's responses to their top strategies. We then collated responses into overarching themes based on the setting recruitment occurred in (e.g., schools) and then organized similar recruitment strategies used under each overarching themes.
In Round 2, participants reviewed the recruitment strategies put forward in Round 1 and rated how effective and resource efficient they believed each strategy to be separately on two different 4-point Likert scales (4 = very effective/resource-efficient, 1 = not effective/ resource-efficient). To rank strategies, summary scores were created in which scores for effectiveness were weighted by a factor of 2. Therefore, the weighted scores for effectiveness ranged between 2 and 8 and the scores for resource efficiency between 1 and 4. Effectiveness was weighted more than resource efficiency as we believed effectiveness was a more important factor related to recruitment strategies. The top 10 strategies were then taken forward to Round 3. In the final round, participants were asked to rank the final 10 recruitment strategies into their individual top 10. Following completion, all rankings were summed to determine an overall rank of the strategy (i.e., lower scores indicated higher ranks).    Table 1 (the reference list for included studies can be found in Table S2). Of the 49 separate studies, the majority were undertaken in the United States (57%), were pilot or feasibility studies (43%), aimed to improve physical activity only (37%), and recruited parentchild dyads (53%) rather than participation of more family members.  newsletters, letters from the head teacher (i.e., principal), research teams presenting study information to students and parents at assemblies, research teams presenting study information at other school events (e.g., parent teacher association meetings), and research teams soliciting parents during pickup/drop-off. Local newspapers and referral-based recruitment (e.g., word-of-mouth) were also popular recruitment strategies. Less commonly reported recruitment strategies included using electronic/digital media (e.g., television, radio, social media, Google AdWords, and Craigslist), face-to-face recruitment (e.g., home visits and community demonstrations), mail, phone calling, and distribution lists (e.g., via marketing companies).

| Phase 2-Delphi study
We invited 107 experts to participate in the Delphi study representing all inhabited continents. Twenty-three experts actively declined as they either were no longer conducting family-based research (n = 3), did not have the time (n = 2), or no reason (n = 18

| Round 1-Overview of experience with recruitment settings and strategies
Twenty-one participants provided information in Round 1; Table 4 summarizes  Table S3 for an overview of the 36 recruitment strategies. Disseminating study information through social media was the strategy that the most participants planned to implement in future.

School-based recruitment
They reported that print and electronic media were wide-reaching and generally inexpensive to use. However, those with experience with this recruitment strategy reported low and slow response rates.
Creating regular content on social media platforms or newsletters (e.g., update posts, quarterly newsletter, and blogs) was considered more beneficial than one-off posts, advertisements, or newsletters.
Caution was raised that some media-driven strategies can be less targeted than others (e.g., posts in social media groups and television advertisements/stories), which can lead to a lot of interest from ineligible participants (and increased staff requirements). Staff time was considered the greatest resource requirement (e.g., searching for online groups/communities, creating content, and increased eligibility checking).

Community settings-based recruitment
The strategies applied in community settings-based recruitment were hard copy leaflets or pull-tab posters, speaking to parents

Employer-based recruitment
Employer-based recruitment strategies included hard copy leaflets displayed in employee common areas (e.g., staff kitchen) or emails to employees from within an organization on behalf of the research team (e.g., an email sent from human resources to employees within an organization and technically challenging.

Referral-based recruitment
Referral-based recruitment (i.e., word-of-mouth) was usually not a method that was explicitly used by participants, but they reported that a modest amount of enrolled families in their studies were recruited through word-of-mouth (10-30% of their total sample). One Delphi participant reported that referral-based recruitment was particularly effective in studies with multiple waves of recruitment.

| Round 3-Ranking recruitment strategies
The 10 highest scoring strategies were ranked by 17 participants in Round 3. Table 5 shows participants' ranking of the top 10 most effective and resource-efficient strategies for recruiting families into intervention studies. Findings between the top-rated strategies in Round 2 and the final ranking of the top 10 strategies in Round 3 were largely similar apart from "speaking to parents" and "attending parent meetings," which were rated more highly in the final ranking, and "word-of-mouth" and "social media posting," which were rated lower.

| DISCUSSION
The aim of this study was to systematically identify effective and resource-efficient strategies for recruiting families into physical activ- Participants' rankings were summed to determine an overall rank of the strategy (i.e., lower scores indicated higher ranks). b Count of number 1 rankings strategy received.
that most parents that expressed interest were mothers may not come as a surprise as historically mothers, compared with fathers, are more likely to be their family's social agent and lead on tasks such as family event preparation. 29,30 It may be prudent to consider this in the recruitment of families in two ways. First, recruitment materials that target mothers and their family may be the most efficient method of attracting expressions of interest; and it may also be an important catalyst for the inclusion of more fathers in family-based research.
Second, separate recruitment materials that explicitly target fathers may also be useful and should be considered; Morgan and colleagues have written extensively on recruiting and engaging with fathers in family-based research. 9,31,32 Related to target sample size, we found that only 38% and 56% of the studies included in our Delphi and review recruited their target sample size, respectively. Similarly, other reviews of publicly funded trials have found that only 33-50% of included trials recruited 80-100% of their target sample size within their pre-agreed timescale. 14,15,33,34 Few studies reported on reach and representativeness, but generally, healthy and affluent families were recruited. Only one study, as far as we are aware, described formative work that consulted with families to inform the development of their recruitment strategy. Although the target sample size was not achieved in that study, public involvement should be encouraged and has been highlighted as a good method for helping with participant recruitment, engagement, and retention. 35 It is also possible that the chance of being randomly allocated into a study arm that was not a families' preferred study arm may have negatively affected recruitment 36 ; however, no study reported that the randomization procedure hindered their recruitment.
Analogous to our Delphi findings, one third of trials received an extension of some kind due to recruitment related issues. 14,15 Our findings showed the planned median recruitment duration to be about 10-11 weeks, and when recruitment extensions were implemented, recruitment was extended for an additional 20 weeks, which would have a substantial impact on a study's timeline. A recent survey on research priorities related to the methodology of trials among direc-

tors of the Clinical Trials Units registered with the NIHR Clinical
Research Network in the United Kingdom identified the recruitment of participants in trials among the top three priorities needing improvement 37 ; overall, our findings reinforce these concerns.
The majority of included studies reported which settings they recruited participants from and our findings indicate that researchers recruited in about two settings per study, on average. In both our review and Delphi, we found the two most common recruitment settings were in schools and in the community. Other recruitment settings included primary-care settings, employers, and social media. Although it was positive to find recruitment occurred in multiple settings, as recommended by others, 17,18,38 it usually was not possible to discern what proportion of a study's sample was recruited by setting.
On average researchers used about three recruitment strategies per study. As with other studies, 38,39 the most commonly used recruitment strategies for family-based recruitment included disseminating study information through leaflets, posters, or newsletters.
Placing advertisements in local newspapers, using electronic media (e.g., social media platforms, radio, and television) and referral-based recruitment (e.g., word-of-mouth) were also popular recruitment strategies. Considering school-based recruitment was the most used recruitment setting, it was unsurprising that this setting had the highest number recruitment strategies. Generally, recruitment strategies were only listed in study manuscripts and not described in any great detail, particularly around how these strategies were actually implemented and by whom. However, among the few studies that recruited a sufficient sample size, many included strategies that targeted adults and children and oftentimes while they were together. [40][41][42][43][44] For example, interacting with parents and children at school drop-off/pickup, study presentations at school events (e.g., parent nights), and announcements from the pulpit are some of the strategies used in the studies that recruited their target sample size. Further, based on the top 10 of recommended recruitment strategies identified here, it appears that leveraging familiar, and perhaps trusted, relationships would be beneficial. For example, disseminating study information via correspondence from head teachers, general practitioners, human resource personnel, and by word-of-mouth (e.g., through family friends) are all strategies that generally indicate a potential participant would have at least some familiarity with the person disseminating the study information. Even those Delphi participants who were most enthusiastic about recruiting through social media platforms appear to be trying to build rapport with their followers (i.e., potential participants). Delphi participants recommended that creating regular content on social media platforms or newsletters (e.g., update posts, quarterly newsletter, and blogs) would be more beneficial than one-off posts, advertisements, or newsletters. The top 10 strategies also include several face-to-face recruitment strategies (e.g., delivering assemblies and attending parent-teacher meetings).
Although face-to-face recruitment can be extremely time (and resource) consuming, studies that included this type of recruitment were more likely to achieve close to their target sample size. 34 In addition, face-to-face recruitment allows for quick rapport building and gives potential participants the ability to receive responses to their questions about the study in real time, while they are still interested. 45

| Strengths and limitations
As far as we are aware, this study is the first comprehensive assessment of recruitment in family-based intervention research. We conducted a systematic review of the available evidence, and the inclusion of the Delphi procedure provides additional credibility and insight to the findings of the review. Also, our Delphi procedure included several strengths including participant blinding, iterative data collection, controlled feedback, and purposive sampling. Despite these strengths, there are some limitations that should be considered. The data available related to recruitment duration, reach, expressions of interest, expressions of interest rate, and enrollment rate were scant; and details were often vague regarding the implementation of recruitment strategies. Despite efforts to contact authors for this information (if available), few responded to our emails. Additionally, we also had a low and variable response rate for our Delphi study.
Generally, Delphi participants reported recruitment strategies that they perceived to be effective and resource efficient; therefore, selfreport could be considered to be a limitation. Also, the majority of included studies and Delphi participants were from North America and Europe. Optimum recruitment strategies and setting may differ by context, and we recognize the lack of global perspective on how best to recruit families and that some recommended recruitment strategies may not be feasible or appropriate everywhere. Lastly, our review was limited to articles published in English and our Delphi was also limited to researchers who were competent in English.
In conclusion, this study highlights that (a) underrecruitment is a major issue in family-based trials and (b) there is a clear need to improve reporting related to recruitment, for example, by following the checklist items in the Consolidated Standards of Reporting Trials. 26,27 Improved reporting around effectiveness of recruitment will give future researchers the ability to better budget their time and resources and provide greater confidence in meeting their target sample size. Our findings suggest that researchers should employ a multifaceted recruitment approach that targets adults and children and provides potential participants with repeated exposure to study information. This study also provides experts' recommendations for recruitment strategies; future research should investigate the effectiveness of these in different settings. In the future, analyses should be conducted to estimate the cost of recruiting families into trials.
Future research should also explore more sophisticated and innovative research strategies which may include, for example, the consultation of experts in data science, marketing, advertising, graphic design, or social media.