The results of this study support the hypothesis that training at regional basic science campuses influences IUSM students to pursue primary care careers near the regional campuses they attended. For the graduating classes of 1988–1997, attending a regional campus was a significant predictor of both medical specialty choice and practice location choice in logistic regression models that incorporated several covariates known to influence these career decisions. According to the regression models, four of the regional campuses were significantly more likely than Indianapolis to produce family practitioners, and five of the regional campuses were significantly more likely than the others to have former students practicing in the region. When analyzed collectively, attendance at any regional campus was a significant predictor of a primary care practice located outside the Indianapolis metropolitan area. These findings suggest that the regional campus environment during the first two years of medical school predisposes some students to pursue different career paths than those exposed only to academic medical centers. In other words, the preclinical training site can influence students' career choice independently of other variables.
Predicting career decisions in primary care medicine is a complex challenge involving numerous interacting variables. According to the Bland-Meurer Model of medical career decision-making, three principal factors influence specialty choice: (1) student characteristics, (2) medical school characteristics, and (3) students' perceptions of medical specialties . Most studies have focused on the student characteristics that portend a primary care residency choice. Lawson and Hoban  reviewed six multivariate studies and found that students who go into primary care tend to be older than their classmates, female, belong to an underserved minority, have parents of lower socioeconomic status, have a rural hometown, lower MCAT scores, lower ratio of educational debt to expected income, and decided their specialty preference before medical school.
Where young physicians choose to establish their practices is likewise affected by the personal characteristics they bring as students. Having a rural or small town background seems especially influential. Laven and Wilkinson  identified 12 case-control or cohort studies that made quantitative comparisons between rural and urban doctors. They found that rural background was associated with rural practice in 10 of the 12 studies, and that students with a rural background were about twice as likely (OR = 2.0 to 2.5) to establish a rural practice compared to other students. In a previous study of IUSM graduates, Indiana physicians from non-metro hometowns were 4.7 times more likely to choose a non-metro practice location compared to their peers from metro hometowns, adjusting for age and sex . The authors concluded that Indiana physicians from small hometowns have a strong preference to practice in regions similar to their hometowns if not actually near their hometowns.
By including several of these student characteristics as covariates in the regression models, we effectively controlled for their confounding influence on career choice to reveal an independent effect of regional campus attendance. As far as we are aware, no other study has examined the role of regional basic science campuses in this regard. Yet placed in their broader context, our findings are consistent with previous studies of medical students' career decisions. A variety of medical school programs designed to foster interest in primary care medicine have been shown to have a beneficial impact on the number of students entering primary care fields, often in medically underserved communities [10–12, 18–20]. A key factor in the success of these programs appears to be primary care experiences during training, especially in community-based practice settings. Even experiences as early as the first year of medical school seem to positively impact residency choice . However, Rabinowitz et al.  noted that students with rural backgrounds and early intentions to enter primary care were almost as likely to become rural primary care physicians as were students with similar characteristics but exposed to additional curricular experiences in primary care (e.g., an elective senior-year rural family medicine preceptorship). These authors concluded that a student's background and early career plans are the most important determinants of a career in primary care medicine, but special curricular experiences and other factors can enhance this outcome.
In the WWAMI program (acronym for Washington, Wyoming, Alaska, Montana, Idaho), students complete the first-year basic science coursework at their home state campus, and then move to the main Seattle campus for their second year. Students complete the 3rd and 4th year clinical rotations at sites of their choosing in the five-state region. Graduates of this program are reported to have higher rates of return to their home states for practice than the national average . Although not directly comparable to the IUSM system, the WWAMI program does share certain features (e.g., preclinical training at regional campuses) and produces similar outcomes.
It is not intuitively obvious why attending a regional campus for the first two years of medical school would be conducive to a primary care career. Unlike schools with clinical branch campuses, where 3rd and 4th year students receive much of their clinical training from community physicians , the clinical education at IUSM's regional campuses is largely limited to the first- and second-year Introduction to Medicine courses. Nevertheless, according to the regression models, IUSM students who attended Evansville, South Bend, Muncie, or Terre Haute had a disproportionate propensity to become family physicians relative to their Indianapolis peers, adjusting for seven covariates known to affect specialty choice. What is it about these four regional campuses that could promote such an outcome? The formal curriculum is unlikely to be responsible because it is essentially the same at all sites. But each regional campus does have its own unique training environment, characterized in part by its network of connections with the local clinical community. This aspect is especially important because IUSM students at the regional campuses tend to have greater exposure to family physicians and other primary care providers than do their 1st and 2nd year counterparts at the Indianapolis campus, who tend to have greater exposure to specialists. Frequent interactions with these physician educators early in medical school may help to shape positive attitudes towards the primary care fields and family medicine in particular. Perhaps the four regional campuses in question have a greater involvement of family physicians in their educational programs than do the other campuses. Further studies will be needed to ascertain whether this possibility or other factors are responsible for the pronounced effect on specialty choice.
Equally intriguing is the pronounced effect of regional campus attendance on practice location choice. Our regression models predicted that IUSM students who attended Terre Haute, South Bend, Bloomington, West Lafayette, or Evansville were much more likely to return and practice in the surrounding area than were the students who attended other regional campuses, adjusting for seven influential covariates, including hometown location. The study cohort completed medical school in the decade 1988–1997, which means that graduates would have finished their specialty training and established their practices in the approximate period 1991–2003, assuming 3–6 years of postgraduate training. Because the information about medical specialty and practice location was drawn from the 2003 AMA Physician Masterfile, our findings should be viewed as a "snapshot" of the cohort as it existed in 2003. The state of affairs before or after this year is unknown. However, given the age of the cohort, it is reasonable to assume that most of the graduates were still in their first practice locations in 2003, and that relatively few had moved or changed their specialties.
Why would a student who attended a regional campus be drawn back to the area several years later to establish a practice? Proximity to the student's hometown cannot be the reason because hometown location was controlled for in the regression models. Other influences must be responsible. Relatively little is known about how new physicians decide initial practice locations, but in their survey of third-year family medicine residents, Costa et al.  found that spousal influence and favorable community characteristics were the most important determinants of location choice. We speculate that some of the IUSM students who attend regional campuses develop an affinity for these smaller communities and the family-friendly lifestyles they engender. As the students become familiar with the local hospitals, physicians, and patients, they come to appreciate what it might be like to practice in the community and envision similar opportunities for themselves. After graduation, these favorable impressions may influence their choice of practice location.
In some cases, a graduate's choice of a particular residency program may signal his or her intention to practice in a given community. For example, it is likely that several of the IUSM graduates who entered family medicine residency programs in South Bend and Evansville had previously decided to establish their practices in those regions. In such instances, residency location choice would be a proxy for practice location choice, which would confound the statistical relationship between independent and dependent variables in the regression models. This may explain why the South Bend and Evansville campuses were no longer significant predictors of practice location choice when residency location was included as a covariate. It is also possible that residency training in those two cities had a strong and independent effect on practice location choice, thus obscuring the effect of regional campus training.
This was an observational study with no provision to randomly assign students to a campus. We therefore had to rely on multivariate regression techniques to adjust for any biases that may have resulted from the non-random assignment. No information was available about the students' campus preferences, so we could not adjust for this variable. In theory, some students may have self-selected to attend certain regional campuses based on their desire to eventually practice primary care medicine in those same regions. In practical terms, however, this seems unlikely to fully account for our findings because relatively few entering students self-select to attend any regional campus except Bloomington. Moreover, the fact that a student preferences a regional campus as his/her first choice assignment does not necessarily imply an early career decision. There are a variety of other reasons why a student might preference a particular regional campus, including its association with the student's undergraduate college, its unique learning environment and small class size (e.g., Problem-Based Learning and other non-traditional teaching modalities are offered at some campuses), and economic incentives (e.g., some of the regional campuses offer scholarship inducements to attend). Evidence that self-selection is unlikely to explain our results can be seen from the outcomes at specific campuses (Additional files 1 and 2). For example, the Terre Haute campus attracts the fewest students of any regional campus (an average of only one student per year) and yet it produces a disproportionate number of graduates who return to practice family medicine in the area. Conversely, the Gary campus attracts a third of its class (an average of seven students per year) and yet it does not have a significant impact on the supply of local primary care providers. Despite these observations, we cannot rule out the possibility that self-selection may have confounded our findings to some degree. No information was available about the students' debt load or their specialty preferences prior to medical school, both of which have been implicated as factors in career decision-making [10, 23].
As mentioned previously, the students' clinical experiences at certain regional campuses, and the particular kinds of interactions they have with local physicians, may account for much of their predilection toward primary care careers. In our judgment, fully understanding how the clinical experiences at regional campuses differ from those at Indianapolis is key to explaining the campus effect.
Our study cohort completed medical school approximately 10–20 years ago, at a time when the economic forces and practice conditions affecting new physicians were somewhat different than they are now. In the intervening years, the composition of the IUSM student body has changed as well, with greater ethnic diversity and gender balance. How the career decisions of today's medical students are being shaped by their regional campus experiences remains to be determined. Other outcomes may manifest in a later cohort of graduates. Finally, IUSM is a large, midwestern medical school with educational traditions and demographic characteristics that may differ substantially from other schools in other locales. Our findings may not generalize to other student populations.