Recruitment of older adults to three preventative lifestyle improvement studies

Overview of projects

Participant recruitment

GP mail-outs were utilised for the Lifestyle Matters RCT and Putting Life In Years RCT. Local GP surgeries were identified through the Primary Care Research Network (PCRN) to ascertain their interest. Interested GPs were often larger practices that were research active and able to identify resources to undertake large mail-outs. Each interested GP surgery was then approached by a member of the study team to arrange a time to discuss the study and arrange site set up. It was requested that GPs send potential eligible participants (patients registered to their surgery who met the age inclusion criteria for the study) an invitation letter and study leaflet along with a pre-paid response card. Potential participants who were interested completed the response card, indicating whether or not they met the initial eligibility criteria as indicated on the card and returning it to the study team. On receipt, the study team telephoned the individual to arrange a screening visit.

Regardless of which method used to approach the individual, a research assistant (RA) visited each potential participant at home for a screening visit. During the visit, the study was discussed with the participant before undertaking the Six-item Cognitive Impairment Test (6CIT) to test for cognitive impairment; those who obtained a score of 8 or more were deemed ineligible and advised to visit their GP [28]. Those deemed eligible could then either consent to the study and complete the baseline measures, have more time to consider whether or not to take part in the study and be contacted again by the research team, or decline to participate. Consenting participants were randomised to either the control or the intervention group.

For the Putting Life In Years RCT, mail-outs were also sent via an existing cohort study (the South Yorkshire Cohort) to individuals who were willing to be contacted regarding future research [29]. In January 2013, recruitment to the Putting Life In Years RCT was halted early due to issues with delivering the intervention by the local branch of a national charity, despite meeting internal feasibility targets for the recruitment and retention of participants [22].

Data collection

Analysis

A framework approach was utilised in qualitative data analysis, which took a case study approach [30]. Framework Analysis is a five-stage process of: familiarisation, forming a thematic framework, indexing, charting and mapping and interpretation. The analysis was undertaken by the same two researchers (SC and KS). Firstly, both researchers independently familiarised themselves with the same sample of interview transcripts for facilitators and participants through reading and rereading, at which point saturation was confirmed. The same sample of transcripts were then coded for emergent themes independently by each researcher before being triangulated for cross-cutting and diverging themes between the participant sample or facilitator sample and across both samples. From this an initial thematic framework, consisting of major nodes and one level of ‘tree’ nodes, was developed by the two researchers. This framework was then applied to the same sample of transcripts. The researchers met to discuss any differences which were managed through consensus before finalising the framework. All transcripts were then coded by either SC or KS using the final index using NVivo 10 software. Matrices were developed, with one major node per matrix. These matrix charts were then examined for cross-cutting themes and patterns in the data were mapped to inform the final level of interpretation. In writing up the analysis, quotations that demonstrated the breadth of opinions were chosen. Participants did not provide feedback on the findings.

Ethical considerations

For the qualitative aspect of the Lifestyle Matters RCT, all participants and facilitators provided written informed consent for participation in the study. Ethical approval for the main trial and this qualitative sub-study was granted by the South Yorkshire Research Ethics Committee (reference number 12/YH/0101). Transcripts were anonymised prior to data analysis in order to preserve confidentiality.

Lifestyle Matters feasibility study

District nurse referrals did not yield any participants; this appeared to be primarily due to a lack of engagement in the study. In addition, requesting that participants telephoned the university to find out information about the study may have been an additional barrier for some individuals, excluding individuals with low confidence or hearing problems. As a result, a programme of community engagement activities were conducted which enabled the recruitment of 25 participants. Community engagement overcame communication barriers allowing participants to gain a ‘taste’ of the intervention, allowing 89.3% of the studies target to be recruited.

Lifestyle Matters and Putting Life In Years RCTs

Principal findings

In the three studies discussed in this article GP mail-outs were successful methods of recruiting the numbers of participants required for large-scale, preventative, lifestyle intervention studies. In the Lifestyle Matters RCT, interviewed participants suggested that mail-outs should contain more information about what taking part in the study would involve. Other recruitment strategies including referral by district nurses and other health professionals, or using media and posters, made a trivial contribution to recruitment as previously reported by the FiT study [16]. Community engagement was successful in a non-randomised feasibility study, but not in the subsequent RCT of the same intervention.

GP mail-outs for both RCTs described in this paper resulted in a low yield of consenting participants (under 3%). Compared to other studies reporting recruitment via GP mail-outs, this figure is moderate; 7% of potential participants that were approached for involvement in the FiT study consented to participate, and 0.04% in the Booster study [15, 16]. These recruitment rates are part of an expected range of recruitment rates that empirical studies demonstrate are lower for prevention RCTs than for therapy RCTs [31].

Interviewed participants in the Lifestyle Matters RCT were happy having received a letter from their GP regarding the trial; but this would be expected as the majority of participants were recruited via this method. A few participants did speak of confusion on receiving such a letter; such a reaction has been documented in another trial [32]. Participants and facilitators felt that individuals who would most benefit from the intervention were not recruited to the study. The apparent socialisation and confidence of those who took part could have been due to the recruitment methods used; interested participants had to opt in to the study and contact the research team. Such an action requires a certain level of confidence, thus resulting in possible volunteer bias. The source and impact of such bias is currently under-researched in trials of behavioural interventions; the effect of participation itself may interact with such interventions [33]. It was hoped that district nurses and community engagement would allow the targeting of those who were isolated and would benefit most from the intervention. Instead, the use of GP mail-outs in the Lifestyle Matters RCT appeared to have resulted in a sample that was not as isolated as the intervention facilitators expected. Despite this, facilitators still witnessed improvements in participants, despite not viewing them as isolated individuals.

Several interviewed Lifestyle Matters RCT participants felt that direct contact with GPs would be the most beneficial method to recruit, although, as far as we are aware, this was not achieved in the studies we undertook. However, other studies have found difficulty in this type of recruitment due to GPs’ levels of understanding of research not being sufficient, time constraints, issues with introducing research to vulnerable populations, or due to emotional concerns that the GPs may have, such as worries over the patients’ eligibility [34, 35]. In other cases GPs can be overly enthusiastic to recruit, resulting in patients being recruited who are not eligible to participate [18].

Other primary care staff can be instrumental in identifying study participants, particularly groups such as older vulnerable people. District nurses may have acted as ‘gatekeepers’ to participants being recruited to the studies; gatekeepers may act paternalistically once they have decided that the experimental intervention is beneficial, potentially not offering the trial to clients who were the target population due to concerns around vulnerability and randomisation outcome (i.e. allocation to the control group) [36]. In addition, it seemed evident during the recruitment phases of the trials that recruitment of patients to a preventative clinical trial was not a priority for the health care services.

Strengths and weaknesses

This paper presents data from three studies that recruited participants with similar characteristics. It adds empirical observations to the theories that prevention studies have comparably low recruitment rates and that recruitment by targeted mail-outs following database searches are more efficient than opportunistic referral by health professionals. Much of the data derive from a single UK region and the two large trials were conducted from the same trials unit, although we reference comparable studies conducted elsewhere. Only one of the three studies included qualitative data of relevance to this topic.

Recommendations for researchers and commissioners

We recommend the use of GP mail-outs for large-scale RCTs as a time-efficient method of recruiting participants. Such mail-outs should include relevant information in order for the individual to understand the research project and nature of the intervention and/or control arm and should make it clear that the communication is from the individual’s GP. Researchers should be wary of the potential volunteer and participation biases that may cause certain populations to be under represented when using recruitment methods that require potential participants to actively make contact with the research team [37]. To reduce volunteer bias it may be necessary for potential trial participants to have direct contact with clinicians. To achieve this, it may be effective to ‘buy out’ recruitment capacity among health professionals who serve hard-to-reach and vulnerable populations or obtain honorary contracts for research nurses to work more effectively with such services. In addition to monetary mechanisms, it is pertinent to ensure clinicians ‘buy in’ to the research – they should understand the need for the study and the need for randomisation in conditions of clinical equipoise, given that social interventions can cause harms to target populations [38]. Training of health care staff may aid recruitment of vulnerable populations to clinical trials; training strategies are required that orientate staff towards perceiving recruitment as ethical and rational given therapeutic uncertainty [39]. In light of the increasing quantities of applied health research being undertaken, plus the significant clinical involvement required to test psychosocial interventions, it is difficult to commit resources to recruitment activities at times when health services are struggling to meet demands for routine clinical services.

The lack of relatedness in terms of participant recruitment between the feasibility study and the Lifestyle Matters RCT adds to a growing body of research around issues around answering important methodological questions in feasibility studies [40]. The feasibility study was not randomised; the addition of the control (usual care) arm to the Lifestyle Matters RCT may have caused difficulties recruiting isolated adults. The size of the studies was also a salient factor – the research team for the feasibility study were focussed on one location only and were able to identify a clear strategy for recruitment from that area which may not have been successful elsewhere; for example, in this case by linking with local health champions and providing taster sessions in well-used community venues. Lessons from the feasibility study helped design the main Lifestyle Matters trial, but research suggests that neither pilot nor feasibility studies can or should be used to estimate recruitment rates to the main trial [41]. We add to the previous recommendations that qualitative research should be used alongside feasibility studies to assess the feasibility of recruitment techniques [42].

Future research

More research is needed into the characteristics of volunteers who enrol and those who do not to better understand the reasons for non-participation so that this problem can be addressed [43]. Participation bias threatens external validity, making it hard to say how generalisable the findings are. There is a lack of evidence regarding the effect on recruitment of adding a control arm when transitioning from a single-arm feasibility study to an RCT; this requires further quantification.

The cost-effectiveness of different recruitment methods would be important to take into consideration in order to assess the most efficient method of recruitment. Huynh et al. undertook a systematic review to identify research that has been undertaken comparing the cost of different recruitment strategies, identifying that only two studies compared costs of recruitment, with monetary incentives in these studies costing more per patient recruited than direct contact with the patient [44].

In order to assess general recruitment method effectiveness, and cost-effectiveness, embedded trials could be undertaken that compare two or more recruitment methods within a study. Such an approach has been recommended by the Systematic Techniques for Assisting Recruitment to Trials (START) programme [45]. One embedded trial compared the use of a £100 incentive to attract socially deprived and elderly patients to the use of no incentive, finding that although it led to an increased patient response, it did not attract the more socially deprived patients [46].

Further research is required to identify effective recruitment strategies for vulnerable older adults. Participants recommended ‘snowballing’ in order to identify other participants, which may be an effective method of identifying older adults who are isolated and difficult to recruit. Other studies have tested this method, finding that snowballing (also known as respondent-driven sampling; RDS) was a cost-effective method of identifying individuals [47]. It needs to be established if such a method could recruit the number and frequency of participants required for a large RCT that aims to recruit participants over a long time-scale. In addition, opt-out methods are becoming increasingly acceptable in cluster trials where outcome data are routinely collected [48]. However, with the new General Data Protection Regulation (which replaces the Data Protection Act) coming into effect from May 2018, opt-out methods of consent have effectively been outlawed due to the requirement for informed consent to be ‘unambiguous’ and involve ‘clear affirmative action’ [49].