Table 2.

Characteristics of included studies

Study	Country	Participants and sample size	Study design	Intervention(s)	Outcomes measured	Main description of results
Carson et al. [32]	Iceland, Ireland, Norway, Scotland, Sweden, Greenland	1046 rural health professionals (doctors, nurses, allied health professionals)	Cross-sectional	Selection of rural background or rurally-schooled students; rural health professional training (rural pipeline)	Intention to stay with same rural organisation for at least the next 2 years (or until retirement if intending to retire within 2 years)	There was a significant relationship (χ2 = 3.98, p < 0.05) between having a rural background and intending to stay with the same rurally located organisation. Spending the majority of school education in a rural area was also significantly associated with intention to stay (χ2 = 8.7, p < 0.01). Health professionals working in outer rural areas who had undertaken some of their health professional training in a rural area were more likely to intend to stay than those who had no rural training (χ2 = 4.22, p < 0.05)
Chauhan et al. [55]	Canada	642 rural physicians	Cross sectional	Any length of rural training during medical school	Intention to leave rural practice within the next 2 years	There was no significant association between having had at least one rural experience of any length during medical school (versus no rural rotations at all) and intention to leave rural practice within 2 years
Cogbill and Bintz [39]	USA	19 rural General Surgeons who worked for Gundersen Health System since it commenced in 1978	Case report	A rural general surgery network designed as a sustainable model for delivery of general surgery services within a large rural region of the midwestern USA (Minnesota, Wisconsin, Iowa)	Mean number of months that currently employed General Surgeons have been practising in Gundersen Health System network	The study showed that of the 19 rural General Surgeons ever employed in the network 21% (n = 4) have retired, 53% (n = 9) continue to practice in the network, and only 26% (n = 5) left the network before retirement. The nine currently employed General Surgeons have been practising in the network for a mean of 88.0 months (SD 83.6; range 24 to 288 months). Six have practiced in one location for more than 20 years
Fleming and Mathews [48]	Canada	391 physicians who were initially licensed to practice in Newfoundland-Labrador (NL) between 1993–2004	Cohort	Trained locally at Memorial University (MMG) versus trained elsewhere in Canada (CMG) versus provisionally licensed international medical graduates (IMGs (Prov)) versus fully licensed international medical graduates (IMGs (Full))	Hazard of remaining in NL	Overall, CMGs, IMGs (Prov) and IMGs (Full) were 1.99 (95% confidence interval [CI] 1.28–3.10), 2.11 (95% CI 1.50–3.00) and 1.86 (95% CI 1.19–2.90) times more likely to leave NL, respectively, than locally trained MMGs. Physician group was the only significant covariate in the Cox regression analysis
Gardiner et al. [40]	Australia	361 rural General Practitioners (GPs) in South Australia	Cohort	Group and individual coaching by qualified psychologists and 6 weeks of email coaching over a 3-year period	Retention rate in rural general practice at two time points, 3 years apart	Despite having a much higher intention to leave rural general practice before coaching, only 6% left after coaching. In other words, 94% of participants stayed compared with 80% of the general rural GP population [χ2 = 4.89, p = 0.027]
Gaski and Abelsen [41]	Norway	388 Graduating medical students	Cohort	Medical internships early sign up versus raffle in the study area versus a comparison area with similar workforce issues	% former interns in the area staying in (rural) county as at April 2014 (Finnmark versus another rural county)	The proportion of interns who signed up early who still worked as physicians in the study area (29%) was twice as high as among the regular interns (15%) and interns in the comparison area (14%). None of the 59 physicians who had been early signup interns worked in any of the 15 remote municipalities in the study area
Gorsche and Woloschuk [42]	Canada	29 rural doctors working in an established rural practice in Alberta with written support from their regional medical director, who enrolled in the skills enrichment program March 2001–March 2005 and their 29 matched controls	Cohort	Skills enrichment program comprising training for acquisition of or maintenance of existing skills of at least 2 weeks’ duration but less than 1 year, with preceptor funding and locum support available and funded up to $80,000 per annum (pro rata)	Rural practice retention 5 years after the program	After 5 years, all 29/29 (100%) physicians in the enrichment group remained in rural practice compared with 22/29 (71%) of physicians who did not participate in the enrichment program: (Relative Risk 1.31; 95% CI 1.06–1.62). Two enrichment participants who left the province and their controls were excluded from analyses
Jamar et al. [56]	Australia	74 University of Adelaide Rural Clinical School domestic medical students 2003–2010	Cohort	Rural Clinical School education exposure comprising completing the whole of fifth (penultimate) year of clinical studies at a rural location	% of graduates who spent no time, 3 to 18 m and 2+ years in a rural area since graduation	Of the 72 survey responses analysed, 44% (n = 32) spent no time in a rural location, 28% (n = 20) spent 3 to 18 months in a rural area and another 28% (n = 20) spent two or more years in a rural area since graduation
Jamieson et al. [53]	Canada	480 University of British Columbia Family Medicine residency graduation cohorts from 1990 to 2007	Cohort	Rurally distributed postgraduate training sites versus training in metropolitan teaching centers	Retention in current community for more than 1 year	Amongst those who had been in their current community for more than 1 year, those who had postgraduate training in smaller rurally distributed sites were 36 times more likely to be working in a rural or regional practice than those trained in metropolitan sites (95% CI 12.2–108.5)
Johnson et al. [31]	Australia	397 dental graduates of University of Sydney 2009–2013	Cohort	Rural Clinical Placement Program comprising an opt in 1-month clinical placement in rural NSW in final year of dental school; previous rural life experience	Working in a rural location in both 2015 and 2017 versus not working in a rural location in both 2015 and 2017	Having had some form of rural experience prior to the rural clinical placement (PR 3.75 95% CI 2.75–5.11) and pre‐placement rural intentions (PR 3.54 95% CI 2.25–5.57) were significant independent predictors of an increased likelihood of working in a rural location in both 2015 and 2017
Kahn et al. [49]	USA	47 IMG physicians participating in Conrad program 1995–2003 who were assigned to a rural community for return-of-service	Cohort	Conrad program which allows thirty J-1 visa waivers each year in Washington state with participants having a ROS obligation of 3 years for primary care physicians and 5 years for specialists	Retention rate in rural areas after completing rural J-1 visa waiver ROS obligation	Of the 47 physicians who undertook their periods of service in a rural area, only 12 (26%) stayed in a rural area following completion of the ROS obligation
Kwan et al. [38]	Australia	729 medical graduates of University of Queensland (2002–2011)	Cohort	Rural clinical exposure for one (RCS-1) or two (RCS-2) years; bonded scholarship; rural background vs metropolitan background	Long term rural practice (LTRP) defined as ≥ 50% of the graduates’ primary place of practice since graduation being rural	Independent predictors of LTRP were Rural Background (OR 2.10 [95% CI 1.37 ± 3.20]), RCS-1 (OR 2.85 [95% CI 1.77 ± 4.58]), RCS-2 (OR 5.38 [95% CI 3.15 ± 9.20]) and having a bonded scholarship (OR 2.11 [95% CI 1.19 ± .76])
Li et al. [57]	Australia	1117 rural GPs who participated in the Medicine in Australia: Balancing Employment and Life (MABEL) survey in 2009	Cross-sectional and discrete choice experiment	Locum relief guarantee, retention payments, rural skills loading, family isolation/secondary school costs or retention grants for existing rural doctors	Probability of attracting rural General Practitioners to (hypothetically) choose to stay in rural practice	Increasing the level of locum relief guarantee, GP retention payments and rural skills loading from zero to the middle and high levels was associated with increased rural GP retention. In order from highest to lowest effect on retention were: guaranteed locum coverage for 6 weeks every 12 months (β = 1.51); 50% increase in retention payments (β = 1.36); guaranteed locum coverage for 4 weeks every 12 months (β = 0.85); rural skills loading payment increase by 20% (β = 0.82); 25% increase in retention payments (β = 0.54); and a rural skills loading payment increase by 10% (β = 0.33). One-quarter of rural GPs were not influenced by the rural incentive packages
Mathews et al. [50]	Canada	60 physicians who were trained at Memorial University Newfoundland and held ROS (Return-of-Service) agreements (1997 to 2009) and started practice in Newfoundland and Labrador (NL) 2000–2005 compared with all 67 other NL physicians who started practice in NL between 2000 and 2005	Cohort	Opt-in ROS agreements offered to medical students and junior doctors which included two types of bursaries: Family Medicine Bursary and Special Funded Residency Position (offering postgraduate training positions which are usually accepted by physicians who were unable to secure a position through the usual application process)	Retention of physicians in the Canadian province of NL at the end of the follow-up period (December 31, 2010)	Whether or not Memorial University Newfoundland-trained (MUN-trained) physicians received a ROS bursary was a significant predictor of leaving NL. ROS physicians were 3.2 (95% CI 1.4–7.1) times less likely to leave NL than non-ROS physicians. Amongst the 60 ROS and 67 non-ROS MUN-trained physicians, 10 (16.9%) ROS versus 28 (41.8%) non-ROS physicians left NL province by 2010 (p = 0.004)
McGrail and Humphreys [36]	Australia	3782 responses from GPs who responded to Medicine in Australia: Balancing Employment and Life (MABEL) survey at least twice between 2008 and 2012 (inclusive)	Cohort	Training location (Australian trained vs non- Australian trained); restrictions on access to provider numbers related to geographic location for International Medical Graduates (IMGs)	Annual location retention rates in regional, rural and remote areas and odds ratios for leaving a rural area	There was no significant difference in the risk of leaving rural practice for IMGs compared to Australian non-restricted graduates. This was true for IMGs whether or not they had restrictions on access to provider numbers limiting where they could practice
McGrail et al. [58]	Australia	610 Medicine in Australia: Balancing Employment and Life (MABEL) survey respondents (2008–2014 inclusive) who had completed GP training and were transitioning to independent practice	Cohort	Vocational training location (rural or metro); rural/metro origin; rural bonding (being contracted to work for part of their early career in a rural location)	Proportions of GPs working in rural and metropolitan locations during each of the first 4 years following completion of GP vocational training; proportions of rurally-trained GPs working in the same or a different rural community from that in which they completed their vocational training	The rural training pathway, regardless of childhood location, was extremely strongly associated with subsequent rural practice (ORs ranged from 29 to 92 in the first 4 years following completing GP training). The odds of rural practice for the rural training cohorts of GPs decreased with time. Rural bonding (OR 3.5–5.1) and rural origin (ORs 2.0–4.1) were also positively and significantly associated with rural practice in each of the first 4 years following completing GP training
Murray et al. [28]	USA	60 nursing staff employed by Bassett Medical Center in rural upstate New York, USA	Cohort	A Partnership for Nursing Opportunities opt-in Program which involved Bassett Medical Center collaborating with two different local colleges to better design postgraduate degree pathways that were attractive to nurse employees. The intervention included: fully paid tuition with return-of-service; flexible scheduling arrangements supporting full and part-time paid work; local teaching; academic advisor	Vacancy rates and annual turnover rates of nurses from the rurally located Bassett Medical Center	Licensed Practice Nurses turnover decreased from 16.8% in 2005 to 6.8% in 2009. There was no trend in reduced annual turnover evident for Registered Nurses, but Registered Nurse vacancy rates fell from 16.5% in 2005 to 4.3% in 2009
Nilsen et al. [29]	Norway	159 nursing students doing a bachelor’s degree course in Finnmark University and who graduated in 2002, 2004 and 2005	Cohort	Bachelor of Nursing program which offered off-campus training in rural areas near the students’ place of residence and using more flexible and team-based learning methods as an alternative to the usual on-campus training at a regional centre	Retention rate in Finnmark county (living and working) for at least 4 years after completing the study program	Off-campus training was associated with a considerably higher retention rate n = 37/40 (92.5%) in Finnmark county for graduates compared to on-campus training n = 83/119 (70%). The majority of the nurses who trained off-campus worked in rural or remote communities where nurse shortages were more pronounced
Norbye and Skaalvik [30]	Norway	233 Registered Nurses training at the Arctic University of Norway by a decentralised nursing education model	Cohort	Decentralized nursing education allowing part-time postgraduate study at proximate study centres to accommodate students’ family and work responsibilities	Nurse retention rates in rural areas after graduation from postgraduate studies	The majority [n = 190/233 (81.6%)] of nurses completing postgraduate education using the decentralised nursing education model continued to work in rural areas
Opoku et al. [51]	USA	240 physicians with an initial rural Nebraska county practice location and enrolled in J-1 visa or Nebraska Loan Repayment Program 1996–2012 inclusive	Cohort	J-1 visa program (waiver of the requirement for IMGs on J-1 visas to return to their home country in exchange for service in a health professional shortage area) and state loan repayment programs (loaning physicians up to $40,000 per annum for up to 3 years in exchange for service)	Hazard of leaving rural Nebraska (hazard ratio), average length of stay (years)	The average length of stay in rural Nebraska for J-1 visa waiver and state loan repayment physicians were 4.1 and 8.1 years, respectively (SD = 0.27 and 0.47, respectively) and for physicians who had completed the minimum obligatory period were 5.6 and 9.7 years (SD = 0.27 and 0.43, respectively) The likelihood of departure from rural Nebraska was higher for beneficiaries of the J-1 visa waiver program (HR = 3.76, 95% CI 2.02–6.98) compared to those receiving state loans
Patterson et al. [59]	USA	Family Medicine “1–2” Rural Training Track (RTT) residency programs that responded to survey providing information about 253 enrolled residents who graduated 2008–2014 inclusive	Cohort	“1–2” RTT Family Medicine residency program which includes up to 1 year of urban training and 2 years of rural training	% of RTT graduates who were practising in a rural location at 1 year, 2 years, and 3 years after finishing residency	Just under one-third (32.8%) of RTT graduates practiced in rural areas in their 1st year post graduation. The percentage practising in a rural location increased thereafter such that the percentage was above 35% in most of the 7 years post-graduation
Pepper et al. [61]	USA	693 physicians practising in the State of Wyoming and responding to a survey	Cross-sectional	Variation in malpractice insurance costs between different USA states	Intention to change practice in Wyoming in next 10 years: relocate within Wyoming, move out of Wyoming, stop patient care	The high cost of malpractice insurance in Wyoming was associated with planning to move out of state instead of moving within the state (OR 22, 95% CI 1.7–287.9). Of those planning to leave Wyoming, just under one half (44%) gave malpractice insurance costs as a reason
Playford et al. [60]	Australia	915 University of Western Australia graduates 2004–2010	Cohort	Training medical students at a Rural Clinical School; selecting 25% of students into medicine based on having a rural background; recipients of a Bonded Medical Place at medical school (a place in medical school in exchange for return of service in a district of workforce shortage) or a Medical Rural Bonded Scholarship (recipients of a bursary during medical school in exchange for return of service in a rural area following graduation)	Mean cumulative duration (in years) in rural practice, ratio of means	Rural origin Rural Clinical School participants had a cumulative duration of rural practice over 5 times higher than the urban origin/urban training reference group (Ratio of means 5.4, 95% CI 4.3–6.8). Urban origin and Rural Clinical School (Ratio of means 2.2, 95% CI 1.8–2.7) and rural origin and no Rural Clinical School (Ratio of means 2.9, 95% CI 2.2–3.8). Bonded graduates had longer mean cumulative duration of rural practice than non-bonded graduates (p < 0.0001)
Rabinowitz et al. [43]	USA	92 Jefferson Medical College graduates 1978–1986 inclusive who initially practised Family Medicine in a rural area	Cohort	Physician Shortage Area Program (PSAP) of Jefferson Medical College which is medical school rural program with features including: selection based on rural background, commitment to rural practice, faculty mentoring and career guidance, 6 week clinical placement in a small town in 3rd year, encouragement to take up rural preceptorship in 4th year and expectation of completing family medicine residency training	Numbers of PSAP and non-PSAP graduates who originally entered rural family medicine and were still practicing family medicine in the same rural area (including in five adjacent counties) in 2011	Of the 37 PSAP graduates who originally entered rural family medicine, 26 (70.3%) were still practicing family medicine in the same rural area in 2011. A significantly smaller proportion of the 52 non-PSAP graduates (n = 24, 46.2%) were practising in the same rural area (p = 0.02)
Renner et al. [33]	USA	66 health professionals who were participants in one of three Colorado loan repayment programs 1992–2007 and had fulfilled terms of service	Cohort	The Colorado loan repayment programs had varying rates of repayment of recipients’ educational debts (maximum of $35,000 per annum), duration of return of service commitments (0, 1 and 2 years) and eligibility based on health profession. Analysis was for participants in any of the 3 programs, with no comparison of outcomes of the different programs	Retention rate in a rural community, retention rate in the same rural community after completing their return of service commitment, percentage retained in the same rural community for 0–1, 2–4 and 5 or more years	Of the loan repayment recipients who had completed their service commitment at the time of the survey, 27 (64%) of the rural participants were still practising in a rural community compared to 23 (96%) of the urban participants who were still practicing in an urban community. Rural retention rates were not associated with past attendance at rural high schools or by intention to practice in a rural community regardless of loan repayment. Of the 36 loan repayment recipients who were still at their original site after completing their terms of service or having their loan paid off, 21 were rural participants. Of these who stayed, 10 (47.6%) had stayed 0–1 additional years, 8 (38.1%) had stayed 2–4 years, and 3 (14.3%) had stayed for 5 years or longer
Robinson and Slaney [44]	Australia	57 GP registrars trained in the rurally located Bogong Regional Training Network 2004–2009, Victoria	Cohort	Decentralised model of GP training in rural Victoria	Retention rate in rural general practice following training in Bogong region; retention rate in Bogong region following training in Bogong region	More than 42% of the GPs who had completed their GP vocational training remained in rural general practice and 32% remained in the Bogong region
Rodney et al. [45]	USA	80 Obstetrics and Gynaecology Family Medicine fellows 1992–2010 inclusive	Cohort	1-year post-residency obstetrics fellowship undertaken in a rural location	Service in a rural community for at least 2 years	Rural service of at least 2 years occurred among 47/74 (64%) of the graduates
Ross [46]	USA	62 graduates of a rural Family Medicine residency program in Oregon (Cascades East Family Medicine Residency Program) 1994–2009 inclusive	Cohort	Rural residency in Oregon undertaken in a community of population size 42,000	Length of stay in first practice location, length of stay in current (2009) practice location	Graduates spent a mean 3.5 years at their first practice location and 3.7 years at their current practice location. Only 40% (n = 25/62) of graduates had relocated their practices at least once since graduation. Half of all program graduates remained currently working in rural communities (defined as having a population size of less than 25,000 and located more than 25 miles from major centres)
Russell et al. [37]	Australia	2783 rural GPs working in New South Wales 2003–2012, (excludes locums, GPs in border towns or offshore and GP Registrars)	Cohort	Holding rights to admit patients to the local hospital as a Visiting Medical Officer, having restrictions on access to a provider number (conditional registration)	Hazard of leaving a rural community for more than 3 months	Rural NSW GPs with conditional registration were more likely to leave (HR 1.49, 95% CI 1.24–1.79) a rural community than those with no such restriction. Having no public hospital admitting rights (HR 1.49, 95% CI 1.30–1.71) was associated with a higher risk of leaving rural NSW communities
Straume et al. [52]	Norway	76 postgraduate trainee physicians in Family Medicine or Public Health 1995–2008 inclusive	Cohort	A 5-year postgraduate training model offered to interns in either a Family Medicine or Public Health training group which included professional support provided through group tutorials for 2–3 years, in-service training opportunities provided in rural areas and completion of compulsory courses	Retention rates in Finnmark 5 years after completion of compulsory group tutorage component of the postgraduate training	Amongst the 15 Public Health doctors who had completed the group tutorage more than 5 years ago, 10 (66.7%) were still working in the county. Overall 28/40 (70%) of the Public Health graduates were still working in Finnmark in 2009. Amongst the 37 Family Medicine doctors who had completed the group tutorage more than 5 years ago, 24 (64.9%) were still working in the county. Overall 53/72 (73.6%) of the Public Health graduates were still working in Finnmark in 2009
Wearne et al. [47]	Australia	24 General Practice Registrars enrolled in the Remote Vocational Training Scheme (RVTS) 1999–2005 inclusive	Cohort	RVTS program which trains doctors in general practice in remote communities using distance education and supervision	Retention rates of RVTS graduates still working in rural areas	The majority [n = 17/21 (81%)] of RVTS graduates continue to work in rural areas (defined as Rural, Remote and Metropolitan Areas 3–7) 2–8 years after completing the program
Woolley et al. [54]	Australia	529 James Cook University (JCU) Queensland graduates 2005–2011 inclusive	Cohort	JCU Rural Clinical School (Darwin), rural generalist (residency) training, outer regional or remote location for internship	Prevalence odds ratios (POR) for practising in a remote location for 1 year or more from Postgraduate Years (PGY) 4–10	Forty-seven (8.9%) of the 529 JCU medical graduates in the first seven cohorts had practised for at least 1 year in a remote location between PGY 4 and 10. The likelihood of JCU medical graduates practising in a ‘remote’ location from PGY 4 to 10 was associated with undertaking rural generalist training (p < 0.001; POR = 17.0), attending the Darwin clinical school in years 5–6 (p = 0.005; POR = 4.7) and undertaking an internship based in an outer-regional or remote community (p = 0.006; POR = 3.5)
Yong et al. [35]	Australia	Australian GPs responding to the Medicine in Australia, Balancing Employment and Life (MABEL) survey 2008–2014 inclusive and providing data for 4,822 exits from rural locations that were eligible for receiving General Practice Rural Incentives Program (GPRIP) incentive payments	Cohort	Changes to the GPRIP policy made some locations newly eligible for rural retention incentives and increased retention incentives for other rural locations already eligible	GP exits (turnover) from always eligible locations versus newly eligible locations under GPRIP	For both newly eligible and always eligible rural locations, GPRIP appeared to have no statistically significant effects on GP relocation exits (turnover)
Zhou [34]	USA	Physicians licensed in North Carolina and working in rural areas. Sample size for the rural retention simulation study was unclear. The overall study sample was 29,908 unique physicians working in North Carolina contributing 165,688 person-year observations	Cohort and simulation	Loan forgiveness programs, increased physician reimbursement rate, change in the composition of health care providers: changes to mid-level practitioner supply, increase supply of RNs, Medicaid and Medicare expansion in rural areas	Retention in rural areas of North Carolina (excludes becoming inactive, moving out of North Carolina)	Simulated loan forgiveness programs (representing an almost doubling of income for rural physicians) was associated with a small (up to 3%) decrease in the probability of them moving away from the same rural area after 1 or 2 years of service, while a 5% increase in rural county physician salary decreased the probability of moving away from the same rural area after 2 years by 7%. There was a decreasing marginal return on the probability of moving away from the same rural area with salary increases of 10 and 20%. A 5% increase in mid-level practitioners in rural counties increased the probability of rural physicians moving away after 2 years by 8%, while a 5% increase in RNs in rural counties decreased the likelihood of physicians leaving rural areas by 3%. Medicaid and Medicare expansion in rural areas were associated with increased physician movement away from their rural county of service after 2 years by 10% and 11% respectively