Measuring Health Sciences Research Capacity in Africa: Mapping The Available Data

Background: In recent years there have been calls to strengthen health sciences research capacity in African countries. This capacity can contribute to improvements in health, social welfare, and poverty reduction through domestic application of research ndings; it is increasingly seen as critical to pandemic preparedness and response. Developing research infrastructure and performance may reduce national economies’ reliance on primary commodity and agricultural production, as countries strive to develop knowledge-based economies to help drive macroeconomic growth. Yet efforts to date to understand health sciences research capacity are limited to output metrics of journal citations and publications, failing to reect the complexity of the health sciences research landscape in many settings. Methods: We map and assess current capacity for health sciences research across all 54 countries of Africa by collecting a range of available data. This included structural indicators (research institutions and research funding), process indicators (clinical trial infrastructures, intellectual property rights, and regulatory capacities) and output indicators (publications and citations). Results: While there are some countries which perform well across the range of indicators used, for most countries the results are varied – suggesting high relative performance in some indicators, but lower in others. Missing data for key measures of capacity or performance is also a key concern. Taken as a whole, existing data suggest a nuanced view of current health sciences research landscape on the African continent. Conclusion: Mapping existing data may enable governments and international organisations to identify where gaps in health sciences research capacity lie, particularly in comparison to other countries in the region. It also highlights gaps where more data are needed. These data can help to inform investment priorities and future system needs.


Introduction
Health sciences research (HSR) has been de ned to include basic, clinical, and applied science on human health and wellbeing, as well as the determinants, prevention, detection, treatment, and management of disease. (1)(2)(3) To date, the majority of HSR takes place in the global north.(4-6) As of 2018, less than 1% of scienti c articles published worldwide each year include at least one author based at an African institution, according to analysis by Elsevier, one of the world's largest publishers of scienti c, technical, and medical information. (7) In the past few years, however, a number of international organizations, including the African Union (8), World Health Organization (9), and World Bank (10), have called for political and economic investment in HSR in Africa. Several high-pro le reports have further raised awareness of the so-called 10/90 gap: only a tenth of global expenditure on health research is targeted to issues that affect the poorest 90% of the world's population. (5) There are two key reasons why investments in HSR in Africa may be particularly important from a developmental perspective. First, the promotion of a strong health science industry can contribute to development goals by reducing national economies' reliance on primary commodity and agricultural production; this can help governments develop knowledge-based economies, which may be important for macroeconomic growth. (11)(12)(13)(14)(15) In a seminal 1990 report, the Commission on Health Research and Development (R&D)(4) stated that strengthening research capacity in low-and middle-income countries (LMICs) is "one of the most powerful, cost-effective and sustainable means of advancing health and development" (p. 71). Bilateral agencies, such as the United States Agency for International Development and the (former) United Kingdom Department for International Development, have highlighted the important role of science and technology in promoting economic development in LMICs. (16,17) Second, HSR may contribute to improvements in health, social welfare, and poverty reduction through domestic application of the ndings of the research itself. (18)(19)(20) The 2013 World Health Report stressed that all nations should be producers, users, and consumers of HSR. (6) Africa is home to nearly one-sixth of the world's population and is estimated to account for about a quarter of the global burden of disease. (21) Yet only a small fraction of global health research currently focuses on diseases which exclusively affect LMICs. (22,23) While there have been developments in the HSR landscape over the past three decades, many LMICs still lack su cient capacity to build a local evidence base to inform policy and improve population health at a local level. (24)(25)(26) The International Vaccines Task Force of the World Bank has highlighted the importance of building research capacity to lower the risk of emergent epidemics. (27) To date, few academic studies have evaluated the HSR capacity of LMICs. The most widely available indicator of health research capacity is the publication of health-related scienti c journal articles. These have been the focus of research in the past, with bibliometric analyses undertaken to map the numbers of African publications related to cardiovascular diseases (28,29), genomics (30), health economic evaluation (31), health policies and systems (32,33), human immunode ciency virus (34), neglected tropical diseases (35), and public health. (36) Four studies have also examined the total number of African publications on any health-related topic (as indexed in major bibliographic databases). (37)(38)(39)(40) Beyond publication outputs, however, researchers have also collected data on investments in health-related R&D (37), clinical trial infrastructures (27,37), healthcare workforce numbers (41), and the numbers of universities and 'centres of excellence' (41,42) in African countries to estimate HSR capacity.
Each of these studies can help to understand individual aspects of HSR capacity in African countries, yet no single piece of information can fully capture the degree of capacity in a country or region. There remains a need for analyses which attempt to collect and analytically combine data on multiple indicators to provide a more comprehensive analysis of HSR capacity across the continent. Background 2.1 The importance of knowledge economies Development has been conceptualized as a process that creates growth, progress, positive change and/or the addition of physical, economic, environmental, social and demographic components. (43,44) Science and innovation, if well-utilized, may play a core role in realizing sustainable development. (1,45) As seen from the experiences of many industrialized nations, scienti c research and linked innovations have been core to economic and social advancement over the past two centuries -be it medical innovations such as vaccines and antibiotics, or industrial innovations in manufacturing, communications, and computation. (46)(47)(48) More recently, questions have been asked as to whether scienti c research supports development, or whether it represents a product of development in itself. (49,50) Both these positions have their justi cations. In terms of science resulting in development, it is research and knowledge generation, linked with subsequent innovation and application of that knowledge, that can be seen as critical to overcoming key development challenges in LMICs. (49) Therefore, the need to invest in capacity for mobilizing and using science and innovation can be viewed as an essential component of strategies for promoting sustainable development. (51)(52)(53) This argument appears to underpin the inclusion of research within the Sustainable Development Goals. Goal 3.B speci cally focusses on health research for LMIC needs, calling for "supporting the development of research and development of vaccines and medicines for health conditions which affect LMICS"; goals 9.5 and 12.A call for increased scienti c, technical and research capacity more generally in developing countries. (54) Many calls for the creation of so-called 'knowledge economies' are linked to thinking of research activity as an end goal of development. It has been argued that the conceptualization of an economy of knowledge reproduces a growth and market-oriented rationale for knowledge production, accumulation, and diffusion which has particularly in uenced the international aid, education, and development agenda. (55,56) For example, the OECD(11) de nes knowledgebased economies as those "which are directly based on the production, distribution and use of knowledge and information'' (p. 7). The World Bank identi es four dimensions of the knowledge economy: economic and institutional regime, education and skills, information and communication infrastructure, and an innovation system -with the agency going so far as to create a Knowledge Economy Index (KEI) as an indicator of a country's 'preparedness' for a knowledge economy. (57) Asongu and colleagues (58) found the overall trends in African countries' performance between 1996 and 2010 differed across the World Bank's KEI dimensions: with Tunisia leading in education, the Seychelles in information and communication technology, South Africa in innovation, and Botswana and Mauritius in institutional regime. Oluwatobi and colleagues (59) have argued that the potential for knowledge production and innovation in Sub-Saharan Africa is mitigated by the level of human capital and quality of institutions. Overall, quality education and strong institutions are held to be imperative for the transformation into a knowledge economy. (59)(60)(61) Both educational and economic institutions may create enabling structures for developing knowledge and innovation and for economic growth, but their in uence varies according to institutional arrangements, income, and development levels in countries. (61)(62)(63) In particular, education plays a vital role in strengthening human capital, which directly in uences the ability to create, absorb, transform, disseminate, and use of knowledge and innovation. (59,(64)(65)(66) Education and training emphasizing the value of traditional knowledge and culture also strengthens human capital to innovate contextually relevant solutions for local development problems. (58,63,67)

The contribution of health sciences research
Within the broader remit of science for development, health sciences research is vital in its own speci c way. There are global examples where health research leads to collective human bene t, development of medical treatments, or better understanding of health risks of activities such as tobacco smoking. At a national level, however, HSR can also speci cally generate evidence that is useful for public service planning and program implementation. Local health research is typically seen to provide policy-relevant information, including disease trends, risk factors, outcomes of interventions, patterns of care, as well as health systems and services costs and outcomes. (68) Authors such as Grant and Buxton have gone so far as to develop a framework to estimate the value that HSR provides to countries.(69) Included as bene ts in the framework are cost reduction in the delivery of existing services; improvements in service delivery processes; increased effectiveness of health services; increased population health and equity (achieved through improved allocation of resources, better targeting and accessibility); and contribution to a healthy workforce.
Finally, HSR evidence has also been argued to improve the health policymaking process itself by identifying new issues for the policy agenda, informing decisions about policy content and direction, and through evaluations of policy impact. (70) In many ways, these examples capture the bene ts widely seen to follow from a system of evidence-informed policymaking, whereby a more systematic and robust use of research evidence in decision making is seen to improve planning effectiveness and e ciency to serve the broader social good. (18,71) Health sciences research input and output by national governments are not uniform, with signi cant disparities between regions or income levels, and also across countries within the same region or at similar levels of income. (72,73) On the African continent, for example, Tanzania and Lesotho had similar levels of GDP (per capita USD 2,365 and USD 2,494 respectively in 2013); however, the percentage of GDP invested in research in Tanzania was more than 3 times higher than in Lesotho (0.28 v. 0.08) while the number of publications per million inhabitants was nearly 50 times higher, at 770 in Tanzania compared to 16 in Lesotho. (74) One of the most critical contextual determinants of HSR outputs is historical evolution of research systems. For those African nations subject to colonial rule, for instance, modern forms of research were often developed in service to the economic interests of the colonising power. The focus of research thus centred on key exports such as agriculture, forestry and mining related activity, with little interest in HSR to bene t local populations. (75,76) After independence, HSR remained embryonic, with governments often choosing to invest in economies based on the commercialisation of cash products and natural resources rather than in the development of research and technology.(77) Moreover, countries which have experienced con ict, instability, and other socio-political crises, have had to direct resources towards reconstruction and peacekeeping investments, rather than towards scienti c research and innovation. (78) For some nations, the catalyst for investment and development of HSR has only been through the emergence of health crises-new diseases such as HIV/AIDS and Ebola, or the rising incidence of tuberculosis and plague.(79) (Ministère de la santé publique, 2015; MTN, 2017). Outbreaks have also at times inspired new policies calling for investment in HSR by global organisations such as the World Health Organization (WHO) and UNESCO. (80,81) These calls for investment have allowed for a more open dialogue and progress on conceptualising the importance of health research, even within low income African states -with several governments now committed to investing in scienti c research in connection with a country's economic and sustainable development priorities. (82-84) Despite these shifts, such as the Bamako initiative, WHO's efforts to regionalise research efforts, and signs of increased attention to domestic HSR, key drivers of research and research funding in the health sector remain exogenous to African states. Indeed, funding largely re ects global HSR priorities, with limited options for investigator-initiated research on local health concerns.

How can we measure health sciences research?
While there is a strong case that HSR in LMICs is important at national and global levels -for improving health, preventing epidemic spread, supporting health policy and systems, and as an in uential factor of national development more broadly -there is no single framework or consensus method to assess HSR capacity across countries. Indicators for measuring and monitoring HSR generally include standard output indicators of knowledge production and innovation, such as scienti c journal articles per million inhabitants or patents per million inhabitants (57), and input and process indicators of health research and development. Such process indicators can include gross domestic research and development expenditure on health as a percentage of GDP (GERD), number of clinical trials per million inhabitants, research grants, and full-time equivalent health researcher per million inhabitants.(37) From the perspective of decision-makers in national agencies, these indicators of knowledge production and human resources for research are helpful for benchmarking performance against regional and global comparators and for informing policy and strategy to strengthen research and development. (85) Researchers and international organisations have attempted to compile indicators and measure HSR capacity in different ways. For example, the WHO has created a Global Observatory on Health Research and Development which aims to "consolidate, monitor, and analyse relevant information on health research and development activities". (86) This uses a logic model perspective to assessing HSR, tracking a range of indicators to monitor health R&D inputs, processes, and outputs as identi ed and de ned by Røttingen and colleagues.(37) While these indicators are useful for monitoring and benchmarking the state of HSR and development activities, funding, and performance at the national level, they do not provide information to assess overall capacity of national health research systems as a set of "people, institutions and processes" for HSR. Moreover, there are incomplete or missing data for many of these indicators.
A second approach takes a systems perspective to assessing HSR capacity, recognising that research and development funds and personnel represent but two components of a nation's HSR capacity. In this vein, Pang and colleagues (19) de ned a national health research system within a conceptual framework of four functions: stewardship, nancing, creating and sustaining resources, and producing and using research. This framework has been operationalised under the Research for Health unit at the WHO Regional O ce for Africa and used to assess the evolution of national health research systems, with data collected In applying this approach, Kirigia and colleagues (88) analysed trends between 2003 and 2014 to show that although there have been positive gains across many functions, there are still considerable gaps in many African countries. For example, fewer than half of the countries in Sub-Saharan Africa have an o cial health research policy, a national strategic health research plan, a law regulating HSR, or a budget line for HSR within the health budget; and only about half of them have a national health research institute or council, a health research programme at the Ministry of Health, or a health research management forum. Public nancing for HSR is also typically measured to be very low, with minimal progress towards the goal of 2% of the national health expenditure allocated to HSR. Instead, most funds for HSR come from external sources such as international organisations, NGOs, or multilateral/bilateral partners. Within the systems approach, Kirigia and colleagues attempted to quantify the performance and capacity of national HSR systems in the African national health research systems barometer. Overall, there have been a variety of attempts to identify various elements of HSR activity, performance, and capacity in Africa. Some have assessed R&D potential, measured funding inputs, or identi ed gaps in national HSR systems. These efforts shed light on where strengths and weaknesses lie, but currently do not provide a comprehensive review and synthesis of data on which to comparatively evaluate HSR knowledge and innovation, HSR and development activities, and HSR systems at the national level across Africa.
The aim of this paper is thus to develop this work by collecting and aggregating data on a range of variables to consider HSR activity, performance, and capacity in all African countries. We propose a framework for evaluating a country's capacity for HSR based on publicly available data sources. This framework incorporates and expands on indicators from previous studies. Using this framework, we present data on HSR capacity in each of the 54 UNrecognised African states to map current capacity across the region for health sciences research -providing one of the rst analyses to systematically outline the contribution of African countries to HSR across such a wide range of indicators.

Data collection
We reviewed data for each of the 54 UN-recognised states in Africa. This excluded any foreign departments (e.g., Mayotte), regions (e.g., Réunion), or territories (e.g., Saint Helena) located in Africa, as well as the disputed territory of Western Sahara. We collected population and gross domestic product (GDP) data from the World Bank(91) for each of these states to be able to benchmark our ndings against broader development metrics.
We sought to identify a range of indicators which could help measure the HSR capacity in each country. We used the indicators selected by the World Health Organization Global Observatory on Health R&D database as a starting point, which comprised gross expenditure on R&D (GERD) as a proportion of GDP, health researchers per million inhabitants, number of institutions, and o cial development assistance for the medical research and basic health sectors as a proportion of gross national income. (92) We then supplemented this with others measures of HSR capacity which we identi ed through discussions between authors and members of a project oversight committee [1], including bibliographic data, data on clinical trial infrastructures, regulatory environment, intellectual property rights, and research funding. All data were acquired between June and September 2018.
To classify and conceptualise the various indicators available, we followed the Donabedian(93) model of health care quality measurement to categorise our indicators into one of three types: structural, process, and output measures related to HSR. Structural measures capture inputs into the system and thus comprised metrics such as workforce numbers, budget allocation to R&D, and numbers of organisations, regulations, and guidelines on human subject protections. Process measures are indicators of ongoing HSR activities, including numbers of clinical trials registered and patent applications. Finally, output measures capture the outputs of research activities including numbers of peer-reviewed publications and citations for these publications. (ii) researchers (per million inhabitants), and (iii) researchers with doctoral or equivalent degrees (as a proportion of total number of researchers). From the same database, we also collected data on gross domestic expenditure on R&D in '000s current purchasing power parity (PPP) $; these gures were also shown as a proportion of GDP and per capita. Whenever possible, we collected expenditure and personnel data speci c to medical and health sciences.

Research institutions
We collected data on the number of universities in each country from the website https://univ.cc/; which compiles a list based in part on information from the International Association of Universities. (94) We recognise that there may be limitations affecting the quality of data from this source, so we also identi ed The data were collected from each funder website. As we are seeking to understand current capacity, we only counted funding allocated to researchers based at institutions in African countries. We excluded funding for research projects in which the principal investigators were based at non-African institutions, even if these projects included collaborators, eld sites, or locations of research in Africa.
All amounts were reported in 2018 U.S. dollars based on consumer price index adjustments to account for in ation. Foreign currencies were converted to dollars based on the yearly average exchange rates published by the World Bank. We also collected information on the number of organisations, regulations, and guidelines on human subjects protections in each country. These data, which are collected annually by the US Department of Health and Human Services (105), re ect protections in each of the following categories: "general (ie, applicable to most or all types of human subjects research)", "drugs and devices", "clinical trial registries", "research injury", "social-behavioural research", "privacy/data protection", "human biological materials", "genetic", and "embryos, stem cells, and cloning". We used the 2018 edition of the compilation of protections. (105) Finally, we collected data from the World Intellectual Property Organization on the numbers of patents issued to residents in each country (2016, or most recent available year). (106)

Output Indicators (publications and citations)
To systematically collect publication data, we searched Scopus, the world's largest abstract and citation database of peer-reviewed literature. (107) Scopus was chosen as it includes a larger volume of non-English language journals than many other major bibliographic databases (e.g. Web of Science). (107) We searched for any articles published in the following Scopus subject areas: health sciences (medicine, nursing, veterinary, dentistry, health professions) and life sciences (agricultural and biological sciences, biochemistry, genetics and molecular biology, immunology and microbiology, neuroscience, and pharmacology, toxicology and pharmaceutics). We included the following types of publications: articles, in press, books, chapters, and conference papers.
We searched for articles published with at least one author based at an institution in each of the 54 countries, using the "A liation country" eld in Scopus. We searched the names of each country in English, French, and Portuguese, as well as variant spellings of country names. We restricted the searches to publications published in the ten-year period from 2008 to 2017. The search strategy, including the country names, can be found in Appendix 1.
For each country, we extracted data on the number of publications with at least one author based in the country, as well as the number of publications rst authored by a local researcher. We also collected citation data for all articles. For publications published in the ve-year period from 2013 to 2017, we collected data on the proportion of publications with international, institutional, and national collaborators; these data were unavailable for articles published before 2013. These data were obtained in SciVal, a research information tool developed by Elsevier to synthesise bibliometric data from Scopus.

Descriptive statistics
Appendix 2: Table 1 Tables 3 and 4 present R&D personnel and spending, respectively. Based on data from 2016, or the nearest available year, the number of researchers per million inhabitants ranged from 7 in the Democratic Republic of Congo to 1,965 in Tunisia; the proportion of researchers with doctoral or equivalent degrees ranged from 10% in Malawi to 72% in Cape Verde. The gross expenditure on R&D as a share of GDP ranged from 0.01% in Madagascar to 0.8% in South Africa. The proportion of gross expenditure on R&D that went to the medical and health sciences ranged from 0% in Lesotho to 30% in Swaziland.
Appendix 2: Table 5 presents data on regulatory capacities. About half the countries had a national public health institute (27/54) and national ethics committee (25/54). The number of institutional review boards ranged from 0 in several countries to 30 in South Africa. Table 6 shows the amount of funding awarded to researchers in each country (2008-2017) from ten large organisations. Table 7 presented in combines our main indicators into a single table. While there are some high-achievers across the board, for most states the results are varied -suggesting high relative performance in some indicators, but lower in others (or lacking data). For example, Libya is a relatively high achiever for publications, rst author publications and number of research institutions, but lags behind this success with the number of clinical trials conducted within the country. Conversely, Burundi has low numbers of publications, rst author publications, number of clinical trials and GERD as a % of GDP, but performs relatively well in number of research institutions. There are a few limitations to this study. First, there is no single indicator for accurately ascertaining HSR. Accordingly, we used a variety of available metrics that serve as proxy indicators. Thus, there is a risk that these proxies do not capture the full landscape we sought to map. For example, we have not accounted for broader nancing and infrastructure which contributes to health sciences research -such as buildings, routine access to electricity, primary and secondary education attainment, etc. Each of these, alongside others, contribute to a country's capacity for HSR, and yet such indicators may not demonstrate impact of HSR and can be di cult to disaggregate.
Second, there was a lack of data for several indicators. The most comprehensive data sources were for publications and clinical trials, but many other categories of data were missing results for numerous countries. In some data categories, there were issues of reliability and comparability between sources. Furthermore, while we aimed to collect data from 2007-2017 across indicators, some data points come from before this time frame if more recent data were not available. This included data on patent applications and human resources, as we deemed it to be better to include data outside the period to get a fuller picture of the HSR landscape.
Third, for the output indicators, it is important to note that research is not always published in peer-reviewed journals, and therefore limiting the analysis to Fourth, we were unable to nd a consistent data source across the African continent to measure government budget allocation to HSR. Instead, we used GERD as a proxy; this captures investments in R&D, which is not disaggregated by health sciences. Similarly, when measuring the number of universities in each state, we were not able to ascertain whether these universities undertake research or those which solely offer degrees or training in health sciences.
Fifth, these metrics are all aggregated at the national level, and thus this crude analysis fails to reveal the sub-national interaction. This may not only produce analysis revealing particular "hubs" of excellence, but it may also reveal institutional capacity or individual capacities which form key components of the national landscape.

Discursive Conclusion
We have sought to assess current capacity for health sciences research across Africa, based on a subset of proxy indicators for which we had more complete data. In doing so we contribute to quantifying current strengths and lack of capacity in the African health sciences research landscape.
Our ndings have raised a number of points for consideration: There are some unsurprising high performers across the variety of indicators South Africa, Egypt, Tunisia, which score highly across broader development metrics, perform well in health sciences research also. However, it is worth noting that it is not simply the level of development (GDP) or international or national nancing for HSR (GERD and international research funding) that leads to a success in HSR. States which have had major donor investment in HSR There are several possible explanations for results such as these. One explanation might be that that reliance of donor funding has limited the sustainability of the health research sector when these collaborations end (108,109), or lacked signi cant improvements in broader infrastructures within the national system. Alternatively, international arrangements may result in research agendas set by the global north, which could imply that they either re ect the need of the funding location (110,111), a focus on spotlight issues (112) or so-called 'parachute research', any of which may be limited in improving health research outcomes in the host location. The importance of local research development, however, has been highlighted as vital to building a knowledge economy and addressing domestic health concerns, as in country researchers have the best understanding of the national agenda and cultural context which increased the likelihood of evidence uptake by policymakers. (24,113) Secondly, we recognise that the experience of con ict, or post-con ict reconstruction, can have profound impact on health sciences research capacity. It is unsurprising, perhaps that the lowest performers, South Sudan, Somalia, Democratic Republic of Congo have all been or continue to be engaged in domestic con ict. Importantly we have not included metrics of con ict into our analyses, so this remains speculative. Yet, this is not conclusive, as other locations which have recently been engaged in con ict or political strife have fared better across a range of indicators, such as Sudan, Nigeria and Central African Republic.
Thirdly, we recognise (as per the limitations) that using these indicators to measure performance does not capture the nuance of what is occurring within each system, and the progress that systems are making more holistically. For example, these metrics are not able to infer political commitment to HSR, the relative importance of the HSR landscape globally, how these systems have developed, where the success stories are and where barriers remain to solidifying knowledge economies. They are also unable to infer the historical contexts which led to the development of these systems, whether rooted in colonial science or post-colonial investments, each of which will lead to different looking HSR environments.
Our hope is that with this mapping exercise, governments will be able to identify where their gaps in HSR lie, particularly in comparison to their regional counterparts, and this may offer meaningful analysis for investment priorities and future system needs.