Preferences of Medical Students for Rural Medical Internships in South Africa: A Discrete Choice Experiment


 Background: Globally the proportion of medical doctors to population in rural areas in low- and middle-income countries remains insufficient to address their health care needs. Therefore, it is imperative to design strategies that attract medical doctors to rural areas to reduce health inequalities and achieve universal health coverage. Methods: This study assessed preferences of medical students for rural internships using a discrete choice experiment. Attributes of rural job were identified through literature and focus group discussions. A D-efficient design was generated with 15 choice sets, each with forced binary, unlabelled, rural hospital alternatives. An online survey was conducted, and data analysed using mixed logit models of main effects only and main effects plus interaction terms. Results: Majority of the respondents were females (130/66.33%) and had urban origin (176/89.80%). The main effects only model showed advanced practical experience, hospital safety, correctly fitting personal protective equipment, and availability of basic resources as the most important attributes influencing take up of rural internship, respectively. Respondents were willing to pay ZAR 2645.92 monthly (95%CI: 1345.90; 3945.94) to gain advanced practical experience (equivalent to 66.15% of current rural allowance). In contrast, increases in rural allowance and the provision of housing were the least important attributes. Based on the interaction model, female respondents and those intending general practise associated higher weight for hospital safety over advanced practical experience. Conclusion: In the context of limited budgets and resource constraints, policy makers and rural health facility managers are advised to prioritise meaningful internship practise environments that offer supervised learning environment, safety from physical and occupational hazards and the provision of basic resources for healthcare system-wide benefits to both staff and rural health facility users alike.


3
Background 52 The health workforce as a critical building block of a functional health system requires the availability, 53 accessibility, acceptability, and quality of health workers to determine the level of health service 54 coverage and attainment of the highest possible standard of health (1,2). The move towards universal 55 health coverage cannot be realized without motivated and empowered health workers with required 56 competencies, equitably distributed to provide needed services of good quality to the population (3). 57 The Demographic and Health Survey revealed that many countries that accelerated the move towards 58 universal health coverage "have left the poor and rural population behind"(1). The struggle for health 59 equity is faced by countries globally, with the delivery of healthcare to those living in remote and rural 60 areas identified as a pressing challenge (4). The lack of adequate skilled personnel in rural areas has 61 been attributed as the top limiting factor to the scale up of health interventions such as life-saving 62 anti-retroviral treatment and the improvement of maternal and child-health outcomes (5). 63 64 Rural medical practice is often seen to be challenging due to social and cultural isolation, lack of 65 infrastructure and transport, electricity, telecommunications and restricted access to goods and 66 services (6). The South African National Department of Health's (DOH) strategies for rural doctor 67 recruitment includes recruiting rural-origin students to be trained in Cuba on condition of fixed-term 68 mandatory rural service and the provision of on-site housing which is both expensive and time-69 consuming to maintain (7). South African medical doctors are trained in undergraduate medical schools 70 (either in South Africa or Cuba), followed by a two-year compulsory internship at an approved 71 government hospital and additional one-year mandatory community service before they can be 72 certified for independent practice (7). 73

74
The minimum recommended doctor-population ratio for middle-income countries; such as, South 75 Africa (SA) is 18 doctors per 10,000 people(8); but in 2017, the medical doctors per 10,000 population 76 4 in SA was only 9.1 (9). Only 2.9% of doctors in the SA public sector practise medicine at rural facilities 77 where an estimated 33% of the population live (10,11). The doctor shortage is both an absolute as well 78 as relative issue, as there is unequal division of doctors along public-private lines, provincial lines, rural-79 urban lines, poor-wealthy lines and state dependant-medically insured lines (12). The overall 80 distribution of public sector post-internship medical posts are approximately 75% urban, 25% rural (13). 81 Despite a desperate need for health workers, in 2003, there was 31% vacancy rate in the SA public 82 health sector (5). In 2010, there were 10,860 unfilled public sector medical practitioner vacancies, with 83 the rural province of Limpopo contributing to 46.5% of these unfilled posts compared to the urbanised 84 province of Gauteng which shared only 10.2% of the total unfilled posts (14). 85

86
The first five years of practice after graduation as a medical doctor are critical for retention in 87 practising medicine (15). This is further illustrated by SA studies of public sector doctors which found 88 that between 6.6%-45% of newly graduated doctors planned to leave medicine, citing lack of 89 equipment at facilities and unbearable workload as push factors (16,17). A qualitative cross-sectional 90 study showed that medical interns were motivated to choose an internship based on proximity of 91 facility location to family and the fulfilment of their provincial bursary obligations, however, that 92 research was not investigating the preferences for rural facilities specifically (16). (18,19)Although 93 there is data available to describe health worker's practise location intentions, there is currently a 94 dearth of knowledge on the job preferences of medical students. This study therefore seeks to 95 uncover preferences for rural internship job among final-year medical students at the authors 96 institute. 97

Study Design 100
The study population was the entire final-year medical students at the authors institute who applied 101 and 'Hospital safety' have two levels each as described in Table 1. The analysis of the DCE responses followed the random utility theory framework in which individuals 155 are assumed to have an indirect utility for choice alternatives and make choices based on their 156 discrimination capabilities (21,29). Given binary choice alternatives of 'Rural Hospital A' and 'Rural 157 Hospital B' as described by the attributes, students choose the alternatives that give them the highest 158 utility. The deterministic part of the utility (V jnt ), which is observable, is defined as a linear function of 159 the job attribute levels and is given by: 160 V jnt = β 0 + β 1 sup_regist nt +β 2 sup_consul nt + β 3 allowance nt + β 4 house_provided nt + β 5 reso_avail nt + β 6 exp_proced nt + β 7 safety_good nt + β 8 mask_poor nt + β 9 mask_correct nt Where, the variables are defined in Table 1  and safety were quantified; they were willing to pay the equivalent of 66.15% in current rural 219 allowance to work in a facility with advanced practical experience compared to a facility which only 220 offered limited practical experience, all other things being equal (Table 4). are ideally replaced on a daily basis, resulting in a monthly cost of approximately ZAR200 (5% of 242 current rural allowance) at the time of this study's data collection. Interestingly, a poorly fitting N95 243 mask was less preferred than having no mask at all, highlighting the priority with which medical 244 students value their health and their understanding that a poorly fitting N95 mask is just as ineffective 245 as having no mask at all. In a survey among SA medical and physiotherapy students they rated 246 themselves at a 4.4 times increased risk of contracting TB compared to the general population (33). In 247 the same study 49% of students reported no access to N95 respirators at the health facilities where 248 they were training(33). Likewise, access to basic resources such as gloves, syringes and needles was a 249 preference that significantly influenced choices both in overall and sub-group analysis. In rural 250 facilities which are situated far from medical supply depots, the budgeting and timely procurement of 251 basic resources is vital for the provision of quality healthcare and achieving positive health outcomes. 252

253
A new threat has emerged necessitating the urgent provision of PPE to healthcare workers. At the 254 time of writing, the COVID-19 pandemic has infected an estimated 55,000 healthcare workers in SA; 255 its rapid spread attributed to the pre-existing shortage of PPE (34). Globally the swift response to roll 256 out protective measures and improve the use of PPE have reduced the infection risk among 257 healthcare workers (35). The resulting increase in demand, has led to rapid price surges with N95 258 masks trebling in price since the pandemic began (36). Therefore, the recruitment cost-effectiveness 259 of PPE and basic resource provision as argued in this study, pre COVID-19, may no longer be tenable. 260 The authors do however remain committed to the continued protection of healthcare workers as an 261 immediate and long-term health priority. The popularity of rural allowance and housing provision as a recruitment strategy is thought to be due 271 to its ability to offset travel expenses, thereby lowering the living expenses associated with living in a 272 rural area (37-44). This study however, showed that both rural allowance and housing provision were 273 less important to the study medical students than the other attributes investigated. This supports the 274 findings of Vujicic et al.(19) who denounced the cost-effectiveness of housing provision as a 275 recruitment strategy in Vietnam. Although higher wages are associated with lower rates of worker 276 attrition, this relationship is inelastic at higher salary levels (as in the case of SA doctors), in which 277 instance other job attributes become a more important influence (45). Pending further research, the 278 DOH should reconsider the implementation of its rural allowance policy for doctors as systematic 279 alternatives, which have been mentioned above, may prove to be impactful and cost-effective in the 280 long-term. 281

282
The sub-population analysis further highlight heterogeneity in preference of these job attributes by 283 gender, career aspiration, and rural medicine exposure. The finding that female medical students 284 were marginally more sensitive to rural allowance and housing provision is supported by studies in 285 Burkina Faso and Indonesia which found that females were twice more likely to choose a job offer 286 with free housing and were more sensitive to the recruitment effect of rural allowance compared to 287 their male counterparts respectively (42,46). For the graduate who intends to specialise, rural health facilities can provide the advance practical 294 skills they seek to learn. For those who prefer general practise, a rural facility's safety and resource 295 track record is more influential. 296

297
Medical students with rural medicine exposure valued hospital safety highly reflecting the safety 298 concerns they may have encountered personally or heard about during their time at the rural facility. 299 That medical students without rural medicine exposure preferred being provided with housing more 300 so than their rural-exposed colleagues contrasts with existing literature (49). This could be due to 301 rural-exposed students feeling more confident to organise their own accommodation