This study described the magnitude of internal and external migration among Mozambicans physicians qualifying between 1980–2006. We found substantial flight from the public sector, especially among more experienced physicians. Overall, internal migration was a larger contributor than external migration to loss of physicians from the public sector, a difference that accelerated at an increased rate over the 26-month observation period as physicians left in increasing numbers to work in the growing NGO sector. When excluding physicians with stronger ties to the colonial power, Portugal (those qualifying within 10 years after independence), internal migration accounted for more than three times as much physician loss than external migration. Notably, the extent of external migration in this sample does not approximate the high level (over 75 percent) previously reported for Mozambique, which included Portuguese nationals who returned to mainland Portugal at independence 
. We believe that the level of external migration reported here more accurately represents the effect of human capital flight on the post-independence NHS.
Previous research on brain drain has focused on external migration, generally ignoring migration from the public sector to other institutions in the same country. It has been argued that internal migration does not weaken health systems as physicians largely leave to work for agencies that support the public sector, resulting in better directed aid 
. We believe that internal migration has a distortive effect on health systems as senior-level managers graduate to well-paying jobs outside the public sector, perpetually leaving junior managers with less training and experience in their place. Given the shortage of physicians, this transition frequently leaves management gaps until replacements are identified, trained and begin their new position. The results of this study support the view that it is the seasoned managers who leave the public sector, including physicians with longer work experience and those with public health training, which aligns with the priorities of international agencies seeking experienced health system managers to provide the inside knowledge and personal connections to meet the policy and implementation expectations of these agencies. This study also found that nearly half of physicians were working in the capital, Maputo, where five percent of the country resides, representing an important impediment to broadly providing quality health services. Given the higher percentage of cases of internal migration residing in the capital city compared with those in the public sector, it is apparent that increased contracting of physicians to work with NGOs and donor institutions does not represent an effective solution to the imbalanced distribution of physicians between urban and rural areas.
An especially worrisome finding of this study is the increased frequency of internal physician migration observed alongside the explosion of the NGO sector in Mozambique, fuelled by the dramatic increase in external assistance from global health initiatives relying on NGOs as the primary channel for aid to meet ambitious targets. There are likely multiple push and pull factors that shape physicians’ decisions to remain within the public sector or pursue outside opportunities. Among important push factors are 1) low salaries, especially for those outside of the capital city who have fewer opportunities to pursue part-time clinical work in the private sector and those with public health training who (with the exception of those with doctoral-level training in public health or a related discipline) are not eligible for an increase in base salary compared with physicians with post graduate clinical specialization; 2) lack of clear career advancement opportunities, particularly among health system managers; and 3) challenging working conditions in a health system with limited resources and great need. Budget ceilings have been highlighted as a factor that contributes to resource constraints in the public sector in Mozambique 
, which may partially explain the rationale for channeling increased external aid through the NGO sector. Though the NGO sector is not a new phenomenon in Mozambique 
, recent funding increases have impacted the sector’s dimensions and relationships with the public sector, and a pluralistic system that brings together NGOs with the public sector will likely continue as long as donors continue to emphasize NGOs as fundamental to their development assistance strategies.
This study has a number of strengths. First, its novel sampling approach provides a comprehensive view of both internal and external migration for physicians trained at all the medical schools that feed medical doctors into Mozambique. Following physicians from training source also captures cases of migration that occur after training but before absorption into the public sector. Our sample also included medical graduates over a 26-year period, providing an indication of how the proportion of physicians leaving the public sector changes along career paths. Finally, by collecting information on employment status at two points over a 26-month period, we were able to describe the impact of increased foreign aid through the NGO sector on migration patterns.
The primary limitation of this study is that key informants were used to develop the list of Mozambican physicians trained outside of Mozambique and to provide information on colleagues’ work status. As a result, the list of physicians trained outside of Mozambique may be incomplete and there may be errors in the classification of the work status and location for physicians in the sample. Because of the relatively high rate of movement in the public sector, misclassification of work status would be expected to lead to underestimates in the degree of physician migration. However, given the extensive triangulation of data from multiple sources, including provincial and hospital human resource records as well as multiple proxy reports, we believe that this bias is unlikely to be substantial and does not greatly affect our conclusions. In addition, we were unable to account for part-time income generation activities outside public sector functions that may be important in explaining why physicians with clinical specialties are less likely to leave the public sector. A second limitation is the focus on physicians, without considering the full range of cadres (such as non-physician clinicians or técnicos de medicina, nurses, pharmacists and laboratory technicians) that play management and clinical roles within the public sector, and who are in high demand for employment in the international aid community.
The MOH has taken a number of steps to ameliorate the impact of physician flight from the public sector. It has expanded pre-service training capacity for physicians to replace those who no longer are active in the public sector by increasing medical school class size and by increasing the number of medical schools in the country from one to four (including two additional public universities – Lúrio University in Nampula province and Zambeze University in Tete province – and the private Catholic University of Mozambique in Sofala province). As a result, the annual production capacity for new physician graduates currently exceeds 100, compared with an average of less than 21 from Eduardo Mondlane University in the 25-year period between 1980 and 2004. Workforce expansion has also accelerated for non-physician cadres to substitute clinical and management tasks currently under the responsibility of physicians. Working conditions have improved through strengthened management and logistics systems, as well as health facility rehabilitation. Both across the board and targeted incentives for key management positions have been provided to reduce the salary differential between the public sector and international NGOs and donor agencies. Finally, codes of conduct have been signed with donor agencies and NGOs to limit hiring skilled staff away from the MOH 
Despite concrete steps taken by the MOH to stem the flow of workers from the public sector, these efforts have failed prevent the loss of these valuable resources. Ultimately these approaches will not succeed until international agencies take responsibility for their role in contributing to physician migration, and in particular internal migration. Our results indicate that the magnitude of internal migration occurs more frequently than external migration, particularly to the NGO sector, and given the now widely recognized urgency to strengthen struggling public sector health systems and the dramatic increase in resources available through global health initiatives, frank reflection by donors and NGOs is needed on how their hiring practices and projects may undermine the very systems they seek to strengthen.