Recruiting and retaining general practitioners to a primary care asthma-intervention study in Australia

Additional keywords: general practice, intervention studies, recruitment.

Research about different recruitment methods (RMs) to engage general practitioners (GPs) in educational interventions is limited (Foy et al. 2003). What is apparent is that delays in recruitment greatly add to the cost of intervention studies and threaten the results by insufficient power from insufficient sample size. Investigators have been urged to incorporate evaluations of recruitment strategies into study appraisals in an effort to obtain more evidence about the effectiveness of recruitment methods (Watson and Torgerson 2006). Such evaluations can help researchers obtain information regarding recruitment for studies similar to theirs (Jones et al. 2011).

Primary care research involving GPs spans a variety of different types of studies. In terms of the level of GP engagement in the research, these studies can be broadly classified into two major types: where the primary role of the GP is to help recruit patients, or where the role of the GP is to participate in an educational intervention program as well as recruit patients (Pearl et al. 2003). While the specific level of involvement may differ within each type, it is evident that a far greater level of participation by GPs is required for intervention studies.

Notwithstanding the importance of intervention studies in general practice, recruiting GPs for interventional research is often challenging, especially when the GP’s active participation is central to the research (van der Wouden et al. 2007). Trials have failed (Tognoni et al. 1991) or have had to extend their recruitment period due to difficulties faced in enrolling GPs (Yallop et al. 2006). For example, a survey of published UK primary care trials found that only a minority of trials successfully recruited the required number of participants within the projected timeline (Bower et al. 2007). Time constraints due to other commitments like audits and mass vaccinations can often hinder doctors’ participation in research (Asch et al. 2000; Rosemann and Szecsenyi 2004; Minas et al. 2005; Goodyear-Smith et al. 2009) Australian GPs do not rate financial rewards as a reason to join a research project; however, financial incentives can support a practice’s loss of time. There are also project-related and topic-related reasons why GPs do not participate, such as an overly complex research design or a disease area that is considered too sensitive, such as childhood obesity (Jones et al. 2011, 2012).

This paper describes an observational study of three different approaches that were sequentially used in a randomised controlled trial to recruit GPs for the Practitioner Asthma Communication and Education (PACE) Australia study (Shah et al. 2011).

The PACE Australia study aimed to recruit 120 GPs within a 6-month period in Western Sydney from a cohort of 953 GPs registered with the four local Divisions of General Practice. All GPs were eligible for recruitment unless they had participated in the earlier pilot study (Shah et al. 2010) or were planning to retire within the next year. Initially only one GP per practice was recruited to reduce the effect of clustering.

GP participation in the trial consisted of: (i) completion of a questionnaire at baseline and at 12 months; (ii) participation in the PACE intervention, either soon after recruitment in the intervention arm or after completing the 12-month questionnaire in the control arm; and (iii) identifying paediatric asthma patients, either through electronic patient records, written records and/or knowledge of patients. GP participation in patient recruitment was limited to patient identification and signing prepared study invitation letters; research staff sent out the letters and made the preliminary screening phone call to the family of the patient.

The PACE intervention consisted of two 3-h interactive workshops facilitated by a community physician, paediatric respiratory physician and a locally known GP. The details of the program have been published elsewhere (Shah et al. 2011). Participating GPs received Continuing Professional Development points from the Royal Australian College of General Practitioners.

All GPs listed in a GP Division database were contacted with a study invitation letter, an information flyer and an Expression of Interest form. A GP was defined as ‘interested’ when the contacted GP faxed back an Expression of Interest form or sent an email to the research team requesting additional information about the study. ‘Interested’ GPs were defined as ‘recruited’ once they formally consented and were subsequently randomised to intervention or control (delayed) arms. ‘Recruited’ GPs were defined as ‘retained’ if they completed a questionnaire about their asthma management practices and communication behaviour at baseline (Roydhouse et al. 2011) and did not notify staff of their intention to withdraw at any time during the study.

Due to slower than anticipated recruitment of GPs in the first 6 months, the recruitment period was extended by a further 12 months. In addition, the inclusion criteria were modified to enable more than one GP per practice to be recruited and the catchment area was extended to Central Sydney after 12 months. Three different approaches were used sequentially to recruit the GPs required for the study. In each approach, initial contact was made using a combination of invitation letter, email and/or fax. GPs faxed back an Expression of Interest form or contacted the research team by telephone to participate in the study. Staff visited the GP practices to further explain the study, obtain GP consent and a list of children aged 2–14 years with asthma. The key difference in each method was who led and delivered the recruitment strategy, specifically the telephone call and face to face GP practice visit.

A list of GP names and addresses were provided by the four local GP Divisions in the Western Sydney region. A Division has GP members from specific geographic areas and offers programs and services for its participants (Williamson et al. 2007). An invitation letter was sent to each GP on the provided list. The study was also advertised in the Division’s newsletter and at local GP meetings. After receiving an expression of interest, PACE staff visited the practice manager and the GP to further explain the study, and to obtain consent and a patient list. Staff commonly had difficulty making an initial appointment with the GP and, despite the appointment time, often required a long wait before meeting the GP. At times the project staff had to leave without meeting the GP.

After 6 months of poor responses from the GPs, the recruitment was outsourced to a Western Sydney GP research group, initiated by three GP leaders, for 5 months. The same invitation letter or fax was sent to GPs listed in their database. As in RM1, following a returned Expression of Interest form, the GP research group’s project manager visited the practice manager and the GP. At that time the recruitment criteria were relaxed to include more than one GP per practice.

After 12 months, the study was extended to include an additional Divisional area, the Central Sydney General Practice Network. As before an initial fax or email invitation was sent to GPs listed in the database of the two Divisions of General Practice. A personalised invitation letter, signed by the heads of the Division of General Practice and the chief investigator (CI), was also sent. As in RM1 and RM2, GPs faxed back an Expression of Interest form or emailed, if interested. However the CI, a medical peer, made the appointments directly with interested GPs and visited each of their practices with the research officer. The CI also contacted the GPs who had previously expressed interest under RM1 but who had not yet enrolled in the study.

The total number of GPs contacted for each approach was calculated by adding the total number of faxes and letters sent (this number was provided by either the Division or the local GP organisation), plus any other recorded numbers contacted (e.g. the number of personal contacts). All statistical analyses were conducted using SPSS, version 15 (SPSS Inc., Chicago, IL, USA).

During RM1, 953 GPs were contacted, of whom 38 were interested and 36 were recruited (Table 1). During RM2, 575 GPs were contacted, of whom 34 were interested and recruited. Lastly, during RM3, 681 GPs were contacted, of whom 94 were interested and 82 were recruited.

Table 1.  Interested, recruited, and retained participants, by recruitment method (RM)

Fourteen interested GPs were turned away during RM1 due to the restriction of recruiting only one GP per practice. Three of these were subsequently recruited during RM3.

The recruitment rate for the three methods was seven GPs per month for RM1 (6-month recruitment period), seven GPs per month for RM2 (5-month recruitment period) and 12 GPs per month for RM3 (7-month recruitment period).

In RM3, the primary area of focus was Central Sydney, from which 70 GPs (81%) were recruited in comparison to only 16 from Western Sydney (19%). In RM3 recruits from Western Sydney were also GPs who had been turned away under RM1, as well as personal GP contacts of the CI. Furthermore, fewer Western Sydney GPs withdrew under RM3 than under the others, with only 1 of 16 (6%) withdrawing under RM3, compared with 7 of 34 (21%) under RM2 and 19 of 37 (51%) under RM1.

The demographic characteristics of the participants across the three different recruitment methods are shown in Table 2.

Table 2.  Characteristics of recruited GPs by recruitment method (RM)

Across all three recruitment methods 57% of GPs were Fellows of the Royal Australian College of General Practitioners and 86% were in group practices. There were more women than men within RM1 and RM3. Fewer GPs who completed the baseline questionnaires within RM1 (62%) were ‘retained’ compared with RM2 (81%) and RM3 (89%).

The recruited PACE GP had been in practice for a mean of around 20 years, which was similar across the three recruitment methods. In this respect, the PACE GPs were similar to the GPs participating in the national Bettering the Evaluation of Care and Health study (Britt et al. 2008; Roydhouse et al. 2011). Within RM3, more GPs graduated in Australia than overseas and had the same median number of continuing professional development hours devoted to asthma.

The slow recruitment of GPs for this randomised controlled trial of an intervention study is consistent with the recruitment difficulties experienced by other studies in primary care where significant engagement was required by GPs. However, there were differences in the rate of recruitment and retention according to which recruitment method was used. The third method, RM3, which included initial contact and visit by the CI, a medical peer, was the most successful in terms of both GP recruitment and retention. RM1, which used project staff for recruitment, was the least successful on both counts. However, direct comparison with the other two methods is difficult because neither was limited to recruiting just one GP from each practice. Under RM1, interested GPs were turned away from the study because of this exclusion criterion, three of whom were subsequently enrolled under RM3. A limitation of this observational study is that the recruitment approaches were not randomised, thus findings from comparisons are subject to bias.

Both RM1 and RM2 exclusively recruited GPs in Western Sydney, where 30 GPs had previously participated in the PACE pilot study in 2004−05, and so were excluded from subsequent study participation. It is thus possible that the most ‘research-interested’ GPs may have already participated in the pilot.

The recruitment area was one of the differences between the three approaches as only RM3 recruited in both Central and Western Sydney. A possible explanation of the success of RM3 could be differences between GPs from Central and Western Sydney. The available data indicate that the Central and Western Sydney recruits from RM3 were similar to each other as well as being similar to those recruited in RM1 and RM2 in terms of practice type, Fellow of the Royal Australian College of General Practitioners status and number of years in practice. As the characteristics of recruited GPs did not vary substantially by recruitment area, differing participant characteristics do not appear to explain the reason for the success of RM3. Recruiting in these two GP divisions alone and the small response rate (out of those contacted with a letter) may limit the generalisability of the findings to other GP divisions and the wider GP population.

In 2007–08 the proportion of female GPs in the Western Sydney Division was 30.8% (165/536) while in the Central Sydney Division it was 46.9% (298/635) (PHC RIS 2012). As all GPs in a Division were contacted, the above findings from the PHC RIS 2007–8 Annual Survey of Divisions can be compared with our data. The proportion of females in RM1 was 57% (12/21), in RM2 it was 46% (12/26) and in RM3 it was 65% (49/75). Overall, 60% of the GPs recruited were female while in the large sample-size Bettering the Evaluation of Care and Health study, 36.8% (351/953) were female (Britt et al. 2008). Thus, in this study, a possible explanation for the high proportions of female GPs recruited could be due to female GPs being more responsive to interventional research, or alternatively this was because all the PACE recruiters were female.

With the caveats outlined above, in the present study, RM3 was the approach associated with the largest proportion of GPs recruited out of those contacted. The greater success of RM3 appears to be attributable to who did the recruiting. Specifically, having a medical peer telephone a GP’s surgery greatly simplified access to the practice, a finding that has been previously noted (Veitch et al. 2001; Sellors et al. 2002; Ellis et al. 2007). Reasons why the medical-peer approach works is the ability to pass the practice ‘gatekeeper’, usually a receptionist, because of their respect and credibility (Heywood et al. 1995) and the strength of their personal networks and relationships (Asch et al. 2000). A potential drawback of this approach is that GPs may find it hard to say no to a peer at the beginning with time then wasted on a participant who later withdraws (Goodyear-Smith et al. 2009). In our study, this did not appear to be the case as RM3 was associated with both higher recruitment and retention of GPs. The perceived cost of recruitment by medical peers is often cited as a major reason for not recruiting GPs in this way. While initial costs may be greater using a medical peer, the recruitment timeframe would potentially be shortened, leading to cost savings that could offset this amount.

As recruitment delays can greatly increase the cost of research, a better understanding of strategies that improve recruitment and retention rates is a critical aspect of primary care research. This paper reinforces that research projects requiring GP participation need to carefully plan their timelines and budgets to account for the challenges of recruiting GPs into research studies. The findings indicate that having a medical peer as the primary contact may have enhanced recruitment of GPs to a trial of a highly participatory educational intervention.

None declared.