The global scientific research response to the public ...Jul 16, 2019 · 11 4 Casa Oswaldo Cruz,...
Transcript of The global scientific research response to the public ...Jul 16, 2019 · 11 4 Casa Oswaldo Cruz,...
1
The global scientific research response to the 1
public health emergency of Zika virus infection 2
3
Juliane Fonseca de Oliveira1, Julia Moreira Pescarini1, Moreno de Souza Rodrigues1,2, Bethania 4 de Araujo Almeida1, Claudio Maierovitch Pessanha Henriques3, Fabio Castro Gouveia4, Elaine 5 Teixeira Rabello5, Gustavo Correa Matta6, Mauricio L. Barreto1,7, Ricardo Barros Sampaio1,3* 6
7 1 Centro de Integração de Dados e Conhecimentos para Saúde, Fiocruz, Salvador, Bahia, Brazil 8 2 Laboratório de Análise e Visualização de dados, Fiocruz, Rondônia, Brazil 9 3 Gerência Regional de Brasília, Fiocruz, Brasília, Brazil 10 4 Casa Oswaldo Cruz, Fiocruz, Rio de Janeiro, Brazil 11 5 Universidade do Estado do Rio de Janeiro, Rio de Janeiro, Brazil 12 6 Escola Nacional de Saúde Pública, Fiocruz, Rio de Janeiro, Brazil 13 7 Instituto de Saúde Coletiva, Universidade Federal da Bahia, Salvador, Bahia, Brazil 14
15
*Corresponding author: 16
Ricardo Barros Sampaio e-mail: [email protected] 17
18
All authors have contributed equaly to this work 19
20
Keywords: 21
Zika Virus Infection; Network Analysis; Scientometrics; Text Analysis 22
23
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.
1
Abstract 24
Background 25
Science studies have been a field of research for different knowledge areas and they have 26
been successfully used to analyse the construction of scientific knowledge, practice and 27
dissemination. In this study, we aimed to verify how the Zika epidemic has moulded scientific 28
production worldwide analysing international collaboration and the knowledge landscape 29
through time, research topics and country involvement. 30
Methodology 31
We searched the Web of Science (WoS) for studies published up to 31st December 2018 32
on Zika using the search terms “zika”, “zkv” or “zikv”. We analysed the scientific production 33
regarding which countries have published the most, on which topics, as well as country level 34
collaboration. We performed a scientometric analysis of research on Zika focusing on 35
knowledge mapping and the scientific research path over time and space. 36
Findings 37
We found two well defined research areas divided into three subtopics accounting for six 38
clusters. With regard to country analysis, the USA followed by Brazil were the leading countries 39
in publications on Zika. China entered as a new player focusing on specific research areas. When 40
we took into consideration the epidemics and reported cases, Brazil and France were the leading 41
research countries on related topics. As for international collaboration, the USA followed by 42
England and France stand out as the main hubs. The research areas most published included 43
public health related topics from 2015 until the very beginning of 2016, followed by an increase 44
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
2
in topics related to the clinical aspects of the disease in 2016 and the beginnings of laboratorial 45
research in 2017/2018. 46
Conclusions 47
Mapping the response to Zika, a public health emergency, demonstrated a clear pattern of 48
the participation of countries in the scientific advances. The pattern of knowledge production 49
found in this study represented the different perspectives and interests of countries based firstly 50
on their level of exposure to the epidemic and secondly on their financial positions with regard to 51
science. 52
Introduction 53
Zika virus infection (ZVI) was first reported in humans in 1954, but only in 2007 did the 54
first outbreak occur in Micronesia [1, 2], and several small outbreaks in Micronesia and the 55
Pacific Islands have been reported since then. However, it was only in 2013 that ZVI started 56
revealing its epidemic potential [3]. In 2015, the virus was linked with associated neurological 57
disorders, including Guillain-Barré syndrome, microcephaly and yet unknown neurological 58
disorders in newborns [4, 5]. It was and remains an event of scientific uncertainties. 59
In February 2016, mostly driven by the Brazilian Zika-associated microcephaly cases, 60
WHO declared a public health emergency of international concern (PHEIC). Different research 61
groups all over the world, as part of separate and collaborative research networks, made a 62
concerted effort to understand the various factors required to cope with the epidemic, including 63
the Zika virus infection, transmission and control, as well as the social impact of severe forms of 64
the disease in the most affected societies [3]. During the outbreak, especially after WHO PHEIC, 65
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
3
international funding and fast-tracking procedures for publishing scientific outcomes stimulated 66
scientific research and publication. The increasing number of publications on Zika after 2015 has 67
been attributed mainly to its association with microcephaly [6], but by 2016, only a few core 68
journals had published on Zika [7]. 69
Since the emergency declaration, science studies' applied to the Zika epidemics 70
publications has come mostly from focusing on scientific networks and their collaboration and 71
co-publication. Some researchers have adopted bibliometric and/or scientometric perspectives, 72
analysing papers and documents to discuss different aspects of scientific production in terms of 73
co-authorship, cooperative networks and innovative potentials on the Zika topic (see [7-10]). 74
Furthermore, the discussion about the influence and centrality of certain groups, 75
researchers and even countries have enriched the debate raised by scientometrics applied to Zika. 76
A first mapping of the scientific networks and their most influential researchers in Zika as a 77
theme was done in 2018 [11]. Using techniques of Social Network Analysis, the authors mapped 78
the co-authorship networks on Zika papers published in 2015 and 2016. The study revealed that, 79
besides the number of publications, two other factors might be connected to explain the 80
prominence of a given researcher in these scientific networks: the diversity of partnerships and 81
the established connections with the pioneers in the field. 82
Another interesting perspective is to address the role of specific regions on the matter. 83
Zika-related publications and patents, investigated from a scientometric perspective, highlighted 84
the strategic role played by the Latin America and Caribbean research network in terms of 85
addressing the science and technology (S&T) challenges related to the outbreak [12]. According 86
to this study, Brazil itself stands out in reporting and researching the clinical manifestations of 87
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
4
the ZVI and also in studies about vector control, the latter related to the filing of patents in 88
particular. 89
State of the art considered, these studies are central to the systematisation of current 90
knowledge in terms of which research areas have been most developed so far. This scenario 91
allows, for instance, policy managers to contribute by allocating and prioritising research to the 92
current prevention and epidemiological combat of the Zika virus. It is also essential to identify 93
interaction and collaboration among countries to help further the development of shared 94
governance in Zika and other research-related fields and increase scientific production and 95
knowledge in the area. Regarding neglected populations and diseases, these studies can support 96
further analysis on equity and sustainability, and science production between low and middle-97
income countries and developed countries, during and after health emergencies. 98
In this study, we aimed to verify how the Zika epidemic has moulded scientific 99
production worldwide analysing international collaboration and the knowledge landscape 100
through time, research topics and country involvement. Therefore, this study explores how the 101
epidemic of Zika has translated into research all over the world and over time, identifying what 102
the main thematic areas of scientific development on Zika are. We also explore how the S&T 103
apparatus of each country responded to the Zika emergency, from the first studies until 2018 and 104
according to knowledge areas. 105
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
5
Materials and methods 106
Data collection, treatment and cleaning 107
Our dataset was obtained by searching terms related to Zika virus infection (“zika” OR 108
“zkv” OR “zikv”) by Topic in Web of Science (WoS). We selected all the articles published 109
from 1945 to the end of December 2018. The search was performed on January 21st, 2019. From 110
the retrieved articles, we extracted the title and abstract to develop a knowledge map and used 111
the metadata variables regarding date published, authors and co-authors’ affiliation countries as 112
well as the ISI subject categories from WoS assigned to them [13]. 113
The selected contents were downloaded as a text file containing all the results of our 114
query. The file was cleaned by (i) converting all words to lowercase, (ii) removing special 115
characters (e.g., “.”, “:” and “/”), and (iii) standardizing similar terms (e.g., “ae aegypti”, “aedes 116
aegypti”). Data cleaning was conducted in Python. The Python code used is available at GitHub 117
[14]. 118
Data Analysis 119
We conducted two main analyses on the Zika publications retrieved. First, a descriptive 120
analysis to understand how countries collaborate towards Zika knowledge; and second, a textual 121
analysis of research on Zika using knowledge mapping over time and space (i.e., countries). 122
Descriptive analysis 123
The descriptive analysis was performed both directly by filtering information from Web 124
of Science homepage and by using VOSviewer software [15] to perform country collaboration 125
network analysis. To summarize data information for each research article, we extracted from 126
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
6
each author his institutional affiliation country. Articles produced by different institutions of the 127
same country were considered a national collaboration, and articles produced by authors 128
affiliated to institutions from more than one country were considered international 129
collaborations. After that, we constructed a co-authorship network at the country level using data 130
exported from VOSviewer and reconstructed and filtered using Gephi software for visualization 131
[16]. 132
In order to measure collaboration in Zika publications, we deduced an international 133
collaboration index (ICI), which is the ratio of the number of countries that a given country 134
collaborated with, over the total number of countries (minus the country itself) which produced 135
results toward Zika knowledge. The closer the index is to one, the more collaborative the country 136
is, and the closer it is to zero, the less internationally collaborative the country is. 137
To address the issue of international collaboration over time as a proportion of published 138
articles, an international collaboration factor (ICF) was calculated as the number of articles with 139
author(s) affiliation from more than one country over the number of articles with author(s) 140
affiliation to just one country. As with the ICI, the more colaborative a country is, the closer it is 141
to one, the less collaborative, the closer it is to zero. 142
Textual analysis 143
We carried out a textual analysis to discover textual data relationships among extracted 144
words. The method reveals thematic trends and identifies publication subjects surrounding 145
specific topics or fields [17]. To do this, we analysed the titles and abstracts of the articles in our 146
sample. More specifically, we analysed words (also called terms or forms), extracted from the 147
titles and abstracts, grouped by text segments comprised of 40 terms each, and applied 148
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
7
Correspondence Factorial Analysis (CFA) to correlate them. The baseline assessment of research 149
areas into clusters relied on the clustering algorithm of IRAMuTeQ v7.2 (Interface for 150
Multidimensional Analysis of Texts and Questionnaires). The correlation process is obtained by 151
testing if the frequency of a given word is statistically associated with another using Pearson 152
Chi-square at a significance level of 5%. Based on the similarity, words are aggregated into 153
clusters of main areas of knowledge. Content analysis was also performed using IRAMuTeQ. 154
The defined number of clusters in IRAMuTeQ required empirical knowledge in the field 155
as well as several tests on how many clusters made a fair and accurate separation of knowledge 156
areas. Therefore, to validate our baseline clusters, we compared the number and content of each 157
cluster obtained using IRAMuTeQ with those obtained using VOSviewer. Moreover, we used 158
the categories defined by WoS for each article, based on the journal in which it was published to 159
further evaluate if the clusterisation process and defined sub-areas of research were aligned. 160
Finally, we showed the results to a panel of specialists on Zika and public health. 161
After we had established statistically significant knowledge area clusters, we evaluated 162
which knowledge areas emerged over time, and stratified them into four groups by publication 163
year: <=2015, 2016, 2017 or 2018; second, we evaluated the participation in each cluster/field of 164
the top 10 countries with regard to the number of publications. 165
Results 166
Descriptive analysis on the countries collaborations 167
The collected data yielded a total of 3,454 articles produced by 18,453 authors, associated 168
with institutions from 150 countries. These articles were published with and without 169
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
8
international collaboration. Fig 1 shows a dispersion matrix using the number of publications and 170
ICI by country. 171
Fig 1. Number of publications by International Collaboration Index for each of the top ten 172
countries. The graph is a dispersion matrix, the y axis is the number of publications on a 173
logarithmic scale base 2 starting at 100 publications and the ICI is on the x axis starting at 0.3. 174
The size of the circles is related to the number of publications and the intensity of the color from 175
dark red to light pink is related to the ICI as shown on the right side scale. 176
According to Fig 1 The most internationally collaborative countries were the USA (ICI = 177
0.67), followed by France (ICI = 0.57), England (ICI = 0.55), Germany (ICI = 0.47) and Brazil 178
(ICI = 0.45). The remaining most productive countries, Italy, India, Australia, China and Canada, 179
presented international collaboration indexes of 0.41, 0.40, 0.35, 0.32 and 0.31, respectively. 180
When we take into consideration the ICI with the number of publications we can see that the 181
USA has the highest ICI and number of publications placing it in the top right corner of the 182
graph. In contrast, China, the third country with the highest number of publications, is the ninth 183
in ICI in the bottom left corner of the graph. Brazil seems to occupy an intermediate position 184
regarding the number of collaborating countries and the number of publications, with the second 185
largest number of publications on Zika, and the fourth in ICI. France and England seems to be 186
well positioned in terms of the number of collaborating countries. The remaining five countries 187
have fewer than 200 publications and show a similar ratio between the logarithmic number of 188
publications and the number of countries with which they have collaborated. 189
Table 1 summarizes the overall information about international collaborations for the ten 190
most productive countries. The countries that published the most on Zika during the study period 191
were the USA and Brazil followed by China, France and England. 192
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
9
Table 1. National and International publications about Zika Virus Infection for the top 10 193
most productive countries. The table depicts the total number of publications. The column 194
“Articles Total” represents the total number of articles published by a country and the percentage 195
of these productions over the total number of articles about Zika found in WoS until 2018. 196
Among these values, we see the number of articles produced by the country itself and the 197
proportion produced with international collaboration. 198
Countries
Articles Total Articles with single country author(s)
Articles with international collaboration
N % N % N %
USA 1,676 48.52 958 57.16 718 42.84
Brazil 631 18.27 299 47.38 332 52.62
China 272 7.87 125 46.0 147 54.0
France 257 7.44 68 26.46 189 73.54
England 230 6.66 35 15.22 195 84.78
Germany 181 5.24 46 25.41 135 74.58
Italy 172 4.98 42 24.42 130 75.58
India 148 4.28 54 36.49 94 63.51
Australia 144 4.17 65 45.14 79 54.86
Canada 139 4.02 44 31.65 95 68.35
The column with the total number of articles and the percentage in Table 1 accounts for 199
more than the total number of articles on Zika due to the collaboration amongst these countries. 200
With regard to the articles written by authors from one country alone and with collaboration we 201
can see that the only country with fewer than half of its articles being the result of collaboration 202
is the USA, in contrast with England, which has published almost double that of the USA, with 203
84.78%. The other three European countries, France, Germany and Italy, also shows higher 204
percentages compared to other non European countries. 205
In order to see how international collaboration among the top 10 publishing countries 206
occurred over time, in Fig 2 we plotted the yearly evolution of their international collaboration 207
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
10
factor - ICF (the number of articles produced with international collaborations over the total 208
production) until 2018 combining papers before 2015 all together. We can see a clear pattern in 209
the USA, Brazil, China and Australia, where the countries have a balanced level of production 210
with and without international collaboration. As a contrast, we can see that England, Germany 211
and France published articles in international collaborations at a higher frequency. Nonetheless, 212
the yearly ICF of the ten most productive countries together with the ICF for Zika publication 213
and WoS overall publication are considered in Fig 2. 214
Fig 2. The production evolution of the top 10 countries plus the overall Zika Production 215
and WoS production. The closer the dots are to 0.5 (highlighted by the black line), the articles 216
produced with collaborations are proportional to the articles produced without collaborations. 217
The value corresponding to the year 2015 represents the sum of all work produced by the 218
selected country up until 2015. Points on the zero line correspond to publications that had no 219
international collaboration. The size of the dots for Global WoS articles are divided by 1,000 in 220
order to fit the scale. 221
Using co-authorship networks between countries, we see that the majority (61.89%) of 222
the articles were produced within one country, 24.17% were collaborations between two 223
countries, 8.88% between three countries and 2.51% between four countries. The remaining 224
2.55% were collaborations between five or more countries. Interestingly, we also found that 225
international collaborations were present on average in 64.47% of articles produced by the top 10 226
countries. The reason for this apparent discrepancy has to do with the fact that an international 227
paper will be considered a collaboration for every country represented in that paper but only 228
once when comparing articles with and without international collaboration. 229
Analysing country-level collaboration, as shown in Table 1 England has the highest 230
number of publications that resulted from international collaboration (84.78%), followed by Italy 231
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
11
(75.58%) and Germany (74.58%). Another collaboration analysis which supports the results for 232
Table 1 and Fig 2 is presented in Fig 3 with an international collaboration network graph. This 233
figure presents all the collaborations amongst the top ten countries. 234
Fig 3: Collaborations among the 10 most productive countries. Each node corresponds to a 235
country and the size of the node is related to the number of international collaborations they have 236
within that group. The thickness of the lines represents the number of collaborations between any 237
two countries and an aproximation of those numbers are presented by the box on the top right. 238
The color scheme emphasizes link differences between countries since the color of a line is a mix 239
of the color of the nodes it connects. The node distribution on the figure has been chosen only to 240
facilitate visualization. The two countries with the fewest number of publications are positioned 241
at the center of the graph and the most productive to the least from top to bottom and left to right. 242
The USA is in the first position in terms of absolute numbers of publications. In contrast, 243
it occupies the 10th position (42.84%) in terms of the percentage of articles produced with 244
international collaboration. Despite being in 10th position, the number of articles that the USA 245
published with international collaboration corresponds to more than three times the total number 246
of articles published by England, the country with the highest international collaboration ratio. 247
Out of the 958 articles published by the USA with international collaboration, over 15% were 248
with Brazil and nearly 10% with China. This strength of collaboration is shown in Fig 3 by the 249
thickness of the line between the nodes. 250
Out of the 299 Brazilian studies published as results of international collaboration, over 251
60% were with the USA, followed by nearly 20% with England and 10% with both France and 252
Germany. For the 147 Chinese studies published with international collaboration, over 75% were 253
with the USA and a few articles with other countries. (Fig 3). 254
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
12
Main research areas of knowledge 255
A total of 24,685 words were analyzed inside 21,055 text segments, of which 17,525 256
were considered active words (i.e., verbs, nouns and adjectives). Using IRAMuTeQ, six clusters 257
were identified and named according to the most frequent words. Cluster 1 represented the 258
“Clinical Aspects” of the disease (13.8% of the words); Cluster 2 the “Diagnosis” (10.6%); 259
Cluster 3 included the “Epidemiology” of the disease (20.1%); Cluster 4 the “Entomology” 260
(14.5%); Cluster 5 “Cellular Biology” (19.9%); and Cluster 6 "Molecular Biology" (21%) (Fig 261
4). 262
Fig 4. Dendrogram of the clusters obtained with IRAMuTeQ. The dendrogram is divided 263
into 6 clusters formed by branch divisions. The first branch divides Group A, and Group B. 264
Group A is further divided into two subgroups called Group A1 and Group A2. The size of each 265
bar corresponds to the number of words in that cluster directly related to the percentage value in 266
the bottom right corner of each bar. The associated name for each cluster appears on its bar, and 267
the colour scheme has been chosen to facilitate visualization. In front of each bar, there is a 268
square word cloud of words with the highest chi-square value to aid the understanding and the 269
representation of each cluster. 270
Using the WoS subject categories, 165 of 252 possible categories were present in the 271
studied articles, and each article is classified in one or more category. To compare WoS 272
categories with the clusters defined using Iramuteq, we used the following approach. First, we 273
selected the ten most frequent categories presented in our studied articles. Second, we analysed if 274
the most significant categories for each one of the 6 clusters independently appeared with a high 275
frequency or not. By doing this, we were able to better classify the clusters by the words which 276
appeared in them. 277
The ten most frequent categories were: Infectious Diseases; Public Environmental 278
Occupational Health; Tropical Medicine; Virology; Multidisciplinary Sciences; Microbiology; 279
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
13
Parasitology; Immunology; Biochemistry Molecular Biology; and Medicine General Internal. 280
Other six categories, among the 10% most frequent, which were not included in the top ten list 281
but were significant to categorize the clusters were: Entomology; Cell Biology; Health Policy 282
and Services; Clinical Neurology; Pediatrics; and Diagnosis. Based on these sixteen categories, 283
we found the following results for Group A: Public Environmental Occupational Health category 284
was well defined in the “Epidemiology” cluster (cluster 3). The clusters "Diagnosis" and 285
"Entomology" (clusters 2 and 4) shared some of the same categories such as Infectious Diseases 286
and Tropical Medicine, with a difference for the Parasitology category which appeared in cluster 287
4 but not in 2. Furthermore, Diagnosis category was in cluster 2 and Entomology category in 288
cluster 4; each cluster was named after these categories. Cluster 1 “Clinical Aspects” had none of 289
the top 10 high-frequency categories of WoS in it. However, it included less frequent categories 290
such as Clinical Neurology and Pediatrics. 291
In Group B, with clusters "Cellular Biology" and "Molecular Biology" (cluster 5 and 6), 292
there were similarities in Virology and Microbiology as main categories. The main difference 293
was that in cluster 6, the most common category was Immunology, and in cluster 2 Virology 294
with Multidisciplinary Sciences were some of the areas. The category Cell Biology was included 295
in cluster 5 which was named after it. Similarly, the category Biochemistry Molecular Biology 296
was very evident in cluster 6. 297
After the validation of the clusters by the categories and their representations, we asked a 298
total of 12 specialists to evaluate the clustering results and the nomenclature. All responses were 299
positive in considering the clusters a well-defined set of research sub-areas within the field of 300
Zika taking into consideration the words available and the categories. Spatial distribution of the 301
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
14
words coloured according to the cluster was presented to the specialists with the inferred 302
nomenclature and positive responses were received. 303
Fig 5. Knowledge map. Words extracted from titles and abstracts of articles on WoS related to 304
Zika Virus infection from January of 1945 through December 2018. The names in the solid line 305
boxes represent the cluster names and the dotted line boxes represent the most frequent periods 306
for the articles represented by the clusters. The size of each word represents their chi-square 307
value; the distance between them represents how statistically related they are based on the results 308
of the correspondence factorial analysis. 309
Year and Country distribution by cluster 310
Another scientometric analysis, after the construction of the knowledge map and the 311
validation of the clusters, concerns the global research path of science over time, based on the 312
year of publication of the articles. Because a large number of papers were published from 2016 313
onward we decided to group all publications before and including 2015 together. These 314
accounted for 3.2%. The publications of our sample yielded four groups (<=2015, 2016, 2017 315
and 2018). We carried out this investigation in the same way as we did with the subject 316
categories, by evaluating how statistically relevant a specific time group were in one or more of 317
the six clusters. 318
We incorporated the findings into Fig 5 with the boxes and using tilde (~) in front of the 319
year to emphasise the fact that the results are an approximation. This means that a particular 320
cluster might have publications or words in all year groups, although its prevalence is in a 321
particular quadrant position or year. 322
Similar to the subject categories, the year of publication analysis presented a well-defined 323
pattern. Publications before and up to 2015 were most prevalent in the bottom right part of the 324
figure with entomology research and related words in years before 2015 and regarding public 325
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
15
health in 2015 and 2016. Publications in 2015 showed a higher than expected presence, by 326
analysing the chi-square, for clusters 3 "Epidemiology" and cluster 2 "Diagnosis". For 327
"Diagnosis", the production surged between 2015 and 2016 the period at the start of the 328
epidemics. The top right quadrant has publications from 2015 to 2016 mostly, where the clusters 329
"Diagnosis" and "Clinical Aspects" are present. In broad terms, epidemiology is defined as the 330
study of the distribution and determinants of health-related states or events in specified 331
populations in order to characterise and determine associations between exposures and outcomes 332
[18]. Taking into account that 2015 through 2016 was the most critical period due to knowledge 333
gaps related to the infection, the three remaining areas in Group A - clinical aspects, diagnosis 334
and entomology - are key factors to be studied in order to better understand epidemiology. 335
Furthermore, entomology is related to the need for a public health quick response during an 336
emergency. 337
In the left part of Fig 5, the clusters from Group B can be seen for both years 2017 on top 338
and 2018 on the bottom. Each year is well established in those quadrants and aligned with the 339
areas of research and clusters "Cellular Biology" and "Molecular Biology" respectively. Taking 340
into consideration the areas with lower than expected values for chi-square for these years, 341
cluster 4 "Entomology" is lower in 2017 and cluster 1 "Clinical Aspects" is lower for 2018, 342
which shows a significant shift in research topics from previous areas to new areas. 343
Finally, we performed a country analysis for the clusters with the top ten countries 344
regarding the number of publications. Fig 6 shows the Chi (not squared) results in order to 345
capture the negative values (lower than expected) for the countries participating in a cluster. The 346
higher or lower than expected results could be interpreted as the amount of research above or 347
below average, represented on the text analysis, carried out by each country. If a country’s 348
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
16
research distribution to a particular cluster is close to zero, the percentage of articles expected for 349
that cluster follows the overall trend. If the number of articles is above or below the percentage 350
expected for that cluster, a bar will appear considering how far from the expected value they are. 351
If a country shows a particular concentration in a cluster, it should show a lower than 352
expected in one or more clusters. From Fig 6, we can infer the presence of a country as well as 353
its contribution to research on the topic or area. 354
Fig 6. Chi values for the top 10 countries for each cluster. The results are calculated taking 355
the observed value (number of text segments) within a cluster minus the expected value for that 356
cluster and country divided by the square root of the expected value. This is the Chi value (not 357
squared) to show negative (or less than expected) values. For each cluster (1 through 6) we have 358
all ten countries and their respective chi value. 359
Starting with the most productive, the USA has a high presence with statistical relevance 360
in cluster 6 focusing on "Cellular Biology". Although the USA has a presence in all the 6 361
clusters, due to its high number of publications and collaborations, the most significant presence, 362
statistically speaking, is in cluster 6 and its least significant presence in cluster 1 "Clinical 363
Aspects" and 4 "Entomology". Brazil, the country with the second highest scientific production 364
is most significant in cluster 1 regarding "Clinical Aspects" of the disease, while it is almost 365
absent in "Cellular Biology" (cluster 5). China is present in two clusters mainly, "Cellular 366
Biology" (cluster 5) and "Molecular Biology" (cluster 6) along with the USA. Clusters 1 through 367
4 are of very little statistical relevance for China. France is present in clusters 1, 2 and 4 playing 368
a role as an important research centre for public health but focusing on the disease as well since 369
French Polynesia had been one of the first places to show endemic signs of the virus. About 370
England, while a non endemic country, its important presence in cluster 3 “Epidemiology” 371
might be related to the fact that it is home to one of the largest centre for tropical and neglected 372
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
17
diseases in the northern hemisphere, the London School of Hygiene & Tropcial Medicine. 373
Australia, Canada and Italy show a very high presence in cluster 4 "Entomology". For Germany, 374
India, and China, cellular biology (cluster 5) appears as the main area of research. 375
Discussion 376
In this study, one of the first analyses on Zika publication that has taken into 377
consideration international collaborations, we have attempted to understand the scientific 378
research path towards Zika science production. The unexpected appearance of the virus in an 379
epidemic form and with devastating effects on pregnancy was followed by the rapid growth of 380
scientific research on ZVI with a great inflow of research capital and institutional interests. It 381
makes research into Zika scientific production at the international context a compelling case in 382
international health and biomedical research perspectives, given its unique characteristics. 383
In general, collaborations in science occur through the need for complementary resources 384
among different actors, institutions and countries, such as knowledge and skills, financial 385
resources and access to the field and key actors in order to conduct relevant research. Co-386
authorship studies indicate that more connected actors tend to attract a higher number of 387
connections over time [19], an aspect that amplifies learning capacities and the absorption of 388
knowledge. This study shows how ZVI related-science has evolved since 2015, as a result of the 389
interest of an international research community focused on solving new and challenging research 390
questions, related to a recognized worldwide health problem. 391
The appearance of ZVI in its epidemic form in the Americas in 2015, particularly in 392
Brazil, initially triggered epidemiological field studies. In 2016, when the outbreak was declared 393
an international public health emergency, the science related to ZVI was intensified in endemic 394
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
18
and non-endemic countries. The evolution of the study field over a short period evolved from 395
epidemiology and entomology to cover clinical and diagnostic aspects of the virus. 396
Subsequently, laboratory-based studies took over, with an emphasis on molecular and cellular 397
biology. This process has mobilized scientific efforts and resources from different countries, 398
institutions and research teams from different scientific areas. In a very short time, complex 399
interactions among these teams from different scientific areas have taken place. An analysis of 400
articles published on this subject in this period provides important insights into our 401
understanding of the path of the scientific response to ZVI. 402
The strong collaboration between the USA and countries from different continents (e.g., 403
Brazil, Canada, Germany and Australia), showed that the USA is the main contributor in terms 404
of collaboration and suggests its key research capacity in the field. Among the possible reasons 405
why the USA is found to be such a prominent publication country as well as a central player in 406
terms of international collaboration in Zika, might be the fact that it is the top R&D funder for 407
neglected diseases [20]. Furthermore, this country has considerable knowledge accumulation in 408
tropical diseases, including the Department of Defence, which has been one of the main 409
supporters of scientific research in this area [21,22]. 410
By using the International Collaboration Index (ICI) and the International Collaboration 411
Factor (ICF), we were able to examine the influence of the countries by taking into consideration 412
the number of countries they have contributed with and their ratio of international collaboration 413
over time. The most collaborative countries in relative terms are in Europe. The high levels of 414
the international collaboration of England may be explained by the significant recent investment 415
made available by the British government and philanthropic funding agencies for the research 416
community aiming to respond to the challenges posed by ZVI [23]. The same can be said for the 417
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
19
European Union, which counts on four large research consortia (ZikaPlan, ZikAction, 418
ZikAlliance and ZikaVax) coordinated by European organizations in association with research 419
institutions from different regions across the world [24]. 420
Also, the potential threat of a Zika epidemic could partially explain the increased country 421
collaboration over time. When the first case of Zika virus infection was confirmed in the 422
Northeast region of Brazil, in May 2015, an epidemiological alert by the Pan American Health 423
Organization (PAHO)/World Health Organization (WHO) was released. The alert recommended 424
Member States to establish and maintain a capacity for ZVI detection, clinical management and 425
an effective public communication strategy [25]. In November 2015, Brazil declared ZVI a 426
national public health emergency and three months later, the World Health Organization 427
declared it a PHEIC [26]. We suggest that a causal effect between the alert/emergency and the 428
increased number of collaborations took place during this period considering the knowledge gaps 429
in understanding ZVI. 430
Brazil is a country with a vast scientific literature on dengue virus transmission, among 431
other arboviruses. However, the results indicate that it had carried out almost no research on Zika 432
virus before the start of the epidemic. Most of the entomological studies had taken place in 433
countries such as Italy, India and Australia. The scarce research resources for research and the 434
virtual inexistence of Zika virus in the Western Hemisphere until 2014 could justify this finding. 435
One of the first steps for health surveillance is to take the kind of disease into 436
consideration. In this sense, research into the clinical aspects of Zika helps in the characterization 437
of the disease. At the end of 2015, Zika infection and especially microcephaly appeared as a new 438
problem to be tackled. Questions regarding the clinical aspects and the epidemiology were the 439
main focus at this moment. Physicians might argue: "What is the Diagnosis?", "What are the 440
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
20
physical pathological mechanisms of the virus/disease?", "What is the case definition?", putting 441
the focus on physiopathology. 442
These questions boosted basic science, since research on the clinical aspects and 443
epidemiology depends basically on people and methods, while applied science builds upon 444
cellular and molecular biology, depending on expensive equipment and other technologies and 445
qualifications. This scenario gives an academic advantage to countries with a higher 446
technological density. We saw Cluster 1 "Clinical Aspects" in 2016 being superseded by Clusters 447
5 and 6 ("Cellular Biology" and "Molecular Biology") in the following years. It also gives clues 448
to understanding the dominant participation of countries such as the USA and China in the 449
clusters related to cellular and molecular biology. 450
Nevertheless, these results have some limitations as our source of information was the 451
Web of Science, which could have excluded articles not indexed in this database. All in all, this 452
study, with its strengths and limitations, shows that science has values, interests and practical 453
applications. The fulfilling of the research objectives are thus strategic in the sense that it could 454
aid the dissemination of research, appropriation and the use of the knowledge produced within 455
the framework of scientific research to support strategies for the control and prevention of the 456
ZVI. 457
However, in the Global Health arena, it is known that inequities in science and 458
technology among high and low and middle-income countries area default in the historical 459
trajectory. Since the Zika outbreak, Brazil has played an important and essential role in 460
developing research and publishing about ZVI and its repercussions. However, regarding ZVI 461
scientific production, we can ask how sustainable the Brazilian prominence in time is. Until 462
when will Brazil be an almost mandatory partner to develop research and publishing on this 463
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
21
matter? How to improve more equitable and sustainable scientific production in the so-called 464
"Global South"? Challenging questions yet to be considered in further studies. 465
Conclusions 466
In this study, we have demonstrated that the scientific global research path in response to 467
the public health emergency of ZVI can be traced with data from indexed articles. It shows the 468
value of the study of scientific networks to help to understand science production in response to a 469
PHEIC. As soon as the public health emergency of ZVI started, our findings showed a first phase 470
of the scientific inquiry focused on entomological and epidemiological studies. It was followed 471
by a second phase with a dominance of studies on the diagnosis and clinical aspects of the ZVI, 472
and by a third and fourth phase focused on molecular and cellular biology. We also found that 473
Brazil has played an important role not only in the rapid response to the epidemic but also in the 474
capacity to produce research and to evoke international collaboration within a short period with 475
high-income countries such as the USA and England. 476
477
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
1
Acknowledements 478
We would like to thank all the specialists that evaluate the clusters established in our study. 479
480
This work is supported by the Center of Data and Knowledge Integration for Health (CIDACS) 481
through the Zika Platform - a long-term surveillance platform for the Zika virus and 482
microcephaly, Unified Health System (SUS) - Brazilian Ministry of Health. 483
484
This work was partially supported by the European Union’s Horizon 2020 Research and 485
Innovation Programme under ZIKAlliance Grant Agreement no. 734548, and by the Oswaldo 486
Cruz Foundation/ Vice-Presidency of Research and Biological Collections-Fiocruz/ VPPCB and 487
the Newton Fund/ British Council. 488
References 489
1. Macnamara FN. Zika virus: a report on three cases of human infection during an 490
epidemic of jaundice in Nigeria. Trans R Soc Trop Med Hyg. 1954 Mar 1;48(2):139-45. 491
doi: 10.1016/0035-9203(54)90006-1 492
2. Duffy MR, Chen TH, Hancock WT, Powers AM, Kool JL, Lanciotti RS, et al.. Zika virus 493
outbreak on Yap Island, federated states of Micronesia. N Engl J Med. 2009 Jun 494
11;360(24):2536-43. doi: 10.1056/NEJMoa0805715 495
3. PAHO. Pan American Health Organization Response to the Epidemic of Zika Virus in 496
the Americas December 2015 - 2016. 497
4. Kindhauser MK, Allen T, Frank V, Santhana RS, Dye C. Zika: the origin and spread of a 498
mosquito-borne virus. Bull World Health Organ. 2016 Sep 1;94(9):675. doi: 499
10.2471/BLT.16.171082 500
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
2
5. Paixão ES, Teixeira MG, Rodrigues LC. Zika, chikungunya and dengue: the causes and 501
threats of new and re-emerging arboviral diseases. BMJ Glob Health. 2018 Jan 1;3(Suppl 502
1):e000530. doi: 10.1136/bmjgh-2017-000530 503
6. Melo MC, Donalisio MR, Freitas AR. Zika virus, from discovery to the present days: 504
Bibliometric review study. Int J Curr Res. 2017;9:46728-40. 505
7. Singh N. Scientometric analysis of research on Zika virus. Virusdisease. 2016 Sep 506
1;27(3):303-6. doi: 10.1007/s13337-016-0339-3 507
8. Martins MD. [Bibliometric analysis of scientific articles on the Zika virus] Análise 508
bibliométrica de artigos científicos sobre o vírus Zika. Rev Electron Comun Inf Inov 509
Saude. 2016 Jan-Mar; 10(1):1-9. Portuguese. doi: 10.29397/reciis.v10i1.1096 510
9. Albuquerque PC, Castro MJ, Santos-Gandelman J, Oliveira AC, Peralta JM, Rodrigues 511
ML. Bibliometric indicators of the Zika outbreak. PLoS Negl Trop Dis. 2017 Jan 512
19;11(1):e0005132. doi: 10.1371/journal.pntd.0005132 513
10. Delwiche FA. Bibliometric Analysis of Scholarly Publications on the Zika Virus, 1952–514
2016. Sci Technol Libr (New York, NY). 2018 Apr 3;37(2):113-29. doi: 515
10.1080/0194262X.2018.1431589 516
11. Maia LF, Lenzi M, Rabello ET, Oliveira J. Colaborações científicas em Zika: 517
identificação dos principais grupos e pesquisadores através da análise de redes sociais. 518
Cad Saude Publica. 2019;35:e00220217. doi: 10.1590/0102-311x00220217 519
12. Machado-Silva A, Guindalini C, Fonseca FL, Pereira-Silva MV, de Paula Fonseca B. 520
Scientific and technological contributions of Latin America and Caribbean countries to 521
the Zika virus outbreak. BMC Public Health. 2019 Dec;19(1):530. doi: 10.1186/s12889-522
019-6842-x 523
13. Leydesdorff L, Rafols I. A global map of science based on the ISI subject categories. J 524
Am Soc Inf Sci Technol. 2009 Feb;60(2):348-62. doi: 10.1002/asi.20967 525
14. Rodrigues M. Zika_sience; 2019 [cited 2019 Jun 12]. Database: GitHub repository 526
[Internet]. Available from: https://github.com/rodriguesmsb/Zika_science 527
15. Van Eck NJ, Waltman L. VOS: A new method for visualizing similarities between 528
objects. In: Decker R, Lenz HJ, editors. Advances in Data Analysis. Studies in 529
Classification, Data Analysis, and Knowledge Organization. . Springer, Berlin, 530
Heidelberg. 2007. pp. 299-306. 531
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
3
16. Bastian, M., Heymann, S. and Jacomy, M. (2009) ‘Gephi: an open source software for 532
exploring and manipulating networks’, Proceedings of the Third International Conference 533
on Weblogs and Social Media, 8: 361-2. 534
17. Sacerdote H. A mediação segundo Feuerstein e o uso da informação em educação on-535
line. D. Sc. Thesis, Universidade de Brasília. 2018. Available from: 536
http://repositorio.unb.br/handle/10482/32414 537
18. Frérot M, Lefebvre A, Aho S, Callier P, Astruc K, Aho Glélé LS. What is epidemiology? 538
Changing definitions of epidemiology 1978-2017. PLoS One. 2018;13(12):e0208442. 539
Published 2018 Dec 10. doi:10.1371/journal.pone.0208442. 540
19. Barabasi AL, Jeong H, Neda Z, Ravasz E, Schubert A, Vicsek T. Evolution of the social 541
network of scientific collaborations. Physica A: Statistical Mechanics and its 542
Applications. v 311, P. 590-614, 2002. 543
20. G-FINDER. Neglected Disease Research and Development: Reflecting on a Decade of 544
Global Investment, 2017. Available at: http://policycuresresearch.org/downloads/Y10_G-545
FINDER_full_report.pdf 546
21. Beaumier CM, Gomez-Rubio AM, Hotez PJ, Weina PJ. United States military tropical 547
medicine: extraordinary legacy, uncertain future. PLoS Negl Trop Dis. 2013 Dec 548
26;7(12):e2448. doi: 10.1371/journal.pntd.0002448. 549
22. Foray D, Mowery DC, Nelson RR. Public R&D and social challenges: What lessons from 550
mission R&D programs? Research Policy, 2012, vol. 41, issue 10, 1697-1702. 551
23. GOV.UK. UK trebles funding to tackle Zika virus. Available at: 552
https://www.gov.uk/government/news/uk-trebles-funding-to-tackle-zika-virus 553
24. European Commission. Commission activities in the area of Zika. Available at: 554
https://ec.europa.eu/research/health/index.cfm?pg=area&areaname=zika 555
25. PAHO/WHO. Epidemiological Alert: Zika Virus Infection. 7 May 2015. Available at: 556
https://www.paho.org/hq/dmdocuments/2015/2015-may-7-cha-zika-virus-epi-alert.pdf 557
26. Lowe R, Barcellos C, Brasil P, Cruz OG, Honório NA, Kuper H, Carvalho MS. The Zika 558
Virus Epidemic in Brazil: From Discovery to Future Implications. Int J Environ Res 559
Public Health. 2018 Jan 9;15(1). pii: E96. doi: 10.3390/ijerph15010096. 560
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint
. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. certified by peer review)
(which was notThe copyright holder for this preprint this version posted July 16, 2019. ; https://doi.org/10.1101/19001313doi: medRxiv preprint