Interventions For Health Workforce Retention In Rural And Remote Areas .

Russell et al. Hum Resour Health(2021) 19:103Page 3 of 24Table 1 Inclusion and exclusion criteriaIncludeExcludeYear published2010 or laterPrior to 2010Participant typesMedical doctors, nurses, midwives, pharmacists, allied healthprofessions, or health professionals generallyOther types of workers in the health sector or in other sectorsCountriesHigh-income country as per World Bank criteria [23], or ifmixed income countries then data for high-income countriesreported separatelyNot high income or not reported separatelyExposureInterventions, support measures or incentive programs imple‑mented (or simulated) with the intention of increasing reten‑tion or reducing turnoverNo intervention (or simulation) designed to impact on retentionor turnoverSettingRural or remote as defined by ASGS (Australia) or equivalentnational classification system or study’s own description ofbeing rural or remoteNot rural or remoteOutcomesPRIMARY: Mean or median length of employment; survival prob‑abilities; hazard, odds or relative risk ratios for staying/leavingrural; stability rates; settlement rates. (profile retention inrural/remote area or community) SECONDARY: Vacancy rates;unfilled positions; turnover numbers or rates; attrition or wast‑age rates; rate of leaving before end of contract; intention tostay/leave; intention to returnLack of quantifiable primary or secondary outcomes, job satis‑factionStudy typesAnalytical observational studies (prospective and retrospectiveQualitative studiescohort studies, case–control studies, cross-sectional studies);descriptive observational studies (case series, descriptive crosssectional studies)LanguageEnglishAssessment of methodological qualityEligible studies were critically appraised and scored bytwo independent reviewers for methodological qualityusing the appropriate JBI critical appraisal instrument[25]. Disagreements were resolved through discussion,or with a third reviewer.Data extractionData were extracted by two independent reviewersusing the standardized data extraction tool from JBISUMARI. Study citation details, study objective, participant information, details of the setting/context, detailsof the retention intervention, and study results for therelevant outcomes were extracted.Data synthesisThe structure of the narrative synthesis of extracteddata was based on categories for rural health workforceinterventions used in the WHO Global Policy Recommendations [26]: education; regulatory interventions;financial incentives; and personal and professional support. A further category—health systems—was added,as proposed by Putri et al. [27].ResultsStudy selectionThe search strategy identified 2592 papers, with a further 57 papers identified from other sources (Fig. 1).Non-EnglishAfter duplicates were removed 2043 papers remained.1901 papers were excluded by title and abstract screening and 142 articles underwent full text assessment. 108were excluded, leaving 34 articles. The main reasons forexclusion on full text review were a lack of quantifiableretention outcome measures, no intervention or ineligible study type (Additional file 2).Methodological qualityMost were cohort studies (29/34, 85%). Methodological quality of included studies was generally low (Additional file 3). The median score for included cohortstudies was 16 out of a maximum of 22 (interquartile range 13–20). Only one-third of included studiesapplied appropriate statistical analysis, with less thanhalf adjusting for key potential confounders. Manystudies had no comparator group.Study characteristicsStudy characteristics, participants and sample size,interventions, outcome measures and main findings areshown in Table 2. Of the 34 included studies, 13 (38%)were from Australia, 11 (32%) from USA, five from Canada and five originated from Nordic and nearby northernEuropean countries. Most (n 28, 82%) studies exclusively examined retention of doctors. Three studies wereexclusively of nurses [28–30], one exclusively of dentists

Iden fica onRussell et al. Hum Resour Health(2021) 19:103Page 4 of 24Records iden fied throughdatabase searching(n 2592)Addi onal records iden fiedthrough other sources(n 57)ScreeningRecords a er duplicates removed(n 2043)Titles and abstractsscreened(n 2043)IncludedEligibilityFull-text ar cles assessedfor eligibility(n 142)Studies included inquan ta ve synthesis(n 34)Records excluded(n 1901)Full-text ar cles excluded(n 108)Reasons (Addi onal file 2):Outcome not reten on (n 49)No interven on(n 26)Study type(n 18)Not rural se ng(n 6)Duplicate cohort/not addingaddi onal useful data (n 6)Full text not available(n 1)Not high-income country (n 1)Par cipant type(n 1)Studies included inquan ta ve synthesis(meta-analysis)(n 0)Fig. 1 PRISMA flow diagram of search and study selection process [24][31] and two studies included mixed health professions[32, 33]. There were a total of 58,188 participants or participant observations in included studies. Four studieswere outliers in terms of their comparatively large samplesize [34–37]. Most (n 29, 85%) studies measured actualretention (or turnover), with the remainder measuringhealth professional preferences, intentions or simulating interventions. Actual turnover and retention weremeasured/defined over very variable periods of time:for example, one study measured only 6 months ruralpractice as being ‘long-term’ [38] whereas another documented up to 38 years of rural retention [39]. The outcome measures in 12 studies were retention rates [28–30,33, 40–47], while five studies used survival probabilities or hazard ratios [37, 48–52], and a further five usedodds or relative risk ratios for staying or leaving [31, 35,36, 53, 54]. The turnover or retention profile used mostfrequently in included studies was at the level of rural orremote practice anywhere within a country [31, 35, 36,38, 46, 47, 54–59]. Next most frequent was at the levelof a largely rural jurisdiction [29, 41, 48, 50, 52] and ruralpractice within a jurisdiction [34, 40, 42, 49, 51], followedby turnover or retention within a rural community [33,37, 45, 53].FindingsA meta-analysis could not be conducted because thestudies were highly heterogeneous, reporting differentinterventions and retention outcomes. Hence, a narrativeapproach was taken.EducationTwenty-one studies investigated the impact of educational interventions, including: selecting university students with rural backgrounds [31, 32, 38, 58, 60]; locationof university training and its duration (rural, within a

CountryIceland, Ireland, Norway,Scotland, Sweden,GreenlandCanadaStudyCarson et al. [32]Chauhan et al. [55]Table 2 Characteristics of included studies642 rural physicians1046 rural health profes‑sionals (doctors, nurses,allied health profes‑sionals)Participants and samplesizeCross sectionalCross-sectionalStudy designOutcomes measuredAny length of rural training Intention to leave ruralduring medical schoolpractice within the next2 yearsSelection of ruralIntention to stay withbackground or rurallysame rural organisationschooled students; ruralfor at least the nexthealth professional train‑2 years (or until retire‑ing (rural pipeline)ment if intending toretire within 2 years)Intervention(s)There was no significantassociation between hav‑ing had at least one ruralexperience of any lengthduring medical school(versus no rural rotationsat all) and intention toleave rural practice within2 yearsThere was a significantrelationship (χ2 3.98,p 0.05) between havinga rural background andintending to stay withthe same rurally locatedorganisation. Spending themajority of school educa‑tion in a rural area was alsosignificantly associatedwith intention to stay(χ2 8.7, p 0.01). Healthprofessionals working inouter rural areas who hadundertaken some of theirhealth professional train‑ing in a rural area weremore likely to intend tostay than those who hadno rural training (χ2 4.22,p 0.05)Main description of resultsRussell et al. Hum Resour Health(2021) 19:103Page 5 of 24

CountryUSACanadaAustraliaStudyCogbill and Bintz [39]Fleming and Mathews [48]Gardiner et al. [40]Table 2 (continued)361 rural General Practi‑tioners (GPs) in SouthAustralia391 physicians who wereinitially licensed to prac‑tice in NewfoundlandLabrador (NL) between1993–200419 rural General Surgeonswho worked for Gun‑dersen Health Systemsince it commenced in1978Participants and samplesizeCohortCohortCase reportStudy designGroup and individualcoaching by quali‑fied psychologists and6 weeks of email coach‑ing over a 3-year periodRetention rate in rural gen‑ Despite having a mucheral practice at two timehigher intention to leavepoints, 3 years apartrural general practicebefore coaching, only 6%left after coaching. In otherwords, 94% of participantsstayed compared with80% of the general ruralGP population [χ2 4.89,p 0.027]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) timesmore likely to leave NL,respectively, than locallytrained MMGs. Physiciangroup was the only signifi‑cant covariate in the Coxregression analysisTrained locally at Memorial Hazard of remaining in NLUniversity (MMG) versustrained elsewhere inCanada (CMG) versusprovisionally licensedinternational medicalgraduates (IMGs (Prov))versus fully licensedinternational medicalgraduates (IMGs (Full))Main description of resultsThe study showed thatof the 19 rural GeneralSurgeons ever employedin the network 21% (n 4)have retired, 53% (n 9)continue to practice in thenetwork, and only 26%(n 5) left the networkbefore retirement. Thenine currently employedGeneral Surgeons havebeen practising in thenetwork for a mean of88.0 months (SD 83.6;range 24 to 288 months).Six have practiced in onelocation for more than20 yearsOutcomes measuredA rural general surgeryMean number of monthsnetwork designed asthat currently employeda sustainable modelGeneral Surgeons havefor delivery of generalbeen practising in Gun‑surgery services within adersen Health Systemlarge rural region of thenetworkmidwestern USA (Minne‑sota, Wisconsin, Iowa)Intervention(s)Russell et al. Hum Resour Health(2021) 19:103Page 6 of 24

CountryNorwayCanadaAustraliaStudyGaski and Abelsen [41]Gorsche and Woloschuk[42]Jamar et al. [56]Table 2 (continued)74 University of AdelaideRural Clinical Schooldomestic medical stu‑dents 2003–201029 rural doctors workingin an established ruralpractice in Alberta withwritten support fromtheir regional medicaldirector, who enrolledin the skills enrichmentprogram March 2001–March 2005 and their 29matched controls388 Graduating medicalstudentsParticipants and samplesizeCohortCohortCohortStudy design% former interns in thearea staying in (rural)county as at April2014 (Finnmark versusanother rural county)Outcomes measuredRural Clinical Schooleducation exposurecomprising complet‑ing the whole of fifth(penultimate) year ofclinical studies at a rurallocation% of graduates who spentno time, 3 to 18 m and2 years in a rural areasince graduationSkills enrichment program Rural practice reten‑comprising training fortion 5 years after theacquisition of or mainte‑programnance of existing skills ofat least 2 weeks’ durationbut less than 1 year,with preceptor fundingand locum supportavailable and funded upto 80,000 per annum(pro rata)Medical internships earlysign up versus raffle inthe study area versus acomparison area withsimilar workforce issuesIntervention(s)Of the 72 survey responsesanalysed, 44% (n 32)spent no time in a rurallocation, 28% (n 20)spent 3 to 18 months in arural area and another 28%(n 20) spent two or moreyears in a rural area sincegraduationAfter 5 years, all 29/29(100%) physicians inthe enrichment groupremained in rural practicecompared with 22/29(71%) of physicians whodid not participate in theenrichment program:(Relative Risk 1.31; 95% CI1.06–1.62). Two enrich‑ment participants wholeft the province and theircontrols were excludedfrom analysesThe proportion of internswho signed up early whostill worked as physiciansin the study area (29%)was twice as high asamong the regular interns(15%) and interns in thecomparison area (14%).None of the 59 physi‑cians who had been earlysignup interns workedin any of the 15 remotemunicipalities in the studyareaMain description of resultsRussell et al. Hum Resour Health(2021) 19:103Page 7 of 24

CountryCanadaAustraliaUSAAustraliaStudyJamieson et al. [53]Johnson et al. [31]Kahn et al. [49]Kwan et al. [38]Table 2 (continued)CohortCohortStudy design729 medical graduates ofCohortUniversity of Queensland(2002–2011)47 IMG physiciansCohortparticipating in Conradprogram 1995–2003 whowere assigned to a ruralcommunity for return-ofservice397 dental graduates ofUniversity of Sydney2009–2013480 University of Brit‑ish Columbia FamilyMedicine residencygraduation cohorts from1990 to 2007Participants and samplesizeRetention in current com‑munity for more than1 yearOutcomes measuredRural clinical exposurefor one (RCS-1) or two(RCS-2) years; bondedscholarship; rural back‑ground vs metropolitanbackgroundOf the 47 physicians whoundertook their periods ofservice in a rural area, only12 (26%) stayed in a ruralarea following completionof the ROS obligationHaving had some form ofrural experience prior tothe rural clinical placement(PR 3.75 95% CI 2.75–5.11)and pre‐placement ruralintentions (PR 3.54 95% CI2.25–5.57) were significantindependent predictors ofan increased likelihood ofworking in a rural locationin both 2015 and 2017Amongst those who hadbeen in their currentcommunity for more than1 year, those who hadpostgraduate training insmaller rurally distributedsites were 36 times morelikely to be working in arural or regional practicethan those trained inmetropolitan sites (95% CI12.2–108.5)Main description of resultsLong term rural practiceIndependent predictors of(LTRP) defined as 50%LTRP were Rural Back‑of the graduates’ primaryground (OR 2.10 [95% CIplace of practice since1.37 3.20]), RCS-1 (ORgraduation being rural2.85 [95% CI 1.77 4.58]),RCS-2 (OR 5.38 [95% CI3.15 9.20]) and having abonded scholarship (OR2.11 [95% CI 1.19 .76])Conrad program whichRetention rate in ruralallows thirty J-1 visaareas after completingwaivers each year inrural J-1 visa waiver ROSWashington state withobligationparticipants having aROS obligation of 3 yearsfor primary care physi‑cians and 5 years forspecialistsRural Clinical PlacementWorking in a rural locationProgram comprising anin both 2015 and 2017opt in 1-month clinicalversus not working inplacement in rural NSWa rural location in bothin final year of dental2015 and 2017school; previous rural lifeexperienceRurally distributedpostgraduate trainingsites versus training inmetropolitan teachingcentersIntervention(s)Russell et al. Hum Resour Health(2021) 19:103Page 8 of 24

CountryAustraliaCanadaStudyLi et al. [57]Mathews et al. [50]Table 2 (continued)Study design60 physicians who wereCohorttrained at Memorial Uni‑versity Newfoundlandand held ROS (Returnof-Service) agreements(1997 to 2009) andstarted practice inNewfoundland andLabrador (NL) 2000–2005compared with all 67other NL physicians whostarted practice in NLbetween 2000 and 20051117 rural GPs who par‑Cross-sectional andticipated in the Medicinediscrete choicein Australia: BalancingexperimentEmployment and Life(MABEL) survey in 2009Participants and samplesizeOutcomes measuredOpt-in ROS agreementsoffered to medicalstudents and juniordoctors which includedtwo types of bursaries:Family Medicine Bursaryand Special FundedResidency Position(offering postgraduatetraining positions whichare usually accepted byphysicians who wereunable to secure a posi‑tion through the usualapplication process)Retention of physicians inthe Canadian provinceof NL at the end ofthe follow-up period(December 31, 2010)Locum relief guarantee,Probability of attractingretention payments,rural General Practition‑rural skills loading, familyers to (hypothetically)isolation/secondarychoose to stay in ruralschool costs or retentionpracticegrants for existing ruraldoctorsIntervention(s)Whether or not MemorialUniversity Newfoundlandtrained (MUN-trained)physicians received a ROSbursary was a significantpredictor of leaving NL.ROS physicians were 3.2(95% CI 1.4–7.1) timesless likely to leave NLthan non-ROS physicians.Amongst the 60 ROS and67 non-ROS MUN-trainedphysicians, 10 (16.9%) ROSversus 28 (41.8%) non-ROSphysicians left NL provinceby 2010 (p 0.004)Increasing the level oflocum relief guarantee, GPretention payments andrural skills loading fromzero to the middle andhigh levels was associatedwith increased rural GPretention. In order fromhighest to lowest effect onretention were: guaran‑teed locum coverage for6 weeks every 12 months(β 1.51); 50% increasein retention payments(β 1.36); guaranteedlocum coverage for4 weeks every 12 months(β 0.85); rural skillsloading payment increaseby 20% (β 0.82); 25%increase in retentionpayments (β 0.54);and a rural skills loadingpayment increase by 10%(β 0.33). One-quarterof rural GPs were notinfluenced by the ruralincentive packagesMain description of resultsRussell et al. Hum Resour Health(2021) 19:103Page 9 of 24

CountryAustraliaAustraliaStudyMcGrail and Humphreys[36]McGrail et al. [58]Table 2 (continued)Study design610 Medicine in Australia: CohortBalancing Employmentand Life (MABEL) surveyrespondents (2008–2014inclusive) who hadcompleted GP trainingand were transitioning toindependent practice3782 responses from GPsCohortwho responded to Medi‑cine in Australia: Balanc‑ing Employment and Life(MABEL) survey at leasttwice between 2008 and2012 (inclusive)Participants and samplesizeOutcomes measuredVocational training loca‑tion (rural or metro);rural/metro origin; ruralbonding (being con‑tracted to work for partof their early career in arural location)Proportions of GPs work‑ing in rural and metro‑politan locations duringeach of the first 4 yearsfollowing completion ofGP vocational training;proportions of rurallytrained GPs working inthe same or a differentrural community fromthat in which they com‑pleted their vocationaltrainingTraining locationAnnual location retention(Australian trained vsrates in regional, ruralnon- Australian trained);and remote areas andrestrictions on accessodds ratios for leaving ato provider numbersrural arearelated to geographiclocation for InternationalMedical Graduates(IMGs)Intervention(s)The rural training pathway,regardless of childhoodlocation, was extremelystrongly associated withsubsequent rural practice(ORs ranged from 29to 92 in the first 4 yearsfollowing completingGP training). The odds ofrural practice for the ruraltraining cohorts of GPsdecreased with time. Ruralbonding (OR 3.5–5.1) andrural origin (ORs 2.0–4.1)were also positively andsignificantly associatedwith rural practice in eachof the first 4 years follow‑ing completing GP trainingThere was no significant dif‑ference in the risk of leav‑ing rural practice for IMGscompared to Australiannon-restricted graduates.This was true for IMGswhether or not they hadrestrictions on access toprovider numbers limitingwhere they could practiceMain description of resultsRussell et al. Hum Resour Health(2021) 19:103Page 10 of 24

CountryUSANorwayNorwayStudyMurray et al. [28]Nilsen et al. [29]Norbye and Skaalvik [30]Table 2 (continued)CohortStudy design233 Registered Nursestraining at the ArcticUniversity of Norway bya decentralised nursingeducation modelCohort159 nursing studentsCohortdoing a bachelor’sdegree course in Finn‑mark University and whograduated in 2002, 2004and 200560 nursing staff employedby Bassett MedicalCenter in rural upstateNew York, USAParticipants and samplesizeOutcomes measuredLicensed Practice Nursesturnover decreased from16.8% in 2005 to 6.8% in2009. There was no trendin reduced annual turno‑ver evident for RegisteredNurses, but RegisteredNurse vacancy rates fellfrom 16.5% in 2005 to 4.3%in 2009Main description of resultsThe majority [n 190/233(81.6%)] of nursescompleting postgradu‑ate education using thedecentralised nursing edu‑cation model continued towork in rural areasRetention rate in Finnmark Off-campus training wascounty (living and work‑associated with a consider‑ing) for at least 4 yearsably higher retention rateafter completing then 37/40 (92.5%) in Finn‑study programmark county for graduatescompared

Prior reviews suggested that health professional edu-cation delivered in rural areas is positively associated with rural retention, although participating in rural training may reflect pre-existing intention and moti-vation for rural practice rather than the intervention itself increasing rural retention [11]. Many of the 10,