While the seven factors identified here may come as no surprise to anyone who has faced the task of recruiting entire medical practices for research studies, they have not previously been either explicitly identified as a group or demonstrated to be so successfully combined in one research study. In fact, most of the prominent studies of organizational behavior do not report much information on the methods used to recruit participant organizations. For example, one of the most well known of such recent studies – the National Survey of Physician Organizations (NSPO) – only reported the organizational response rate to its survey (70%), but nothing about the methods involved in obtaining agreement to participate [14
]. This study was limited to medical groups or IPAs (Independent Practice Associations) with more than 20 physicians, and found that medical groups were less likely than IPAs to respond to the survey (66% vs. 79%, P < .001).
Another study in the Minnesota region obtained a 90% response rate to surveys of the medical director and administrator of 172 individual clinics about their organizational structure [19
]. Perhaps this means that there is an additional R factor for Region
of the country, but the same research group more recently obtained only a 71% response rate from the administrators of 127 group practices for a survey about practice structure [9
]. Even a strongly hierarchical medical care organization such as the Veteran's Health Administration wasn't able to obtain high response rates to surveys of VA medical center quality managers and primary care administrators about their efforts to improve quality of care [20
]. Although the latter article at least reported some of the details of their survey methods, none of these or other studies of care delivery organizations provide enough information about their recruitment and survey methods to allow others to know, for example, whether any of the R-factors reported here were used.
The few studies reporting on recruitment of group practices note the benefit of recruitment through the group's physician leader or medical director, as was done in this study. McBride recruited 65% of eligible practices in the Midwest by dealing with the practice leader, but 54% were recruited through mailings to individual physicians [6
]. He recommended phone calls from study physicians to practice medical directors followed by recruitment meetings at the practice site. Kottke also compared different methods, finding that only 6% of individual family physicians and 2.7% of internists and cardiologists recruited by mail, with a follow-up phone call if interested, ended up participating in a smoking cessation trial [4
]. In both cases, the project had been endorsed by the respective local professional associations. However, when 11 groups were approached through their medical director on behalf of a local health plan, all 11 groups participated and a mail survey of physicians in these groups achieved an 86% response rate. Again, practice informational meetings were held to familiarize all personnel with the project. Although neither of these reports specifically discussed the R-factors noted in this study, most of them appear to have been involved, at least to some extent.
Two other reports provide some information to corroborate these observations and recommendations. Carey, et al [7
] report on a variety of aspects of conducting research in community practices in North Carolina and note several components that contributed to success:
1. "Direct recruitment of clinicians by clinicians,
2. Ongoing personal contact to maintain the relationship, and
3. Recognition of the value of the community clinicians' time."
Ganz, et al describe recruitment of what they call 'provider organizations' in California, although most of these groups appear to have been much less integrated than the medical groups described here [21
]. They recruited 71% of 174 provider organizations for a medical director survey and 71% of a subset of 51 for a randomized trial, reporting an average of five calls and 37 days to get initial agreement to participate in the trial (compared to two and 14 in this study).
Our experience and the literature suggest that it is very important to have a physician recruiter for physician subjects. Researchers without that degree would be well-advised to partner with a physician to do this recruitment, ideally one with a good local reputation and established relationships. Lacking those R's, however, an unknown physician will at least facilitate access and credibility.
This report of an apparently successful approach to recruiting entire medical groups for a research study does have some limitations. The practice of medicine in this region is unusually collective, both in having most physicians in relatively large groups and in having a relatively high degree of integration of the practices within most groups. There also may be a greater sense of community cooperation here. However, other than being of sufficient size to have enough depressed patients for this study, there is nothing about these 41 medical groups that would make them more responsive to recruitment for this study. Even "small" medical groups in this region have a designated medical director as a focus for recruitment and study coordination, perhaps in part because the high managed care penetration in this region virtually requires such an organized management. This local characteristic of medical groups may affect generalizability to some other regions with mostly small practices, although we find that even in such areas there is usually at least an informal physician leader, often the practice founder. Requirement of time for participation will be a similar issue for all groups, large or small, because of the pressure on any primary care practice and its leaders.
A greater limitation for generalizability may be the reputation and pre-existing relationship of the recruiting physician with many of the medical group leaders. The impending appearance of pay-for-performance may have contributed to an increased willingness of medical groups to participate in studies that will inform those efforts, but there was nothing specific about this study or the measures used that were tied directly to such efforts in the minds of recruitment subjects.
A discussion of practice recruitment for participation in research would not be complete without mentioning practice-based research networks (PBRNs). These existing aggregations of physicians and/or practices were developed in the 70's and have increased to the point of there being 111 identifiable PBRNs throughout the U.S. in 2003 [22
]. According to a report of a survey of 87 PBRN's from the AHRQ-funded PBRN Resource Center at the University of Indiana, they contained 2,724 practices caring for 14.7 million patients in 44 states and Puerto Rico. While these networks represent a valuable resource, they are usually small (average size of 4.7 physicians per practice), and many began or continue as aggregations of research-interested individual physicians rather than whole medical groups. They also may not fit geographically or demographically with the needs of many research studies, and they may not be willing or able to participate. Finally, this study is an example of a project that could not have used a PBRN, since eligibility required that they have outcome data in a public accountability set, and most were not members of the local PBRN.