Rankings of physician migration based on the number of physician émigrés from African, emigration fraction and physician density produce different results. Only the physician migration density and, to a lesser extent, the number of physician émigrés show a systematic pattern of associations with health workforce, health status, health spending and social-economic profiles of African countries. This study highlights an important but often neglected problem in studies and reports which quantify the magnitude and patterns of health workforce migration: metrics tell tales and quite often different ones, depending on the perspectives adopted.
Although this study is limited by its focus on one continental – African – experience, as far as the author knows, it is novel in looking at the impact of choice of migration metrics on the perception of which source countries may suffer relatively more physician emigration. Like other health workforce studies, this study is also limited by its use of metrics which quantify stock, rather than actual flow over time [22, 23]. Unfortunately, few studies can reliably have the luxury of data on time-dependent flow of health-workers. That said, this study is among the first to use the new database which accounts for bilateral net flows among source and other countries. The quality of these data is not necessarily comparable or completely reliable across countries although it has been improving over the last few years [1, 4, 13, 17, 20]. Differential bias in the results could occur if the quality of the health workforce and migration data is shown to be systematically associated with the observed levels of the countries' profiles. Partly based on how well these data have worked in global health analyses and the intuitive nature of the results, I have little reason to suspect any substantial or differential bias in the findings which could arise from the varying data quality.
Other limitations are given in Clemens and Pettersson [13], although their data which were used here tried to overcome some of the standard problems of existing health workforce data such as focusing only on physicians trained in own birth country. As discussed elsewhere, using birth country to classify physicians may reflect the extent of "Africa-ness" of the physicians although this need not be suitable for every health workforce research [13]. Considering my experience with a recent analysis of a different global database of more than one hundred and forty countries which lost physicians to the US, Canada, Australia and UK [2] and in which the physician migration metric was originally proposed [20], classifying the physician émigré according to country of medical training yields similar migration correlates as the current study. As one of the reviewers of the current study thoughtfully pointed out, using only African-born physicians in the denominator of the emigration fraction might overstate the magnitude of migration or yield misleading results because foreign-born physicians who remained active in Africa would not be counted. Including foreign-born physicians in the denominator of the emigration fraction, however, also implies that they should be included in the numerator whenever they emigrate. Otherwise, the emigration fraction might paint the wrong picture since the foreign-born physicians would – inappropriately and – statistically 'not be allowed to be at risk' of emigration.
Given its scope and ecological design, this paper does not and cannot address the correlates of why individual physicians emigrate. For such analysis, researchers would need coupled hierarchical data, nesting individual physician émigrés within both destination and source countries, to avoid cross-level inferential fallacies [20]. Furthermore, this study does not pretend to answer the question of who an African physician émigré should be [13]. Is it a doctor born in Africa? Or is it a doctor who just holds an African citizenship or a physician trained in Africa? This study made use of a database which classified the African physician as someone born in Africa, currently employed as a medical doctor, and had been residing in the destination country long enough to be included in the country's recent census [13]. This definitional choice does not detract from the central thrust of this study which is to show how the extent and patterns of migration might be dependent on the type of metric used.
Unlike other studies which have also addressed the African migration crisis [1–3, 13, 24], this paper emphasizes that, although the emigration fraction is useful for indicating the extent of workforce losses through migration, it is not designed to account for the importance of the population size or to pattern migration according to national contextual profiles of the source countries [20]. By relating to the size of the physician pool, the emigration fraction intuitively outperforms the total number of émigrés metric. Nonetheless, the emigration fraction differs from the physician migration density which adjusts for source population size in its ability to depict the macro-patterns of migration. At first glance, the correlates of migration might seem counterintuitive [20], but a closer look reveals that somewhat richer African countries like Seychelles (1.51), Mauritius (1.06) and Tunisia (1.34) also have higher physician densities per 1000 population than the average African country (0.27). Also, higher physician capacity and wealth are usually seen in countries with higher health spending, less poverty and better overall development [1, 4, 19, 20, 23]. It is, therefore, not surprising that physician migration density is also positively associated with development-related profiles. Previous studies have tended to allude qualitatively to the poorer profiles of countries with higher emigration fractions. This study goes further and assesses the actual correlations and finds that the emigration fraction was not patterned according to common national profiles. Like Mejia's landmark study in 1978 [25], this paper shows that migration has a positive gradient with source countries' capacity [20, 21]. This study suggests that the emigration fraction may be more appropriate for depicting physician stock depletion while the migration density is more appropriate for understanding country-level patterns in emigration [20]. Recent analysis reveals that the methods used in this study work well on nurse migration data and yield similar findings [26].
So, what do these findings mean for policy and future research? Policies [8, 11, 27] being suggested for solving the migration threats to the health workforce in Africa and other poor areas might be barking up the wrong tree [28]. If wealthier North American and European countries draw relatively more physicians from less poor countries with which they may have better visa prospects, recognition of educational qualifications, and foreign relations [20, 29], in the long-term, it is possible that migration reversal and retention policies might benefit the 'rich' but not necessarily the very poor source countries which have absolutely and relatively insufficient physicians to begin with [20]. This does not imply that every physician who returned to a physician-poor setting would not improve the supply of that country. Treating more patients could make a big difference to the suffering patients and their families but the impact would be hard to gauge at the population level in countries with very low physician densities but high disease burden [20]. Unfortunately, this scenario is not far-fetched in many African countries. Policymakers need to be careful about seeing migration reversal as a long-term strategic solution to health workforce shortages.
Comments
View archived comments (1)