The DEPS-GP project has been successful in recruiting a large sample of GPs and their older patients. However, the retention rate of participants to date highlights the challenges of maintaining GPs' participation and commitment. Critical reflection on the recruitment processes, strategies to maintain participation, and the reasons given by the GP at the time of withdrawal provides valuable insight to other researchers experiencing the challenges of recruiting within the general practice setting. Various aspects of these processes, strategies and reasons are considered in turn below, and a summary of these can be found in Table .
Summary of barriers and enablers to the recruitment and retention of general practitioners
Recruitment relied on the use of an existing database to identify the sampling frame. There may have been several alternative approaches. One approach would have been to use an alternative database, but the options were limited. For example, the 119 Divisions of General Practice in Australia each hold information about their own networks of GPs and have local knowledge that assists in maintaining the accuracy of their lists, but these data are not aggregated nationally so drawing on Divisional data was impractical [13
]. Another approach would have been to generate our own list of GPs and/or to actively recruit GPs through our own networks, but this would have been labour-intensive and impractical given our required sample size, and would have potentially introduced sampling bias [6
]. Veitch et al [8
] have cautioned against using existing databases for recruitment because of their inherent inaccuracies, and it is true that the Australasian Medical Publishing Company Proprietary Limited database proved to be somewhat over-inclusive. However, we felt that the advantages of this approach outweighed the disadvantages. It resulted in a high absolute number of recruited GPs (n = 772) but a low overall response rate (4.1%).
Having identified the sampling frame, we pursued a recruitment strategy that involved minor promotion through Infonet
and a letter of invitation, both of which have been cited in the literature as successful recruitment methods [2
]. Our large sample size target dictated our choice; these methods were practical and less costly and labour-intensive than others that have been described in the literature, including telephone contact [2
], practice visits by the study team [2
], and the use of physician and/or peer recruiters [6
Although we did not explicitly collect data on why our study GPs chose to participate, some inferences can be made. Studies that have looked at reasons for participation and non-participation have consistently found an interest in the research topic, minimal time commitment and professional recognition as influential [1
]. Anecdotally, a number of our participants indicated that they had a particular interest in mental health and/or that they were keen to avail themselves of the CPD points available. Some also questioned the time commitment, and agreed to participate when they were satisfied that it was not too onerous. It is interesting that the highest response rate was recorded in Western Australia where the study originated. The investigators in this state may be known to the general practice population due to a previous study [15
]. Victoria was the most difficult state to recruit from and this may be due to the high volume of research undertaken in general practice there.
Once GPs agreed to participate, establishing a relationship with them and their reception staff was crucial, as were providing clear instructions and maintaining regular contact and support. Frequent calls were made to check on progress and provide encouragement. Often the GPs had varying preferences with regard to communication, with some preferring fax or email communication over telephone contact and vice versa, and some choosing ad hoc contact and others preferring to set specific times for teleconferences. As it was not always possible to speak to the GP directly, the reception staff became an important point of contact. Our research assistants were dedicated to establishing rapport with both GPs and reception staff, providing them with clear instructions and protocols, and 'fitting in' with their stated preferences.
In addition, we endeavoured to make the tasks required of the GPs and reception staff as simple as possible. For example, a number of GPs had difficulty generating lists of patients eligible to receive the postal questionnaire (either directly or via the study team). Consistent with technological barriers identified in other studies [5
], this was largely due to lack of knowledge of the relevant computerised system. We produced a step-by-step guide on how to interrogate standard databases in a manner that produced listings of patients in the relevant age group. In addition, research assistants visited practices to assist with the generation of lists and other aspects of the questionnaire process, such as sticking labels on envelopes. In most cases, this overcame technological problems, but in some instances (e.g., where the GP used a non-standard computerised medical records system, or worked in a non-computerised practice) residual difficulties remained.
The slightly higher response rate of questionnaires sent directly by the GP (29.8%) compared to the response rate of those posted by the research team on behalf of the GP (26.0%) suggests that a direct method would be preferable in future studies. This would have to be weighed against the extra time required by the research assistant in ensuring the GP completed the mailout within project timeframes and whether or not the GP would have participated had the indirect method not been offered. As recommended by Edwards et al [16
], we included the covering letter from the GP and reply paid envelope and used coloured ink for the survey to increase the response rate of the questionnaire.
Analysis of telephone logs from each state showed the reasons for withdrawal stated in Table . In addition, informal feedback from participating GPs suggests that the above strategies were successful in keeping them 'on board'. Nonetheless, we acknowledge that our retention rates were worse than our initial recruitment rates, and that there were barriers to ongoing participation that we were not always able to address successfully.
Most notably, and consistent with other studies [2
], withdrawing GPs reported that they were unable to complete the required tasks in the time available due to competing demands that were sometimes unforseen when they 'signed up'. This was particularly the case with the practice audit. Tasks associated with the practice audit created a greater burden for the GP than tasks associated with the postal questionnaire because the latter could often be completed by reception staff. Some GPs reported being too busy to turn their attention to the audit, and did not manage to open the package of audit materials; others were unable to find the time to set the audit up; and still others began the task, but found that their schedules were too hectic to allow them to identify their allotted 20 patients and conduct the audit with them. The time issues associated with the audit were exacerbated by the fact that the audit was delayed by 3 months, so some GPs who had anticipated that it would occur at a time of year when their load was relatively light found that it actually took place when they were particularly busy.
Some GPs were 'lost' to the study by virtue of ineligibility, despite the study team's best efforts to provide clear instructions regarding eligibility at the outset. Some GPs were found to be ineligible once the project began because it transpired that they had over-estimated the numbers of English-speaking patients aged 60 or over attending their practice. Others ruled themselves out by not adhering to study protocols (e.g., handing out questionnaires to patients when they presented for a consultation, rather than posting them out).
Further retention difficulties arose because of issues with personnel in the given GP's practice. In some instances reception staff acted as 'gate keepers', making it difficult for the study team to establish and maintain contact with the GP directly. In other cases, lack of available reception staff was a problem, particularly for GPs in solo practices. In addition, non-participating GPs in the participating GP's practice directly or indirectly raised barriers either by actively objecting to their colleague's participation, or because the shared patient record system rendered it impossible to identify patients of the participating GP.
In a small minority of cases, GPs withdrew during the postal questionnaire stage because they were concerned the project could have a negative impact on some of their patients. In one case, the situation arose because recipients of the postal questionnaire (or their carers) raised concerns about the content of the questionnaire (specifically the items related to suicidality). This GP felt that continued participation would interfere with their practice and therefore withdrew.