Research on Neonatal Conditions in Africa: Volume, Impact, Thematic Spectrum, and Collaboration from a Bibliometric Perspective

I searched the keywords related to NC, validated by peers in Confraria & Wang, 2020, in titles, abstracts and keywords to find the relevant articles indexed in the Web of Science Core Collection (WoS) and published between 2000 and 2019 (see the search query in the Support Information (SI)). I extracted the metadata of 1784 articles published by at least one scientist (not all authors of a publication are scientists in the strict sense of the word, but for simplicity, I use this word to refer to the authors in the publications) whose affiliation is with an African institution (hereafter African scientists). The use of discipline–specific terms imposes limitations that affect the accuracy of the methodology due to the multiple meanings and variety of terms that can be used to represent a topic. I then analyzed each article and retrieved those referring to NC in Africa. The final set comprises 1250 articles. The limitations of the methodology are described in the SI.

The knowledge on NC has increased over time. In 2000, there are 15 articles indexed in WoS and in 2019, 196 (Figure 1). The growth can be described as exponential, with an annual growth rate of around 19%.

The production is concentrated in the period between 2010 and 2019, mainly in 2015–2019 when about 61% of the 1250 articles were published (Figure 1).

A large proportion of the scientific knowledge produced is the responsibility of a few African countries: Egypt, South Africa, Nigeria, Tanzania and Kenya were involved in 787 articles, representing about 63% of all articles (SI, Table 1). This suggests that knowledge may be concentrated in a few places, which is undesirable as the burden greatly affects the whole continent.

These countries are also consistently among the African leaders in the production of knowledge in general (e.g., Cerdeira, Mesquita, & Vieira, 2023; Mouton & Blanckenberg, 2018; Sooryamoorthy, 2021). Therefore, their national scientific systems are expected to have characteristics that put them in a satisfactory position regarding knowledge transfer to other African countries.

Over time, there have been no significant changes in the African countries with the highest contribution (SI, Table 1). Egypt has significantly increased its production; two papers were published in 2000–2004 and 150 in 2015–2019, the country with the highest number. The changes in 2015–2019 indicate that other countries might play a crucial role in NC research in the future. While Egypt, South Africa and Nigeria remain in the top 5, Tanzania and Kenya have been overtaken by Ethiopia. Uganda and Ghana have a similar number of articles to Tanzania.

Collaborative activities are common in NC research (Figure 4) and have increased. In 2000–2019, there are 555 articles with DRC and 669 with IRC (SI, Table 5). The growth can be characterized as exponential, with an annual growth rate of about 23% for IRC and 17% for DRC. In 2000, collaborative activities are present in all articles and in 2019, in 98%. Considering 5-year periods, we see that DRC is more frequent than IRC in 2000–2004; it is about two times the frequency of IRC (Figure 4), although the number of articles is small. Over time, IRC has become more frequent and accounts for 54% of all articles in 2015–2019, while DRC accounts for 45% (SI, Table 5).

Among the types of IRC, collaboration between African scientists without the participation of non-African scientists (IRC_intra) was rare. We find 19 articles in 2000–2019, and 16 in 2015–2019 (SI, Table 5).

Articles in which African scientists from various countries and non-African scientists (IRC_{inter_intra}) contributed were practically non-existent until 2009. After that, the number of these articles began to increase; in 2015–2019 they represent 18.5 % of all articles (SI, Table 5).

This rare pattern of collaboration (IRC_intra and IRC_{inter_intra}) may not allow the development and establishment of networks among African scientists that permit them to address their local challenges. These collaborations provide insights to answer current research questions and offer African scientists the opportunity to share their previous experiences, acquired knowledge and the challenges in their countries through informal communication. As they interact and realize each other’s capabilities and interests, trusting relationships are formed. These relationships and previous collaborations can lead to new joint projects that address local issues (Cerdeira, Mesquita, & Vieira, 2023; Owusu-Nimo & Boshoff, 2017). In addition, trusting relationships may spread among the collaborators of African scientists involved in IRC_intra and IRC_{intra_inter}, increasing opportunities for collaboration with these scientists’ partners (Newman, 2001). In short, it is expected that all mechanisms involved in collaborative activities will contribute to building a stronger research network between African scientists from different countries with or without the presence of non-African scientists.

Collaboration between African scientists, all from the same country, and non-African scientists (IRC_inter) is the most common, in 88.2% of all articles in 2000–2004 and 77.6% in 2015–2019 (SI, Table 5). The benefits of these joint activities can be maximized if the African scientists involved have well-established national and intra/interregional African networks. In this scenario, they can serve as a bridge between other African scientists representing their professional connections in Africa and the non-African scientists involved in IRC_inter.

If we look at the network behind NC research, we can speculate on the potential benefits of IRC from this perspective.

NC research is characterized by high geographical dispersion as the network consists of 129 countries (89 non-African and 40 African countries, SI, Figure 5). There are 5059 links, which account for 61% of all possible interactions given the number of countries involved. The interactions between non-African countries (the links connecting two non-African countries) dominate the network as they account for 57.4% of all links (SI, Figure 5), followed by the interactions between African and non-African countries, which account for 36.3% of all links. Finally, interactions between African countries account for 6.3% of all links. In the network, 2192 links (43.3% of all links) represent a single publication in 20 years.

Collaborations between African countries appear to be fragile, as the subnetwork related to these interactions has 60% of its links representing only one article in 20 years (Figure 5). When these links are removed, the number of African countries with links to another African country drops from 40 to 25.

The subnetwork of the collaborations between African and non-African countries (SI, Figure 6) also shows weaknesses, as the links representing one publication in 20 years account for 49% of all its links. The Republic of Congo, Liberia, South Sudan, and Togo, which also have a fragile position in Figure 5, albeit Togo to a lesser extent, lost all their connections to non-African countries when the links representing one publication are deleted (SI, Figure 7). In addition, African countries lost connections with 15 non-African countries (Albania, Armenia, Bhutan, Bosnia and Herzegovina, Haiti, Hungary, Jamaica, Kazakhstan, Latvia, Malta, Montenegro, Myanmar, North Macedonia, Syria and Yemen).

The subnetworks indicate that Ghana, Nigeria, Tanzania and Kenya can play an important role in the flow of knowledge, in the continent, and in establishing professional links between African countries. They are quite well connected with other African countries within and outside their region (28 countries in the case of Kenya and 29 for the others). South Africa can also be crucial but is not well connected within Southern Africa. In 20 years, it co-authored a single publication with Botswana. If we ignore the links representing one article, the top 5 consist of the same countries and Uganda also appears at the top.

These countries, in addition to Egypt and Zambia, have the highest number of connections to non-African countries. They are influential in the network, as they are connected to countries with many connections also as supported by the high values of eigencentrality (SI, Table 8).

As for non-African countries, the USA and the UK are key players in the flow and access to knowledge as shown by the high number of connections and centrality (they have the highest values for degree and eigencentrality, SI, Table 8).

The role of these countries becomes even clearer when analyzing the number of jointly authored articles by African countries (SI, Table 7). They are among the 5 countries that collaborated the most with African countries in 2000–2019 and in each 5 years. In 2000–2019, about 57% and 28% of the articles with IRC have the collaboration of scientists from the USA and the UK, respectively.

Sweden, the Philippines, and Switzerland lost their positions in this top. Their contribution is limited to two and three articles respectively in 2000–2004 (SI, Table 7), while others such as India, Canada, Pakistan and France emerged, although Pakistan and France are no longer at the top in 2015–2019. The research with multinational clinical trials conducted in India, Pakistan, African countries and other regions could justify the high presence of India and Pakistan at the top.

627 universities participated in IRC (SI, Table 9), and they are present in 96% of all articles with IRC. The distribution of universities by region is asymmetrical, as non-African universities predominate, although those from Africa increased over time (SI, Table 10). In 2015–2019, there are 126 African universities, accounting for 22% of all universities. The distribution of universities by country is also skewed, as those from the USA represent 18.5% of all universities in 2000–2019, followed by the UK with 4.8%.

The University of London (UoL-GBR) is the leader given the number of articles in which it was involved in the entire period, followed by the University of California (UC-System-USA) (Table 1). Other well-known universities are present in the top 10, such as Harvard University (HU-USA), Johns Hopkins University (JHU-USA), the University of North Carolina (UNC-System-USA) and the University of Oxford (OX-GBR). As for African universities, the University of Malawi (UNIMA-MWI), the University of Makerere (Mak-UGA) and the University of Witwatersrand (Wits Univ-ZAF) published more articles with IRC in 2000–2019. Over time, the changes in the top 10 are not significant. We see changes from the first to the second 5-year period, but the number of articles we are dealing with is so small that an in-depth discussion is impossible.

The collaboration network has 627 universities and 48468 links (SI, Figure 8). The visible links represent 25% of the possible interactions between the participating universities. The interactions between non-African universities dominate the network as they account for 77% of all links (SI, Figure 8), followed by the interactions between African and non-African universities, which are 21% of all links. Finally, interactions between African universities account for 2% of the observed links.

The network is characterized by a dense core dominated by non-African universities. There are clusters of universities that are closely connected to this core. If we move from the core to the periphery, the network consists of small groups of universities with weak connections to each other and fragile connections to the core.

This network can be considered too fragile as 67% of the links represent a publication in 20 years. If these links are deleted, the network consists of 477 universities (397 and 80 African and non-African universities, respectively).

Regarding the subnetwork among African universities, it cannot be said that it is a mature and well-established network, as several universities have a very fragile position (SI, Figure 9). 17 out of 138 universities have no connections to other African universities; they are part of the network because they are only linked to non-African universities. 557 out of 804 links represent one publication in 20 years. If we delete these links, only 48 African universities will still be connected and most of the small groups of African universities, which are at the network periphery, disappear.

The University of Cairo in Egypt (CU-EGY), MUHAS-TZA in Tanzania, Wits Univ-ZAF in South Africa, University of Kinshasa in the Democratic Republic of Congo (UNIKIN-ZAR) and University of Cape Coast in Ghana (UCC-GHA), from different African regions, can play an important role in the flow of knowledge between African universities as they are linked to several universities in other countries from their region and also from other African regions, except Wits Univ-ZAF and UNIKIN-ZAR. Although they are linked to universities in other African regions, they have no joint publications with countries within their region.

It is important to know which African universities can play a crucial role in the flow of knowledge and the expansion of professional relationships within the continent and to identify those that stand out for their links with non-African universities. The role of the above universities is diminished by the absence or presence of a small number of links with non-African universities.

The subnetwork between African and non-African universities, which includes 600 universities (SI, Figure 10), is also fragile, as the number of links representing one publication is 6896 out of its 10409 links. Most African universities have at least one publication with non-African universities, as 127 out of 138 are in the subnetwork. However, the presence of African universities seems delicate as only 74 remain in the subnetwork if the links representing one publication are deleted. In addition, many connections with non-African universities are lost as the number of these universities drops to 239.

Despite these weaknesses, we find African universities that occupy an outstanding position. The universities highlighted above (SI, Figure 9), as well as the University of Ahmadu Bello (ABU-NGA), the University of Stellenbosch (SU-ZAF) and the University of Cape Town (UCT-ZAF) (from Nigeria and South Africa) have the most connections with non-African universities. They are also among the universities with the highest centrality in the network (except for UNIKIN-ZAR), as shown by the eigencentrality values (SI, Table 11).

Finally, as far as non-African universities is concerned, outstanding universities such as UoL-GBR, OX-GBR and UW-USA are important in NC research, as shown by their high number of connections and their centrality in the network (SI, Table 11). These institutions have the most connections to African universities (about 14% of their total connections).

As for DRC, Egypt, Nigeria, South Africa and Tunisia are among the top 3 countries with the highest number of documents in each period, although, Tunisia’s presence is marginal as only two articles were identified in 2000–2004 (SI, Table 6). These countries account for 74% of all articles with DRC. Although the changes concerning DRC are positive for these countries, this concentration indicates a low level of interaction between African scientists within the boundaries of the remaining countries.

As for Egypt, Nigeria and South Africa, it seems that each country is trying to solve its local problems by drawing on local knowledge, as the number of articles with DRC increased, albeit at different rates.

South Africa, which had the highest number of articles with DRC in 2000–2004, about 47% of all articles in that period, was displaced in the following period by Nigeria, which has about 39% of all articles in 2005–2009. In 2010–2014, Nigeria lost its first place to Egypt, which retained the lead in 2015–2019. The relative weight of articles with DRC from South Africa decreased over time, while articles from Egypt increased, accounting for about 34% of all articles with DRC in 2015–2019.

The data indicates that Ethiopia has made a significant commitment to DRC over the last five years. While there were no interactions within the borders in the previous periods (SI, Table 6), articles with DRC in 2015–2019 account for 14% of all articles with DRC in that period.

Research involving scientists from several countries leads to results with greater visibility than research conducted only within borders, as the CNCI values show (SI, Table 12). The median value is 0.5 of the world average for articles with DRC, and the value has not changed much over time. As for the articles with IRC, the median values are very close to the world average, and the highest value is recorded for articles published in 2010–2014; 1.272 the world average (SI, Table 12). The highest visibility of articles with IRC was also highlighted in previous studies Vieira, 2023.

Although the articles with IRC_inter are prevalent in NC research, articles with IRC_{inter_intra} have a higher visibility (articles published in 2010–2014 and 2015–2019, SI, Table 13). Articles with IRC_{inter_intra} and published in 2010–2014 have a median value that is twice as high as the world average and 1.5 times as high as for articles with IRC_inter in the same period. For the most recent articles (2015–2019), visibility decreased for both types of IRC. -

Maternal Mortality, Group B Streptococcus, Bilirubin, Gestational Diabetes Mellitus and Preterm Labor collaborated the most in the entire period (SI, Table 14). Therefore, the distribution is very similar to Figure 3.

Maternal Mortality and Group B Streptococcus have the largest number of collaborative articles in each 5 years. However, the thematic spectrum is dynamic, with Bilirubin and Fetal Heart Rate emerging at the top in 2005–2009, Bilirubin maintaining its position and Preterm Labor appearing in the last period (SI, Table 14).

It has been argued that the scientific production of African countries is dependent on the contribution of foreign scientists Vieira & Cerdeira, 2022. However, the data show that different patterns can be found depending on the level of analysis.

Among the topics for which there are documents with IRC or DRC, IRC is the most common in 25 topics, and in the top 5, we have Maternal Mortality, Gestational Diabetes Mellitus, HIV Prevalence & Prophylaxis, and Malaria. The number of articles with IRC is about three times as high as those with DRC (SI, Table 15). As for Fetal Heart Rate, the distribution is more balanced, and the difference is one document.

DRC is the most frequent in 24 topics, and in the top 5 we find Group B Streptococcus, Bilirubin, Preeclampsia, Preterm Labor and Patent Ductus Arteriosus (SI, Table 15). Still, the differences between DRC and IRC are not as pronounced as in the previous case, except for Patent Ductus Arteriosus.

The IRC articles’ visibility is above that of articles with DRC for the topics with 20 or more articles (SI, Table 16). Except for Group B Streptococcus and Gestational Diabetes Mellitus, the visibility of articles with IRC is twice as high as that of articles with DRC. For these topics, the visibility regarding the presence of IRC is also close to or slightly above the world average, except for Bilirubin with a median value of 1.5 times the world average.

The search query used to get the publications from the WoS Core Collection is as follows:

((ts="Preterm birth" OR ts="Birth asphyxia" OR ts="birth trauma" OR ts="Neonatal sepsis" OR tss="neonatal infection*" OR ts="Gastroschisis" OR ts="Jaundice" OR ts="Necrotizing enterocolitis" OR ts="Persistent pulmonary hypertension of the newborn" OR ts="Intrauterine growth restriction" OR ts="Bronchopulmonary dysplasia" OR ts="infant apnoea" OR ts="infant respiratory distress syndrome" OR ts="asphyxia at birth" OR ts="anaemia in neonates" OR ts="neonatal alloimmune thrombocytopenia" OR ts="bronchopulmonary dysplasia" OR ts="cardiac failure in neonates" OR ts="hyaline membrane disease" OR ts="hypocalcaemia in neonates" OR ts="hypoglycaemia of the newborn" OR ts="hyponatraemia in neonates" OR ts="hypothermia in neonates" OR ts="intestinal obstruction in neonates" OR ts="pulmonar interstitial emphysema”) and (cu="South Africa" OR cu="Egypt" OR cu="Nigeria" OR cu="Ethiopia" OR cu=” Rep Congo” OR cu=”DEM Rep Congo” OR cu="Tanzania" OR cu="Kenya" OR cu="Algeria" OR cu="Sudan" OR cu=”south sudan” OR cu="MORocco" OR cu="Uganda" OR cu="Mozambique" OR cu="Ghana" OR cu="angola" OR cu="somalia" OR cu="Ivory Coast" OR cu=”cote ivoire” OR cu="Madagascar" OR cu="Cameroon" OR cu="Burkina Faso" OR cu="Niger" OR cu="Malawi" OR cu="zambia" OR cu="mali" OR cu="Senegal" OR cu="Zimbabwe" OR cu="Chad" OR cu="Tunisia" OR cu="Guinea" OR cu="Guinea bissau" OR cu=”Equat Guinea” OR cu="Rwanda" OR cu=" Benin" OR cu="Burundi" OR cu="Eritrea" OR cu="Sierra Leone" OR cu="Togo" OR cu="Libya" OR cu="CENT AFR REPUBL" OR cu="Mauritania" OR cu="Liberia" OR cu="Namibia" OR cu="Botswana" OR cu="Lesotho" OR cu="Gambia" OR cu="Gabon" OR cu="Mauritius" OR cu="Eswatini" OR cu=”swaziland” OR cu="Djibouti" OR cu="Comoros" OR cu="Cape Verde" OR cu=" SAO TOME & PRIN " OR cu="Seychelles") and py=2000–2019)

As for the method used to extract the articles on NC, the less positive points must be emphasized.

Non-African scientists, mainly those from high-income countries (HICs) might conduct research on NC in Africa without the participation of African scientists, a phenomenon known in the literature as parachute science (e.g., Stefanoudis et al., 2021). However, we should emphasize the results from the literature of the research done by scientists from HICs on the greatest burden of disease. Most of the research on Medical and Health Sciences (M&HS), which includes NC, has been done by HICs and they have addressed their local scientific problems, which are very different from those of Africa (Atal et al., 2018; Evans, Shim, & Ioannidis, 2014). As for neonatal disorders, results have shown that disease-specific gaps exist in Sub-Saharan Africa (Atal et al., 2018) and that the burden is very prevalent in Africa in general (Ou et al., 2022).

In M&HS, in specific, research involving African and non-African scientists unveiled that while 67% of the total 520 clinical trials conducted in Africa addressed local scientific problems, in 14% of cases the research addressed the needs of developed countries. Therefore, I can not guarantee that all the NC research identified is related to the local needs in Africa.

Normally, when determined for a set of documents, the CNCI is the average value of the CNCI of all documents. However, when using this indicator, we must consider that the citation distribution of documents is typically skewed (Vieira & Gomes, 2010). We normally find a small number of documents highly cited and many documents with relatively few citations (Figure 1 in this document). The mean value is affected by highly cited documents therefore the average values might not represent well the impact/visibility. Thus, I decided to present the median, which is not affected by the presence of highly cited documents. The distribution of the values of CNCI (Figure 1, Figure 3, Figure 4 in this document) shows that indeed the average value is not a good choice.

When dealing with small sets of documents, the probability of finding a median value of zero is high given that many documents are never cited, and others given their recent publication did not have the time to be acknowledged by the scientific community. Therefore, I only discuss these indicators for sets with 20 or more articles. However, I present the values for sets with less than 20 documents only for information purposes.

Lastly, in calculating the CNCI, I used an open window, i.e. the count of citations represents the cumulative citations between the publication year and September 2023 (the moment when I collected the data from the WoS).

The citation topics schema names clusters of documents based on cited and citing relationships between documents (Potter, 2020). This is a three-level hierarchical document classification system. The macro, meso and micro level include 10, 326, and 2449 scientific fields, respectively. At the micro-level, each paper is categorized into only one topic. The schema is a dynamic one, meaning that every year the number of scientific fields can change. Some topics might require a change in the name, and others might be added. Therefore, every year, we have an update of the schema. The data presented here is based on the schema available for 2023.

As for collaboration, I followed the following heuristics: