We accessed publicly available data.²³ To estimate data on primary and secondary school spending we used Unesco data from two published reports on public expenditure for students in primary and secondary schools as a percentage of gross domestic product per capita.²⁴ Data on gross domestic product were obtained from the United Nations and included the gross domestic product and gross domestic product per capita in current US dollars.²⁴

The number of medical schools and contact information were identified through the Institute for International Medical Education website (a comprehensive database of international medical schools) and the sub-Saharan Africa Medical Schools Study.^{19, 25} For each country we obtained the costs of medical schools up to July 2011 through information posted on the medical school or university website. In the absence of available data for the largest medical university in a given country, we chose the largest with available data. University costs were gathered from the Addis Ababa University (Ethiopia), University of Nairobi (Kenya), University of Malawi (Malawi), Igbinedion University Okada (Nigeria), University of Stellenbosch (South Africa), Makerere University (Uganda), Hubert Kairuli Memorial University (Tanzania), University of Zambia (Zambia), and the University of Zimbabwe (Zimbabwe). When medical school costs subsidised by government were not available from our contacts for public universities, we used the costs from large not for profit private medical schools.

We obtained statistics on HIV prevalence (percentage in those aged 15-49) from the UNAIDS 2008 report on the global AIDS epidemic and the 2009 AIDS epidemic update.¹³ Data on the density of doctors was obtained from the WHO Global Health Observatory.²⁶ Currency conversion was done through the XE universal converter website in July 2011.²⁷ We derived data on the number of African trained doctors currently practising in the destination countries through communication with respondents at the Australian Medical Association (Masterfile), the Canadian Medical Association (Masterfile), the American Medical Association (Masterfile), and the UK General Medical Council (through an access to information request). All data were accurate up to September 2010.

We took the average age of retirement for doctors in the destination countries from various sources, including direct contact with the medical councils of the destination countries.^{28, 29, 30, 31} The age distribution of active doctors for Australia, Canada, the United Kingdom, and the United States was taken from, respectively, the Australian Institute of Health and Welfare, the Canadian Medical Association, the British Medical Association and General Medical Council, and the American Medical Association, respectively.^{27, 32, 33, 34}

We obtained the most recent interest rates on savings for each African country from the websites of their respective major national banks. Given the economic instability of Zimbabwe, we derived a conservative inflation rate from neighbouring countries. The average cost of medical education was obtained from the medical association of each destination country.

A sensitivity analysis was carried out in two steps by considering factors that reduce the estimated loss in investment, a best scenario, and its counterpart, the worst scenario. We obtained the best scenario estimate by reducing the years over which the principal is compounded, by reducing the public’s cost in the case when students go to private schools and reducing medical school costs by removing non-tuition fees. The mean time from end of school to emigration is conservatively estimated as 7.2 years.³⁸ We used a weighted average to account for the various ages of the doctors working abroad. Rather than use the midpoint of each age category to calculate the amount of time over which to compound, we used the extremity of the intervals to minimise the compounding time. Only Nigeria and Tanzania have private medical schools, so the number of doctors in these countries was reduced by the percentage of private medical schools. For the remaining countries, the cost of primary and secondary schools was reduced by the percentage of students attending private schools as reported by Unesco. The worst scenario compounds costs over the full length of a doctor’s career using the deposit rate for each country (see web extra of data for our analyses).

The estimated compounded lost investment for South Africa, accounting for more than 50% of the estimated compounded lost investment for all nine countries, was much larger than that observed in the other countries. Figure 2fig2 displays the loss in relative terms. The size of each bubble represents the ratio of the estimated compounded lost investment over gross domestic product. From this perspective, Zimbabwe and South Africa seem to have the largest losses. The y axis corresponds to the ratio of (domestically trained) doctors working abroad in the target countries and those currently working domestically. Ethiopia, Zambia, and South Africa faired the worst.

Table 5tbl5 shows the results for the sensitivity analysis. Applying the best scenario assumption, the conditions decreased the estimated investment costs to $1.41bn. Compounding over the full length of a doctor’s career and using the deposit rate for each country (the worst scenario) led to substantial increases in investment loss. Half the estimated losses increased by more than 10-fold. In absolute numbers, South Africa had the largest change, from $1.41bn to $9.76bn. In total the countries would have lost $13.53bn.

Destination countries do not have to provide medical school training to doctors who successfully pass licensing examinations. Therefore destination countries benefit from not having trained recruited doctors. Based on the number of doctors working from the nine source countries and the average cost of medical education in these countries, this equals a saving of at least $621m for Australia, $384m for Canada, $2.7bn for the United Kingdom, and $846m for the United States; $4.55bn in total. As the United Kingdom had the largest number of African doctors practising, its savings were the largest.

Our study has several strengths and limitations. Strengths included the use of conservative estimates of costs and lost investments compared with interest rates often reported by differing international financial institutions; we chose interest rates on the lower end of available data to avoid the overestimation of lost investments. When, for example, we applied a sensitivity analysis examining deposit rates, the lost investment increased to over $10bn. We do not know the number of doctors who emigrated to the destination countries but never entered medical practice nor did we quantify the number of doctors practising outside the four destination countries in settings such as Saudi Arabia, a popular destination for new graduates. Both of these limitations may result in underestimates of the true loss to the source countries under study. However, we did not consider the number of doctors who return to their source countries nor examine the benefits of doctors sending financial resources back to families in the home countries. While these factors may mitigate the losses, little is known about how widespread or systematic they may be. Remittance by all professionals to sub-Saharan Africa in 2010 is estimated at $21.5bn, a growth of 5.5% from the previous year.⁴¹ A recent survey estimated that doctors typically remit $4500 yearly to their source countries.³⁸ Remittances typically go to family members rather than the state and so it is impossible to quantify the impact of remittances on the local economy.

To the best of our ability we assessed whether education in the source countries is government supported, private, or a combination of the two, but recognise that this may change over differing time periods and differing governments. For example, Uganda has recently changed its public university coverage to focus solely on science, thereby increasing the number of medical school attendees with government coverage. Our study assumes that students go directly from secondary school to medical school and does not account for those who have received previous medical training, including former nurses and clinical officers. We used the current gross domestic product as a proxy of costs for primary and secondary schools. Gross domestic product in sub-Saharan African countries has fluctuated over the past four decades and it is possible that a different gross domestic product would alter our study findings for pre-university education. Finally, although confidence intervals are provided, these assume a fixed interest rate through time. Given the current poor economic climate, these results are conservative and we acknowledge a higher degree of variance.

Three previous studies attempted to quantify the economic value of losing health staff, but these analyses were limited to particular countries (Ghana, Kenya, and Malawi).^{39, 40, 42} Other studies found that doctors typically migrate from African countries to more developed countries—namely, Australia, Canada, the United Kingdom, and the United States^{4, 12, 18, 43, 44}—but do not attempt to quantify the economic implications of such migration patterns. Our study focused on the direct costs of educating doctors in the source countries. Additional economic costs to the loss of doctors from these source countries occur, including the lost investment on the education of other health workers. A previous analysis that examined a case study of only Ghanaian doctors in the United Kingdom estimated a saving in current training cost to the United Kingdom of about £65m from the employment of 293 Ghanaian doctors working in 2004.⁴² A recent survey of African doctors working in Canada and the United States showed that most doctors emigrate immediately after training, but, when considering all respondents to the survey, the average number of years working in the source countries was 7.2.³⁸ The author implies that during this period doctors may have already repaid their debt to the source country.⁴⁵ Many of the countries we surveyed employ a lower cadre of health workers for many common health provisions, including the care of people with HIV/AIDS, including non-doctor clinicians,⁴⁶ nurses,⁴⁷ and community health workers.⁴⁸ Other notable economic issues relevant to the lost investment in doctors include the lack of specialised medical care available and the morbidity and mortality associated with it. Although our study examined only doctors, the emigration of nurses and pharmacists from the source countries has also been important. We also recognise that developing countries experience an out-migration of health workers to other developing countries, and future research should attempt to estimate losses in other regions.

The new code of practice on the international recruitment of health personnel suggests that source and destination countries could benefit by crafting bilateral agreements that acknowledge the transfer of staff from developing countries to developed ones, and provide technical assistance and other support to countries that are losing trained health professionals. Our study highlights that the loss to developing countries is substantial and that any compensation should be more than token: the lost return on investments in medical education is one way to attach a value to the amount of technical or other compensatory assistance that recipient countries should provide. In 2008, the United States’ President’s Emergency Plan for AIDS Relief (PEPFAR) re-authorisation legislation recognised the need to build an infrastructure for health workers and committed to the expansion of medical training and research capabilities through African academic centres, by contributing $130m to improve and achieve large numbers of health staff over the next five years.⁴⁹ Contributions from Australia, Canada, and the United Kingdom to medical education have been substantially lower.^{50, 51, 52}

Our results indicate that South Africa incurs the highest costs for medical education and the greatest lost returns on investment for all doctors currently working in destination countries. These findings are supported by statistics on human health resources. South Africa has the highest density of doctors per population. However, the distribution of doctors in South Africa is unbalanced and there is a 14-fold difference in density of doctors between urban and rural settings.¹⁷ Previous estimates indicate that up to 30% of South African doctors have emigrated to the destination countries we examined, many during apartheid.⁵³ In addition, interviews with health workers revealed that 58% were intending to emigrate to these countries.⁵³ Thus, South Africa, although producing a large number of doctors, also loses the most to developed countries. Conservative estimates indicate that South Africa requires three times its current workforce to meet the requirements of providing care for AIDS.⁵⁴ Any future approaches to improving the numbers of doctors will need to recognise that additional educational opportunities may lead to additional lost investments.

Previous research has focused on the number of health workers working in the destination countries^{4, 12, 44} and on the ethics of recruitment of health staff.^{55, 56, 57, 58} Less research has examined the impact of the density of health workers on morbidity and mortality in the source countries. With a new emphasis on strengthening health systems by major international donors, questions are arising about what investments should be made to strengthen a health system and what measures should be used to determine the strength of a health system. Canada, the United Kingdom, and the United States have clearly stated that maternal and child health will be the focus of present investments, steering away from disease specific investments such as HIV/AIDS.⁵⁹ There is a clear need to recognise that measuring the effectiveness of a health system is a complex endeavour that may result in unclear findings. The support for education and retention of health staff represents one major way to ensure that general and specialty healthcare are available in these source countries. With the exception of the United States, our chosen destination countries have not targeted medical or health training as a focused supportive role. Recent studies have indicated that although the capacity for medical education is expanding in Africa, substantial support is needed to improve weaknesses in infrastructure and that retention strategies need to be developed to reinforce the number of teaching staff, who are also among those medical staff who emigrate.¹⁹

Countries in sub-Saharan Africa are losing considerable investments in medical education through the emigration of doctors to wealthier destination countries. The new voluntary code urges the government, private agencies, and non-government agencies that benefit from the immigration of doctors to increase their technical and financial support to enhance the strengthening of health systems in developing countries with critical shortages in health workforce. Efforts to increase support can include training, financial compensation, and population specific interventions. These should be commensurate with the benefits enjoyed by recipient countries.

We thank Peter Arnold (Australian Medical Council) for access to data; Meredith Fordyce and Mark Doescher (Department of Family Medicine Research Section, University of Washington) for access to data on African medical graduates practising in the United States; and Niyi Awofeso, Ike Anya, Eben Mouton, and Jimmy Volmink for access to country costs. EJM is supported by a Canada Research Chair from the Canadian Institutes of Health Research. IB is supported by a CIHR-Health Canada Research Chair in Health Human Resources Policy. JN is supported through a National Institutes of Health mentored patient oriented research career award (K23 AI068582-01).

Contributors: EJM and NF conceived the study. EJM, CGA-Y, MA, NF, ILB, NB, and AH designed the study. EJM, AH, JN, CGA-Y, MA, AM, and SK identified the data and obtained the data. SK carried out the statistical analysis. NF, RSH, NB, and EJM supervised the project. CGA-Y, MA, AM, ILB, NF, AH, SL, and EJM prepared the first draft of the manuscript. All authors revised the manuscript for important intellectual content. All authors approved the final version for publication. EM will act as guarantor.

Funding: This study received no direct funding. No funding agency has seen this study.

Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; and no other relationships or activities that could appear to have influenced the submitted work. JN and SL work at universities that have received Medical Education Partnership Initiatives (MEPI) funding from the President’s Emergency Plan for AIDS Relief.

Ethical approval: This study was approved by the University of British Columbia. No informed consent was required as no patients were enrolled and only secondary data were used.

Data sharing: The technical appendix, statistical code, and dataset are available from the corresponding author at Edward.mills@Uottawa.ca.

Data from UNAIDS 2010 update, sub-Saharan African Medical Schools Study, and World Health Report.

*Adults and children.

GDP=gross domestic product; NA=not available.

*As percentage of GDP per capita.

*Inflation rate.