Recruiting participants to walking intervention studies: a systematic review

Purpose Most researchers who are conducting physical activity trials face difficulties in recruiting participants who are representative of the population or from specific population groups. Participants who are often the hardest to recruit are often those who stand to benefit most (the least active, from ethnic and other minority groups, from neighbourhoods with high levels of deprivation, or have poor health). The aim of our study was to conduct a systematic review of published literature of walking interventions, in order to identify the impact, characteristics, and differential effects of recruitment strategies among particular population groups. Methods We conducted standard searches for studies from four sources, (i) electronic literature databases and websites, (ii) grey literature from internet sources, (iii) contact with experts to identify additional "grey" and other literature, and (iv) snowballing from reference lists of retrieved articles. Included studies were randomised controlled trials, controlled before-and-after experimental or observational qualitative studies, examining the effects of an intervention to encourage people to walk independently or in a group setting, and detailing methods of recruitment. Results Forty seven studies met the inclusion criteria. The overall quality of the descriptions of recruitment in the studies was poor with little detail reported on who undertook recruitment, or how long was spent planning/preparing and implementing the recruitment phase. Recruitment was conducted at locations that either matched where the intervention was delivered, or where the potential participants were asked to attend for the screening and signing up process. We identified a lack of conceptual clarity about the recruitment process and no standard metric to evaluate the effectiveness of recruitment. Conclusion Recruitment concepts, methods, and reporting in walking intervention trials are poorly developed, adding to other limitations in the literature, such as limited generalisability. The lack of understanding of optimal and equitable recruitment strategies evident from this review limits the impact of interventions to promote walking to particular social groups. To improve the delivery of walking interventions to groups which can benefit most, specific attention to developing and evaluating targeted recruitment approaches is recommended.


Introduction
It is over a decade since Professors Jerry Morris and Adrienne Hardman described walking as the 'nearest activity to perfect exercise' (Hardman & Morris, p328, 1997) [1]. The epidemiological research underpinning their statement has rapidly increased, so that the promotion of walking is now a central pillar in many international physical activity strategies and national plans, e. g. 2010 Toronto Charter for Physical Activity [2]. Regular walking, independent of other physical activity, can reduce the risk of overall mortality, of cardiovascular disease (CVD) and improve risk factors for CVD, including diastolic blood pressure and lipid profiles [3][4][5]. Regular walking is associated with a reduction in body mass index and body weight, with reduced risk of type 2 diabetes [6] and is suggested to improve self esteem, relieve symptoms of depression and anxiety, and improve mood [7,8]. From a public health perspective, enabling an increase in overall population levels of physical activity through walking will produce an effective reduction in risk of all cause mortality [9].
A systematic review of the effectiveness of walking interventions found evidence for a range of approaches [10]. These included brief advice to individuals, remote support to individuals, group-based approaches, active travel and community level approaches. Recent reviews have provided evidence to support environmental and school based travel interventions [10][11][12]. Despite the evidence for the benefits of walking for health, population rates of walking and overall physical activity remain low and below recommended levels [13][14][15]. Population surveys report that walking behaviour is socially patterned by gender, age, socio-economic status (SES) and by the purpose of walking i.e. for leisure or transport. For example, in the UK long brisk paced walks are more common among affluent groups, whereas walking for transport is more common among less affluent groups [14,16].
One criticism of the evidence base for walking interventions is a failure to recruit specific groups of the population and further studies are needed to broaden the reach of walking interventions [10][11][12]. Intervention reach, or recruiting specific population sub-groups, is only partially reflected in public health and clinical research. For example the RE-AIM framework is designed to guide the implementation of behaviour change interventions [17]. It recommends assessing both an intervention's effectiveness and ability to reach a targeted group. Similarly, recent CONSORT (2010) guidelines [18] recommend clearly displaying the flow of participants throughout a study. Despite identifying recruitment as part of their framework, the guidelines do not define the actions needed to identify and recruit potential populations of participants. There is an absence of conceptual frameworks for recruitment to intervention studies and also a lack of procedural models and systems for recruitment. There is a need to identify what factors are effective in engaging participation at the recruitment phase [19][20][21].
Research examining recruitment practice has focused on drug or medical interventions rather than public health interventions [22]. Little is known about recruitment to physical activity interventions. A Cochrane review identified three stages of recruitment (invitation, screening, intervention starting) for potential participants into physical activity randomised control trials (RCTs). The authors noted a considerable loss of participants across each stage limiting the effectiveness of interventions [23]. The CONSORT (2010) guidelines, suggest that studies report the number of eligible participants prior to randomisation but do not insist on the need to report the original overall number of responders invited to participate (prior to eligibility) [18].
Clearly the effectiveness of a walking programme is limited by not only its efficacy of dose (how well the intervention works on its participants) but also by its recruitment (maximising the numbers who will participate and receive the intervention dose). In response to frequent research calls to evaluate effective approaches to the recruitment of individuals to walking studies, the Scottish Physical Activity Research Collaboration http:// www.sparcoll.org.uk undertook a series of studies to examine recruitment strategies for research and community based programmes of walking promotion. We defined recruitment for such walking studies or programmes as the process of inviting participation to a formal activity including the invitation, informing and facilitation of interested parties to take part in an organised study, activity or event. This paper reports the results of a systematic review to examine the reported recruitment procedures of walking studies, in order to identify the characteristics of effective recruitment, and the impact and differential effects of recruitment strategies among particular population groups.

Identification of studies
We used The Quality of Reporting of Meta-analysis statement (QUOROM) to provide the structure for our review [24]. We identified four possible sources of potential studies, (i) electronic literature databases and websites, (ii) grey literature from internet sources, (iii) contact with experts to identify additional "grey" and other literature, and (iv) snowballing from reference lists of retrieved articles. In the first stage of the literature search, titles and abstracts of identified articles were checked for relevance. In the second stage, full-text articles were retrieved and considered for inclusion. In the final stage, the reference lists of retrieved full-text articles were searched and additional articles known to the authors were assessed for possible inclusion. We conducted a systematic search of electronic databases including OVID MEDLINE, EMBASE, PsychINFO, PubMed, Scopus, SIGLE and SPORTDiscus. We searched a number of web based databases including National Institute of Health and Clinical Excellence (NICE), Effective Public Health Project (EPHP Hamilton), Health Evidence Canada, and the Evidence for Policy and Practice Information and Co-ordinating Centre (EPPI)). We conducted searches of internet sites of key international walking promotion agencies including Walk England, the Centers for Disease Control and Prevention (CDC) and the World Health Organisation (WHO).
Studies published from the end of 2000 up to and including the search date (05/2009) were considered for inclusion. Individualized search strategies for the different databases included combinations of the following key words: (walk*) AND (recruit* OR participat* OR market*). Articles published or accepted for publication in refereed journals were considered for the review. Articles reported in UK grey and web based literature including any evidence of types of recruitment approaches and strategies, any evidence of effectiveness, economic costs, and evidence of any differential response to recruitment approaches were also considered in the review. Conference proceedings and abstracts were included if further searching of the databases or contact with the author was able to retrieve a full article from the study presented in the original piece of literature. We sent emails to international experts, identified in a previous systematic review on walking promotion [10].

Criteria for study inclusion/exclusion
Titles, abstracts and reports were independently assessed (by AM, CF and GB) for inclusion. Studies were considered to be eligible for inclusion according to the following criteria: (i) participants were of any age and were not trained athletes or sports students, (ii) studies of any type including randomised controlled trials, controlled before-and-after experimental or observational studies, (iii) studies that examined the effects of an intervention to encourage people to walk independently or in a group setting, (iv) interventions of any kind and in any field, whether targeted on individuals, communities, settings, groups or whole populations, (v) details of methods of recruitment were reported or were retrievable through correspondence with the authors, (vi) qualitative studies that examined the experiences of the participants during recruitment and which aimed to assess the effectiveness of the recruitment methods used, and (vii) studies published in English.
Included studies were categorised by study design using standardised criteria for quantitative experimental or observational studies (e.g. RCT, non-Randomised Control Trials (NRCT), before-and-after, cross-sectional), or qualitative studies (e.g. focus groups) [25].
Criteria for assessment of study quality in relation to recruitment Two authors (GB and CF) independently assessed the quality of the studies in relation to recruitment description that met the inclusion criteria. The criteria for assessing the recruitment reporting quality of each study were adapted from Jadad (1998) [26], and in consultation with experts. A formal quality score for each study was completed on a 5-point scale by assigning a value of 0 (absent or inadequately described) or 1 (explicitly described and present) to each of the following questions listed: (i) did the study report where the population was recruited? (ii) did the study report who conducted the recruitment? (iii) did the study report the time spent planning/preparing the recruitment? (iv) did the study report the time spent conducting the recruitment? (v) did the study target a specific population? Studies that scored 4-5 were considered as high quality studies while studies that scored 1-3 were considered low quality.

Criteria for assessing efficiency and effectiveness
Where possible we calculated recruitment rates and efficiency ratios for each study, based on a previous systematic review of interventions to promote physical activity [23]. We defined four terms, (i) "pool"-the total number of potential participants who could be eligible for study, (ii) "invited"-the total number of potential participants invited to participate in the study, (iii) "responded"-the total number of potential participants who responded to the invitation, (iv) "started"-the number of participants who were assessed as eligible to participate and began the programme. If data were reported we calculated ratios for each stage, e.g. started/ pool-by dividing the number of participants who started into the study by the total reported in the pool, and expressed as proportions. If possible we attempted to calculate a weekly rate of recruitment for those studies on the number of weeks/months spent recruiting per participant.
Passive recruitment methods, which require no interaction with the potential participants, were popular (Figure 2). Flyers/posters/advertisements/mail drops were the most cited approach used, appearing in 31 studies.   This was almost twice as prevalent as the second most popular approach, newsletters/newspaper articles (n = 18) and was nearly three times more frequently used than word of mouth. Word of mouth appeared in 12 studies, but we were unable to identify whether this was a proactive recruitment strategy or a reactive strategy, responding to low recruitment numbers. Less popular methods included medical and health insurance referral, invitations derived from clinical or employment data, study information sessions, resident listings, announcements at group meetings or community events and information stands. Table 6 presents data on the setting and location of recruitment and the study. We observed some studies that "matched" where the recruitment was conducted with where the intervention was delivered. Culos-Reed et al, 2008 reported recruiting participants for a mall walking study at the mall where the intervention was going to be delivered [71]. Other studies did not match in this way, and recruited in many different locations, often relying on print material alone, and requiring potential participants to attend a location which may not be easily accessible to them. Studies reported that they were "community-based" (n = 25) [27][28][29][30][31]35,36,42,[48][49][50][51][52][54][55][56][57][58]61,63,68,69,[71][72][73] but asked community members to travel into a research setting to begin the process of participation; for example medical centres or universities (n = 20) [29,30,[33][34][35][36][37][38]41,43,46,47,49,56,[62][63][64][65][66][67]. These interventions used a mixture of recruitment approaches including media events and led walking groups, face to face interventions (e.g. counselling, pedometers) or mediated interventions, such as internet, e-health and mobile phone technology [74].

Recruitment rates and efficiencies
We originally planned to calculate recruitment rates and efficiency ratios for each study but we were unable to do so due to missing data (Table 7-Recruitment rates and efficiency ratios). Only three studies provided all the data points [33,36,65]. We were able to calculate a weekly recruitment rate using the final numbers of participants divided by the time spent recruiting in weeks for eleven studies (mean 38 participants per week, range 1 to 268 participants per week). We were not able to see any pattern between recruitment approaches and weekly rates. Two studies reported some data on the efforts needed to undertake recruitment. Jancey et al (2008) reported that after potential participants had received invitation cards it took approximately 9 calls to recruit one participant [53].

Developing Recruitment Approaches
We identified factors that may have helped or hindered recruitment from qualitative [57,64,69] and protocol [27,28,30,35] papers. These factors emerged as possible principles of recruitment and were related to training, engaging possible participants in the recruitment process and allowing sufficient time to pilot-test approaches. Watson et al (2009) used trained post-natal health care staff to actively recruit participants during their first home and health centres visits, and at group meetings for new mothers [51]. Recruitment approaches used by Banks-Wallace et al (2004) were based on a 5 month needs assessment study of the concerns and priorities of their target group [27]. The authors reported this process established trust between the research team and participants and ensured active participation in the study and in fact over-recruited from this population. Nguyen et al (2002) reported promoting participation via word of mouth, e.g. one participant tells/recruits another participant [69]. These appeared to have more impact on recruitment than passive approaches like posters or media stories [69]. These data suggest that developing recruitment approaches is a time and resource intensive activity, requiring skilled research and recruitment staff.

Discussion
We conducted a systematic review to examine the reported recruitment procedures of walking studies, in order to identify the characteristics of effective recruitment and the impact and differential effects of recruitment strategies among particular population groups. We identified the need for a common understanding of the recruitment process for walking studies in terms of conceptual definition, defining effectiveness and more detailed reporting. Due to the heterogeneity of studies we were not able to identify what specific recruitment approaches were most successful with particular population groups.
We identified eighteen recruitment strategies from 47 studies but did not see any relationship between one   particular strategy or group of strategies and recruitment rates. Many studies blended different recruitment approaches and strategies, adopting an almost "trial and error" approach. Only two studies reported the effectiveness of their approaches to recruitment [28,35]. We were able to distinguish active and passive recruitment approaches. Further research is needed to directly compare specific recruitment strategies. Very few studies examined the successes of recruitment approaches to physical activity interventions. Harris et al (2008) conducted a randomized controlled trial of four recruitment strategies in their physical activity promotion intervention study for older adults [75]. The authors reported that telephone follow up a week post invitation significantly increased recruitment compared to invitation only. Certainly the principle of follow up was found in a number of our included studies [53,63,72] but we could not assess the efficacy of these strategies. The efficacy of phone recruitment has been questioned by Margitic et al (1999) [76] who compared three recruitment strategies for Project ACT: patient mailings, office-based questionnaires and telephone contact. However their participants were not randomized to a particular strategy. The authors also reported that despite telephone recruitment appearing to be productive this strategy was dropped in two out of eight recruitment sites on cost grounds. This behavioural approach of using phone follow up has previously been reported to be more effective than no follow up in changing physical activity and walking behaviour [10,23] and certainly warrants further testing in terms of a possible recruitment strategy.
Tai and Iliffe's (2000) experiences of conducting physical activity studies also support our observation that piloting and pre-testing of recruitment methods would improve rates of recruitment and precision in recruiting specific target groups [77]. Our review clearly shows that current recruitment strategies resulted in recruiting mostly white, well-educated, middle aged women. The attraction of walking projects to particular social groups has also been reported in previous evaluation studies of community walking programmes both in the UK [78] and USA [27,31]. Our review found that recruitment rates were poorer for men, especially within workplace or community settings but we were unable to determine if it is a fault of the recruitment, or the offer of walking or a combination of both that is at fault.
We identified a number of studies that "matched" where the recruitment was conducted, with where the intervention was delivered. This principle supports the notion that connecting the place of recruitment and intervention may be more efficient both for the participants, recruiters and interventions teams. We found studies that did not effectively match these aspects and perhaps this was reflected in the total number of participants recruited and the longer time it took to conduct their recruitment phase. For example,  reported that participants were expected to travel to the university to receive their intervention. We found little data on the time spent planning/preparing and implementing recruitment so any potential learning from recruitment remains unreported [63].
We identified a number of studies that also aimed to match those recruiting with those being recruited. Banks-Wallace et al (2004) reported in detail their use of a recruitment mediator [27]. The mediator was the same gender as the target group, was a prominent local figure, trained in delivering community interventions and female. Her role was to introduce the study to key significant figures in the area and increase awareness, to assist directly with the recruitment phase and to introduce the researchers to the potential participants at an information session. Banks-Wallace et al (2004) described this approach as increasing trust and decreasing differences between the recruiters and recruited [27].
Our review clearly found there was very little consistency in the definition or reporting of recruitment. We found many different interpretations of (i) what is the recruitment process? and (ii) what is an appropriate metric for evaluating the effectiveness of recruitment?   The lack of conceptual clarity about recruitment as a process is surprising and potentially impacts on costeffectiveness. The RE-AIM framework emphasises the need to judge the success of an intervention from both the reach and uptake of an intervention [79]. In light of this we constructed a conceptual framework for our review by defining the stages of recruitment and potential pool of participants ( Figure 3). This framework offers a starting point for further debate and refinement. The framework offers a clear concept of the stages and steps of recruitment and the chance to record the numbers of participants at each stage and action. Our framework divided recruitment into two phases, planning/preparing and implementation, with four stages involving discrete actions by researchers/recruiters, (i) identification of participant pools, (ii) invitation and monitor response and uptake, (iii) assessment, screening and facilitation and (iv) re-invitation of responders, before the delivery of intervention to starters. This framework highlights the actions needed at the start of a recruitment process, i.e. planning/preparing the recruitment process. It also emphasises the importance of the reach of an intervention i.e. the pool of participants used to provide recruits. This differs from the recent new CONSORT framework which asks for dates of recruitment period (i.e. delivery) and enrolment stage [18]. CONSORT stipulates data must be reported for numbers of participants eligible for study which we feel not only ignores the overall pool of possible participants, especially in community based studies of walking interventions, but also ignores the population deemed ineligible, as seen in pre-screenings of patient lists for existing conditions [18]. The "pool" of participants perhaps provides a more realistic denominator for assessing overall recruitment rates. This metric will allow new studies to (i) consider if the recruitment was efficient (i.e. study recruited expected numbers of participants) and/or (ii) consider if it was effective (i.e. study recruited the right target group), and/or (iii) reflect the true costs of all recruitment actions within overall cost benefit calculations. The need for better reporting of recruitment actions and numbers is essential to improve the assessment of present recruitment strategies. This view is mirrored in recruitment studies of other health behaviours, and better reporting must begin before we can start to identify which strategies provide the best recruitment rates [80].
The results of our review were limited to walking intervention studies. We were limited by only including studies written in the English language. We were limited by what was reported in papers but our consistent application of inclusion, quality and data extraction criteria have illustrated the need for improvement in both the reporting and science of recruitment. As journals look to keep research reports within word limits, it is likely that there will continue to be a lack of journal space to report recruitment details, and we would like to call on editors and authors to report recruitment details or provide short methods papers for the insight of future researchers. As far as we are aware this review is the first of its kind focusing on one domain of physical activity behaviour. The lack of understanding and studies into recruitment may reflect some of the findings about the existing weaknesses of the evidence base for walking interventions, e.g. lack of generalisability of interventions across different social groups [10].
The evidence base for the benefits of walking is now expanding but until it is clearer what strategies are effective in both recruiting and initiating people to begin walking, such benefits may be out of reach for particular population groups. Practitioners would benefit from the assurance of having an evidence based best  Computer assisted telephone interviews (CATI) was initially conducted by researchers. A database of potential participants was screened for telephone numbers. If this was not successful in recruiting the sample size needed the direct mailing was used. Finally, to complete the sample size quota canvassing in the local community (including face to face, door to door, posters and flyers at churches and senior housing units, snowballing, utilising 'community brokers', and newspapers) was conducted. Recruitment was systematic, purposeful and carried out in the order described but was somewhat inequitable as the first screening criterion was the availability of a phone number. It also required significant community assistance to reach those harder to engage. Researchers designed a flyer with community input and received advice on where to place it. Emails and newspaper announcements were also used. Recruitment staff distributed print material at specified schools, churches, grocery shops, libraries, clinics, community agencies and community fairs and at 10 presentations in community agencies, clinics, and churches. Email announcement at local medical centre workplaces and an announcement in the community newspaper were used. A good aim of matching the invitation to the invitee and finding the best place to distribute it was a positive here. Unfortunately word of mouth wasn't actively used or reported and only the research team recruitment staff acted as recruiters for face to face recruitment.

Baker et al, 2008b
4 Mail drops were carried out and adverts were placed in local papers and posters in GP surgeries and shops. Manned community stalls were also set up. This approach was modified and expanded throughout the recruitment phase as the researchers identified their lack of impact on the target group. However, the methods were mainly passive and not altered to be more engaging or mediating with the target group. It is not specifically stated who conducted the recruitment.  A three phased sequenced approach was used. Researchers, assisted by staff nurses sent an eligibility questionnaire to a randomly selected group from a GP client list (GP letters included). This was followed by a letter explaining the study to those expressing an interest and then a phone call to the responders to arrange which info session they could attend.

Lee et al, 1997
4 Researchers and trained female students conducted telephone calls, face to face approaches at supermarkets, direct mailing and flyers.

Matthews et al, 2007
3 Clinical staff recruited women by letter and phone follow up in two health centres. The paper also states that in another centre clinical populations were recruited, but this is not clearly explained. Women who were also past participants in a case control study and had agreed to take part in future research. Pastors who attended luncheons regarding health promotion and disease prevention strategies among African Americans were recruited to help introduce the intervention and aid recruitment of participants. Following this, researchers placed flyers in churches, bulletins in newsletters, announcements at Sunday services and held information meetings.  This qualitative paper did not clearly describe the processes behind their recruitment approach. It emphasises the need for the walk leaders and organisers to become actively engaged in the process and how interpersonal approaches are highly necessary and more effective in engaging a broader range of participants or specific target groups. practice model which details how best to recruit participants as well as what is the best intervention to promote walking. Our conceptual framework offers researchers, practitioners and policy makers a way forward to develop and assess the success of a recruitment strategy to target particular groups. The model offers options through the four stages to assess how many people are responding and engaging in a walking intervention, but also whether any bias is occurring and if efforts need to be refined towards a specific group. It could also provide a true picture of the costs of the intervention as the inclusion of recruitment development and implementation should be included in economic evaluations. The results of our review could translate into a series of recruitment principles for further evaluation by researchers. These principles include (i) form recruitment plans and strategies on evidence of what the target group feels would be appropriate, based on formative research, (ii) conduct a pilot phase of testing, (iii) recruit in places where the participants are located, (iv) allow sufficient time to recruit participants and monitor the uptake, (v) provide training in recruitment methods for recruitment staff, (vi) monitor the participants response to recruitment approaches and use different recruitment strategies where necessary.
The future of walking and physical activity promotion will lie not only in establishing the effectiveness of different interventions but also in improved recruitment practice. Currently, generalisability is limited by reach within studies; but while the current methods being used are applied, the current limited reach will prevail. We offer principles for recruitment that require further        Figure 3 Conceptual framework for the stages and steps of recruitment with actions for researchers and participants. evaluation, (i.e. matching "where to where" and "who to who"). Future research to identify "what is effective recruitment?" may best lie in identifying approaches that reflect the needs and expectations of hard to recruit target groups. This will allow researchers the opportunity to investigate the strategic use of the right recruitment methods, for the right group, in the right order.