Relationship of Volunteer Student-Run Clinic Experience with Future Primary Care and Underserved Practice

,


Introduction
The United States (US) faces a shortfall of primary care (family medicine, general internal medicine, and general pediatrics) physicians that will exceed 45,000 by the year 2025. 1 This shortage disproportionately affects the uninsured, racial or ethnic minorities, those living in rural or inner-city areas, and people of lower socio-economic status. 2 As of January 2013, there were 57 million people living in health professional shortage areas (HPSAs) -geographic areas with 3,500 or more people per primary care physician. 3ccess to primary care (PC) remains a critical component of health.While the Affordable Care Act has reduced the number of uninsured Americans, the influx of patients newly insured has further taxed an already burdened PC system. 2 There are currently not enough PC physicians in the medical education pipeline to meet the need.While many residency programs focus on PC careers and medically underserved communities, [4][5][6] medical student interest in PC remains low.The number of US medical school graduates matching into PC residencies in 2010 was 16-18%, 2 with as few as 2% of graduating students planning to enter PC internal medicine. 7Two national surveys found that while students may enter medical school intending to practice PC, their interest erodes as they progress through their medical education and training. 8,90][21] SRCs are also increasingly common; as of 2012, at least 72 medical schools had been reported to support at least one SRC. 22The effects of SRC participation and associations with career outcomes is unclear.Although an early observational study found a positive relationship between SRC participation and PC residency choice, 23 a more recent study found no relationship. 22For this study, we established a longitudinal student database for tracking of medical student residency choice and eventual specialty and practice setting.Using this cohort, we sought to examine the relationship between SRC participation during medical school and post-residency PC careers and practice in HPSAs.

Setting
The University of California, Davis (UC Davis) School of Medicine is a public school located in Sacramento, California.UC Davis has an established mission of promoting PC and improving workforce diversity with multiple programs and initiatives to meet that goal.UC Davis started its first SRC in 1972 and now supports seven SRCs that provide free or low-cost primary health services in Sacramento.The majority of first and second year students participate in the SRCs, with most students volunteering for one clinic per month (intermittent volunteers).Third and fourth year students occasionally participate but not as commonly; traditionally at UC Davis the SRC leadership is also left to the first and second year students in order to provide them with clinical experiences and extra-curricular opportunities during their pre-clinical years.A smaller subset devotes more time as clinic leaders.Volunteer faculty provide staffing, mentorship, and sponsorship at the clinics with some volunteering at different clinics and others mainly serving at one clinic.Undergraduate volunteers fill the roles of interpreters, patient advocates, receptionists, and lab workers while first and second year medical students provide medical care.The clinics operate on the weekends and provide a range of services including PC, acute, prenatal, child, and preventive health care.The seven clinics are located in different communities with some focusing on a specific population or at-risk groups.

Participants and Study Design
In order to examine SRC participation, we conducted a retrospective cohort study of all UC Davis School of Medicine graduates from 2004 to 2008 (N=454).We chose this timeframe because PC residency programs (family medicine, general internal medicine, and general pediatrics) are three years in duration, meaning graduates will have finished residency from 2007 to 2011.As a result, current outcome data is two to five years after the cohort has completed residency.
The UC Davis Institutional Review Board has provided exempt status to this study.

Data Collection Methods
A complete list of 2004 to 2008 UC Davis graduates was obtained from the registrar.Practice data were collected on all graduates as of the summer of 2012, the conclusion of the academic calendar year.All data were assessed by two independent reviewers.A third reviewer reconciled discordant findings and repeated data collection on a random sample of 20% of graduates.Variables were entered into a research database and identifying information was removed.A coded key was stored separately.Data on current career specialty and current practice setting were collected using a systematic internet-based search described below.
PC vs. Non-PC Careers: First, we obtained board certification information from online, publicly available records. 24,25This information was cross-referenced with UC Davis records on residency match specialty.Then, publicly available physician search websites (healthgrades.com,findadoctor.com)were used to locate individual graduates. 26,27When no match was found, we used a general internet search.The information was then re-cross-referenced with board certification data. 25Residency websites, and in some cases the graduates themselves, were contacted directly via email or phone to verify practice location because we were unable to find a practice location with the general search (n=7).
Practice Setting: We searched for practice location (address of clinic or hospital with which the graduate was currently affiliated) using physician search websites. 26,27When no match was found, a general internet search was used.The practice location address was then entered into the Department of Health and Human Services HPSA tool to categorize into HPSA or non-HPSA. 3raduate Characteristics: Demographic data (gender, race/ethnicity) and performance data (Medical College Admission Test [MCAT] scores, first time United States Medical Licensing Examination [USMLE] Steps 1 and 2 first time pass/fail status) were obtained from school records archived by the UC Davis registrar's office.Data on research experience during medical school, level of participation in SRCs, and languages spoken were abstracted from each graduate's Medical School Performance Evaluation (MSPE).The UC Davis Financial Aid Office provided debt at graduation and it was included as a continuous variable.
Completeness of Data: We attempted to obtain complete data for all variables.Missing data resulted from incomplete registrar records, item non-response (in the case of race/ethnicity), or missing MSPEs.Demographic data (gender or race/ethnicity) were missing for 18 of 454 graduates (4%).Clinic participation or research experience data were missing for 37 of 454 graduates (8.1%).Performance data (MCAT scores, USMLE Steps 1 and 2 first time pass/fail status) and dollar debt at graduation was 100% complete.We were unable to confirm practice specialty for 8 of 454 graduates (1.7%) and practice setting for 46 of 454 graduates (10.1%).All missing variables were excluded from analysis.

Main Measures
Two categorical outcome measures were constructed for each primary outcome: post-residency practice specialty and post-residency practice setting.Practice specialty was categorized as PC if the graduate was currently practicing family medicine, PC internal medicine, or general pediatrics.All other specialties were categorized as not PC.Post-residency practice setting was categorized as a HPSA or not.Although the focus of this study was post-residency practice, residency program type was also evaluated (family medicine, general internal medicine, general pediatrics, or specialty) in order to compare the percentage of graduates in PC practice.
We chose predictor variables based on previous studies and available data in medical school records.The key variable, SRC participation, was defined by the student's self-reported level of participation and categorized as: 1) none; 2) intermittent volunteer; or 3) committed volunteer.Committed volunteers devote more time (10-20 hours per month with at least two clinics per month) and gain greater exposure to the clinic and ambulatory role models.They also assist in clinic administration and are more engaged in patient and fellow student education compared to intermittent volunteers.The no participation group (none) was used as the reference.Other variables including gender, race, more than one language spoken, MCAT scores, USMLE Steps 1 and 2 first time pass/fail status, total dollar debt at graduation, and research experience during medical school were also recorded.Race was evaluated as both a categorical variable and as a binary variable but was included in the final model as a binary variable (under-represented minority status [African Americans, Mexican-Americans, Native Americans, and Pacific Islanders] or not).Gender, USMLE Steps 1 and 2 first time pass/fail status, more than one language spoken, and research experience were binary variables.MCAT score and debt at graduation were evaluated as continuous variables but only debt was included in the final model described below.

Statistical Analysis
In order to independently examine both primary outcome variables, we built two different logistic regression models.We started with models that included gender, race, MCAT score, USMLE Steps 1 and 2 first time pass/fail status, dollar debt at graduation, and research experience.We then performed manual backward selections, retaining variables if they were associated with PC careers or HPSAs at p<0.05 or if the variable was considered important based on past literature (e.g.debt).A goodness-of-fit test was performed on the full models and chi-square statistics were compared.A p-value of <0.05 was required for statistical significance in all cases.All statistical analysis was performed with statistical software package SPSS v. 20.

Overall Graduate Characteristics
A total of 454 medical students graduated from UC Davis from 2004 to 2008 (Table 1).This cohort included 207 male (46.4%) and 234 female (52.5%) graduates; 13 graduates had missing gender data.Graduates were 45.1% white and 15.6% under-represented minorities.Nearly one quarter of the graduates (23.1%) spoke more than one language.The mean MCAT score was 31 and was normally distributed; the first time USMLE Step 1 pass rate was 96.6% and the first time USMLE Step 2 pass rate was 95.7%.Total debt at graduation was not normally distributed; median debt was $113,139 with a range from $0 to $284,087.Almost half of the graduates (47.1%) participated in research during medical school.
Most graduates (68%) participated in at least one SRC during medical school ranging from 62-80% across all five years.Of all the graduates, 28.8% were committed SRC volunteers and 39.2% were intermittent SRC volunteers.Of the volunteers, 65.4% were women, 72.1% were from underrepresented minority groups.

Career Outcomes
Committed SRC participation was associated with post-residency PC practice (odds ratio [OR] 2.15, 95% confidence interval [CI] 1.18-3.75),while intermittent SRC participation was not (Table 2).Female gender was positively associated with PC practice (OR 1.93, 95% CI 1.2-3.11),while research during medical school was negatively associated (OR 0.48, 95% CI 0.32-0.77).Under-represented minority status, more than one language spoken, MCAT score, USMLE Steps 1 or 2 first time pass/fail status, and debt at graduation were not associated with PC practice.We attempted to examine the composite outcome of PC practice and practice in HPSAs, but the small number of these graduates (N=38) limited our ability to interpret the relationship.More than one language spoken was the only variable associated with practice in HPSAs (adjusted OR [AOR] 1.97, 95% CI 1.2-3.25,p=0.01) (Table 3).There was no association between practice in HPSAs and any level of SRC participation or other factors (gender, race, debt at graduation, participation in research, MCAT score, and USMLE Steps 1 or 2 first time pass/fail status).Because many fellowships are located in HPSAs, we adjusted for this variable in the final model but the AORs did not change significantly (Table 3).
At graduation from medical school, 200 of the 454 students (44%) entered a PC residency: 49 (10.7%) in family medicine, 81 (17.6%) in internal medicine, 70 (15.4%) in pediatrics, and one (0.2%) in medicine/pediatrics.Two to five years after residency, 120 of the 454 graduates (28%) remained in PC practice.The percentage of graduates in PC practice after residency varied across specialties and was lowest in internal medicine (Figure 1).Of the 80 graduates who completed internal medicine residency, only 27 (33%) remained in PC practice.By comparison, 44 (63%) of the graduates who completed a pediatrics residency and 47 (96%) of the graduates who completed a residency in family medicine remained in PC practice.After completing residency, 149 (37.2%) of UC Davis graduates were practicing in HPSAs; however, 13.9% of these were in subspecialty fellowship (Table 3).Thirty eight of the 120 graduates in PC practice (31%) were practicing in HPSAs.There was no statistical difference in the percentage of graduates in PC or HPSAs across the 5 years of the cohort (data not shown).

Discussion
This longitudinal cohort study demonstrates that UC Davis graduates who were committed SRC volunteers were more likely to remain in PC practice after residency.Contrary to our expectation, we did not find an association between SRC participation and eventual practice in HPSAs.Only more than one language spoken was associated with eventual practice in HPSAs, consistent with prior studies. 28,29Unlike prior literature, we found no association between underrepresented minority status and practice in HPSAs. 30,31We also found no association between debt and eventual PC practice.
In contrast to the considerable literature examining medical student residency choice and student surveys, 10,16,[32][33][34] we report actual practice patterns after residency from a longitudinal cohort of graduates.This and the use of self-reported   participation in SRCs as opposed to using only the presence of a SRC in a particular medical school may explain why our findings differ from Tong's cross-sectional study which found no association between the presence of a SRC and PC residency choice. 22As in prior studies, we found an association between female gender and PC practice, and an inverse relationship with research experience. 12While other studies have reported that under-represented minorities are more likely to choose PC residencies, 7,34 we observed only a trend with eventual PC practice that did not reach statistical significance, possibly due to a smaller sample size.We found no association between debt and eventual practice which while counterintuitive, is consistent with prior literature. 11,33,35Of note, however, most debt studies examining debt and career choice were completed in early 2000s, and median debt has continued to rise; the effects of this have not been robustly evaluated. 36More recent survey data from the United Kingdom, Australia, and New Zealand has more mixed findings. 37More research in this arena in the US is definitely needed, especially that evaluating actual practice patterns and not just residency choice.Not surprisingly, the percentage of UC Davis graduates in PC practice varies dramatically across the three PC disciplines and is lowest in graduates choosing internal medicine, consistent with previous literature. 12his highlights the importance of using actual post-residency practice patterns rather than residency choice when assessing medical student career outcomes.
Our study has limitations.The decision to practice in PC or in an underserved area is a complicated one which may be difficult to capture in a predictive model.In addition, the observed relationship between committed SRC participation and PC practice is statistically significant but modest.Our model for practice in HPSAs was also statistically less robust perhaps because we did not include factors such as graduate's family background or expected income which have been previously associated with practice patterns. 17,34Although our data collection was relatively complete (missing data < 10.1%), this database could contain some errors.For example, despite the multi-step process used to confirm practice address, some addresses may be mailing addresses and not actual practice location.This would underestimate the number of graduates practicing in HPSAs.Additionally, the location of a practice in an underserved area does not mean that the patient population is underserved.SRC participation was by self-report from MSPEs, which may overestimate the amount of true participation.We also measured career outcomes as close as two years post-residency which may miss later attrition from PC specialization greater than three years post-residency. 38,39Finally, the percentage of UC Davis graduates in PC is relatively high: 120 of 200 graduates who entered a PC residency (60%) remained in PC practice.In comparison, a recent analysis of the National Graduate Medical Education Census estimates that the national average is closer to 32%. 40This difference may affect the generalizability of our findings.
Why is SRC participation associated with eventual PC practice?Our observational design cannot determine whether the findings are due to unmeasured student differences or due to the experience as a committed SRC volunteer.The early exposure to rewarding longitudinal care experiences and PC mentors in SRCs may sustain a student's predisposition towards PC.PC practice is often perceived as having poor work-life balance and associated with a lack of prestige.Many graduates may also feel poorly equipped to enter PC careers due to primarily inpatient rotations during medical school. 39A lack of exposure to ambulatory experiences may not provide an opportunity for an interest to develop in the first place.We believe that exposure to SRCs can buffer against these strong influences. 14,16We believe our results could inform the development or further incorporation of existing SRCs into the preclinical experience of medical school.Institutions interested in promoting primary care and care for the underserved could also invest resources in supporting SRCs.
The results of this study have important implications for future research involving medical student career outcomes.They reinforce the notion that PC residency choice is only a surrogate marker for eventual PC practice.More research is needed to elucidate the influences on career choice, but future studies should examine practice patterns rather than residency choice alone. 21s one of the few studies to examine post-residency outcomes, [41][42] our findings suggest that intense exposure to rewarding PC experiences with underserved populations may be part of the strategy to strengthen the PC workforce, in addition to selecting incoming students inclined to PC careers.Supporting student interest in primary care and caring for the underserved may also help reinforce students entering primary care practice.Our methods include a longitudinal career database, a tool that could easily be developed by other institutions to understand the impact of their medical education program on the workforce.Such research would guide future efforts by policy makers and medical schools to strengthen the PC pipeline and workforce.
Journal of Student-Run Clinics | Relationship of Volunteer Student-Run Clinic Experience with Future Primary Care and Underserved Practice journalsrc.org| J Stud Run Clin 4;1 | 7

Table 1 .
Demographic information, performance data, and activities of UC Davis graduates 2004 to 2008 as of the summer of 2012 (N=454)

Table 2 .
Results of logistic regression examining PC practice (family medicine, general internal medicine, or general pediatrics) in UC Davis 2004 to 2008 after adjustment for selected graduate characteristics (N=427)

73 (0.96-3.12)
MCAT: Medical College Admissions Test; USMLE: United States Medical Licensing Examination; SRC: student-run clinic.*Bolded ORs represent those included in final model; PC model chi-square 28.8, degrees of freedom 6, p<0.05.†Self-identified as African American, Hispanic/Latino, Native American, or Pacific Islander.‡Type of participation was categorized as: none, intermittent volunteer (<10 hours/month), and committed volunteer (≥10 hours/month) with "none" as the reference category.Journal of Student-Run Clinics | Relationship of Volunteer Student-Run Clinic Experience with Future Primary Care and Underserved Practice journalsrc.org| J Stud Run Clin 4;1 | 6

Table 3 .
Results of logistic regression analysis examining practice in medically underserved areas in UC Davis 2004 to 2008 graduates after adjusting for selected graduate characteristic (N=401) UC Davis: University of California, Davis; MUA: medically underserved area; OR: odds ratio; CI: confidence interval; AOR: adjusted odds ratio; MCAT: Medical College Admissions Test; USMLE: United States Medical Licensing Examination; SRC: studentrun clinic.*Bolded ORs represent those included in final model; MUAs model chi-square 7.8, degrees of freedom 3, p <0.05.†Type of participation was categorized as: none, intermittent volunteer (<10 hours/month), or committed volunteer (≥10 hours/month) with "none" as the reference category.