research paper example about covid 19

• Research article
• Open access
• Published: 04 June 2021

Coronavirus disease (COVID-19) pandemic: an overview of systematic reviews

• Israel Júnior Borges do Nascimento 1 , 2 ,
• Dónal P. O’Mathúna 3 , 4 ,
• Thilo Caspar von Groote 5 ,
• Hebatullah Mohamed Abdulazeem 6 ,
• Ishanka Weerasekara 7 , 8 ,
• Ana Marusic 9 ,
• Livia Puljak   ORCID: orcid.org/0000-0002-8467-6061 10 ,
• Vinicius Tassoni Civile 11 ,
• Irena Zakarija-Grkovic 9 ,
• Tina Poklepovic Pericic 9 ,
• Alvaro Nagib Atallah 11 ,
• Santino Filoso 12 ,
• Nicola Luigi Bragazzi 13 &
• Milena Soriano Marcolino 1

On behalf of the International Network of Coronavirus Disease 2019 (InterNetCOVID-19)

BMC Infectious Diseases volume  21 , Article number:  525 ( 2021 ) Cite this article

18k Accesses

35 Citations

14 Altmetric

Metrics details

Navigating the rapidly growing body of scientific literature on the SARS-CoV-2 pandemic is challenging, and ongoing critical appraisal of this output is essential. We aimed to summarize and critically appraise systematic reviews of coronavirus disease (COVID-19) in humans that were available at the beginning of the pandemic.

Nine databases (Medline, EMBASE, Cochrane Library, CINAHL, Web of Sciences, PDQ-Evidence, WHO’s Global Research, LILACS, and Epistemonikos) were searched from December 1, 2019, to March 24, 2020. Systematic reviews analyzing primary studies of COVID-19 were included. Two authors independently undertook screening, selection, extraction (data on clinical symptoms, prevalence, pharmacological and non-pharmacological interventions, diagnostic test assessment, laboratory, and radiological findings), and quality assessment (AMSTAR 2). A meta-analysis was performed of the prevalence of clinical outcomes.

Eighteen systematic reviews were included; one was empty (did not identify any relevant study). Using AMSTAR 2, confidence in the results of all 18 reviews was rated as “critically low”. Identified symptoms of COVID-19 were (range values of point estimates): fever (82–95%), cough with or without sputum (58–72%), dyspnea (26–59%), myalgia or muscle fatigue (29–51%), sore throat (10–13%), headache (8–12%) and gastrointestinal complaints (5–9%). Severe symptoms were more common in men. Elevated C-reactive protein and lactate dehydrogenase, and slightly elevated aspartate and alanine aminotransferase, were commonly described. Thrombocytopenia and elevated levels of procalcitonin and cardiac troponin I were associated with severe disease. A frequent finding on chest imaging was uni- or bilateral multilobar ground-glass opacity. A single review investigated the impact of medication (chloroquine) but found no verifiable clinical data. All-cause mortality ranged from 0.3 to 13.9%.

Conclusions

In this overview of systematic reviews, we analyzed evidence from the first 18 systematic reviews that were published after the emergence of COVID-19. However, confidence in the results of all reviews was “critically low”. Thus, systematic reviews that were published early on in the pandemic were of questionable usefulness. Even during public health emergencies, studies and systematic reviews should adhere to established methodological standards.

Peer Review reports

The spread of the “Severe Acute Respiratory Coronavirus 2” (SARS-CoV-2), the causal agent of COVID-19, was characterized as a pandemic by the World Health Organization (WHO) in March 2020 and has triggered an international public health emergency [ 1 ]. The numbers of confirmed cases and deaths due to COVID-19 are rapidly escalating, counting in millions [ 2 ], causing massive economic strain, and escalating healthcare and public health expenses [ 3 , 4 ].

The research community has responded by publishing an impressive number of scientific reports related to COVID-19. The world was alerted to the new disease at the beginning of 2020 [ 1 ], and by mid-March 2020, more than 2000 articles had been published on COVID-19 in scholarly journals, with 25% of them containing original data [ 5 ]. The living map of COVID-19 evidence, curated by the Evidence for Policy and Practice Information and Co-ordinating Centre (EPPI-Centre), contained more than 40,000 records by February 2021 [ 6 ]. More than 100,000 records on PubMed were labeled as “SARS-CoV-2 literature, sequence, and clinical content” by February 2021 [ 7 ].

Due to publication speed, the research community has voiced concerns regarding the quality and reproducibility of evidence produced during the COVID-19 pandemic, warning of the potential damaging approach of “publish first, retract later” [ 8 ]. It appears that these concerns are not unfounded, as it has been reported that COVID-19 articles were overrepresented in the pool of retracted articles in 2020 [ 9 ]. These concerns about inadequate evidence are of major importance because they can lead to poor clinical practice and inappropriate policies [ 10 ].

Systematic reviews are a cornerstone of today’s evidence-informed decision-making. By synthesizing all relevant evidence regarding a particular topic, systematic reviews reflect the current scientific knowledge. Systematic reviews are considered to be at the highest level in the hierarchy of evidence and should be used to make informed decisions. However, with high numbers of systematic reviews of different scope and methodological quality being published, overviews of multiple systematic reviews that assess their methodological quality are essential [ 11 , 12 , 13 ]. An overview of systematic reviews helps identify and organize the literature and highlights areas of priority in decision-making.

In this overview of systematic reviews, we aimed to summarize and critically appraise systematic reviews of coronavirus disease (COVID-19) in humans that were available at the beginning of the pandemic.

Methodology

Research question.

This overview’s primary objective was to summarize and critically appraise systematic reviews that assessed any type of primary clinical data from patients infected with SARS-CoV-2. Our research question was purposefully broad because we wanted to analyze as many systematic reviews as possible that were available early following the COVID-19 outbreak.

Study design

We conducted an overview of systematic reviews. The idea for this overview originated in a protocol for a systematic review submitted to PROSPERO (CRD42020170623), which indicated a plan to conduct an overview.

Overviews of systematic reviews use explicit and systematic methods for searching and identifying multiple systematic reviews addressing related research questions in the same field to extract and analyze evidence across important outcomes. Overviews of systematic reviews are in principle similar to systematic reviews of interventions, but the unit of analysis is a systematic review [ 14 , 15 , 16 ].

We used the overview methodology instead of other evidence synthesis methods to allow us to collate and appraise multiple systematic reviews on this topic, and to extract and analyze their results across relevant topics [ 17 ]. The overview and meta-analysis of systematic reviews allowed us to investigate the methodological quality of included studies, summarize results, and identify specific areas of available or limited evidence, thereby strengthening the current understanding of this novel disease and guiding future research [ 13 ].

A reporting guideline for overviews of reviews is currently under development, i.e., Preferred Reporting Items for Overviews of Reviews (PRIOR) [ 18 ]. As the PRIOR checklist is still not published, this study was reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2009 statement [ 19 ]. The methodology used in this review was adapted from the Cochrane Handbook for Systematic Reviews of Interventions and also followed established methodological considerations for analyzing existing systematic reviews [ 14 ].

Approval of a research ethics committee was not necessary as the study analyzed only publicly available articles.

Eligibility criteria

Systematic reviews were included if they analyzed primary data from patients infected with SARS-CoV-2 as confirmed by RT-PCR or another pre-specified diagnostic technique. Eligible reviews covered all topics related to COVID-19 including, but not limited to, those that reported clinical symptoms, diagnostic methods, therapeutic interventions, laboratory findings, or radiological results. Both full manuscripts and abbreviated versions, such as letters, were eligible.

No restrictions were imposed on the design of the primary studies included within the systematic reviews, the last search date, whether the review included meta-analyses or language. Reviews related to SARS-CoV-2 and other coronaviruses were eligible, but from those reviews, we analyzed only data related to SARS-CoV-2.

No consensus definition exists for a systematic review [ 20 ], and debates continue about the defining characteristics of a systematic review [ 21 ]. Cochrane’s guidance for overviews of reviews recommends setting pre-established criteria for making decisions around inclusion [ 14 ]. That is supported by a recent scoping review about guidance for overviews of systematic reviews [ 22 ].

Thus, for this study, we defined a systematic review as a research report which searched for primary research studies on a specific topic using an explicit search strategy, had a detailed description of the methods with explicit inclusion criteria provided, and provided a summary of the included studies either in narrative or quantitative format (such as a meta-analysis). Cochrane and non-Cochrane systematic reviews were considered eligible for inclusion, with or without meta-analysis, and regardless of the study design, language restriction and methodology of the included primary studies. To be eligible for inclusion, reviews had to be clearly analyzing data related to SARS-CoV-2 (associated or not with other viruses). We excluded narrative reviews without those characteristics as these are less likely to be replicable and are more prone to bias.

Scoping reviews and rapid reviews were eligible for inclusion in this overview if they met our pre-defined inclusion criteria noted above. We included reviews that addressed SARS-CoV-2 and other coronaviruses if they reported separate data regarding SARS-CoV-2.

Information sources

Nine databases were searched for eligible records published between December 1, 2019, and March 24, 2020: Cochrane Database of Systematic Reviews via Cochrane Library, PubMed, EMBASE, CINAHL (Cumulative Index to Nursing and Allied Health Literature), Web of Sciences, LILACS (Latin American and Caribbean Health Sciences Literature), PDQ-Evidence, WHO’s Global Research on Coronavirus Disease (COVID-19), and Epistemonikos.

The comprehensive search strategy for each database is provided in Additional file 1 and was designed and conducted in collaboration with an information specialist. All retrieved records were primarily processed in EndNote, where duplicates were removed, and records were then imported into the Covidence platform [ 23 ]. In addition to database searches, we screened reference lists of reviews included after screening records retrieved via databases.

Study selection

All searches, screening of titles and abstracts, and record selection, were performed independently by two investigators using the Covidence platform [ 23 ]. Articles deemed potentially eligible were retrieved for full-text screening carried out independently by two investigators. Discrepancies at all stages were resolved by consensus. During the screening, records published in languages other than English were translated by a native/fluent speaker.

Data collection process

We custom designed a data extraction table for this study, which was piloted by two authors independently. Data extraction was performed independently by two authors. Conflicts were resolved by consensus or by consulting a third researcher.

We extracted the following data: article identification data (authors’ name and journal of publication), search period, number of databases searched, population or settings considered, main results and outcomes observed, and number of participants. From Web of Science (Clarivate Analytics, Philadelphia, PA, USA), we extracted journal rank (quartile) and Journal Impact Factor (JIF).

We categorized the following as primary outcomes: all-cause mortality, need for and length of mechanical ventilation, length of hospitalization (in days), admission to intensive care unit (yes/no), and length of stay in the intensive care unit.

The following outcomes were categorized as exploratory: diagnostic methods used for detection of the virus, male to female ratio, clinical symptoms, pharmacological and non-pharmacological interventions, laboratory findings (full blood count, liver enzymes, C-reactive protein, d-dimer, albumin, lipid profile, serum electrolytes, blood vitamin levels, glucose levels, and any other important biomarkers), and radiological findings (using radiography, computed tomography, magnetic resonance imaging or ultrasound).

We also collected data on reporting guidelines and requirements for the publication of systematic reviews and meta-analyses from journal websites where included reviews were published.

Quality assessment in individual reviews

Two researchers independently assessed the reviews’ quality using the “A MeaSurement Tool to Assess Systematic Reviews 2 (AMSTAR 2)”. We acknowledge that the AMSTAR 2 was created as “a critical appraisal tool for systematic reviews that include randomized or non-randomized studies of healthcare interventions, or both” [ 24 ]. However, since AMSTAR 2 was designed for systematic reviews of intervention trials, and we included additional types of systematic reviews, we adjusted some AMSTAR 2 ratings and reported these in Additional file 2 .

Adherence to each item was rated as follows: yes, partial yes, no, or not applicable (such as when a meta-analysis was not conducted). The overall confidence in the results of the review is rated as “critically low”, “low”, “moderate” or “high”, according to the AMSTAR 2 guidance based on seven critical domains, which are items 2, 4, 7, 9, 11, 13, 15 as defined by AMSTAR 2 authors [ 24 ]. We reported our adherence ratings for transparency of our decision with accompanying explanations, for each item, in each included review.

One of the included systematic reviews was conducted by some members of this author team [ 25 ]. This review was initially assessed independently by two authors who were not co-authors of that review to prevent the risk of bias in assessing this study.

Synthesis of results

For data synthesis, we prepared a table summarizing each systematic review. Graphs illustrating the mortality rate and clinical symptoms were created. We then prepared a narrative summary of the methods, findings, study strengths, and limitations.

For analysis of the prevalence of clinical outcomes, we extracted data on the number of events and the total number of patients to perform proportional meta-analysis using RStudio© software, with the “meta” package (version 4.9–6), using the “metaprop” function for reviews that did not perform a meta-analysis, excluding case studies because of the absence of variance. For reviews that did not perform a meta-analysis, we presented pooled results of proportions with their respective confidence intervals (95%) by the inverse variance method with a random-effects model, using the DerSimonian-Laird estimator for τ 2 . We adjusted data using Freeman-Tukey double arcosen transformation. Confidence intervals were calculated using the Clopper-Pearson method for individual studies. We created forest plots using the RStudio© software, with the “metafor” package (version 2.1–0) and “forest” function.

Managing overlapping systematic reviews

Some of the included systematic reviews that address the same or similar research questions may include the same primary studies in overviews. Including such overlapping reviews may introduce bias when outcome data from the same primary study are included in the analyses of an overview multiple times. Thus, in summaries of evidence, multiple-counting of the same outcome data will give data from some primary studies too much influence [ 14 ]. In this overview, we did not exclude overlapping systematic reviews because, according to Cochrane’s guidance, it may be appropriate to include all relevant reviews’ results if the purpose of the overview is to present and describe the current body of evidence on a topic [ 14 ]. To avoid any bias in summary estimates associated with overlapping reviews, we generated forest plots showing data from individual systematic reviews, but the results were not pooled because some primary studies were included in multiple reviews.

Our search retrieved 1063 publications, of which 175 were duplicates. Most publications were excluded after the title and abstract analysis ( n = 860). Among the 28 studies selected for full-text screening, 10 were excluded for the reasons described in Additional file 3 , and 18 were included in the final analysis (Fig. 1 ) [ 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 , 41 , 42 ]. Reference list screening did not retrieve any additional systematic reviews.

PRISMA flow diagram

Characteristics of included reviews

Summary features of 18 systematic reviews are presented in Table 1 . They were published in 14 different journals. Only four of these journals had specific requirements for systematic reviews (with or without meta-analysis): European Journal of Internal Medicine, Journal of Clinical Medicine, Ultrasound in Obstetrics and Gynecology, and Clinical Research in Cardiology . Two journals reported that they published only invited reviews ( Journal of Medical Virology and Clinica Chimica Acta ). Three systematic reviews in our study were published as letters; one was labeled as a scoping review and another as a rapid review (Table 2 ).

All reviews were published in English, in first quartile (Q1) journals, with JIF ranging from 1.692 to 6.062. One review was empty, meaning that its search did not identify any relevant studies; i.e., no primary studies were included [ 36 ]. The remaining 17 reviews included 269 unique studies; the majority ( N = 211; 78%) were included in only a single review included in our study (range: 1 to 12). Primary studies included in the reviews were published between December 2019 and March 18, 2020, and comprised case reports, case series, cohorts, and other observational studies. We found only one review that included randomized clinical trials [ 38 ]. In the included reviews, systematic literature searches were performed from 2019 (entire year) up to March 9, 2020. Ten systematic reviews included meta-analyses. The list of primary studies found in the included systematic reviews is shown in Additional file 4 , as well as the number of reviews in which each primary study was included.

Population and study designs

Most of the reviews analyzed data from patients with COVID-19 who developed pneumonia, acute respiratory distress syndrome (ARDS), or any other correlated complication. One review aimed to evaluate the effectiveness of using surgical masks on preventing transmission of the virus [ 36 ], one review was focused on pediatric patients [ 34 ], and one review investigated COVID-19 in pregnant women [ 37 ]. Most reviews assessed clinical symptoms, laboratory findings, or radiological results.

Systematic review findings

The summary of findings from individual reviews is shown in Table 2 . Overall, all-cause mortality ranged from 0.3 to 13.9% (Fig. 2 ).

A meta-analysis of the prevalence of mortality

Clinical symptoms

Seven reviews described the main clinical manifestations of COVID-19 [ 26 , 28 , 29 , 34 , 35 , 39 , 41 ]. Three of them provided only a narrative discussion of symptoms [ 26 , 34 , 35 ]. In the reviews that performed a statistical analysis of the incidence of different clinical symptoms, symptoms in patients with COVID-19 were (range values of point estimates): fever (82–95%), cough with or without sputum (58–72%), dyspnea (26–59%), myalgia or muscle fatigue (29–51%), sore throat (10–13%), headache (8–12%), gastrointestinal disorders, such as diarrhea, nausea or vomiting (5.0–9.0%), and others (including, in one study only: dizziness 12.1%) (Figs. 3 , 4 , 5 , 6 , 7 , 8 and 9 ). Three reviews assessed cough with and without sputum together; only one review assessed sputum production itself (28.5%).

A meta-analysis of the prevalence of fever

A meta-analysis of the prevalence of cough

A meta-analysis of the prevalence of dyspnea

A meta-analysis of the prevalence of fatigue or myalgia

A meta-analysis of the prevalence of headache

A meta-analysis of the prevalence of gastrointestinal disorders

A meta-analysis of the prevalence of sore throat

Diagnostic aspects

Three reviews described methodologies, protocols, and tools used for establishing the diagnosis of COVID-19 [ 26 , 34 , 38 ]. The use of respiratory swabs (nasal or pharyngeal) or blood specimens to assess the presence of SARS-CoV-2 nucleic acid using RT-PCR assays was the most commonly used diagnostic method mentioned in the included studies. These diagnostic tests have been widely used, but their precise sensitivity and specificity remain unknown. One review included a Chinese study with clinical diagnosis with no confirmation of SARS-CoV-2 infection (patients were diagnosed with COVID-19 if they presented with at least two symptoms suggestive of COVID-19, together with laboratory and chest radiography abnormalities) [ 34 ].

Therapeutic possibilities

Pharmacological and non-pharmacological interventions (supportive therapies) used in treating patients with COVID-19 were reported in five reviews [ 25 , 27 , 34 , 35 , 38 ]. Antivirals used empirically for COVID-19 treatment were reported in seven reviews [ 25 , 27 , 34 , 35 , 37 , 38 , 41 ]; most commonly used were protease inhibitors (lopinavir, ritonavir, darunavir), nucleoside reverse transcriptase inhibitor (tenofovir), nucleotide analogs (remdesivir, galidesivir, ganciclovir), and neuraminidase inhibitors (oseltamivir). Umifenovir, a membrane fusion inhibitor, was investigated in two studies [ 25 , 35 ]. Possible supportive interventions analyzed were different types of oxygen supplementation and breathing support (invasive or non-invasive ventilation) [ 25 ]. The use of antibiotics, both empirically and to treat secondary pneumonia, was reported in six studies [ 25 , 26 , 27 , 34 , 35 , 38 ]. One review specifically assessed evidence on the efficacy and safety of the anti-malaria drug chloroquine [ 27 ]. It identified 23 ongoing trials investigating the potential of chloroquine as a therapeutic option for COVID-19, but no verifiable clinical outcomes data. The use of mesenchymal stem cells, antifungals, and glucocorticoids were described in four reviews [ 25 , 34 , 35 , 38 ].

Laboratory and radiological findings

Of the 18 reviews included in this overview, eight analyzed laboratory parameters in patients with COVID-19 [ 25 , 29 , 30 , 32 , 33 , 34 , 35 , 39 ]; elevated C-reactive protein levels, associated with lymphocytopenia, elevated lactate dehydrogenase, as well as slightly elevated aspartate and alanine aminotransferase (AST, ALT) were commonly described in those eight reviews. Lippi et al. assessed cardiac troponin I (cTnI) [ 25 ], procalcitonin [ 32 ], and platelet count [ 33 ] in COVID-19 patients. Elevated levels of procalcitonin [ 32 ] and cTnI [ 30 ] were more likely to be associated with a severe disease course (requiring intensive care unit admission and intubation). Furthermore, thrombocytopenia was frequently observed in patients with complicated COVID-19 infections [ 33 ].

Chest imaging (chest radiography and/or computed tomography) features were assessed in six reviews, all of which described a frequent pattern of local or bilateral multilobar ground-glass opacity [ 25 , 34 , 35 , 39 , 40 , 41 ]. Those six reviews showed that septal thickening, bronchiectasis, pleural and cardiac effusions, halo signs, and pneumothorax were observed in patients suffering from COVID-19.

Quality of evidence in individual systematic reviews

Table 3 shows the detailed results of the quality assessment of 18 systematic reviews, including the assessment of individual items and summary assessment. A detailed explanation for each decision in each review is available in Additional file 5 .

Using AMSTAR 2 criteria, confidence in the results of all 18 reviews was rated as “critically low” (Table 3 ). Common methodological drawbacks were: omission of prospective protocol submission or publication; use of inappropriate search strategy: lack of independent and dual literature screening and data-extraction (or methodology unclear); absence of an explanation for heterogeneity among the studies included; lack of reasons for study exclusion (or rationale unclear).

Risk of bias assessment, based on a reported methodological tool, and quality of evidence appraisal, in line with the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) method, were reported only in one review [ 25 ]. Five reviews presented a table summarizing bias, using various risk of bias tools [ 25 , 29 , 39 , 40 , 41 ]. One review analyzed “study quality” [ 37 ]. One review mentioned the risk of bias assessment in the methodology but did not provide any related analysis [ 28 ].

This overview of systematic reviews analyzed the first 18 systematic reviews published after the onset of the COVID-19 pandemic, up to March 24, 2020, with primary studies involving more than 60,000 patients. Using AMSTAR-2, we judged that our confidence in all those reviews was “critically low”. Ten reviews included meta-analyses. The reviews presented data on clinical manifestations, laboratory and radiological findings, and interventions. We found no systematic reviews on the utility of diagnostic tests.

Symptoms were reported in seven reviews; most of the patients had a fever, cough, dyspnea, myalgia or muscle fatigue, and gastrointestinal disorders such as diarrhea, nausea, or vomiting. Olfactory dysfunction (anosmia or dysosmia) has been described in patients infected with COVID-19 [ 43 ]; however, this was not reported in any of the reviews included in this overview. During the SARS outbreak in 2002, there were reports of impairment of the sense of smell associated with the disease [ 44 , 45 ].

The reported mortality rates ranged from 0.3 to 14% in the included reviews. Mortality estimates are influenced by the transmissibility rate (basic reproduction number), availability of diagnostic tools, notification policies, asymptomatic presentations of the disease, resources for disease prevention and control, and treatment facilities; variability in the mortality rate fits the pattern of emerging infectious diseases [ 46 ]. Furthermore, the reported cases did not consider asymptomatic cases, mild cases where individuals have not sought medical treatment, and the fact that many countries had limited access to diagnostic tests or have implemented testing policies later than the others. Considering the lack of reviews assessing diagnostic testing (sensitivity, specificity, and predictive values of RT-PCT or immunoglobulin tests), and the preponderance of studies that assessed only symptomatic individuals, considerable imprecision around the calculated mortality rates existed in the early stage of the COVID-19 pandemic.

Few reviews included treatment data. Those reviews described studies considered to be at a very low level of evidence: usually small, retrospective studies with very heterogeneous populations. Seven reviews analyzed laboratory parameters; those reviews could have been useful for clinicians who attend patients suspected of COVID-19 in emergency services worldwide, such as assessing which patients need to be reassessed more frequently.

All systematic reviews scored poorly on the AMSTAR 2 critical appraisal tool for systematic reviews. Most of the original studies included in the reviews were case series and case reports, impacting the quality of evidence. Such evidence has major implications for clinical practice and the use of these reviews in evidence-based practice and policy. Clinicians, patients, and policymakers can only have the highest confidence in systematic review findings if high-quality systematic review methodologies are employed. The urgent need for information during a pandemic does not justify poor quality reporting.

We acknowledge that there are numerous challenges associated with analyzing COVID-19 data during a pandemic [ 47 ]. High-quality evidence syntheses are needed for decision-making, but each type of evidence syntheses is associated with its inherent challenges.

The creation of classic systematic reviews requires considerable time and effort; with massive research output, they quickly become outdated, and preparing updated versions also requires considerable time. A recent study showed that updates of non-Cochrane systematic reviews are published a median of 5 years after the publication of the previous version [ 48 ].

Authors may register a review and then abandon it [ 49 ], but the existence of a public record that is not updated may lead other authors to believe that the review is still ongoing. A quarter of Cochrane review protocols remains unpublished as completed systematic reviews 8 years after protocol publication [ 50 ].

Rapid reviews can be used to summarize the evidence, but they involve methodological sacrifices and simplifications to produce information promptly, with inconsistent methodological approaches [ 51 ]. However, rapid reviews are justified in times of public health emergencies, and even Cochrane has resorted to publishing rapid reviews in response to the COVID-19 crisis [ 52 ]. Rapid reviews were eligible for inclusion in this overview, but only one of the 18 reviews included in this study was labeled as a rapid review.

Ideally, COVID-19 evidence would be continually summarized in a series of high-quality living systematic reviews, types of evidence synthesis defined as “ a systematic review which is continually updated, incorporating relevant new evidence as it becomes available ” [ 53 ]. However, conducting living systematic reviews requires considerable resources, calling into question the sustainability of such evidence synthesis over long periods [ 54 ].

Research reports about COVID-19 will contribute to research waste if they are poorly designed, poorly reported, or simply not necessary. In principle, systematic reviews should help reduce research waste as they usually provide recommendations for further research that is needed or may advise that sufficient evidence exists on a particular topic [ 55 ]. However, systematic reviews can also contribute to growing research waste when they are not needed, or poorly conducted and reported. Our present study clearly shows that most of the systematic reviews that were published early on in the COVID-19 pandemic could be categorized as research waste, as our confidence in their results is critically low.

Our study has some limitations. One is that for AMSTAR 2 assessment we relied on information available in publications; we did not attempt to contact study authors for clarifications or additional data. In three reviews, the methodological quality appraisal was challenging because they were published as letters, or labeled as rapid communications. As a result, various details about their review process were not included, leading to AMSTAR 2 questions being answered as “not reported”, resulting in low confidence scores. Full manuscripts might have provided additional information that could have led to higher confidence in the results. In other words, low scores could reflect incomplete reporting, not necessarily low-quality review methods. To make their review available more rapidly and more concisely, the authors may have omitted methodological details. A general issue during a crisis is that speed and completeness must be balanced. However, maintaining high standards requires proper resourcing and commitment to ensure that the users of systematic reviews can have high confidence in the results.

Furthermore, we used adjusted AMSTAR 2 scoring, as the tool was designed for critical appraisal of reviews of interventions. Some reviews may have received lower scores than actually warranted in spite of these adjustments.

Another limitation of our study may be the inclusion of multiple overlapping reviews, as some included reviews included the same primary studies. According to the Cochrane Handbook, including overlapping reviews may be appropriate when the review’s aim is “ to present and describe the current body of systematic review evidence on a topic ” [ 12 ], which was our aim. To avoid bias with summarizing evidence from overlapping reviews, we presented the forest plots without summary estimates. The forest plots serve to inform readers about the effect sizes for outcomes that were reported in each review.

Several authors from this study have contributed to one of the reviews identified [ 25 ]. To reduce the risk of any bias, two authors who did not co-author the review in question initially assessed its quality and limitations.

Finally, we note that the systematic reviews included in our overview may have had issues that our analysis did not identify because we did not analyze their primary studies to verify the accuracy of the data and information they presented. We give two examples to substantiate this possibility. Lovato et al. wrote a commentary on the review of Sun et al. [ 41 ], in which they criticized the authors’ conclusion that sore throat is rare in COVID-19 patients [ 56 ]. Lovato et al. highlighted that multiple studies included in Sun et al. did not accurately describe participants’ clinical presentations, warning that only three studies clearly reported data on sore throat [ 56 ].

In another example, Leung [ 57 ] warned about the review of Li, L.Q. et al. [ 29 ]: “ it is possible that this statistic was computed using overlapped samples, therefore some patients were double counted ”. Li et al. responded to Leung that it is uncertain whether the data overlapped, as they used data from published articles and did not have access to the original data; they also reported that they requested original data and that they plan to re-do their analyses once they receive them; they also urged readers to treat the data with caution [ 58 ]. This points to the evolving nature of evidence during a crisis.

Our study’s strength is that this overview adds to the current knowledge by providing a comprehensive summary of all the evidence synthesis about COVID-19 available early after the onset of the pandemic. This overview followed strict methodological criteria, including a comprehensive and sensitive search strategy and a standard tool for methodological appraisal of systematic reviews.

In conclusion, in this overview of systematic reviews, we analyzed evidence from the first 18 systematic reviews that were published after the emergence of COVID-19. However, confidence in the results of all the reviews was “critically low”. Thus, systematic reviews that were published early on in the pandemic could be categorized as research waste. Even during public health emergencies, studies and systematic reviews should adhere to established methodological standards to provide patients, clinicians, and decision-makers trustworthy evidence.

Availability of data and materials

All data collected and analyzed within this study are available from the corresponding author on reasonable request.

World Health Organization. Timeline - COVID-19: Available at: https://www.who.int/news/item/29-06-2020-covidtimeline . Accessed 1 June 2021.

COVID-19 Dashboard by the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University (JHU). Available at: https://coronavirus.jhu.edu/map.html . Accessed 1 June 2021.

Anzai A, Kobayashi T, Linton NM, Kinoshita R, Hayashi K, Suzuki A, et al. Assessing the Impact of Reduced Travel on Exportation Dynamics of Novel Coronavirus Infection (COVID-19). J Clin Med. 2020;9(2):601.

Chinazzi M, Davis JT, Ajelli M, Gioannini C, Litvinova M, Merler S, et al. The effect of travel restrictions on the spread of the 2019 novel coronavirus (COVID-19) outbreak. Science. 2020;368(6489):395–400. https://doi.org/10.1126/science.aba9757 .

Article   CAS   PubMed   PubMed Central   Google Scholar  

Fidahic M, Nujic D, Runjic R, Civljak M, Markotic F, Lovric Makaric Z, et al. Research methodology and characteristics of journal articles with original data, preprint articles and registered clinical trial protocols about COVID-19. BMC Med Res Methodol. 2020;20(1):161. https://doi.org/10.1186/s12874-020-01047-2 .

EPPI Centre . COVID-19: a living systematic map of the evidence. Available at: http://eppi.ioe.ac.uk/cms/Projects/DepartmentofHealthandSocialCare/Publishedreviews/COVID-19Livingsystematicmapoftheevidence/tabid/3765/Default.aspx . Accessed 1 June 2021.

NCBI SARS-CoV-2 Resources. Available at: https://www.ncbi.nlm.nih.gov/sars-cov-2/ . Accessed 1 June 2021.

Gustot T. Quality and reproducibility during the COVID-19 pandemic. JHEP Rep. 2020;2(4):100141. https://doi.org/10.1016/j.jhepr.2020.100141 .

Article   PubMed   PubMed Central   Google Scholar  

Kodvanj, I., et al., Publishing of COVID-19 Preprints in Peer-reviewed Journals, Preprinting Trends, Public Discussion and Quality Issues. Preprint article. bioRxiv 2020.11.23.394577; doi: https://doi.org/10.1101/2020.11.23.394577 .

Dobler CC. Poor quality research and clinical practice during COVID-19. Breathe (Sheff). 2020;16(2):200112. https://doi.org/10.1183/20734735.0112-2020 .

Article   Google Scholar  

Bastian H, Glasziou P, Chalmers I. Seventy-five trials and eleven systematic reviews a day: how will we ever keep up? PLoS Med. 2010;7(9):e1000326. https://doi.org/10.1371/journal.pmed.1000326 .

Lunny C, Brennan SE, McDonald S, McKenzie JE. Toward a comprehensive evidence map of overview of systematic review methods: paper 1-purpose, eligibility, search and data extraction. Syst Rev. 2017;6(1):231. https://doi.org/10.1186/s13643-017-0617-1 .

Pollock M, Fernandes RM, Becker LA, Pieper D, Hartling L. Chapter V: Overviews of Reviews. In: Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, Welch VA (editors). Cochrane Handbook for Systematic Reviews of Interventions version 6.1 (updated September 2020). Cochrane. 2020. Available from www.training.cochrane.org/handbook .

Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al. Cochrane handbook for systematic reviews of interventions version 6.1 (updated September 2020). Cochrane. 2020; Available from www.training.cochrane.org/handbook .

Pollock M, Fernandes RM, Newton AS, Scott SD, Hartling L. The impact of different inclusion decisions on the comprehensiveness and complexity of overviews of reviews of healthcare interventions. Syst Rev. 2019;8(1):18. https://doi.org/10.1186/s13643-018-0914-3 .

Pollock M, Fernandes RM, Newton AS, Scott SD, Hartling L. A decision tool to help researchers make decisions about including systematic reviews in overviews of reviews of healthcare interventions. Syst Rev. 2019;8(1):29. https://doi.org/10.1186/s13643-018-0768-8 .

Hunt H, Pollock A, Campbell P, Estcourt L, Brunton G. An introduction to overviews of reviews: planning a relevant research question and objective for an overview. Syst Rev. 2018;7(1):39. https://doi.org/10.1186/s13643-018-0695-8 .

Pollock M, Fernandes RM, Pieper D, Tricco AC, Gates M, Gates A, et al. Preferred reporting items for overviews of reviews (PRIOR): a protocol for development of a reporting guideline for overviews of reviews of healthcare interventions. Syst Rev. 2019;8(1):335. https://doi.org/10.1186/s13643-019-1252-9 .

Moher D, Liberati A, Tetzlaff J, Altman DG, PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Open Med. 2009;3(3):e123–30.

Krnic Martinic M, Pieper D, Glatt A, Puljak L. Definition of a systematic review used in overviews of systematic reviews, meta-epidemiological studies and textbooks. BMC Med Res Methodol. 2019;19(1):203. https://doi.org/10.1186/s12874-019-0855-0 .

Puljak L. If there is only one author or only one database was searched, a study should not be called a systematic review. J Clin Epidemiol. 2017;91:4–5. https://doi.org/10.1016/j.jclinepi.2017.08.002 .

Article   PubMed   Google Scholar  

Gates M, Gates A, Guitard S, Pollock M, Hartling L. Guidance for overviews of reviews continues to accumulate, but important challenges remain: a scoping review. Syst Rev. 2020;9(1):254. https://doi.org/10.1186/s13643-020-01509-0 .

Covidence - systematic review software. Available at: https://www.covidence.org/ . Accessed 1 June 2021.

Shea BJ, Reeves BC, Wells G, Thuku M, Hamel C, Moran J, et al. AMSTAR 2: a critical appraisal tool for systematic reviews that include randomised or non-randomised studies of healthcare interventions, or both. BMJ. 2017;358:j4008.

Borges do Nascimento IJ, et al. Novel Coronavirus Infection (COVID-19) in Humans: A Scoping Review and Meta-Analysis. J Clin Med. 2020;9(4):941.

Article   PubMed Central   Google Scholar  

Adhikari SP, Meng S, Wu YJ, Mao YP, Ye RX, Wang QZ, et al. Epidemiology, causes, clinical manifestation and diagnosis, prevention and control of coronavirus disease (COVID-19) during the early outbreak period: a scoping review. Infect Dis Poverty. 2020;9(1):29. https://doi.org/10.1186/s40249-020-00646-x .

Cortegiani A, Ingoglia G, Ippolito M, Giarratano A, Einav S. A systematic review on the efficacy and safety of chloroquine for the treatment of COVID-19. J Crit Care. 2020;57:279–83. https://doi.org/10.1016/j.jcrc.2020.03.005 .

Li B, Yang J, Zhao F, Zhi L, Wang X, Liu L, et al. Prevalence and impact of cardiovascular metabolic diseases on COVID-19 in China. Clin Res Cardiol. 2020;109(5):531–8. https://doi.org/10.1007/s00392-020-01626-9 .

Article   CAS   PubMed   Google Scholar  

Li LQ, Huang T, Wang YQ, Wang ZP, Liang Y, Huang TB, et al. COVID-19 patients’ clinical characteristics, discharge rate, and fatality rate of meta-analysis. J Med Virol. 2020;92(6):577–83. https://doi.org/10.1002/jmv.25757 .

Lippi G, Lavie CJ, Sanchis-Gomar F. Cardiac troponin I in patients with coronavirus disease 2019 (COVID-19): evidence from a meta-analysis. Prog Cardiovasc Dis. 2020;63(3):390–1. https://doi.org/10.1016/j.pcad.2020.03.001 .

Lippi G, Henry BM. Active smoking is not associated with severity of coronavirus disease 2019 (COVID-19). Eur J Intern Med. 2020;75:107–8. https://doi.org/10.1016/j.ejim.2020.03.014 .

Lippi G, Plebani M. Procalcitonin in patients with severe coronavirus disease 2019 (COVID-19): a meta-analysis. Clin Chim Acta. 2020;505:190–1. https://doi.org/10.1016/j.cca.2020.03.004 .

Lippi G, Plebani M, Henry BM. Thrombocytopenia is associated with severe coronavirus disease 2019 (COVID-19) infections: a meta-analysis. Clin Chim Acta. 2020;506:145–8. https://doi.org/10.1016/j.cca.2020.03.022 .

Ludvigsson JF. Systematic review of COVID-19 in children shows milder cases and a better prognosis than adults. Acta Paediatr. 2020;109(6):1088–95. https://doi.org/10.1111/apa.15270 .

Lupia T, Scabini S, Mornese Pinna S, di Perri G, de Rosa FG, Corcione S. 2019 novel coronavirus (2019-nCoV) outbreak: a new challenge. J Glob Antimicrob Resist. 2020;21:22–7. https://doi.org/10.1016/j.jgar.2020.02.021 .

Marasinghe, K.M., A systematic review investigating the effectiveness of face mask use in limiting the spread of COVID-19 among medically not diagnosed individuals: shedding light on current recommendations provided to individuals not medically diagnosed with COVID-19. Research Square. Preprint article. doi : https://doi.org/10.21203/rs.3.rs-16701/v1 . 2020 .

Mullins E, Evans D, Viner RM, O’Brien P, Morris E. Coronavirus in pregnancy and delivery: rapid review. Ultrasound Obstet Gynecol. 2020;55(5):586–92. https://doi.org/10.1002/uog.22014 .

Pang J, Wang MX, Ang IYH, Tan SHX, Lewis RF, Chen JIP, et al. Potential Rapid Diagnostics, Vaccine and Therapeutics for 2019 Novel coronavirus (2019-nCoV): a systematic review. J Clin Med. 2020;9(3):623.

Rodriguez-Morales AJ, Cardona-Ospina JA, Gutiérrez-Ocampo E, Villamizar-Peña R, Holguin-Rivera Y, Escalera-Antezana JP, et al. Clinical, laboratory and imaging features of COVID-19: a systematic review and meta-analysis. Travel Med Infect Dis. 2020;34:101623. https://doi.org/10.1016/j.tmaid.2020.101623 .

Salehi S, Abedi A, Balakrishnan S, Gholamrezanezhad A. Coronavirus disease 2019 (COVID-19): a systematic review of imaging findings in 919 patients. AJR Am J Roentgenol. 2020;215(1):87–93. https://doi.org/10.2214/AJR.20.23034 .

Sun P, Qie S, Liu Z, Ren J, Li K, Xi J. Clinical characteristics of hospitalized patients with SARS-CoV-2 infection: a single arm meta-analysis. J Med Virol. 2020;92(6):612–7. https://doi.org/10.1002/jmv.25735 .

Yang J, Zheng Y, Gou X, Pu K, Chen Z, Guo Q, et al. Prevalence of comorbidities and its effects in patients infected with SARS-CoV-2: a systematic review and meta-analysis. Int J Infect Dis. 2020;94:91–5. https://doi.org/10.1016/j.ijid.2020.03.017 .

Bassetti M, Vena A, Giacobbe DR. The novel Chinese coronavirus (2019-nCoV) infections: challenges for fighting the storm. Eur J Clin Investig. 2020;50(3):e13209. https://doi.org/10.1111/eci.13209 .

Article   CAS   Google Scholar  

Hwang CS. Olfactory neuropathy in severe acute respiratory syndrome: report of a case. Acta Neurol Taiwanica. 2006;15(1):26–8.

Google Scholar  

Suzuki M, Saito K, Min WP, Vladau C, Toida K, Itoh H, et al. Identification of viruses in patients with postviral olfactory dysfunction. Laryngoscope. 2007;117(2):272–7. https://doi.org/10.1097/01.mlg.0000249922.37381.1e .

Rajgor DD, Lee MH, Archuleta S, Bagdasarian N, Quek SC. The many estimates of the COVID-19 case fatality rate. Lancet Infect Dis. 2020;20(7):776–7. https://doi.org/10.1016/S1473-3099(20)30244-9 .

Wolkewitz M, Puljak L. Methodological challenges of analysing COVID-19 data during the pandemic. BMC Med Res Methodol. 2020;20(1):81. https://doi.org/10.1186/s12874-020-00972-6 .

Rombey T, Lochner V, Puljak L, Könsgen N, Mathes T, Pieper D. Epidemiology and reporting characteristics of non-Cochrane updates of systematic reviews: a cross-sectional study. Res Synth Methods. 2020;11(3):471–83. https://doi.org/10.1002/jrsm.1409 .

Runjic E, Rombey T, Pieper D, Puljak L. Half of systematic reviews about pain registered in PROSPERO were not published and the majority had inaccurate status. J Clin Epidemiol. 2019;116:114–21. https://doi.org/10.1016/j.jclinepi.2019.08.010 .

Runjic E, Behmen D, Pieper D, Mathes T, Tricco AC, Moher D, et al. Following Cochrane review protocols to completion 10 years later: a retrospective cohort study and author survey. J Clin Epidemiol. 2019;111:41–8. https://doi.org/10.1016/j.jclinepi.2019.03.006 .

Tricco AC, Antony J, Zarin W, Strifler L, Ghassemi M, Ivory J, et al. A scoping review of rapid review methods. BMC Med. 2015;13(1):224. https://doi.org/10.1186/s12916-015-0465-6 .

COVID-19 Rapid Reviews: Cochrane’s response so far. Available at: https://training.cochrane.org/resource/covid-19-rapid-reviews-cochrane-response-so-far . Accessed 1 June 2021.

Cochrane. Living systematic reviews. Available at: https://community.cochrane.org/review-production/production-resources/living-systematic-reviews . Accessed 1 June 2021.

Millard T, Synnot A, Elliott J, Green S, McDonald S, Turner T. Feasibility and acceptability of living systematic reviews: results from a mixed-methods evaluation. Syst Rev. 2019;8(1):325. https://doi.org/10.1186/s13643-019-1248-5 .

Babic A, Poklepovic Pericic T, Pieper D, Puljak L. How to decide whether a systematic review is stable and not in need of updating: analysis of Cochrane reviews. Res Synth Methods. 2020;11(6):884–90. https://doi.org/10.1002/jrsm.1451 .

Lovato A, Rossettini G, de Filippis C. Sore throat in COVID-19: comment on “clinical characteristics of hospitalized patients with SARS-CoV-2 infection: a single arm meta-analysis”. J Med Virol. 2020;92(7):714–5. https://doi.org/10.1002/jmv.25815 .

Leung C. Comment on Li et al: COVID-19 patients’ clinical characteristics, discharge rate, and fatality rate of meta-analysis. J Med Virol. 2020;92(9):1431–2. https://doi.org/10.1002/jmv.25912 .

Li LQ, Huang T, Wang YQ, Wang ZP, Liang Y, Huang TB, et al. Response to Char’s comment: comment on Li et al: COVID-19 patients’ clinical characteristics, discharge rate, and fatality rate of meta-analysis. J Med Virol. 2020;92(9):1433. https://doi.org/10.1002/jmv.25924 .

Download references

Acknowledgments

We thank Catherine Henderson DPhil from Swanscoe Communications for pro bono medical writing and editing support. We acknowledge support from the Covidence Team, specifically Anneliese Arno. We thank the whole International Network of Coronavirus Disease 2019 (InterNetCOVID-19) for their commitment and involvement. Members of the InterNetCOVID-19 are listed in Additional file 6 . We thank Pavel Cerny and Roger Crosthwaite for guiding the team supervisor (IJBN) on human resources management.

This research received no external funding.

Author information

Authors and affiliations.

University Hospital and School of Medicine, Universidade Federal de Minas Gerais, Belo Horizonte, Minas Gerais, Brazil

Israel Júnior Borges do Nascimento & Milena Soriano Marcolino

Medical College of Wisconsin, Milwaukee, WI, USA

Israel Júnior Borges do Nascimento

Helene Fuld Health Trust National Institute for Evidence-based Practice in Nursing and Healthcare, College of Nursing, The Ohio State University, Columbus, OH, USA

Dónal P. O’Mathúna

School of Nursing, Psychotherapy and Community Health, Dublin City University, Dublin, Ireland

Department of Anesthesiology, Intensive Care and Pain Medicine, University of Münster, Münster, Germany

Thilo Caspar von Groote

Department of Sport and Health Science, Technische Universität München, Munich, Germany

Hebatullah Mohamed Abdulazeem

School of Health Sciences, Faculty of Health and Medicine, The University of Newcastle, Callaghan, Australia

Ishanka Weerasekara

Department of Physiotherapy, Faculty of Allied Health Sciences, University of Peradeniya, Peradeniya, Sri Lanka

Cochrane Croatia, University of Split, School of Medicine, Split, Croatia

Ana Marusic, Irena Zakarija-Grkovic & Tina Poklepovic Pericic

Center for Evidence-Based Medicine and Health Care, Catholic University of Croatia, Ilica 242, 10000, Zagreb, Croatia

Livia Puljak

Cochrane Brazil, Evidence-Based Health Program, Universidade Federal de São Paulo, São Paulo, Brazil

Vinicius Tassoni Civile & Alvaro Nagib Atallah

Yorkville University, Fredericton, New Brunswick, Canada

Santino Filoso

Laboratory for Industrial and Applied Mathematics (LIAM), Department of Mathematics and Statistics, York University, Toronto, Ontario, Canada

Nicola Luigi Bragazzi

You can also search for this author in PubMed   Google Scholar

Contributions

IJBN conceived the research idea and worked as a project coordinator. DPOM, TCVG, HMA, IW, AM, LP, VTC, IZG, TPP, ANA, SF, NLB and MSM were involved in data curation, formal analysis, investigation, methodology, and initial draft writing. All authors revised the manuscript critically for the content. The author(s) read and approved the final manuscript.

Corresponding author

Correspondence to Livia Puljak .

Ethics declarations

Ethics approval and consent to participate.

Not required as data was based on published studies.

Consent for publication

Not applicable.

Competing interests

The authors declare no conflict of interest.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1: appendix 1..

Search strategies used in the study.

Additional file 2: Appendix 2.

Adjusted scoring of AMSTAR 2 used in this study for systematic reviews of studies that did not analyze interventions.

Additional file 3: Appendix 3.

List of excluded studies, with reasons.

Additional file 4: Appendix 4.

Table of overlapping studies, containing the list of primary studies included, their visual overlap in individual systematic reviews, and the number in how many reviews each primary study was included.

Additional file 5: Appendix 5.

A detailed explanation of AMSTAR scoring for each item in each review.

Additional file 6: Appendix 6.

List of members and affiliates of International Network of Coronavirus Disease 2019 (InterNetCOVID-19).

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Borges do Nascimento, I.J., O’Mathúna, D.P., von Groote, T.C. et al. Coronavirus disease (COVID-19) pandemic: an overview of systematic reviews. BMC Infect Dis 21 , 525 (2021). https://doi.org/10.1186/s12879-021-06214-4

Download citation

Received : 12 April 2020

Accepted : 19 May 2021

Published : 04 June 2021

DOI : https://doi.org/10.1186/s12879-021-06214-4

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Coronavirus
• Evidence-based medicine
• Infectious diseases

BMC Infectious Diseases

ISSN: 1471-2334

• General enquiries: [email protected]

• Fact sheets
• Facts in pictures
• Publications
• Questions and answers
• Tools and toolkits
• Endometriosis
• Excessive heat
• Mental disorders
• Polycystic ovary syndrome
• All countries
• Eastern Mediterranean
• South-East Asia
• Western Pacific
• Data by country
• Country presence 
• Country strengthening 
• Country cooperation strategies 
• News releases
• Feature stories
• Press conferences
• Commentaries
• Photo library
• Afghanistan
• Cholera 
• Coronavirus disease (COVID-19)
• Greater Horn of Africa
• Israel and occupied Palestinian territory
• Disease Outbreak News
• Situation reports
• Weekly Epidemiological Record
• Surveillance
• Health emergency appeal
• International Health Regulations
• Independent Oversight and Advisory Committee
• Classifications
• Data collections
• Global Health Observatory
• Global Health Estimates
• Mortality Database
• Sustainable Development Goals
• Health Inequality Monitor
• Global Progress
• World Health Statistics
• Partnerships
• Committees and advisory groups
• Collaborating centres
• Technical teams
• Organizational structure
• Initiatives
• General Programme of Work
• WHO Academy
• Investment in WHO
• WHO Foundation
• External audit
• Financial statements
• Internal audit and investigations 
• Programme Budget
• Results reports
• Governing bodies
• World Health Assembly
• Executive Board
• Member States Portal
• Coronavirus disease (COVID-19) /

Global research on coronavirus disease (COVID-19)

Section navigation.

• WHO COVID-19 Solidarity Therapeutics Trial
• "Solidarity II" global serologic study for COVID-19
• Accelerating a safe and effective COVID-19 vaccine
• Solidarity Trial Vaccines 

WHO is bringing the world’s scientists and global health professionals together to accelerate the research and development process, and develop new norms and standards to contain the spread of the coronavirus pandemic and help care for those affected.

The  R&D Blueprint  has been activated to accelerate diagnostics, vaccines and therapeutics for this novel coronavirus. 

The solidarity of all countries will be essential to ensure equitable access to COVID-19 health products.

WHO COVID-19 Research Database

The WHO COVID-19 Research Database was a resource created in response to the Public Health Emergency of International Concern (PHEIC). It contained citations with abstracts to scientific articles, reports, books, preprints, and clinical trials on COVID-19 and related literature. The WHO Covid-19 Research Database was maintained by the WHO Library & Digital Information Networks and  was funded by COVID-19 emergency funds. The database was built by BIREME, the Specialized Center of PAHO/AMRO. Its content spanned the time period March 2020 to June 2023. It has now been archived, and no longer searchable since January 2024. 

EDITORIAL article

Editorial: coronavirus disease (covid-19): the impact and role of mass media during the pandemic.

• 1 Department of Social and Organizational Psychology, Iscte-University Institute of Lisbon, CIS-IUL, Lisbon, Portugal
• 2 Department of Psychology and Social Work, Mid Sweden University, Östersund, Sweden
• 3 Department of Psychiatry and Psychotherapy, Medical School and University Hospital, Eberhard Karls University of Tübingen, Tübingen, Germany

Editorial on the Research Topic Coronavirus Disease (COVID-19): The Impact and Role of Mass Media During the Pandemic

The outbreak of the coronavirus disease 2019 (COVID-19) has created a global health crisis that had a deep impact on the way we perceive our world and our everyday lives. Not only has the rate of contagion and patterns of transmission threatened our sense of agency, but the safety measures to contain the spread of the virus also required social and physical distancing, preventing us from finding solace in the company of others. Within this context, we launched our Research Topic on March 27th, 2020, and invited researchers to address the Impact and Role of Mass Media During the Pandemic on our lives at individual and social levels.

Despite all the hardships, disruption, and uncertainty brought by the pandemic, we received diverse and insightful manuscript proposals. Frontiers in Psychology published 15 articles, involving 61 authors from 8 countries, which were included in distinct specialized sections, including Health Psychology, Personality and Social Psychology, Emotion Science, and Organizational Psychology. Despite the diversity of this collective endeavor, the contributions fall into four areas of research: (1) the use of media in public health communication; (2) the diffusion of false information; (3) the compliance with the health recommendations; and (4) how media use relates to mental health and well-being.

A first line of research includes contributions examining the use of media in public health communication. Drawing on media messages used in previous health crises, such as Ebola and Zika, Hauer and Sood describe how health organizations use media. They offer a set of recommendations for COVID-19 related media messages, including the importance of message framing, interactive public forums with up-to-date information, and an honest communication about what is known and unknown about the pandemic and the virus. Following a content analysis approach, Parvin et al. studied the representations of COVID-19 in the opinion section of five Asian e-newspapers. The authors identified eight main issues (health and drugs, preparedness and awareness, social welfare and humanity, governance and institutions, the environment and wildlife, politics, innovation and technology, and the economy) and examined how e-newspapers from these countries attributed different weights to these issues and how this relates to the countries' cultural specificity. Raccanello et al. show how the internet can be a platform to disseminate a public campaign devised to inform adults about coping strategies that could help children and teenagers deal with the challenges of the pandemic. The authors examined the dissemination of the program through the analysis of website traffic, showing that in the 40 days following publication, the website reached 6,090 visits.

A second related line of research that drew the concern of researchers was the diffusion of false information about COVID-19 through the media. Lobato et al. examined the role of distinct individual differences (political orientation, social dominance orientation, traditionalism, conspiracy ideation, attitudes about science) on the willingness to share misinformation about COVID-19 over social media. The misinformation topics varied between the severity and spread of COVID-19, treatment and prevention, conspiracy theories, and miscellaneous unverifiable claims. Their results from 296 adult participants (Mage = 36.23; 117 women) suggest two different profiles. One indicating that those reporting more liberal positions and lower social dominance were less willing to share conspiracy misinformation. The other profile indicated that participants scoring high on social dominance and low in traditionalism were more willing to share both conspiracy and other miscellaneous claims, but less willing to share misinformation about the severity and spread of COVID-19. Their findings can have relevant contributions for the identification of specific individual profiles related to the widespread of distinct types of misinformation. Dhanani and Franz examined a sample of 1,141 adults (Mage = 44.66; 46.9% female, 74.7% White ethnic identity) living in the United States in March 2020. The authors examined how media consumption and information source were related to knowledge about COVID-19, the endorsement of misinformation about COVID-19, and prejudice toward Asian Americans. Higher levels of trust in informational sources such as public health organizations (e.g., Center for Disease Control) was associated with greater knowledge, lower endorsement of misinformation, and less prejudice toward Asian Americans. Media source was associated with distinct levels of knowledge, willingness to endorsement misinformation and prejudice toward American Asians, with social media use (e.g., Twitter, Facebook) being related with a lower knowledge about COVID-19, higher endorsement of misinformation, and stronger prejudice toward Asian Americans.

A third line of research addressed the factors that could contribute to compliance with the health recommendations to avoid the spread of the disease. Vai et al. studied early pre-lockdown risk perceptions about COVID-19 and the trust in media sources among 2,223 Italians (Mage = 36.4, 69.2% female). They found that the perceived usefulness of the containment measures (e.g., social distancing) was related to threat perception and efficacy beliefs. Lower threat perception was associated with less perception of utility of the containment measures. Although most participants considered themselves and others capable of taking preventive measures, they saw the measures as generally ineffective. Participants acknowledged using the internet as their main source of information and considered health organizations' websites as the most trustworthy source. Albeit frequently used, social media was in general considered an unreliable source of information. Tomczyk et al. studied knowledge about preventive behaviors, risk perception, stigmatizing attitudes (support for discrimination and blame), and sociodemographic data (e.g., age, gender, country of origin, education level, region, persons per household) as predictors of compliance with the behavioral recommendations among 157 Germans, (age range: 18–77 years, 80% female). Low compliance was associated with male gender, younger age, and lower public stigma. Regarding stigmatizing attitudes, the authors only found a relation between support for discrimination (i.e., support for compulsory measures) and higher intention to comply with recommendations. Mahmood et al. studied the relation between social media use, risk perception, preventive behaviors, and self-efficacy in a sample of 310 Pakistani adults (54.2% female). The authors found social media use to be positively related to self-efficacy and perceived threat, which were both positively related to preventive behaviors (e.g., hand hygiene, social distancing). Information credibility was also related to compliance with health recommendations. Lep et al. examined the relationship between information source perceived credibility and trust, and participants' levels of self-protective behavior among 1,718 Slovenians (age range: 18–81 years, 81.7% female). The authors found that scientists, general practitioners (family doctors), and the National Institute of Public Health were perceived as the more credible source of information, while social media and government officials received the lowest ratings. Perceived information credibility was found to be associated with lower levels of negative emotional responses (e.g., nervousness, helplessness) and a higher level of observance of self-protective measures (e.g., hand washing). Siebenhaar et al. also studied the link between compliance, distress by information, and information avoidance. They examined the online survey responses of 1,059 adults living in Germany (Mage = 39.53, 79.4% female). Their results suggested that distress by information could lead to higher compliance with preventive measures. Distress by information was also associated with higher information avoidance, which in turn is related to less compliance. Gantiva et al. studied the effectiveness of different messages regarding the intentions toward self-care behaviors, perceived efficacy to motivate self-care behaviors in others, perceived risk, and perceived message strength, in a sample of 319 Colombians (age range: 18–60 years, 69.9% female). Their experiment included the manipulation of message framing (gain vs. loss) and message content (economy vs. health). Participants judged gain-frame health related messages to be stronger and more effective in changing self-behavior, whereas loss-framed health messages resulted in increased perceived risk. Rahn et al. offer a comparative view of compliance and risk perception, examining three hazard types: COVID-19 pandemic, violent acts, and severe weather. With a sample of 403 Germans (age range: 18–89 years, 72% female), they studied how age, gender, previous hazard experience and different components of risk appraisal (perceived severity, anticipated negative emotions, anticipatory worry, and risk perception) were related to the intention to comply with behavioral recommendations. They found that higher age predicted compliance with health recommendations to prevent COVID-19, anticipatory worry predicted compliance with warning messages regarding violent acts, and women complied more often with severe weather recommendations than men.

A fourth line of research examined media use, mental health and well-being during the COVID-19 pandemic. Gabbiadini et al. addressed the use of digital technology (e.g., voice/video calls, online games, watching movies in party mode) to stay connected with others during lockdown. Participants, 465 Italians (age range: 18–73 years, 348 female), reported more perceived social support associated with the use of these digital technologies, which in turn was associated with fewer feelings of loneliness, boredom, anger, and higher sense of belongingness. Muñiz-Velázquez et al. compared the media habits of 249 Spanish adults (Mage = 42.06, 53.8% female) before and during confinement. They compared the type of media consumed (e.g., watching TV series, listening to radio, watching news) and found the increased consumption of TV and social networking sites during confinement to be negatively associated with reported level of happiness. People who reported higher levels of well-being also reported watching less TV and less use of social networking sites. Majeed et al. , on the other hand, examined the relation between problematic social media use, fear of COVID-19, depression, and mindfulness. Their study, involving 267 Pakistani adults (90 female), suggested trait mindfulness had a buffer effect, reducing the impact of problematic media use and fear of COVID-19 on depression.

Taken together, these findings highlight how using different frames for mass media gives a more expansive view of its positive and negative roles, but also showcase the major concerns in the context of a pandemic crisis. As limitations we highlight the use of cross-sectional designs in most studies, not allowing to establish true inferences of causal relationships. The outcome of some studies may also be limited by the unbalanced number of female and male participants, by the non-probability sampling method used, and by the restricted time frame in which the research occurred. Nevertheless, we are confident that all the selected studies in our Research Topic bring important and enduring contributions to the understanding of how media, individual differences, and social factors intertwine to shape our lives, which can also be useful to guide public policies during these challenging times.

Author Contributions

PA: conceptualization, writing the original draft, funding acquisition, writing—review, and editing. FE: conceptualization, writing—review, and editing. MP: writing—review and editing. NP: conceptualization, writing the original draft, writing—review, and editing. All authors approved the submitted version.

PA and NP received partial support to work on this Research Topic through Fundação para a Ciência e Tecnologia (FCT) with reference to the project PTDC/CCI-INF/29234/2017. MP contribution was supported by the German Research Foundation (DFG, PA847/22-1 and PA847/25-1). The authors are independent of the funders.

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher's Note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Acknowledgments

We would like to express our gratitude to all the authors who proposed their work, all the researchers who reviewed the submissions to this Research Topic, and to Rob Richards for proofreading the Editorial manuscript.

Keywords: COVID-19, coronavirus disease, mass media, health communication, prevention, intervention, social behavioral changes

Citation: Arriaga P, Esteves F, Pavlova MA and Piçarra N (2021) Editorial: Coronavirus Disease (COVID-19): The Impact and Role of Mass Media During the Pandemic. Front. Psychol. 12:729238. doi: 10.3389/fpsyg.2021.729238

Received: 22 June 2021; Accepted: 30 July 2021; Published: 23 August 2021.

Edited and reviewed by: Eduard Brandstätter , Johannes Kepler University of Linz, Austria

Copyright © 2021 Arriaga, Esteves, Pavlova and Piçarra. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Patrícia Arriaga, patricia.arriaga@iscte-iul.pt

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 11 February 2021

Methodological quality of COVID-19 clinical research

• Richard G. Jung   ORCID: orcid.org/0000-0002-8570-6736 1 , 2 , 3   na1 ,
• Pietro Di Santo 1 , 2 , 4 , 5   na1 ,
• Cole Clifford 6 ,
• Graeme Prosperi-Porta 7 ,
• Stephanie Skanes 6 ,
• Annie Hung 8 ,
• Simon Parlow 4 ,
• Sarah Visintini   ORCID: orcid.org/0000-0001-6966-1753 9 ,
• F. Daniel Ramirez   ORCID: orcid.org/0000-0002-4350-1652 1 , 4 , 10 , 11 ,
• Trevor Simard 1 , 2 , 3 , 4 , 12 &
• Benjamin Hibbert   ORCID: orcid.org/0000-0003-0906-1363 2 , 3 , 4  

Nature Communications volume  12 , Article number:  943 ( 2021 ) Cite this article

14k Accesses

91 Citations

236 Altmetric

Metrics details

• Infectious diseases
• Public health

The COVID-19 pandemic began in early 2020 with major health consequences. While a need to disseminate information to the medical community and general public was paramount, concerns have been raised regarding the scientific rigor in published reports. We performed a systematic review to evaluate the methodological quality of currently available COVID-19 studies compared to historical controls. A total of 9895 titles and abstracts were screened and 686 COVID-19 articles were included in the final analysis. Comparative analysis of COVID-19 to historical articles reveals a shorter time to acceptance (13.0[IQR, 5.0–25.0] days vs. 110.0[IQR, 71.0–156.0] days in COVID-19 and control articles, respectively; p  < 0.0001). Furthermore, methodological quality scores are lower in COVID-19 articles across all study designs. COVID-19 clinical studies have a shorter time to publication and have lower methodological quality scores than control studies in the same journal. These studies should be revisited with the emergence of stronger evidence.

Similar content being viewed by others

Clinical presentations, laboratory and radiological findings, and treatments for 11,028 COVID-19 patients: a systematic review and meta-analysis

Improving clinical paediatric research and learning from COVID-19: recommendations by the Conect4Children expert advice group

The effect of influenza vaccine in reducing the severity of clinical outcomes in patients with COVID-19: a systematic review and meta-analysis

Introduction.

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic spread globally in early 2020 with substantial health and economic consequences. This was associated with an exponential increase in scientific publications related to the coronavirus disease 2019 (COVID-19) in order to rapidly elucidate the natural history and identify diagnostic and therapeutic tools 1 .

While a need to rapidly disseminate information to the medical community, governmental agencies, and general public was paramount—major concerns have been raised regarding the scientific rigor in the literature 2 . Poorly conducted studies may originate from failure at any of the four consecutive research stages: (1) choice of research question relevant to patient care, (2) quality of research design 3 , (3) adequacy of publication, and (4) quality of research reports. Furthermore, evidence-based medicine relies on a hierarchy of evidence, ranging from the highest level of randomized controlled trials (RCT) to the lowest level of case series and case reports 4 .

Given the implications for clinical care, policy decision making, and concerns regarding methodological and peer-review standards for COVID-19 research 5 , we performed a formal evaluation of the methodological quality of published COVID-19 literature. Specifically, we undertook a systematic review to identify COVID-19 clinical literature and matched them to historical controls to formally evaluate the following: (1) the methodological quality of COVID-19 studies using established quality tools and checklists, (2) the methodological quality of COVID-19 studies, stratified by median time to acceptance, geographical regions, and journal impact factor and (3) a comparison of COVID-19 methodological quality to matched controls.

Herein, we show that COVID-19 articles are associated with lower methodological quality scores. Moreover, in a matched cohort analysis with control articles from the same journal, we reveal that COVID-19 articles are associated with lower quality scores and shorter time from submission to acceptance. Ultimately, COVID-19 clinical studies should be revisited with the emergence of stronger evidence.

Article selection

A total of 14787 COVID-19 papers were identified as of May 14, 2020 and 4892 duplicate articles were removed. In total, 9895 titles and abstracts were screened, and 9101 articles were excluded due to the study being pre-clinical in nature, case report, case series <5 patients, in a language other than English, reviews (including systematic reviews), study protocols or methods, and other coronavirus variants with an overall inter-rater study inclusion agreement of 96.7% ( κ  = 0.81; 95% CI, 0.79–0.83). A total number of 794 full texts were reviewed for eligibility. Over 108 articles were excluded for ineligible study design or publication type (such as letter to the editors, editorials, case reports or case series <5 patients), wrong patient population, non-English language, duplicate articles, wrong outcomes and publication in a non-peer-reviewed journal. Ultimately, 686 articles were identified with an inter-rater agreement of 86.5% ( κ  = 0.68; 95% CI, 0.67–0.70) (Fig.  1 ).

A total of 14787 articles were identified and 4892 duplicate articles were removed. Overall, 9895 articles were screened by title and abstract leaving 794 articles for full-text screening. Over 108 articles were excluded, leaving a total of 686 articles that underwent methodological quality assessment.

COVID-19 literature methodological quality

Most studies originated from Asia/Oceania with 469 (68.4%) studies followed by Europe with 139 (20.3%) studies, and the Americas with 78 (11.4%) studies. Of included studies, 380 (55.4%) were case series, 199 (29.0%) were cohort, 63 (9.2%) were diagnostic, 38 (5.5%) were case–control, and 6 (0.9%) were RCTs. Most studies (590, 86.0%) were retrospective in nature, 620 (90.4%) reported the sex of patients, and 7 (2.3%) studies excluding case series calculated their sample size a priori. The method of SARS-CoV-2 diagnosis was reported in 558 studies (81.3%) and ethics approval was obtained in 556 studies (81.0%). Finally, journal impact factor of COVID-19 manuscripts was 4.7 (IQR, 2.9–7.6) with a time to acceptance of 13.0 (IQR, 5.0–25.0) days (Table  1 ).

Overall, when COVID-19 articles were stratified by study design, a mean case series score (out of 5) (SD) of 3.3 (1.1), mean NOS cohort study score (out of 8) of 5.8 (1.5), mean NOS case–control study score (out of 8) of 5.5 (1.9), and low bias present in 4 (6.4%) diagnostic studies was observed (Table  2 and Fig.  2 ). Furthermore, in the 6 RCTs in the COVID-19 literature, there was a high risk of bias with little consideration for sequence generation, allocation concealment, blinding, incomplete outcome data, and selective outcome reporting (Table  2 ).

A Distribution of COVID-19 case series studies scored using the Murad tool ( n  = 380). B Distribution of COVID-19 cohort studies scored using the Newcastle–Ottawa Scale ( n  = 199). C Distribution of COVID-19 case–control studies scored using the Newcastle–Ottawa Scale ( n  = 38). D Distribution of COVID-19 diagnostic studies scored using the QUADAS-2 tool ( n  = 63). In panel D , blue represents low risk of bias and orange represents high risk of bias.

For secondary outcomes, rapid time from submission to acceptance (stratified by median time of acceptance of <13.0 days) was associated with lower methodological quality scores for case series and cohort study designs but not for case–control nor diagnostic studies (Fig.  3A–D ). Low journal impact factor (<10) was associated with lower methodological quality scores for case series, cohort, and case–control designs (Fig.  3E–H ). Finally, studies originating from different geographical regions had no differences in methodological quality scores with the exception of cohort studies (Fig.  3I–L ). When dichotomized by high vs. low methodological quality scores, a similar trend was observed with rapid time from submission to acceptance (34.4% vs. 46.3%, p  = 0.01, Supplementary Fig.  1B ), low impact factor journals (<10) was associated with lower methodological quality score (38.8% vs. 68.0%, p  < 0.0001, Supplementary Fig.  1C ). Finally, studies originating in either Americas or Asia/Oceania was associated with higher methodological quality scores than Europe (Supplementary Fig.  1D ).

A When stratified by time of acceptance (13.0 days), increased time of acceptance was associated with higher case series score ( n  = 186 for <13 days and n  = 193 for >=13 days; p  = 0.02). B Increased time of acceptance was associated with higher NOS cohort score ( n  = 112 for <13 days and n  = 144 for >=13 days; p  = 0.003). C No difference in time of acceptance and case–control score was observed ( n  = 18 for <13 days and n  = 27 for >=13 days; p  = 0.34). D No difference in time of acceptance and diagnostic risk of bias (QUADAS-2) was observed ( n  = 43 for <13 days and n  = 33 for >=13 days; p  = 0.23). E When stratified by impact factor (IF ≥10), high IF was associated with higher case series score ( n  = 466 for low IF and n  = 60 for high IF; p  < 0.0001). F High IF was associated with higher NOS cohort score ( n  = 262 for low IF and n  = 68 for high IF; p  = 0.01). G No difference in IF and case–control score was observed ( n  = 62 for low IF and n  = 2 for high IF; p  = 0.052). H No difference in IF and QUADAS-2 was observed ( n  = 101 for low IF and n  = 2 for high IF; p  = 0.93). I When stratified by geographical region, no difference in geographical region and case series score was observed ( n  = 276 Asia/Oceania, n  = 135 Americas, and n  = 143 Europe/Africa; p  = 0.10). J Geographical region was associated with differences in cohort score ( n  = 177 Asia/Oceania, n  = 81 Americas, and n  = 89 Europe/Africa; p  = 0.01). K No difference in geographical region and case–control score was observed ( n  = 37 Asia/Oceania, n  = 13 Americas, and n  = 14 Europe/Africa; p  = 0.81). L No difference in geographical region and QUADAS-2 was observed ( n  = 49 Asia/Oceania, n  = 28 Americas, and n  = 28 Europe/Africa; p  = 0.34). In panels A – D , orange represents lower median time of acceptance and blue represents high median time of acceptance. In panels E – H , red is low impact factor and blue is high impact factor. In panels I – L , orange represents Asia/Oceania, blue represents Americas, and brown represents Europe. Differences in distributions were analysed by two-sided Kruskal–Wallis test. Differences in diagnostic risk of bias were quantified by Chi-squares test. p  < 0.05 was considered statistically significant.

Methodological quality score differences in COVID-19 versus historical control

We matched 539 historical control articles to COVID-19 articles from the same journal with identical study designs in the previous year for a final analysis of 1078 articles (Table  1 ). Overall, 554 (51.4%) case series, 348 (32.3%) cohort, 64 (5.9%) case–control, 106 (9.8%) diagnostic and 6 (0.6%) RCTs were identified from the 1078 total articles. Differences exist between COVID-19 and historical control articles in geographical region of publication, retrospective study design, and sample size calculation (Table  1 ). Time of acceptance was 13.0 (IQR, 5.0–25.0) days in COVID-19 articles vs. 110.0 (IQR, 71.0–156.0) days in control articles (Table  1 and Fig.  4A , p  < 0.0001). Case-series methodological quality score was lower in COVID-19 articles compared to the historical control (3.3 (1.1) vs. 4.3 (0.8); n  = 554; p  < 0.0001; Table  2 and Fig.  4B ). Furthermore, NOS score was lower in COVID-19 cohort studies (5.8 (1.6) vs. 7.1 (1.0); n  = 348; p  < 0.0001; Table  2 and Fig.  4C ) and case–control studies (5.4 (1.9) vs. 6.6 (1.0); n  = 64; p  = 0.003; Table  2 and Fig.  4D ). Finally, lower risk of bias in diagnostic studies was in 12 COVID-19 articles (23%; n  = 53) compared to 24 control articles (45%; n  = 53; p  = 0.02; Table  2 and Fig.  4E ). A similar trend was observed between COVID-19 and historical control articles when dichotomized by good vs. low methodological quality scores (Supplementary Fig.  2 ).

A Time to acceptance was reduced in COVID-19 articles compared to control articles (13.0 [IQR, 5.0–25.0] days vs. 110.0 [IQR, 71.0–156.0] days, n  = 347 for COVID-19 and n  = 414 for controls; p  < 0.0001). B When compared to historical control articles, COVID-19 articles were associated with lower case series score ( n  = 277 for COVID-19 and n  = 277 for controls; p  < 0.0001). C COVID-19 articles were associated with lower NOS cohort score compared to historical control articles ( n  = 174 for COVID-19 and n  = 174 for controls; p  < 0.0001). D COVID-19 articles were associated with lower NOS case–control score compared to historical control articles ( n  = 32 for COVID-19 and n  = 32 for controls; p  = 0.003). E COVID-19 articles were associated with higher diagnostic risk of bias (QUADAS-2) compared to historical control articles ( n  = 53 for COVID-19 and n  = 53 for controls; p  = 0.02). For panel A , boxplot captures 5, 25, 50, 75 and 95% from the first to last whisker. Orange represents COVID-19 articles and blue represents control articles. Two-sided Mann–Whitney U-test was conducted to evaluate differences in time to acceptance between COVID-19 and control articles. Differences in study quality scores were evaluated by two-sided Kruskal–Wallis test. Differences in diagnostic risk of bias were quantified by Chi-squares test. p  < 0.05 was considered statistically significant.

In this systematic evaluation of methodological quality, COVID-19 clinical research was primarily observational in nature with modest methodological quality scores. Not only were the study designs low in the hierarchy of scientific evidence, we found that COVID-19 articles were associated with a lower methodological quality scores when published with a shorter time of publication and in lower impact factor journals. Furthermore, in a matched cohort analysis with historical control articles identified from the same journal of the same study design, we demonstrated that COVID-19 articles were associated with lower quality scores and shorter time from submission to acceptance.

The present study demonstrates comparative differences in methodological quality scores between COVID-19 literature and historical control articles. Overall, the accelerated publication of COVID-19 research was associated with lower study quality scores compared to previously published historical control studies. Our research highlights major differences in study quality between COVID-19 and control articles, possibly driven in part by a combination of more thorough editorial and/or peer-review process as suggested by the time to publication, and robust study design with questions which are pertinent for clinicians and patient management 3 , 6 , 7 , 8 , 9 , 10 , 11 .

In the early stages of the COVID-19 pandemic, we speculate that an urgent need for scientific data to inform clinical, social and economic decisions led to shorter time to publication and explosion in publication of COVID-19 studies in both traditional peer-reviewed journals and preprint servers 1 , 12 . The accelerated scientific process in the COVID-19 pandemic allowed a rapid understanding of natural history of COVID-19 symptomology and prognosis, identification of tools including RT-PCR to diagnose SARS-CoV-2 13 , and identification of potential therapeutic options such as tocilizumab and convalescent plasma which laid the foundation for future RCTs 14 , 15 , 16 . A delay in publication of COVID-19 articles due to a slower peer-review process may potentially delay dissemination of pertinent information against the pandemic. Despite concerns of slow peer review, major landmark trials (i.e. RECOVERY and ACTT-1 trial) 17 , 18 published their findings in preprint servers and media releases to allow for rapid dissemination. Importantly, the data obtained in these initial studies should be revisited as stronger data emerges as lower quality studies may fundamentally risk patient safety, resource allocation and future scientific research 19 .

Unfortunately, poor evidence begets poor clinical decisions 20 . Furthermore, lower quality scientific evidence potentially undermines the public’s trust in science during this time and has been evident through misleading information and high-profile retractions 12 , 21 , 22 , 23 . For example, the benefits of hydroxychloroquine, which were touted early in the pandemic based on limited data, have subsequently failed to be replicated in multiple observational studies and RCTs 5 , 24 , 25 , 26 , 27 , 28 , 29 , 30 . One poorly designed study combined with rapid publication led to considerable investment of both the scientific and medical community—akin to quinine being sold to the public as a miracle drug during the 1918 Spanish Influenza 31 , 32 . Moreover, as of June 30, 2020, ClinicalTrials.gov listed an astonishing 230 COVID-19 trials with hydroxychloroquine/plaquenil, and a recent living systematic review of observational studies and RCTs of hydroxychloroquine or chloroquine for COVID-19 demonstrated no evidence of benefit nor harm with concerns of severe methodological flaws in the included studies 33 .

Our study has important limitations. We evaluated the methodological quality of existing studies using established checklists and tools. While it is tempting to associate methodological quality scores with reproducibility or causal inferences of the intervention, it is not possible to ascertain the impact on the study design and conduct of research nor results or conclusions in the identified reports 34 . Second, although the methodological quality scales and checklists used for the manuscript are commonly used for quality assessment in systematic reviews and meta-analyses 35 , 36 , 37 , 38 , they can only assess the methodology without consideration for causal language and are prone to limitations 39 , 40 . Other tools such as the ROBINS-I and GRADE exist to evaluate methodological quality of identified manuscripts, although no consensus currently exists for critical appraisal of non-randomized studies 41 , 42 , 43 . Furthermore, other considerations of quality such as sample size calculation, sex reporting or ethics approval are not considered in these quality scores. As such, the quality scores measured using these checklists only reflect the patient selection, comparability, diagnostic reference standard and methods to ascertain the outcome of the study. Third, the 1:1 ratio to identify our historical control articles may affect the precision estimates of our findings. Interestingly, a simulation of an increase from 1:1 to 1:4 control ratio tightened the precision estimates but did not significantly alter the point estimate 44 . Furthermore, the decision for 1:1 ratio in our study exists due to limitations of available historical control articles from the identical journal in the restricted time period combined with a large effect size and sample size in the analysis. Finally, our analysis includes early publications on COVID-19 and there is likely to be an improvement in quality of related studies and study design as the field matures and higher-quality studies. Accordingly, our findings are limited to the early body of research as it pertains to the pandemic and it is likely that over time research quality will improve over time.

In summary, the early body of peer-reviewed COVID-19 literature was composed primarily of observational studies that underwent shorter peer-review evaluation and were associated with lower methodological quality scores than comparable studies. COVID-19 clinical studies should be revisited with the emergence of stronger evidence.

A systematic literature search was conducted on May 14, 2020 (registered on June 3, 2020 at PROSPERO: CRD42020187318) and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses. Furthermore, the cohort study was reported according to the Strengthening The Reporting of Observational Studies in Epidemiology checklist. The data supporting the findings of this study is available as Supplementary Data  1 – 2 .

Data sources and searches

The search was created in MEDLINE by a medical librarian with expertise in systematic reviews (S.V.) using a combination of key terms and index headings related to COVID-19 and translated to the remaining bibliographic databases (Supplementary Tables  1 – 3 ). The searches were conducted in MEDLINE (Ovid MEDLINE(R) ALL 1946–), Embase (Ovid Embase Classic + Embase 1947–) and the Cochrane Central Register of Controlled Trials (from inception). Search results were limited to English-only publications, and a publication date limit of January 1, 2019 to present was applied. In addition, a Canadian Agency for Drugs and Technologies in Health search filter was applied in MEDLINE and Embase to remove animal studies, and commentary, newspaper article, editorial, letter and note publication types were also eliminated. Search results were exported to Covidence (Veritas Health Innovation, Melbourne, Australia) and duplicates were eliminated using the platform’s duplicate identification feature.

Study selection, data extraction and methodological quality assessment

We included all types of COVID-19 clinical studies, including case series, observational studies, diagnostic studies and RCTs. For diagnostic studies, the reference standard for COVID-19 diagnosis was defined as a nasopharyngeal swab followed by reverse transcriptase-polymerase chain reaction in order to detect SARS-CoV-2. We excluded studies that were exploratory or pre-clinical in nature (i.e. in vitro or animal studies), case reports or case series of <5 patients, studies published in a language other than English, reviews, methods or protocols, and other coronavirus variants such as the Middle East respiratory syndrome.

The review team consisted of trained research staff with expertise in systematic reviews and one trainee. Title and abstracts were evaluated by two independent reviewers using Covidence and all discrepancies were resolved by consensus. Articles that were selected for full review were independently evaluated by two reviewers for quality assessment using a standardized case report form following the completion of a training period where all reviewers were trained with the original manuscripts which derived the tools or checklists along with examples for what were deemed high scores 35 , 36 , 37 , 38 . Following this, reviewers completed thirty full-text extractions and the two reviewers had to reach consensus and the process was repeated for the remaining manuscripts independently. When two independent reviewers were not able reach consensus, a third reviewer (principal investigator) provided oversight in the process to resolve the conflicted scores.

First and corresponding author names, date of publication, title of manuscript and journal of publication were collected for all included full-text articles. Journal impact factor was obtained from the 2018 InCites Journal Citation Reports from Clarivate Analytics. Submission and acceptance dates were collected in manuscripts when available. Other information such as study type, prospective or retrospective study, sex reporting, sample size calculation, method of SARS-CoV-2 diagnosis and ethics approval was collected by the authors. Methodological quality assessment was conducted using the Newcastle–Ottawa Scale (NOS) for case–control and cohort studies 37 , QUADAS-2 tool for diagnostic studies 38 , Cochrane risk of bias for RCTs 35 and a score derived by Murad et al. for case series studies 36 .

Identification of historical control from identified COVID-19 articles

Following the completion of full-text extraction of COVID-19 articles, we obtained a historical control group by identifying reports matched in a 1:1 fashion. From the eligible COVID-19 article, historical controls were identified by searching the same journal in a systematic fashion by matching the same study design (“case series”, “cohort”, “case control” or “diagnostic”) starting in the journal edition 12 months prior to the COVID-19 article publication on the publisher website (i.e. COVID-19 article published on April 2020, going backwards to April 2019) and proceeding forward (or backward if a specific article type was not identified) in a temporal fashion until the first matched study was identified following abstract screening by two independent reviewers. If no comparison article was found by either reviewers, the corresponding COVID-19 article was excluded from the comparison analysis. Following the identification of the historical control, data extraction and quality assessment was conducted on the identified articles using the standardized case report forms by two independent reviewers and conflicts resolved by consensus. The full dataset has been made available as Supplementary Data  1 – 2 .

Data synthesis and statistical analysis

Continuous variables were reported as mean (SD) or median (IQR) as appropriate, and categorical variables were reported as proportions (%). Continuous variables were compared using Student t -test or Mann–Whitney U-test and categorical variables including quality scores were compared by χ 2 , Fisher’s exact test, or Kruskal–Wallis test.

The primary outcome of interest was to evaluate the methodological quality of COVID-19 clinical literature by study design using the Newcastle–Ottawa Scale (NOS) for case–control and cohort studies, QUADAS-2 tool for diagnostic studies 38 , Cochrane risk of bias for RCTs 35 , and a score derived by Murad et al. for case series studies 36 . Pre-specified secondary outcomes were comparison of methodological quality scores of COVID-19 articles by (i) median time to acceptance, (ii) impact factor, (iii) geographical region and (iv) historical comparator. Time of acceptance was defined as the time between submission to acceptance which captures peer review and editorial decisions. Geographical region was stratified into continents including Asia/Oceania, Europe/Africa and Americas (North and South America). Post hoc comparison analysis between COVID-19 and historical control article quality scores were evaluated using Kruskal–Wallis test. Furthermore, good quality of NOS was defined as 3+ on selection and 1+ on comparability, and 2+ on outcome/exposure domains and high-quality case series scores was defined as a score ≥3.5. Due to a small sample size of identified RCTs, they were not included in the comparison analysis.

The finalized dataset was collected on Microsoft Excel v16.44. All statistical analyses were performed using SAS v9.4 (SAS Institute, Inc., Cary, NC, USA). Statistical significance was defined as P  < 0.05. All figures were generated using GraphPad Prism v8 (GraphPad Software, La Jolla, CA, USA).

Reporting summary

Further information on research design is available in the  Nature Research Reporting Summary linked to this article.

Data availability

The authors can confirm that all relevant data are included in the paper and in Supplementary Data  1 – 2 . The original search was conducted on MEDLINE, Embase and Cochrane Central Register of Controlled Trials.

Chen, Q., Allot, A. & Lu, Z. Keep up with the latest coronavirus research. Nature 579 , 193 (2020).

Article   ADS   CAS   Google Scholar  

Mahase, E. Covid-19: 146 researchers raise concerns over chloroquine study that halted WHO trial. BMJ https://doi.org/10.1136/bmj.m2197 (2020).

Chalmers, I. & Glasziou, P. Avoidable waste in the production and reporting of research evidence. Lancet 374 , 86–89 (2009).

Article   Google Scholar  

Burns, P. B., Rohrich, R. J. & Chung, K. C. The levels of evidence and their role in evidence-based medicine. Plast. Reconstr. Surg. 128 , 305–310 (2011).

Article   CAS   Google Scholar  

Alexander, P. E. et al. COVID-19 coronavirus research has overall low methodological quality thus far: case in point for chloroquine/hydroxychloroquine. J. Clin. Epidemiol. 123 , 120–126 (2020).

Barakat, A. F., Shokr, M., Ibrahim, J., Mandrola, J. & Elgendy, I. Y. Timeline from receipt to online publication of COVID-19 original research articles. Preprint at medRxiv https://doi.org/10.1101/2020.06.22.20137653 (2020).

Chan, A.-W. et al. Increasing value and reducing waste: addressing inaccessible research. Lancet 383 , 257–266 (2014).

Ioannidis, J. P. A. et al. Increasing value and reducing waste in research design, conduct, and analysis. Lancet 383 , 166–175 (2014).

Chalmers, I. et al. How to increase value and reduce waste when research priorities are set. Lancet 383 , 156–165 (2014).

Salman, R. A.-S. et al. Increasing value and reducing waste in biomedical research regulation and management. Lancet 383 , 176–185 (2014).

Glasziou, P. et al. Reducing waste from incomplete or unusable reports of biomedical research. Lancet 383 , 267–276 (2014).

Bauchner, H. The rush to publication: an editorial and scientific mistake. JAMA 318 , 1109–1110 (2017).

He, X. et al. Temporal dynamics in viral shedding and transmissibility of COVID-19. Nat. Med. 26 , 672–675 (2020).

Guaraldi, G. et al. Tocilizumab in patients with severe COVID-19: a retrospective cohort study. Lancet Rheumatol. 2 , e474–e484 (2020).

Duan, K. et al. Effectiveness of convalescent plasma therapy in severe COVID-19 patients. Proc. Natl Acad. Sci. USA 117 , 9490–9496 (2020).

Shen, C. et al. Treatment of 5 critically Ill patients with COVID-19 with convalescent plasma. JAMA 323 , 1582–1589 (2020).

Beigel, J. H. et al. Remdesivir for the treatment of covid-19—final report. N. Engl. J. Med. 383 , 1813–1826 (2020).

Group, R. C. et al. Dexamethasone in hospitalized patients with Covid-19—preliminary report. N. Engl. J. Med. https://doi.org/10.1056/NEJMoa2021436 (2020).

Ramirez, F. D. et al. Methodological rigor in preclinical cardiovascular studies: targets to enhance reproducibility and promote research translation. Circ. Res 120 , 1916–1926 (2017).

Heneghan, C. et al. Evidence based medicine manifesto for better healthcare. BMJ 357 , j2973 (2017).

Mehra, M. R., Desai, S. S., Ruschitzka, F. & Patel, A. N. RETRACTED: hydroxychloroquine or chloroquine with or without a macrolide for treatment of COVID-19: a multinational registry analysis. Lancet https://doi.org/10.1016/S0140-6736(20)31180-6 (2020).

Servick, K. & Enserink, M. The pandemic’s first major research scandal erupts. Science 368 , 1041–1042 (2020).

Mehra, M. R., Desai, S. S., Kuy, S., Henry, T. D. & Patel, A. N. Retraction: Cardiovascular disease, drug therapy, and mortality in Covid-19. N. Engl. J. Med. 382 , 2582–2582, https://doi.org/10.1056/NEJMoa2007621. (2020).

Article   PubMed   Google Scholar  

Boulware, D. R. et al. A randomized trial of hydroxychloroquine as postexposure prophylaxis for Covid-19. N. Engl. J. Med. 383 , 517–525 (2020).

Gautret, P. et al. Clinical and microbiological effect of a combination of hydroxychloroquine and azithromycin in 80 COVID-19 patients with at least a six-day follow up: a pilot observational study. Travel Med. Infect. Dis. 34 , 101663–101663 (2020).

Geleris, J. et al. Observational study of hydroxychloroquine in hospitalized patients with Covid-19. N. Engl. J. Med. 382 , 2411–2418 (2020).

Borba, M. G. S. et al. Effect of high vs low doses of chloroquine diphosphate as adjunctive therapy for patients hospitalized with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection: a randomized clinical trial. JAMA Netw. Open 3 , e208857–e208857 (2020).

Mercuro, N. J. et al. Risk of QT interval prolongation associated with use of hydroxychloroquine with or without concomitant azithromycin among hospitalized patients testing positive for coronavirus disease 2019 (COVID-19). JAMA Cardiol. 5 , 1036–1041 (2020).

Molina, J. M. et al. No evidence of rapid antiviral clearance or clinical benefit with the combination of hydroxychloroquine and azithromycin in patients with severe COVID-19 infection. Médecine et. Maladies Infectieuses 50 , 384 (2020).

Group, R. C. et al. Effect of hydroxychloroquine in hospitalized patients with Covid-19. N. Engl. J. Med . 383, 2030–2040 (2020).

Shors, T. & McFadden, S. H. 1918 influenza: a Winnebago County, Wisconsin perspective. Clin. Med. Res. 7 , 147–156 (2009).

Stolberg, S. A Mad Scramble to Stock Millions of Malaria Pills, Likely for Nothing (The New York Times, 2020).

Hernandez, A. V., Roman, Y. M., Pasupuleti, V., Barboza, J. J. & White, C. M. Hydroxychloroquine or chloroquine for treatment or prophylaxis of COVID-19: a living systematic review. Ann. Int. Med. 173 , 287–296 (2020).

Glasziou, P. & Chalmers, I. Research waste is still a scandal—an essay by Paul Glasziou and Iain Chalmers. BMJ 363 , k4645 (2018).

Higgins, J. P. T. et al. The Cochrane Collaboration’s tool for assessing risk of bias in randomised trials. BMJ 343 , d5928 (2011).

Murad, M. H., Sultan, S., Haffar, S. & Bazerbachi, F. Methodological quality and synthesis of case series and case reports. BMJ Evid. Based Med. 23 , 60–63 (2018).

Wells, G. S. B. et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analysis. http://wwwohrica/programs/clinical_epidemiology/oxfordasp (2004).

Whiting, P. F. et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann. Intern. Med. 155 , 529–536 (2011).

Sanderson, S., Tatt, I. D. & Higgins, J. P. Tools for assessing quality and susceptibility to bias in observational studies in epidemiology: a systematic review and annotated bibliography. Int. J. Epidemiol. 36 , 666–676 (2007).

Stang, A. Critical evaluation of the Newcastle-Ottawa scale for the assessment of the quality of nonrandomized studies in meta-analyses. Eur. J. Epidemiol. 25 , 603–605 (2010).

Guyatt, G. et al. GRADE guidelines: 1. Introduction-GRADE evidence profiles and summary of findings tables. J. Clin. Epidemiol. 64 , 383–394 (2011).

Quigley, J. M., Thompson, J. C., Halfpenny, N. J. & Scott, D. A. Critical appraisal of nonrandomized studies-A review of recommended and commonly used tools. J. Evaluation Clin. Pract. 25 , 44–52 (2019).

Sterne, J. A. et al. ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions. BMJ 355 , i4919 (2016).

Hamajima, N. et al. Case-control studies: matched controls or all available controls? J. Clin. Epidemiol. 47 , 971–975 (1994).

Download references

Acknowledgements

This study received no specific funding or grant from any agency in the public, commercial, or not-for-profit sectors. R.G.J. was supported by the Vanier CIHR Canada Graduate Scholarship. F.D.R. was supported by a CIHR Banting Postdoctoral Fellowship and a Royal College of Physicians and Surgeons of Canada Detweiler Travelling Fellowship. The funder/sponsor(s) had no role in design and conduct of the study, collection, analysis and interpretation of the data.

Author information

These authors contributed equally: Richard G. Jung, Pietro Di Santo.

Authors and Affiliations

CAPITAL Research Group, University of Ottawa Heart Institute, Ottawa, Ontario, Canada

Richard G. Jung, Pietro Di Santo, F. Daniel Ramirez & Trevor Simard

Vascular Biology and Experimental Medicine Laboratory, University of Ottawa Heart Institute, Ottawa, Ontario, Canada

Richard G. Jung, Pietro Di Santo, Trevor Simard & Benjamin Hibbert

Department of Cellular and Molecular Medicine, Faculty of Medicine, University of Ottawa, Ottawa, Ontario, Canada

Richard G. Jung, Trevor Simard & Benjamin Hibbert

Division of Cardiology, University of Ottawa Heart Institute, Ottawa, Ontario, Canada

Pietro Di Santo, Simon Parlow, F. Daniel Ramirez, Trevor Simard & Benjamin Hibbert

School of Epidemiology and Public Health, University of Ottawa, Ottawa, Ontario, Canada

Pietro Di Santo

Faculty of Medicine, University of Ottawa, Ontario, Canada

Cole Clifford & Stephanie Skanes

Department of Medicine, Cumming School of Medicine, Calgary, Alberta, Canada

Graeme Prosperi-Porta

Division of Internal Medicine, The Ottawa Hospital, Ottawa, Ontario, Canada

Berkman Library, University of Ottawa Heart Institute, Ottawa, Ontario, Canada

Sarah Visintini

Hôpital Cardiologique du Haut-Lévêque, CHU Bordeaux, Bordeaux-Pessac, France

F. Daniel Ramirez

L’Institut de Rythmologie et Modélisation Cardiaque (LIRYC), University of Bordeaux, Bordeaux, France

Department of Cardiovascular Medicine, Mayo Clinic, Rochester, MN, USA

Trevor Simard

You can also search for this author in PubMed   Google Scholar

Contributions

R.G.J., P.D.S., S.V., F.D.R., T.S. and B.H. participated in the study conception and design. Data acquisition, analysis and interpretation were performed by R.G.J., P.D.S., C.C., G.P.P., S.P., S.S., A.H., F.D.R., T.S. and B.H. Statistical analysis was performed by R.G.J., P.D.S. and B.H. The manuscript was drafted by R.G.J., P.D.S., F.D.R., T.S. and B.H. All authors approved the final version of the manuscript and agree to be accountable to all aspects of the work.

Corresponding author

Correspondence to Benjamin Hibbert .

Ethics declarations

Competing interests.

B.H. reports funding as a clinical trial investigator from Abbott, Boston Scientific and Edwards Lifesciences outside of the submitted work. The remaining authors declare no competing interests.

Additional information

Peer review information Nature Communications Ian White and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary information, peer review file, description of additional supplementary files, supplementary data 1, supplementary data 2, reporting summary, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Jung, R.G., Di Santo, P., Clifford, C. et al. Methodological quality of COVID-19 clinical research. Nat Commun 12 , 943 (2021). https://doi.org/10.1038/s41467-021-21220-5

Download citation

Received : 16 July 2020

Accepted : 13 January 2021

Published : 11 February 2021

DOI : https://doi.org/10.1038/s41467-021-21220-5

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

The quality of covid-19 systematic reviews during the coronavirus 2019 pandemic: an exploratory comparison.

• Kevin T. McDermott
• Jos Kleijnen

Systematic Reviews (2024)

Gatekeeping should be conserved in the open science era

• Hugh Desmond

Synthese (2024)

Exploring COVID-19 research credibility among Spanish scientists

• Eduardo Garcia-Garzon
• Ariadna Angulo-Brunet
• Guido Corradi

Current Psychology (2024)

Primary health care research in COVID-19: analysis of the protocols reviewed by the ethics committee of IDIAPJGol, Catalonia

• Anna Moleras-Serra
• Rosa Morros-Pedros
• Ainhoa Gómez-Lumbreras

BMC Primary Care (2023)

Identifying patterns of reported findings on long-term cardiac complications of COVID-19: a systematic review and meta-analysis

• Chenya Zhao

BMC Medicine (2023)

By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Official websites use .gov

A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS

A lock ( ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

COVID-19 Research Articles Downloadable Database

March 19, 2020

Updated January 12, 2024

COVID-19 Research Guide Home

• Research Articles Downloadable Database
• COVID-19 Science Updates
• Databases and Journals
• Secondary Data and Statistics

Important announcement:  

The CDC Database of COVID-19 Research Articles became a collaboration with the WHO to create the  WHO COVID-19 database  during the pandemic to make it easier for results to be searched, downloaded, and used by researchers worldwide.

The last version of the CDC COVID-19 database was archived and remain available on this website.  Please note that it has stopped updating as of October 9, 2020 and all new articles were integrated into the  WHO COVID-19 database .  The WHO Covid-19 Research Database was a resource created in response to the Public Health Emergency of International Concern (PHEIC). Its content remains searchable and spans the time period March 2020 to June 2023. Since June 2023, manual updates to the database have been discontinued.

If you have any questions, concerns, or problems accessing the WHO COVID-19 Database please email the CDC Library for assistance.

Materials listed in these guides are selected to provide awareness of quality public health literature and resources. A material’s inclusion does not necessarily represent the views of the U.S. Department of Health and Human Services (HHS), the Public Health Service (PHS), or the Centers for Disease Control and Prevention (CDC), nor does it imply endorsement of the material’s methods or findings.

Below are options to download the archive of COVID-19 research articles.  You can search the database of citations by author, keyword (in title, author, abstract, subject headings fields), journal, or abstract when available.  DOI, PMID, and URL links are included when available.

This database was last updated on October 9, 2020 .

• The CDC Database of COVID-19 Research Articles is now a part of the WHO COVID-19 database .  Our new  search results are now being sent to the WHO COVID-19 Database to make it easier for them to be searched, downloaded, and used by researchers worldwide. The WHO Covid-19 Research Database was a resource created in response to the Public Health Emergency of International Concern (PHEIC). Its content remains searchable and spans the time period March 2020 to June 2023. Since June 2023, manual updates to the database have been discontinued.
• To help inform CDC’s COVID-19 Response, as well as to help CDC staff stay up to date on the latest COVID-19 research, the Response’s Office of the Chief Medical Officer has collaborated with the CDC Office of Library Science to create a series called COVID-19 Science Update . This series, the first of its kind for a CDC emergency response, provides brief summaries of new COVID-19-related studies on many topics, including epidemiology, clinical treatment and management, laboratory science, and modeling. As of December 18, 2021, CDC has stopped production of the weekly COVID-19 Science Update.

Excel download:

• Articles from August until October 9 2020 [XLS – 29 MB]
• Articles from December 2019 through July 2020 [XLS – 45 MB]
• The CDC Database of COVID-19 Research Articles is now a part of the WHO COVID-19 database .  Our new search results are now being sent to the WHO COVID-19 Database to make it easier for them to be searched, downloaded, and used by researchers worldwide.
• October 8 in Excel [XLS – 1 MB]
• October 7 in Excel [XLS – 1 MB]
• October 6 in Excel [XLS – 1 MB]
• Note the main Excel file can also be sorted by date added.

Citation Management Software (EndNote, Mendeley, Zotero, Refman, etc.)  download:

• Part 1 [ZIP – 38 MB]
• Part 2 [ZIP – 43 MB]
• October 8 in citation management software format [RIS – 2 MB]
• October 7 in citation management software format [RIS – 2 MB]
• October 6 in citation management software format [RIS – 2 MB]
• Note the main RIS file can also be sorted by date added.

The COVID-19 pandemic is a rapidly changing situation.  Some of the research included above is preliminary.  Materials listed in this database are selected to provide awareness of quality public health literature and resources. A material’s inclusion does not necessarily represent the views of the U.S. Department of Health and Human Services (HHS), the Public Health Service (PHS), or the Centers for Disease Control and Prevention (CDC), nor does it imply endorsement of the material’s methods or findings.

To access the full text, click on the DOI, PMID, or URL links.  While most publishers are making their COVID-19 content Open Access, some articles are accessible only to those with a CDC user id and password. Find a library near you that may be able to help you get access to articles by clicking the following links: https://www.worldcat.org/libraries OR https://www.usa.gov/libraries .

CDC users can use EndNote’s Find Full Text feature to attach the full text PDFs within their EndNote Library.  CDC users, please email Martha Knuth for an EndNote file of all citations.  Once you have your EndNote file downloaded, to get the full-text of journal articles listed in the search results you can do the following steps:

• First, try using EndNote’s “Find Full-Text” feature to attach full-text articles to your EndNote Library.
• Next, check for full-text availability, via the E-Journals list, at: http://sfxhosted.exlibrisgroup.com/cdc/az   .
• If you can’t find full-text online, you can request articles via DocExpress, at: https://docexpress.cdc.gov/illiad/

The following databases were searched from Dec. 2019-Oct. 9 2020 for articles related to COVID-19: Medline (Ovid and PubMed), PubMed Central, Embase, CAB Abstracts, Global Health, PsycInfo, Cochrane Library, Scopus, Academic Search Complete, Africa Wide Information, CINAHL, ProQuest Central, SciFinder, the Virtual Health Library, and LitCovid.  Selected grey literature sources were searched as well, including the WHO COVID-19 website, CDC COVID-19 website, Eurosurveillance, China CDC Weekly, Homeland Security Digital Library, ClinicalTrials.gov, bioRxiv (preprints), medRxiv (preprints), chemRxiv (preprints), and SSRN (preprints).

Detailed search strings with synonyms used for COVID-19 are below.

Detailed search strategy for gathering COVID-19 articles, updated October 9, 2020 [PDF – 135 KB]

Note on preprints:   Preprints have not been peer-reviewed. They should not be regarded as conclusive, guide clinical practice/health-related behavior, or be reported in news media as established information.

Materials listed in these guides are selected to provide awareness of quality public health literature and resources. A material’s inclusion does not necessarily represent the views of the U.S. Department of Health and Human Services (HHS), the Public Health Service (PHS), or the Centers for Disease Control and Prevention (CDC), nor does it imply endorsement of the material’s methods or findings. HHS, PHS, and CDC assume no responsibility for the factual accuracy of the items presented. The selection, omission, or content of items does not imply any endorsement or other position taken by HHS, PHS, and CDC. Opinion, findings, and conclusions expressed by the original authors of items included in these materials, or persons quoted therein, are strictly their own and are in no way meant to represent the opinion or views of HHS, PHS, or CDC. References to publications, news sources, and non-CDC Websites are provided solely for informational purposes and do not imply endorsement by HHS, PHS, or CDC.

To receive the COVID-19 Science Update, please enter your email address to subscribe today.

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

The impact of the COVID-19 pandemic on scientific research in the life sciences

Roles Conceptualization, Formal analysis, Methodology, Writing – original draft, Writing – review & editing

Affiliation AXES, IMT School for Advanced Studies Lucca, Lucca, Italy

Roles Conceptualization, Data curation, Formal analysis, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing

* E-mail: [email protected]

Affiliation Chair of Systems Design D-MTEC, ETH Zürich, Zurich, Switzerland

• Massimo Riccaboni, 
• Luca Verginer

• Published: February 9, 2022
• https://doi.org/10.1371/journal.pone.0263001
• Reader Comments

The COVID-19 outbreak has posed an unprecedented challenge to humanity and science. On the one side, public and private incentives have been put in place to promptly allocate resources toward research areas strictly related to the COVID-19 emergency. However, research in many fields not directly related to the pandemic has been displaced. In this paper, we assess the impact of COVID-19 on world scientific production in the life sciences and find indications that the usage of medical subject headings (MeSH) has changed following the outbreak. We estimate through a difference-in-differences approach the impact of the start of the COVID-19 pandemic on scientific production using the PubMed database (3.6 Million research papers). We find that COVID-19-related MeSH terms have experienced a 6.5 fold increase in output on average, while publications on unrelated MeSH terms dropped by 10 to 12%. The publication weighted impact has an even more pronounced negative effect (-16% to -19%). Moreover, COVID-19 has displaced clinical trial publications (-24%) and diverted grants from research areas not closely related to COVID-19. Note that since COVID-19 publications may have been fast-tracked, the sudden surge in COVID-19 publications might be driven by editorial policy.

Citation: Riccaboni M, Verginer L (2022) The impact of the COVID-19 pandemic on scientific research in the life sciences. PLoS ONE 17(2): e0263001. https://doi.org/10.1371/journal.pone.0263001

Editor: Florian Naudet, University of Rennes 1, FRANCE

Received: April 28, 2021; Accepted: January 10, 2022; Published: February 9, 2022

Copyright: © 2022 Riccaboni, Verginer. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The processed data, instructions on how to process the raw PubMed dataset as well as all code are available via Zenodo at https://doi.org/10.5281/zenodo.5121216 .

Funding: The author(s) received no specific funding for this work.

Competing interests: The authors have declared that no competing interests exist.

Introduction

The COVID-19 pandemic has mobilized the world scientific community in 2020, especially in the life sciences [ 1 , 2 ]. In the first three months after the pandemic, the number of scientific papers about COVID-19 was fivefold the number of articles on H1N1 swine influenza [ 3 ]. Similarly, the number of clinical trials related to COVID-19 prophylaxis and treatments skyrocketed [ 4 ]. Thanks to the rapid mobilization of the world scientific community, COVID-19 vaccines have been developed in record time. Despite this undeniable success, there is a rising concern about the negative consequences of COVID-19 on clinical trial research, with many projects being postponed [ 5 – 7 ]. According to Evaluate Pharma, clinical trials were one of the pandemic’s first casualties, with a record number of 160 studies suspended for reasons related to COVID-19 in April 2020 [ 8 , 9 ] reporting a total of 1,200 trials suspended as of July 2020. As a consequence, clinical researchers have been impaired by reduced access to healthcare research infrastructures. Particularly, the COVID-19 outbreak took a tall on women and early-career scientists [ 10 – 13 ]. On a different ground, Shan and colleagues found that non-COVID-19-related articles decreased as COVID-19-related articles increased in top clinical research journals [ 14 ]. Fraser and coworker found that COVID-19 preprints received more attention and citations than non-COVID-19 preprints [ 1 ]. More recently, Hook and Porter have found some early evidence of ‘covidisation’ of academic research, with research grants and output diverted to COVID-19 research in 2020 [ 15 ]. How much should scientists switch their efforts toward SARS-CoV-2 prevention, treatment, or mitigation? There is a growing consensus that the current level of ‘covidisation’ of research can be wasteful [ 4 , 5 , 16 ].

Against this background, in this paper, we investigate if the COVID-19 pandemic has induced a shift in biomedical publications toward COVID-19-related scientific production. The objective of the study is to show that scientific articles listing covid-related Medical Subject Headings (MeSH) when compared against covid-unrelated MeSH have been partially displaced. Specifically, we look at several indicators of scientific production in the life sciences before and after the start of the COVID-19 pandemic: (1) number of papers published, (2) impact factor weighted number of papers, (3) opens access, (4) number of publications related to clinical trials, (5) number of papers listing grants, (6) number of papers listing grants existing before the pandemic. Through a natural experiment approach, we analyze the impact of the pandemic on scientific production in the life sciences. We consider COVID-19 an unexpected and unprecedented exogenous source of variation with heterogeneous effects across biomedical research fields (i.e., MeSH terms).

Based on the difference in difference results, we document the displacement effect that the pandemic has had on several aspects of scientific publishing. The overall picture that emerges from this analysis is that there has been a profound realignment of priorities and research efforts. This shift has displaced biomedical research in fields not related to COVID-19.

The rest of the paper is structured as follows. First, we describe the data and our measure of relatedness to COVID-19. Next, we illustrate the difference-in-differences specification we rely on to identify the impact of the pandemic on scientific output. In the results section, we present the results of the difference-in-differences and network analyses. We document the sudden shift in publications, grants and trials towards COVID-19-related MeSH terms. Finally, we discuss the findings and highlight several policy implications.

Materials and methods

The present analysis is based primarily on PubMed and the Medical Subject Headings (MeSH) terminology. This data is used to estimate the effect of the start of the COVID 19 pandemic via a difference in difference approach. This section is structured as follows. We first introduce the data and then the econometric methodology. This analysis is not based on a pre-registered protocol.

Selection of biomedical publications.

We rely on PubMed, a repository with more than 34 million biomedical citations, for the analysis. Specifically, we analyze the daily updated files up to 31/06/2021, extracting all publications of type ‘Journal Article’. For the principal analysis, we consider 3,638,584 papers published from January 2019 to December 2020. We also analyze 11,122,017 papers published from 2010 onwards to identify the earliest usage of a grant and infer if it was new in 2020. We use the SCImago journal ranking statistics to compute the impact factor weighted number (IFWN) of papers in a given field of research. To assign the publication date, we use the ‘electronically published’ dates and, if missing, the ‘print published’ dates.

Medical subject headings.

We rely on the Medical Subject Headings (MeSH) terminology to approximate narrowly defined biomedical research fields. This terminology is a curated medical vocabulary, which is manually added to papers in the PubMed corpus. The fact that MeSH terms are manually annotated makes this terminology ideal for classification purposes. However, there is a delay between publication and annotation, on the order of several months. To address this delay and have the most recent classification, we search for all 28 425 MeSH terms using PubMed’s ESearch utility and classify paper by the results. The specific API endpoint is https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi , the relevant scripts are available with the code. For example, we assign the term ‘Ageusia’ (MeSH ID D000370) to all papers listed in the results of the ESearch API. We apply this method to the whole period (January 2019—December 2020) and obtain a mapping from papers to the MeSH terms. For every MeSH term, we keep track of the year they have been established. For instance, COVID-19 terms were established in 2020 (see Table 1 ): in January 2020, the WHO recommended 2019-nCoV and 2019-nCoV acute respiratory disease as provisional names for the virus and disease. The WHO issued the official terms COVID-19 and SARS-CoV-2 at the beginning of February 2020. By manually annotating publications, all publications referring to COVID-19 and SARS-CoV-2 since January 2020 have been labelled with the related MeSH terms. Other MeSH terms related to COVID-19, such as coronavirus, for instance, have been established years before the pandemic (see Table 2 ). We proxy MeSH term usage via search terms using the PubMed EUtilities API; this means that we are not using the hand-labelled MeSH terms but rather the PubMed search results. This means that the accuracy of the MeSH term we assign to a given paper is not perfect. In practice, this means that we have assigned more MeSH terms to a given term than a human annotator would have.

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

https://doi.org/10.1371/journal.pone.0263001.t001

The list contains only terms with at least 100 publications in 2020.

https://doi.org/10.1371/journal.pone.0263001.t002

Clinical trials and publication types.

We classify publications using PubMed’s ‘PublicationType’ field in the XML baseline files (There are 187 publication types, see https://www.nlm.nih.gov/mesh/pubtypes.html ). We consider a publication to be related to a clinical trial if it lists any of the following descriptors:

• D016430: Clinical Trial
• D017426: Clinical Trial, Phase I
• D017427: Clinical Trial, Phase II
• D017428: Clinical Trial, Phase III
• D017429: Clinical Trial, Phase IV
• D018848: Controlled Clinical Trial
• D065007: Pragmatic Clinical Trial
• D000076362: Adaptive Clinical Trial
• D000077522: Clinical Trial, Veterinary

In our analysis of the impact of COVID-19 on publications related to clinical trials, we only consider MeSH terms that are associated at least once with a clinical trial publication over the two years. We apply this restriction to filter out MeSH terms that are very unlikely to be relevant for clinical trial types of research.

Open access.

We proxy the availability of a journal article to the public, i.e., open access, if it is available from PubMed Central. PubMed Central archives full-text journal articles and provides free access to the public. Note that the copyright license may vary across participating publishers. However, the text of the paper is for all effects and purposes freely available without requiring subscriptions or special affiliation.

We infer if a publication has been funded by checking if it lists any grants. We classify grants as either ‘old’, i.e. existed before 2019, or ‘new’, i.e. first observed afterwards. To do so, we collect all grant IDs for 11,122,017 papers from 2010 on-wards and record their first appearance. This procedure is an indirect inference of the year the grant has been granted. The basic assumption is that if a grant number has not been listed in any publication since 2010, it is very likely a new grant. Specifically, an old grant is a grant listed since 2019 observed at least once from 2010 to 2018.

Note that this procedure is only approximate and has a few shortcomings. Mistyped grant numbers (e.g. ‘1234-M JPN’ and ‘1234-M-JPN’) could appear as new grants, even though they existed before, or new grants might be classified as old grants if they have a common ID (e.g. ‘Grant 1’). Unfortunately, there is no central repository of grant numbers and the associated metadata; however, there are plans to assign DOI numbers to grants to alleviate this problem (See https://gitlab.com/crossref/open_funder_registry for the project).

Impact factor weighted publication numbers (IFWN).

In our analysis, we consider two measures of scientific output. First, we simply count the number of publications by MeSH term. However, since journals vary considerably in terms of impact factor, we also weigh the number of publications by the impact factor of the venue (e.g., journal) where it was published. Specifically, we use the SCImago journal ranking statistics to weigh a paper by the impact factor of the journal it appears in. We use the ‘citation per document in the past two years’ for 45,230 ISSNs. Note that a journal may and often has more than one ISSN, i.e., one for the printed edition and one for the online edition. SCImago applies the same score for a venue across linked ISSNs.

For the impact factor weighted number (IFWN) of publication per MeSH terms, this means that all publications are replaced by the impact score of the journal they appear in and summed up.

COVID-19-relatedness.

To measure how closely related to COVID-19 is a MeSH term, we introduce an index of relatedness to COVID-19. First, we identify the focal COVID-19 terms, which appeared in the literature in 2020 (see Table 1 ). Next, for all other pre-existing MeSH terms, we measure how closely related to COVID-19 they end up being.

Our aim is to show that MeSH terms that existed before and are related have experienced a sudden increase in the number of (impact factor weighted) papers.

Intuitively we can read this measure as: what is the probability in 2020 that a COVID-19 MeSH term is present given that we chose a paper with MeSH term i ? For example, given that in 2020 we choose a paper dealing with “Ageusia” (i.e., Complete or severe loss of the subjective sense of taste), there is a 96% probability that this paper also lists COVID-19, see Table 1 .

Note that a paper listing a related MeSH term does not imply that that paper is doing COVID-19 research, but it implies that one of the MeSH terms listed is often used in COVID-19 research.

In sum, in our analysis, we use the following variables:

• Papers: Number of papers by MeSH term;
• Impact: Impact factor weighted number of papers by MeSH term;
• PMC: Papers listed in PubMed central by MeSH term, as a measure of Open Access publications;
• Trials: number of publications of type “Clinical Trial” by MeSH term;
• Grants: number of papers with at least one grant by MeSH term;
• Old Grants: number of papers listing a grant that has been observed between 2010 and 2018, by MeSH term;

Difference-in-differences

The difference-in-differences (DiD) method is an econometric technique to imitate an experimental research design from observation data, sometimes referred to as a quasi-experimental setup. In a randomized controlled trial, subjects are randomly assigned either to the treated or the control group. Analogously, in this natural experiment, we assume that medical subject headings (MeSH) have been randomly assigned to be either treated (related) or not treated (unrelated) by the pandemic crisis.

Before the COVID, for a future health crisis, the set of potentially impacted medical knowledge was not predictable since it depended on the specifics of the emergency. For instance, ageusia (loss of taste), a medical concept existing since 1991, became known to be a specific symptom of COVID-19 only after the pandemic.

Specifically, we exploit the COVID-19 as an unpredictable and exogenous shock that has deeply affected the publication priorities for biomedical scientific production, as compared to the situation before the pandemic. In this setting, COVID-19 is the treatment, and the identification of this new human coronavirus is the event. We claim that treated MeSH terms, i.e., MeSH terms related to COVID-19, have experienced a sudden increase in terms of scientific production and attention. In contrast, research on untreated MeSH terms, i.e., MeSH terms not related to COVID-19, has been displaced by COVID-19. Our analysis compares the scientific output of COVID-19 related and unrelated MeSH terms before and after January 2020.

In our case, some of the terms turn out to be related to COVID-19 in 2020, whereas most of the MeSH terms are not closely related to COVID-19.

Thus β 1 identifies the overall effect on the control group after the event, β 2 the difference across treated and control groups before the event (i.e. the first difference in DiD) and finally the effect on the treated group after the event, net of the first difference, β 3 . This last parameter identifies the treatment effect on the treated group netting out the pre-treatment difference.

For the DiD to have a causal interpretation, it must be noted that pre-event, the trends of the two groups should be parallel, i.e., the common trend assumption (CTA) must be satisfied. We will show that the CTA holds in the results section.

To specify the DiD model, we need to define a period before and after the event and assign a treatment status or level of exposure to each term.

Before and after.

The pre-treatment period is defined as January 2019 to December 2019. The post-treatment period is defined as the months from January 2020 to December 2020. We argue that the state of biomedical research was similar in those two years, apart from the effect of the pandemic.

Treatment status and exposure.

The treatment is determined by the COVID-19 relatedness index σ i introduced earlier. Specifically, this number indicates the likelihood that COVID-19 will be a listed MeSH term, given that we observe the focal MeSH term i . To show that the effect becomes even stronger the closer related the subject is, and for ease of interpretation, we also discretize the relatedness value into three levels of treatment. Namely, we group MeSH terms with a σ between, 0% to 20%, 20% to 80% and 80% to 100%. The choice of alternative grouping strategies does not significantly affect our results. Results for alternative thresholds of relatedness can be computed using the available source code. We complement the dichotomized analysis by using the treatment intensity (relatedness measure σ ) to show that the result persists.

Panel regression.

In this work, we estimate a random effects panel regression where the units of analysis are 28 318 biomedical research fields (i.e. MeSH terms) observed over time before and after the COVID-19 pandemic. The time resolution is at the monthly level, meaning that for each MeSH term, we have 24 observations from January 2019 to December 2020.

The outcome variable Y it identifies the outcome at time t (i.e., month), for MeSH term i . As before, P t identifies the period with P t = 0 if the month is before January 2020 and P t = 1 if it is on or after this date. In (3) , the treatment level is measure by the relatedness to COVID-19 ( σ i ), where again the γ 1 identifies pre-trend (constant) differences and δ 1 the overall effect.

In total, we estimate six coefficients. As before, the δ l coefficient identifies the DiD effect.

Verifying the Common Trend Assumption (CTA).

We show that the CTA holds for this model by comparing the pre-event trends of the control group to the treated groups (COVID-19 related MeSH terms). Namely, we show that the pre-event trends of the control group are the same as the pre-event trends of the treated group.

Co-occurrence analysis

To investigate if the pandemic has caused a reconfiguration of research priorities, we look at the MeSH term co-occurrence network. Precisely, we extract the co-occurrence network of all 28,318 MeSH terms as they appear in the 3.3 million papers. We considered the co-occurrence networks of 2018, 2019 and 2020. Each node represents a MeSH term in these networks, and a link between them indicates that they have been observed at least once together. The weight of the edge between the MeSH terms is given by the number of times those terms have been jointly observed in the same publications.

Medical language is hugely complicated, and this simple representation does not capture the intricacies, subtle nuances and, in fact, meaning of the terms. Therefore, we do not claim that we can identify how the actual usage of MeSH terms has changed from this object, but rather that it has. Nevertheless, the co-occurrence graph captures rudimentary relations between concepts. We argue that absent a shock to the system, their basic usage patterns, change in importance (within the network) would essentially be the same from year to year. However, if we find that the importance of terms changes more than expected in 2020, it stands to reason that there have been some significant changes.

To show that that MeSH usage has been affected, we compute for each term in the years 2018, 2019 and 2020 their PageRank centrality [ 17 ]. The PageRank centrality tells us how likely a random walker traversing a network would be found at a given node if she follows the weights of the empirical edges (i.e., co-usage probability). Specifically, for the case of the MeSH co-occurrence network, this number represents how often an annotator at the National Library of Medicine would assign that MeSH term following the observed general usage patterns. It is a simplistic measure to capture the complexities of biomedical research. Nevertheless, it captures far-reaching interdependence across MeSH terms as the measure uses the whole network to determine the centrality of every MeSH term. A sudden change in the rankings and thus the position of MeSH terms in this network suggests that a given research subject has risen as it is used more often with other important MeSH terms (or vice versa).

We then compare the growth for each MeSH i term in g i (2019), i.e. before the the COVID-19 pandemic, with the growth after the event ( g i (2020)).

Publication growth

Changes in output and COVID-19 relatedness

Before we show the regression results, we provide descriptive evidence that publications from 2019 to 2020 have drastically increased. By showing that this growth correlates strongly with a MeSH term’s COVID-19 relatedness ( σ ), we demonstrate that (1) σ captures an essential aspect of the growth dynamics and (2) highlight the meteoric rise of highly related terms.

We look at the year over year growth in the number of the impact weighted number of publications per MeSH term from 2018 to 2019 and 2019 to 2020 as defined in the methods section.

Fig 1 shows the yearly growth of the impact weighted number of publications per MeSH term. By comparing the growth of the number of publications from the years 2018, 2019 and 2020, we find that the impact factor weighted number of publications has increased by up to a factor of 100 compared to the previous year for Betacoronavirus, one of the most closely related to COVID-19 MeSH term.

Each dot represents, a MeSH term. The y axis (growth) is in symmetric log scale. The x axis shows the COVID-19 relatedness, σ . Note that the position of the dots on the x-axis is the same in the two plots. Below: MeSH term importance gain (PageRank) and their COVID-19 relatedness.

https://doi.org/10.1371/journal.pone.0263001.g001

Fig 1 , first row, reveals how strongly correlated the growth in the IFWN of publication is to the term’s COVID-19 relatedness. For instance, we see that the term ‘Betacoronavirus’ skyrocketed from 2019 to 2020, which is expected given that SARS-CoV-2 is a species of the genus. Conversely, the term ‘Alphacoronavirus’ has not experienced any growth given that it is twin a genus of the Coronaviridae family, but SARS-CoV-2 is not one of its species. Note also the fast growth in the number of publications dealing with ‘Quarantine’. Moreover, MeSH terms that grew significantly from 2018 to 2019 and were not closely related to COVID-19, like ‘Vaping’, slowed down in 2020. From the graph, the picture emerges that publication growth is correlated with COVID-19 relatedness σ and that the growth for less related terms slowed down.

To show that the usage pattern of MeSH terms has changed following the pandemic, we compute the PageRank centrality using graph-tool [ 18 ] as discussed in the Methods section.

Fig 1 , second row, shows the change in the PageRank centrality of the MeSH terms after the pandemic (2019 to 2020, right plot) and before (2018 to 2019, left plot). If there were no change in the general usage pattern, we would expect the variance in PageRank changes to be narrow across the two periods, see (left plot). However, PageRank scores changed significantly more from 2019 to 2020 than from 2018 to 2019, suggesting that there has been a reconfiguration of the network.

To further support this argument, we carry out a DiD regression analysis.

Common trends assumption

As discussed in the Methods section, we need to show that the CTA assumption holds for the DiD to be defined appropriately. We do this by estimating for each month the number of publications and comparing it across treatment groups. This exercise also serves the purpose of a placebo test. By assuming that each month could have potentially been the event’s timing (i.e., the outbreak), we show that January 2020 is the most likely timing of the event. The regression table, as noted earlier, contains over 70 estimated coefficients, hence for ease of reading, we will only show the predicted outcome per month by group (see Fig 2 ). The full regression table with all coefficients is available in the S1 Table .

The y axis is in log scale. The dashed vertical line identifies January 2020. The dashed horizontal line shows the publications in January 2019 for the 0–20% group before the event. This line highlights that the drop happens after the event. The bands around the lines indicate the 95% confidence interval of the predicted values. The results are the output of the Stata margins command.

https://doi.org/10.1371/journal.pone.0263001.g002

Fig 2 shows the predicted number per outcome variable obtained from the panel regression model. These predictions correspond to the predicted value per relatedness group using the regression parameters estimated via the linear panel regression. The bands around the curves are the 95% confidence intervals.

All outcome measures depict a similar trend per month. Before the event (i.e., January 2020), there is a common trend across all groups. In contrast, after the event, we observe a sudden rise for the outcomes of the COVID-19 related treated groups (green and red lines) and a decline in the outcomes for the unrelated group (blue line). Therefore, we can conclude that the CTA assumption holds.

Regression results

Table 3 shows the DiD regression results (see Eq (3) ) for the selected outcome measures: number of publications (Papers), impact factor weighted number of publications (Impact), open access (OA) publications, clinical trial related publications, and publications with existing grants.

https://doi.org/10.1371/journal.pone.0263001.t003

Table 3 shows results for the discrete treatment level version of the DiD model (see Eq (4) ).

Note that the outcome variable is in natural log scale; hence to get the effect of the independent variable, we need to exponentiate the coefficient. For values close to 0, the effect is well approximated by the percentage change of that magnitude.

In both specifications we see that the least related group, drops in the number of publications between 10% and 13%, respectively (first row of Tables 3 and 4 , exp(−0.102) ≈ 0.87). In line with our expectations, the increase in the number of papers published by MeSH term is positively affected by the relatedness to COVID-19. In the discrete model (row 2), we note that the number of documents with MeSH terms with a COVID-19 relatedness between 20 and 80% grows by 18% and highly related terms by a factor of approximately 6.6 (exp(1.88)). The same general pattern can be observed for the impact weighted publication number, i.e., Model (2). Note, however, that the drop in the impact factor weighted output is more significant, reaching -19% for COVID-19 unrelated publications, and related publications growing by a factor of 8.7. This difference suggests that there might be a bias to publish papers on COVID-19 related subjects in high impact factor journals.

https://doi.org/10.1371/journal.pone.0263001.t004

By looking at the number of open access publications (PMC), we note that the least related group has not been affected negatively by the pandemic. However, the number of COVID-19 related publications has drastically increased for the most COVID-19 related group by a factor of 6.2. Note that the substantial increase in the number of papers available through open access is in large part due to journal and editorial policies to make preferentially COVID research immediately available to the public.

Regarding the number of clinical trial publications, we note that the least related group has been affected negatively, with the number of publications on clinical trials dropping by a staggering 24%. At the same time, publications on clinical trials for COVID-19-related MeSH have increased by a factor of 2.1. Note, however, that the effect on clinical trials is not significant in the continuous regression. The discrepancy across Tables 3 and 4 highlights that, especially for trials, the effect is not linear, where only the publications on clinical trials closely related to COVID-19 experiencing a boost.

It has been reported [ 19 ] that while the number of clinical trials registered to treat or prevent COVID-19 has surged with 179 new registrations in the second week of April 2020 alone. Only a few of these have led to publishable results in the 12 months since [ 20 ]. On the other hand, we find that clinical trial publications, considering related MeSH (but not COVID-19 directly), have had significant growth from the beginning of the pandemic. These results are not contradictory. Indeed counting the number of clinical trial publications listing the exact COVID-19 MeSH term (D000086382), we find 212 publications. While this might seem like a small number, consider that in 2020 only 8,485 publications were classified as clinical trials; thus, targeted trials still made up 2.5% of all clinical trials in 2020 . So while one might doubt the effectiveness of these research efforts, it is still the case that by sheer number, they represent a significant proportion of all publications on clinical trials in 2020. Moreover, COVID-19 specific Clinical trial publications in 2020, being a delayed signal of the actual trials, are a lower bound estimate on the true number of such clinical trials being conducted. This is because COVID-19 studies could only have commenced in 2020, whereas other studies had a head start. Thus our reported estimates are conservative, meaning that the true effect on actual clinical trials is likely larger, not smaller.

Research funding, as proxied by the number of publications with grants, follows a similar pattern, but notably, COVID-19-related MeSH terms list the same proportion of grants established before 2019 as other unrelated MeSH terms, suggesting that grants which were not designated for COVID-19 research have been used to support COVID-19 related research. Overall, the number of publications listing a grant has dropped. Note that this should be because the number of publications overall in the unrelated group has dropped. However, we note that the drop in publications is 10% while the decline in publications with at least one grant is 15%. This difference suggests that publications listing grants, which should have more funding, are disproportionately COVID-19 related papers. To further investigate this aspect, we look at whether the grant was old (pre-2019) or appeared for the first time in or after 2019. It stands to reason that an old grant (pre-2019) would not have been granted for a project dealing with the pandemic. Hence we would expect that COVID-19 related MeSH terms to have a lower proportion of old grants than the unrelated group. In models (6) in Table 4 we show that the number of old grants for the unrelated group drops by 13%. At the same time, the number of papers listing old grants (i.e., pre-2019) among the most related group increased by a factor of 3.1. Overall, these results suggest that COVID-19 related research has been funded largely by pre-existing grants, even though a specific mandate tied to the grants for this use is unlikely.

The scientific community has swiftly reallocated research efforts to cope with the COVID-19 pandemic, mobilizing knowledge across disciplines to find innovative solutions in record time. We document this both in terms of changing trends in the biomedical scientific output and the usage of MeSH terms by the scientific community. The flip side of this sudden and energetic prioritization of effort to fight COVID-19 has been a sudden contraction of scientific production in other relevant research areas. All in all, we find strong support to the hypotheses that the COVID-19 crisis has induced a sudden increase of research output in COVID-19 related areas of biomedical research. Conversely, research in areas not related to COVID-19 has experienced a significant drop in overall publishing rates and funding.

Our paper contributes to the literature on the impact of COVID-19 on scientific research: we corroborate previous findings about the surge of COVID-19 related publications [ 1 – 3 ], partially displacing research in COVID-19 unrelated fields of research [ 4 , 14 ], particularly research related to clinical trials [ 5 – 7 ]. The drop in trial research might have severe consequences for patients affected by life-threatening diseases since it will delay access to new and better treatments. We also confirm the impact of COVID-19 on open access publication output [ 1 ]; also, this is milder than traditional outlets. On top of this, we provide more robust evidence on the impact weighted effect of COVID-19 and grant financed research, highlighting the strong displacement effect of COVID-19 on the allocation of financial resources [ 15 ]. We document a substantial change in the usage patterns of MeSH terms, suggesting that there has been a reconfiguration in the way research terms are being combined. MeSH terms highly related to COVID-19 were peripheral in the MeSH usage networks before the pandemic but have become central since 2020. We conclude that the usage patterns have changed, with COVID-19 related MeSH terms occupying a much more prominent role in 2020 than they did in the previous years.

We also contribute to the literature by estimating the effect of COVID-19 on biomedical research in a natural experiment framework, isolating the specific effects of the COVID-19 pandemic on the biomedical scientific landscape. This is crucial to identify areas of public intervention to sustain areas of biomedical research which have been neglected during the COVID-19 crisis. Moreover, the exploratory analysis on the changes in usage patterns of MeSH terms, points to an increase in the importance of covid-related topics in the broader biomedical research landscape.

Our results provide compelling evidence that research related to COVID-19 has indeed displaced scientific production in other biomedical fields of research not related to COVID-19, with a significant drop in (impact weighted) scientific output related to non-COVID-19 and a marked reduction of financial support for publications not related to COVID-19 [ 4 , 5 , 16 ]. The displacement effect is persistent to the end of 2020. As vaccination progresses, we highlight the urgent need for science policy to re-balance support for research activity that was put on pause because of the COVID-19 pandemic.

We find that COVID-19 dramatically impacted clinical research. Reactivation of clinical trials activities that have been postponed or suspended for reasons related to COVID-19 is a priority that should be considered in the national vaccination plans. Moreover, since grants have been diverted and financial incentives have been targeted to sustain COVID-19 research leading to an excessive entry in COVID-19-related clinical trials and the ‘covidisation’ of research, there is a need to reorient incentives to basic research and otherwise neglected or temporally abandoned areas of biomedical research. Without dedicated support in the recovery plans for neglected research of the COVID-19 era, there is a risk that more medical needs will be unmet in the future, possibly exacerbating the shortage of scientific research for orphan and neglected diseases, which do not belong to COVID-19-related research areas.

Limitations

Our empirical approach has some limits. First, we proxy MeSH term usage via search terms using the PubMed EUtilities API. This means that the accuracy of the MeSH term we assign to a given paper is not fully validated. More time is needed for the completion of manually annotated MeSH terms. Second, the timing of publication is not the moment the research has been carried out. There is a lead time between inception, analysis, write-up, review, revision, and final publication. This delay varies across disciplines. Nevertheless, given that the surge in publications happens around the alleged event date, January 2020, we are confident that the publication date is a reasonable yet imperfect estimate of the timing of the research. Third, several journals have publicly declared to fast-track COVID-19 research. This discrepancy in the speed of publication of COVID-19 related research and other research could affect our results. Specifically, a surge or displacement could be overestimated due to a lag in the publication of COVID-19 unrelated research. We alleviate this bias by estimating the effect considering a considerable time after the event (January 2020 to December 2020). Forth, on the one hand, clinical Trials may lead to multiple publications. Therefore we might overestimate the impact of COVID-19 on the number of clinical trials. On the other hand, COVID-19 publications on clinical trials lag behind, so the number of papers related COVID-19 trials is likely underestimated. Therefore, we note that the focus of this paper is scientific publications on clinical trials rather than on actual clinical trials. Fifth, regarding grants, unfortunately, there is no unique centralized repository mapping grant numbers to years, so we have to proxy old grants with grants that appeared in publications from 2010 to 2018. Besides, grant numbers are free-form entries, meaning that PubMed has no validation step to disambiguate or verify that the grant number has been entered correctly. This has the effect of classifying a grant as new even though it has appeared under a different name. We mitigate this problem by using a long period to collect grant numbers and catch many spellings of the same grant, thereby reducing the likelihood of miss-identifying a grant as new when it existed before. Still, unless unique identifiers are widely used, there is no way to verify this.

So far, there is no conclusive evidence on whether entry into COVID-19 has been excessive. However, there is a growing consensus that COVID-19 has displaced, at least temporally, scientific research in COVID-19 unrelated biomedical research areas. Even though it is certainly expected that more attention will be devoted to the emergency during a pandemic, the displacement of biomedical research in other fields is concerning. Future research is needed to investigate the long-run structural consequences of the COVID-19 crisis on biomedical research.

Supporting information

S1 table. common trend assumption (cta) regression table..

Full regression table with all controls and interactions.

https://doi.org/10.1371/journal.pone.0263001.s001

• View Article
• Google Scholar
• PubMed/NCBI
• 8. Brown A, Edwin E, Fagg J. Evaluate Pharma 2021 Preview; 2020. https://www.evaluate.com/thought-leadership/vantage/evaluate-vantage-2021-preview .
• 15. Hook D, Porter S. The COVID Brain Drain; 2020. https://www.natureindex.com/news-blog/covid-brain-drain-scientific-research .
• 17. Page L, Brin S, Motwani R, Winograd T. The PageRank citation ranking: Bringing order to the web. Stanford InfoLab; 1999.

Aid for Incumbents: The Electoral Consequences of COVID-19 Relief

The COVID-19 pandemic led to unprecedented levels of federal transfers to state governments. Did this funding increase benefit incumbent politicians electorally? Identifying the effect of revenue windfalls on voting is challenging because whatever conditions led to the influx of cash might also benefit or harm incumbents for other reasons. We develop an instrument that allows us to predict allocations to states based on variation in congressional representation. We find that incumbents in state-wide races in 2020, 2021, and 2022 performed significantly better in states that received more relief funding due to their overrepresentation in Congress. These results are robust across specifications and after adjusting for a variety of economic and political controls. We consistently find that the pandemic-period electoral advantage of incumbent politicians in states receiving more aid substantially exceeds the more modest advantage politicians in these states enjoyed before the pandemic. This paper contributes to our understanding of economic voting and the incumbency advantage during times of crisis as well as the downstream electoral consequences of both the COVID-19 pandemic and of unequal political representation at the federal level.

Thanks to our excellent research assistants Jensen Ahokovi and Beatrice Lee. We are grateful to Jesús Fernández-Villaverde, Marko Klašnja, Vladimir Kogan, Nolan McCarty, Maria Polyakova, Ron Rogowski, and Hye Young You as well as attendees at the American Political Science Association's 120th Annual Meeting, the 13th Annual Conference of the American Society of Health Economists, Georgetown University, Harvard University, the Hoover Institution, and the University of Queensland for helpful comments and suggestions. Clemens thanks the Hoover Institution for support in his capacity as an adjunct Senior Fellow. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.

MARC RIS BibTeΧ

Download Citation Data

More from NBER

In addition to working papers , the NBER disseminates affiliates’ latest findings through a range of free periodicals — the NBER Reporter , the NBER Digest , the Bulletin on Retirement and Disability , the Bulletin on Health , and the Bulletin on Entrepreneurship  — as well as online conference reports , video lectures , and interviews .

Peer Reviewed

GPT-fabricated scientific papers on Google Scholar: Key features, spread, and implications for preempting evidence manipulation

Article metrics.

CrossRef Citations

Altmetric Score

PDF Downloads

Academic journals, archives, and repositories are seeing an increasing number of questionable research papers clearly produced using generative AI. They are often created with widely available, general-purpose AI applications, most likely ChatGPT, and mimic scientific writing. Google Scholar easily locates and lists these questionable papers alongside reputable, quality-controlled research. Our analysis of a selection of questionable GPT-fabricated scientific papers found in Google Scholar shows that many are about applied, often controversial topics susceptible to disinformation: the environment, health, and computing. The resulting enhanced potential for malicious manipulation of society’s evidence base, particularly in politically divisive domains, is a growing concern.

Swedish School of Library and Information Science, University of Borås, Sweden

Department of Arts and Cultural Sciences, Lund University, Sweden

Division of Environmental Communication, Swedish University of Agricultural Sciences, Sweden

Research Questions

• Where are questionable publications produced with generative pre-trained transformers (GPTs) that can be found via Google Scholar published or deposited?
• What are the main characteristics of these publications in relation to predominant subject categories?
• How are these publications spread in the research infrastructure for scholarly communication?
• How is the role of the scholarly communication infrastructure challenged in maintaining public trust in science and evidence through inappropriate use of generative AI?

research note Summary

• A sample of scientific papers with signs of GPT-use found on Google Scholar was retrieved, downloaded, and analyzed using a combination of qualitative coding and descriptive statistics. All papers contained at least one of two common phrases returned by conversational agents that use large language models (LLM) like OpenAI’s ChatGPT. Google Search was then used to determine the extent to which copies of questionable, GPT-fabricated papers were available in various repositories, archives, citation databases, and social media platforms.
• Roughly two-thirds of the retrieved papers were found to have been produced, at least in part, through undisclosed, potentially deceptive use of GPT. The majority (57%) of these questionable papers dealt with policy-relevant subjects (i.e., environment, health, computing), susceptible to influence operations. Most were available in several copies on different domains (e.g., social media, archives, and repositories).
• Two main risks arise from the increasingly common use of GPT to (mass-)produce fake, scientific publications. First, the abundance of fabricated “studies” seeping into all areas of the research infrastructure threatens to overwhelm the scholarly communication system and jeopardize the integrity of the scientific record. A second risk lies in the increased possibility that convincingly scientific-looking content was in fact deceitfully created with AI tools and is also optimized to be retrieved by publicly available academic search engines, particularly Google Scholar. However small, this possibility and awareness of it risks undermining the basis for trust in scientific knowledge and poses serious societal risks.

Implications

The use of ChatGPT to generate text for academic papers has raised concerns about research integrity. Discussion of this phenomenon is ongoing in editorials, commentaries, opinion pieces, and on social media (Bom, 2023; Stokel-Walker, 2024; Thorp, 2023). There are now several lists of papers suspected of GPT misuse, and new papers are constantly being added. 1 See for example Academ-AI, https://www.academ-ai.info/ , and Retraction Watch, https://retractionwatch.com/papers-and-peer-reviews-with-evidence-of-chatgpt-writing/ . While many legitimate uses of GPT for research and academic writing exist (Huang & Tan, 2023; Kitamura, 2023; Lund et al., 2023), its undeclared use—beyond proofreading—has potentially far-reaching implications for both science and society, but especially for their relationship. It, therefore, seems important to extend the discussion to one of the most accessible and well-known intermediaries between science, but also certain types of misinformation, and the public, namely Google Scholar, also in response to the legitimate concerns that the discussion of generative AI and misinformation needs to be more nuanced and empirically substantiated  (Simon et al., 2023).

Google Scholar, https://scholar.google.com , is an easy-to-use academic search engine. It is available for free, and its index is extensive (Gusenbauer & Haddaway, 2020). It is also often touted as a credible source for academic literature and even recommended in library guides, by media and information literacy initiatives, and fact checkers (Tripodi et al., 2023). However, Google Scholar lacks the transparency and adherence to standards that usually characterize citation databases. Instead, Google Scholar uses automated crawlers, like Google’s web search engine (Martín-Martín et al., 2021), and the inclusion criteria are based on primarily technical standards, allowing any individual author—with or without scientific affiliation—to upload papers to be indexed (Google Scholar Help, n.d.). It has been shown that Google Scholar is susceptible to manipulation through citation exploits (Antkare, 2020) and by providing access to fake scientific papers (Dadkhah et al., 2017). A large part of Google Scholar’s index consists of publications from established scientific journals or other forms of quality-controlled, scholarly literature. However, the index also contains a large amount of gray literature, including student papers, working papers, reports, preprint servers, and academic networking sites, as well as material from so-called “questionable” academic journals, including paper mills. The search interface does not offer the possibility to filter the results meaningfully by material type, publication status, or form of quality control, such as limiting the search to peer-reviewed material.

To understand the occurrence of ChatGPT (co-)authored work in Google Scholar’s index, we scraped it for publications, including one of two common ChatGPT responses (see Appendix A) that we encountered on social media and in media reports (DeGeurin, 2024). The results of our descriptive statistical analyses showed that around 62% did not declare the use of GPTs. Most of these GPT-fabricated papers were found in non-indexed journals and working papers, but some cases included research published in mainstream scientific journals and conference proceedings. 2 Indexed journals mean scholarly journals indexed by abstract and citation databases such as Scopus and Web of Science, where the indexation implies journals with high scientific quality. Non-indexed journals are journals that fall outside of this indexation. More than half (57%) of these GPT-fabricated papers concerned policy-relevant subject areas susceptible to influence operations. To avoid increasing the visibility of these publications, we abstained from referencing them in this research note. However, we have made the data available in the Harvard Dataverse repository.

The publications were related to three issue areas—health (14.5%), environment (19.5%) and computing (23%)—with key terms such “healthcare,” “COVID-19,” or “infection”for health-related papers, and “analysis,” “sustainable,” and “global” for environment-related papers. In several cases, the papers had titles that strung together general keywords and buzzwords, thus alluding to very broad and current research. These terms included “biology,” “telehealth,” “climate policy,” “diversity,” and “disrupting,” to name just a few.  While the study’s scope and design did not include a detailed analysis of which parts of the articles included fabricated text, our dataset did contain the surrounding sentences for each occurrence of the suspicious phrases that formed the basis for our search and subsequent selection. Based on that, we can say that the phrases occurred in most sections typically found in scientific publications, including the literature review, methods, conceptual and theoretical frameworks, background, motivation or societal relevance, and even discussion. This was confirmed during the joint coding, where we read and discussed all articles. It became clear that not just the text related to the telltale phrases was created by GPT, but that almost all articles in our sample of questionable articles likely contained traces of GPT-fabricated text everywhere.

Evidence hacking and backfiring effects

Generative pre-trained transformers (GPTs) can be used to produce texts that mimic scientific writing. These texts, when made available online—as we demonstrate—leak into the databases of academic search engines and other parts of the research infrastructure for scholarly communication. This development exacerbates problems that were already present with less sophisticated text generators (Antkare, 2020; Cabanac & Labbé, 2021). Yet, the public release of ChatGPT in 2022, together with the way Google Scholar works, has increased the likelihood of lay people (e.g., media, politicians, patients, students) coming across questionable (or even entirely GPT-fabricated) papers and other problematic research findings. Previous research has emphasized that the ability to determine the value and status of scientific publications for lay people is at stake when misleading articles are passed off as reputable (Haider & Åström, 2017) and that systematic literature reviews risk being compromised (Dadkhah et al., 2017). It has also been highlighted that Google Scholar, in particular, can be and has been exploited for manipulating the evidence base for politically charged issues and to fuel conspiracy narratives (Tripodi et al., 2023). Both concerns are likely to be magnified in the future, increasing the risk of what we suggest calling evidence hacking —the strategic and coordinated malicious manipulation of society’s evidence base.

The authority of quality-controlled research as evidence to support legislation, policy, politics, and other forms of decision-making is undermined by the presence of undeclared GPT-fabricated content in publications professing to be scientific. Due to the large number of archives, repositories, mirror sites, and shadow libraries to which they spread, there is a clear risk that GPT-fabricated, questionable papers will reach audiences even after a possible retraction. There are considerable technical difficulties involved in identifying and tracing computer-fabricated papers (Cabanac & Labbé, 2021; Dadkhah et al., 2023; Jones, 2024), not to mention preventing and curbing their spread and uptake.

However, as the rise of the so-called anti-vaxx movement during the COVID-19 pandemic and the ongoing obstruction and denial of climate change show, retracting erroneous publications often fuels conspiracies and increases the following of these movements rather than stopping them. To illustrate this mechanism, climate deniers frequently question established scientific consensus by pointing to other, supposedly scientific, studies that support their claims. Usually, these are poorly executed, not peer-reviewed, based on obsolete data, or even fraudulent (Dunlap & Brulle, 2020). A similar strategy is successful in the alternative epistemic world of the global anti-vaccination movement (Carrion, 2018) and the persistence of flawed and questionable publications in the scientific record already poses significant problems for health research, policy, and lawmakers, and thus for society as a whole (Littell et al., 2024). Considering that a person’s support for “doing your own research” is associated with increased mistrust in scientific institutions (Chinn & Hasell, 2023), it will be of utmost importance to anticipate and consider such backfiring effects already when designing a technical solution, when suggesting industry or legal regulation, and in the planning of educational measures.

Recommendations

Solutions should be based on simultaneous considerations of technical, educational, and regulatory approaches, as well as incentives, including social ones, across the entire research infrastructure. Paying attention to how these approaches and incentives relate to each other can help identify points and mechanisms for disruption. Recognizing fraudulent academic papers must happen alongside understanding how they reach their audiences and what reasons there might be for some of these papers successfully “sticking around.” A possible way to mitigate some of the risks associated with GPT-fabricated scholarly texts finding their way into academic search engine results would be to provide filtering options for facets such as indexed journals, gray literature, peer-review, and similar on the interface of publicly available academic search engines. Furthermore, evaluation tools for indexed journals 3 Such as LiU Journal CheckUp, https://ep.liu.se/JournalCheckup/default.aspx?lang=eng . could be integrated into the graphical user interfaces and the crawlers of these academic search engines. To enable accountability, it is important that the index (database) of such a search engine is populated according to criteria that are transparent, open to scrutiny, and appropriate to the workings of  science and other forms of academic research. Moreover, considering that Google Scholar has no real competitor, there is a strong case for establishing a freely accessible, non-specialized academic search engine that is not run for commercial reasons but for reasons of public interest. Such measures, together with educational initiatives aimed particularly at policymakers, science communicators, journalists, and other media workers, will be crucial to reducing the possibilities for and effects of malicious manipulation or evidence hacking. It is important not to present this as a technical problem that exists only because of AI text generators but to relate it to the wider concerns in which it is embedded. These range from a largely dysfunctional scholarly publishing system (Haider & Åström, 2017) and academia’s “publish or perish” paradigm to Google’s near-monopoly and ideological battles over the control of information and ultimately knowledge. Any intervention is likely to have systemic effects; these effects need to be considered and assessed in advance and, ideally, followed up on.

Our study focused on a selection of papers that were easily recognizable as fraudulent. We used this relatively small sample as a magnifying glass to examine, delineate, and understand a problem that goes beyond the scope of the sample itself, which however points towards larger concerns that require further investigation. The work of ongoing whistleblowing initiatives 4 Such as Academ-AI, https://www.academ-ai.info/ , and Retraction Watch, https://retractionwatch.com/papers-and-peer-reviews-with-evidence-of-chatgpt-writing/ . , recent media reports of journal closures (Subbaraman, 2024), or GPT-related changes in word use and writing style (Cabanac et al., 2021; Stokel-Walker, 2024) suggest that we only see the tip of the iceberg. There are already more sophisticated cases (Dadkhah et al., 2023) as well as cases involving fabricated images (Gu et al., 2022). Our analysis shows that questionable and potentially manipulative GPT-fabricated papers permeate the research infrastructure and are likely to become a widespread phenomenon. Our findings underline that the risk of fake scientific papers being used to maliciously manipulate evidence (see Dadkhah et al., 2017) must be taken seriously. Manipulation may involve undeclared automatic summaries of texts, inclusion in literature reviews, explicit scientific claims, or the concealment of errors in studies so that they are difficult to detect in peer review. However, the mere possibility of these things happening is a significant risk in its own right that can be strategically exploited and will have ramifications for trust in and perception of science. Society’s methods of evaluating sources and the foundations of media and information literacy are under threat and public trust in science is at risk of further erosion, with far-reaching consequences for society in dealing with information disorders. To address this multifaceted problem, we first need to understand why it exists and proliferates.

Finding 1: 139 GPT-fabricated, questionable papers were found and listed as regular results on the Google Scholar results page. Non-indexed journals dominate.

Most questionable papers we found were in non-indexed journals or were working papers, but we did also find some in established journals, publications, conferences, and repositories. We found a total of 139 papers with a suspected deceptive use of ChatGPT or similar LLM applications (see Table 1). Out of these, 19 were in indexed journals, 89 were in non-indexed journals, 19 were student papers found in university databases, and 12 were working papers (mostly in preprint databases). Table 1 divides these papers into categories. Health and environment papers made up around 34% (47) of the sample. Of these, 66% were present in non-indexed journals.

Finding 2: GPT-fabricated, questionable papers are disseminated online, permeating the research infrastructure for scholarly communication, often in multiple copies. Applied topics with practical implications dominate.

The 20 papers concerning health-related issues are distributed across 20 unique domains, accounting for 46 URLs. The 27 papers dealing with environmental issues can be found across 26 unique domains, accounting for 56 URLs.  Most of the identified papers exist in multiple copies and have already spread to several archives, repositories, and social media. It would be difficult, or impossible, to remove them from the scientific record.

As apparent from Table 2, GPT-fabricated, questionable papers are seeping into most parts of the online research infrastructure for scholarly communication. Platforms on which identified papers have appeared include ResearchGate, ORCiD, Journal of Population Therapeutics and Clinical Pharmacology (JPTCP), Easychair, Frontiers, the Institute of Electrical and Electronics Engineer (IEEE), and X/Twitter. Thus, even if they are retracted from their original source, it will prove very difficult to track, remove, or even just mark them up on other platforms. Moreover, unless regulated, Google Scholar will enable their continued and most likely unlabeled discoverability.

A word rain visualization (Centre for Digital Humanities Uppsala, 2023), which combines word prominences through TF-IDF 5 Term frequency–inverse document frequency , a method for measuring the significance of a word in a document compared to its frequency across all documents in a collection. scores with semantic similarity of the full texts of our sample of GPT-generated articles that fall into the “Environment” and “Health” categories, reflects the two categories in question. However, as can be seen in Figure 1, it also reveals overlap and sub-areas. The y-axis shows word prominences through word positions and font sizes, while the x-axis indicates semantic similarity. In addition to a certain amount of overlap, this reveals sub-areas, which are best described as two distinct events within the word rain. The event on the left bundles terms related to the development and management of health and healthcare with “challenges,” “impact,” and “potential of artificial intelligence”emerging as semantically related terms. Terms related to research infrastructures, environmental, epistemic, and technological concepts are arranged further down in the same event (e.g., “system,” “climate,” “understanding,” “knowledge,” “learning,” “education,” “sustainable”). A second distinct event further to the right bundles terms associated with fish farming and aquatic medicinal plants, highlighting the presence of an aquaculture cluster.  Here, the prominence of groups of terms such as “used,” “model,” “-based,” and “traditional” suggests the presence of applied research on these topics. The two events making up the word rain visualization, are linked by a less dominant but overlapping cluster of terms related to “energy” and “water.”

The bar chart of the terms in the paper subset (see Figure 2) complements the word rain visualization by depicting the most prominent terms in the full texts along the y-axis. Here, word prominences across health and environment papers are arranged descendingly, where values outside parentheses are TF-IDF values (relative frequencies) and values inside parentheses are raw term frequencies (absolute frequencies).

Finding 3: Google Scholar presents results from quality-controlled and non-controlled citation databases on the same interface, providing unfiltered access to GPT-fabricated questionable papers.

Google Scholar’s central position in the publicly accessible scholarly communication infrastructure, as well as its lack of standards, transparency, and accountability in terms of inclusion criteria, has potentially serious implications for public trust in science. This is likely to exacerbate the already-known potential to exploit Google Scholar for evidence hacking (Tripodi et al., 2023) and will have implications for any attempts to retract or remove fraudulent papers from their original publication venues. Any solution must consider the entirety of the research infrastructure for scholarly communication and the interplay of different actors, interests, and incentives.

We searched and scraped Google Scholar using the Python library Scholarly (Cholewiak et al., 2023) for papers that included specific phrases known to be common responses from ChatGPT and similar applications with the same underlying model (GPT3.5 or GPT4): “as of my last knowledge update” and/or “I don’t have access to real-time data” (see Appendix A). This facilitated the identification of papers that likely used generative AI to produce text, resulting in 227 retrieved papers. The papers’ bibliographic information was automatically added to a spreadsheet and downloaded into Zotero. 6 An open-source reference manager, https://zotero.org .

We employed multiple coding (Barbour, 2001) to classify the papers based on their content. First, we jointly assessed whether the paper was suspected of fraudulent use of ChatGPT (or similar) based on how the text was integrated into the papers and whether the paper was presented as original research output or the AI tool’s role was acknowledged. Second, in analyzing the content of the papers, we continued the multiple coding by classifying the fraudulent papers into four categories identified during an initial round of analysis—health, environment, computing, and others—and then determining which subjects were most affected by this issue (see Table 1). Out of the 227 retrieved papers, 88 papers were written with legitimate and/or declared use of GPTs (i.e., false positives, which were excluded from further analysis), and 139 papers were written with undeclared and/or fraudulent use (i.e., true positives, which were included in further analysis). The multiple coding was conducted jointly by all authors of the present article, who collaboratively coded and cross-checked each other’s interpretation of the data simultaneously in a shared spreadsheet file. This was done to single out coding discrepancies and settle coding disagreements, which in turn ensured methodological thoroughness and analytical consensus (see Barbour, 2001). Redoing the category coding later based on our established coding schedule, we achieved an intercoder reliability (Cohen’s kappa) of 0.806 after eradicating obvious differences.

The ranking algorithm of Google Scholar prioritizes highly cited and older publications (Martín-Martín et al., 2016). Therefore, the position of the articles on the search engine results pages was not particularly informative, considering the relatively small number of results in combination with the recency of the publications. Only the query “as of my last knowledge update” had more than two search engine result pages. On those, questionable articles with undeclared use of GPTs were evenly distributed across all result pages (min: 4, max: 9, mode: 8), with the proportion of undeclared use being slightly higher on average on later search result pages.

To understand how the papers making fraudulent use of generative AI were disseminated online, we programmatically searched for the paper titles (with exact string matching) in Google Search from our local IP address (see Appendix B) using the googlesearch – python library(Vikramaditya, 2020). We manually verified each search result to filter out false positives—results that were not related to the paper—and then compiled the most prominent URLs by field. This enabled the identification of other platforms through which the papers had been spread. We did not, however, investigate whether copies had spread into SciHub or other shadow libraries, or if they were referenced in Wikipedia.

We used descriptive statistics to count the prevalence of the number of GPT-fabricated papers across topics and venues and top domains by subject. The pandas software library for the Python programming language (The pandas development team, 2024) was used for this part of the analysis. Based on the multiple coding, paper occurrences were counted in relation to their categories, divided into indexed journals, non-indexed journals, student papers, and working papers. The schemes, subdomains, and subdirectories of the URL strings were filtered out while top-level domains and second-level domains were kept, which led to normalizing domain names. This, in turn, allowed the counting of domain frequencies in the environment and health categories. To distinguish word prominences and meanings in the environment and health-related GPT-fabricated questionable papers, a semantically-aware word cloud visualization was produced through the use of a word rain (Centre for Digital Humanities Uppsala, 2023) for full-text versions of the papers. Font size and y-axis positions indicate word prominences through TF-IDF scores for the environment and health papers (also visualized in a separate bar chart with raw term frequencies in parentheses), and words are positioned along the x-axis to reflect semantic similarity (Skeppstedt et al., 2024), with an English Word2vec skip gram model space (Fares et al., 2017). An English stop word list was used, along with a manually produced list including terms such as “https,” “volume,” or “years.”

• Artificial Intelligence
• / Search engines

Cite this Essay

Haider, J., Söderström, K. R., Ekström, B., & Rödl, M. (2024). GPT-fabricated scientific papers on Google Scholar: Key features, spread, and implications for preempting evidence manipulation. Harvard Kennedy School (HKS) Misinformation Review . https://doi.org/10.37016/mr-2020-156

• / Appendix B

Bibliography

Antkare, I. (2020). Ike Antkare, his publications, and those of his disciples. In M. Biagioli & A. Lippman (Eds.), Gaming the metrics (pp. 177–200). The MIT Press. https://doi.org/10.7551/mitpress/11087.003.0018

Barbour, R. S. (2001). Checklists for improving rigour in qualitative research: A case of the tail wagging the dog? BMJ , 322 (7294), 1115–1117. https://doi.org/10.1136/bmj.322.7294.1115

Bom, H.-S. H. (2023). Exploring the opportunities and challenges of ChatGPT in academic writing: A roundtable discussion. Nuclear Medicine and Molecular Imaging , 57 (4), 165–167. https://doi.org/10.1007/s13139-023-00809-2

Cabanac, G., & Labbé, C. (2021). Prevalence of nonsensical algorithmically generated papers in the scientific literature. Journal of the Association for Information Science and Technology , 72 (12), 1461–1476. https://doi.org/10.1002/asi.24495

Cabanac, G., Labbé, C., & Magazinov, A. (2021). Tortured phrases: A dubious writing style emerging in science. Evidence of critical issues affecting established journals . arXiv. https://doi.org/10.48550/arXiv.2107.06751

Carrion, M. L. (2018). “You need to do your research”: Vaccines, contestable science, and maternal epistemology. Public Understanding of Science , 27 (3), 310–324. https://doi.org/10.1177/0963662517728024

Centre for Digital Humanities Uppsala (2023). CDHUppsala/word-rain [Computer software]. https://github.com/CDHUppsala/word-rain

Chinn, S., & Hasell, A. (2023). Support for “doing your own research” is associated with COVID-19 misperceptions and scientific mistrust. Harvard Kennedy School (HSK) Misinformation Review, 4 (3). https://doi.org/10.37016/mr-2020-117

Cholewiak, S. A., Ipeirotis, P., Silva, V., & Kannawadi, A. (2023). SCHOLARLY: Simple access to Google Scholar authors and citation using Python (1.5.0) [Computer software]. https://doi.org/10.5281/zenodo.5764801

Dadkhah, M., Lagzian, M., & Borchardt, G. (2017). Questionable papers in citation databases as an issue for literature review. Journal of Cell Communication and Signaling , 11 (2), 181–185. https://doi.org/10.1007/s12079-016-0370-6

Dadkhah, M., Oermann, M. H., Hegedüs, M., Raman, R., & Dávid, L. D. (2023). Detection of fake papers in the era of artificial intelligence. Diagnosis , 10 (4), 390–397. https://doi.org/10.1515/dx-2023-0090

DeGeurin, M. (2024, March 19). AI-generated nonsense is leaking into scientific journals. Popular Science. https://www.popsci.com/technology/ai-generated-text-scientific-journals/

Dunlap, R. E., & Brulle, R. J. (2020). Sources and amplifiers of climate change denial. In D.C. Holmes & L. M. Richardson (Eds.), Research handbook on communicating climate change (pp. 49–61). Edward Elgar Publishing. https://doi.org/10.4337/9781789900408.00013

Fares, M., Kutuzov, A., Oepen, S., & Velldal, E. (2017). Word vectors, reuse, and replicability: Towards a community repository of large-text resources. In J. Tiedemann & N. Tahmasebi (Eds.), Proceedings of the 21st Nordic Conference on Computational Linguistics (pp. 271–276). Association for Computational Linguistics. https://aclanthology.org/W17-0237

Google Scholar Help. (n.d.). Inclusion guidelines for webmasters . https://scholar.google.com/intl/en/scholar/inclusion.html

Gu, J., Wang, X., Li, C., Zhao, J., Fu, W., Liang, G., & Qiu, J. (2022). AI-enabled image fraud in scientific publications. Patterns , 3 (7), 100511. https://doi.org/10.1016/j.patter.2022.100511

Gusenbauer, M., & Haddaway, N. R. (2020). Which academic search systems are suitable for systematic reviews or meta-analyses? Evaluating retrieval qualities of Google Scholar, PubMed, and 26 other resources. Research Synthesis Methods , 11 (2), 181–217.   https://doi.org/10.1002/jrsm.1378

Haider, J., & Åström, F. (2017). Dimensions of trust in scholarly communication: Problematizing peer review in the aftermath of John Bohannon’s “Sting” in science. Journal of the Association for Information Science and Technology , 68 (2), 450–467. https://doi.org/10.1002/asi.23669

Huang, J., & Tan, M. (2023). The role of ChatGPT in scientific communication: Writing better scientific review articles. American Journal of Cancer Research , 13 (4), 1148–1154. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10164801/

Jones, N. (2024). How journals are fighting back against a wave of questionable images. Nature , 626 (8000), 697–698. https://doi.org/10.1038/d41586-024-00372-6

Kitamura, F. C. (2023). ChatGPT is shaping the future of medical writing but still requires human judgment. Radiology , 307 (2), e230171. https://doi.org/10.1148/radiol.230171

Littell, J. H., Abel, K. M., Biggs, M. A., Blum, R. W., Foster, D. G., Haddad, L. B., Major, B., Munk-Olsen, T., Polis, C. B., Robinson, G. E., Rocca, C. H., Russo, N. F., Steinberg, J. R., Stewart, D. E., Stotland, N. L., Upadhyay, U. D., & Ditzhuijzen, J. van. (2024). Correcting the scientific record on abortion and mental health outcomes. BMJ , 384 , e076518. https://doi.org/10.1136/bmj-2023-076518

Lund, B. D., Wang, T., Mannuru, N. R., Nie, B., Shimray, S., & Wang, Z. (2023). ChatGPT and a new academic reality: Artificial Intelligence-written research papers and the ethics of the large language models in scholarly publishing. Journal of the Association for Information Science and Technology, 74 (5), 570–581. https://doi.org/10.1002/asi.24750

Martín-Martín, A., Orduna-Malea, E., Ayllón, J. M., & Delgado López-Cózar, E. (2016). Back to the past: On the shoulders of an academic search engine giant. Scientometrics , 107 , 1477–1487. https://doi.org/10.1007/s11192-016-1917-2

Martín-Martín, A., Thelwall, M., Orduna-Malea, E., & Delgado López-Cózar, E. (2021). Google Scholar, Microsoft Academic, Scopus, Dimensions, Web of Science, and OpenCitations’ COCI: A multidisciplinary comparison of coverage via citations. Scientometrics , 126 (1), 871–906. https://doi.org/10.1007/s11192-020-03690-4

Simon, F. M., Altay, S., & Mercier, H. (2023). Misinformation reloaded? Fears about the impact of generative AI on misinformation are overblown. Harvard Kennedy School (HKS) Misinformation Review, 4 (5). https://doi.org/10.37016/mr-2020-127

Skeppstedt, M., Ahltorp, M., Kucher, K., & Lindström, M. (2024). From word clouds to Word Rain: Revisiting the classic word cloud to visualize climate change texts. Information Visualization , 23 (3), 217–238. https://doi.org/10.1177/14738716241236188

Swedish Research Council. (2017). Good research practice. Vetenskapsrådet.

Stokel-Walker, C. (2024, May 1.). AI Chatbots Have Thoroughly Infiltrated Scientific Publishing . Scientific American. https://www.scientificamerican.com/article/chatbots-have-thoroughly-infiltrated-scientific-publishing/

Subbaraman, N. (2024, May 14). Flood of fake science forces multiple journal closures: Wiley to shutter 19 more journals, some tainted by fraud. The Wall Street Journal . https://www.wsj.com/science/academic-studies-research-paper-mills-journals-publishing-f5a3d4bc

The pandas development team. (2024). pandas-dev/pandas: Pandas (v2.2.2) [Computer software]. Zenodo. https://doi.org/10.5281/zenodo.10957263

Thorp, H. H. (2023). ChatGPT is fun, but not an author. Science , 379 (6630), 313–313. https://doi.org/10.1126/science.adg7879

Tripodi, F. B., Garcia, L. C., & Marwick, A. E. (2023). ‘Do your own research’: Affordance activation and disinformation spread. Information, Communication & Society , 27 (6), 1212–1228. https://doi.org/10.1080/1369118X.2023.2245869

Vikramaditya, N. (2020). Nv7-GitHub/googlesearch [Computer software]. https://github.com/Nv7-GitHub/googlesearch

This research has been supported by Mistra, the Swedish Foundation for Strategic Environmental Research, through the research program Mistra Environmental Communication (Haider, Ekström, Rödl) and the Marcus and Amalia Wallenberg Foundation [2020.0004] (Söderström).

Competing Interests

The authors declare no competing interests.

The research described in this article was carried out under Swedish legislation. According to the relevant EU and Swedish legislation (2003:460) on the ethical review of research involving humans (“Ethical Review Act”), the research reported on here is not subject to authorization by the Swedish Ethical Review Authority (“etikprövningsmyndigheten”) (SRC, 2017).

This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided that the original author and source are properly credited.

Data Availability

All data needed to replicate this study are available at the Harvard Dataverse: https://doi.org/10.7910/DVN/WUVD8X

Acknowledgements

The authors wish to thank two anonymous reviewers for their valuable comments on the article manuscript as well as the editorial group of Harvard Kennedy School (HKS) Misinformation Review for their thoughtful feedback and input.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

The PMC website is updating on October 15, 2024. Learn More or Try it out now .

• Advanced Search
• Journal List

The impact of the COVID-19 pandemic on scientific research in the life sciences

Massimo riccaboni.

1 AXES, IMT School for Advanced Studies Lucca, Lucca, Italy

Luca Verginer

2 Chair of Systems Design D-MTEC, ETH Zürich, Zurich, Switzerland

Associated Data

The processed data, instructions on how to process the raw PubMed dataset as well as all code are available via Zenodo at https://doi.org/10.5281/zenodo.5121216 .

The COVID-19 outbreak has posed an unprecedented challenge to humanity and science. On the one side, public and private incentives have been put in place to promptly allocate resources toward research areas strictly related to the COVID-19 emergency. However, research in many fields not directly related to the pandemic has been displaced. In this paper, we assess the impact of COVID-19 on world scientific production in the life sciences and find indications that the usage of medical subject headings (MeSH) has changed following the outbreak. We estimate through a difference-in-differences approach the impact of the start of the COVID-19 pandemic on scientific production using the PubMed database (3.6 Million research papers). We find that COVID-19-related MeSH terms have experienced a 6.5 fold increase in output on average, while publications on unrelated MeSH terms dropped by 10 to 12%. The publication weighted impact has an even more pronounced negative effect (-16% to -19%). Moreover, COVID-19 has displaced clinical trial publications (-24%) and diverted grants from research areas not closely related to COVID-19. Note that since COVID-19 publications may have been fast-tracked, the sudden surge in COVID-19 publications might be driven by editorial policy.

Introduction

The COVID-19 pandemic has mobilized the world scientific community in 2020, especially in the life sciences [ 1 , 2 ]. In the first three months after the pandemic, the number of scientific papers about COVID-19 was fivefold the number of articles on H1N1 swine influenza [ 3 ]. Similarly, the number of clinical trials related to COVID-19 prophylaxis and treatments skyrocketed [ 4 ]. Thanks to the rapid mobilization of the world scientific community, COVID-19 vaccines have been developed in record time. Despite this undeniable success, there is a rising concern about the negative consequences of COVID-19 on clinical trial research, with many projects being postponed [ 5 – 7 ]. According to Evaluate Pharma, clinical trials were one of the pandemic’s first casualties, with a record number of 160 studies suspended for reasons related to COVID-19 in April 2020 [ 8 , 9 ] reporting a total of 1,200 trials suspended as of July 2020. As a consequence, clinical researchers have been impaired by reduced access to healthcare research infrastructures. Particularly, the COVID-19 outbreak took a tall on women and early-career scientists [ 10 – 13 ]. On a different ground, Shan and colleagues found that non-COVID-19-related articles decreased as COVID-19-related articles increased in top clinical research journals [ 14 ]. Fraser and coworker found that COVID-19 preprints received more attention and citations than non-COVID-19 preprints [ 1 ]. More recently, Hook and Porter have found some early evidence of ‘covidisation’ of academic research, with research grants and output diverted to COVID-19 research in 2020 [ 15 ]. How much should scientists switch their efforts toward SARS-CoV-2 prevention, treatment, or mitigation? There is a growing consensus that the current level of ‘covidisation’ of research can be wasteful [ 4 , 5 , 16 ].

Against this background, in this paper, we investigate if the COVID-19 pandemic has induced a shift in biomedical publications toward COVID-19-related scientific production. The objective of the study is to show that scientific articles listing covid-related Medical Subject Headings (MeSH) when compared against covid-unrelated MeSH have been partially displaced. Specifically, we look at several indicators of scientific production in the life sciences before and after the start of the COVID-19 pandemic: (1) number of papers published, (2) impact factor weighted number of papers, (3) opens access, (4) number of publications related to clinical trials, (5) number of papers listing grants, (6) number of papers listing grants existing before the pandemic. Through a natural experiment approach, we analyze the impact of the pandemic on scientific production in the life sciences. We consider COVID-19 an unexpected and unprecedented exogenous source of variation with heterogeneous effects across biomedical research fields (i.e., MeSH terms).

Based on the difference in difference results, we document the displacement effect that the pandemic has had on several aspects of scientific publishing. The overall picture that emerges from this analysis is that there has been a profound realignment of priorities and research efforts. This shift has displaced biomedical research in fields not related to COVID-19.

The rest of the paper is structured as follows. First, we describe the data and our measure of relatedness to COVID-19. Next, we illustrate the difference-in-differences specification we rely on to identify the impact of the pandemic on scientific output. In the results section, we present the results of the difference-in-differences and network analyses. We document the sudden shift in publications, grants and trials towards COVID-19-related MeSH terms. Finally, we discuss the findings and highlight several policy implications.

Materials and methods

The present analysis is based primarily on PubMed and the Medical Subject Headings (MeSH) terminology. This data is used to estimate the effect of the start of the COVID 19 pandemic via a difference in difference approach. This section is structured as follows. We first introduce the data and then the econometric methodology. This analysis is not based on a pre-registered protocol.

Selection of biomedical publications

We rely on PubMed, a repository with more than 34 million biomedical citations, for the analysis. Specifically, we analyze the daily updated files up to 31/06/2021, extracting all publications of type ‘Journal Article’. For the principal analysis, we consider 3,638,584 papers published from January 2019 to December 2020. We also analyze 11,122,017 papers published from 2010 onwards to identify the earliest usage of a grant and infer if it was new in 2020. We use the SCImago journal ranking statistics to compute the impact factor weighted number (IFWN) of papers in a given field of research. To assign the publication date, we use the ‘electronically published’ dates and, if missing, the ‘print published’ dates.

Medical subject headings

We rely on the Medical Subject Headings (MeSH) terminology to approximate narrowly defined biomedical research fields. This terminology is a curated medical vocabulary, which is manually added to papers in the PubMed corpus. The fact that MeSH terms are manually annotated makes this terminology ideal for classification purposes. However, there is a delay between publication and annotation, on the order of several months. To address this delay and have the most recent classification, we search for all 28 425 MeSH terms using PubMed’s ESearch utility and classify paper by the results. The specific API endpoint is https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi , the relevant scripts are available with the code. For example, we assign the term ‘Ageusia’ (MeSH ID D000370) to all papers listed in the results of the ESearch API. We apply this method to the whole period (January 2019—December 2020) and obtain a mapping from papers to the MeSH terms. For every MeSH term, we keep track of the year they have been established. For instance, COVID-19 terms were established in 2020 (see Table 1 ): in January 2020, the WHO recommended 2019-nCoV and 2019-nCoV acute respiratory disease as provisional names for the virus and disease. The WHO issued the official terms COVID-19 and SARS-CoV-2 at the beginning of February 2020. By manually annotating publications, all publications referring to COVID-19 and SARS-CoV-2 since January 2020 have been labelled with the related MeSH terms. Other MeSH terms related to COVID-19, such as coronavirus, for instance, have been established years before the pandemic (see Table 2 ). We proxy MeSH term usage via search terms using the PubMed EUtilities API; this means that we are not using the hand-labelled MeSH terms but rather the PubMed search results. This means that the accuracy of the MeSH term we assign to a given paper is not perfect. In practice, this means that we have assigned more MeSH terms to a given term than a human annotator would have.

The list contains only terms with at least 100 publications in 2020.

Clinical trials and publication types

We classify publications using PubMed’s ‘PublicationType’ field in the XML baseline files (There are 187 publication types, see https://www.nlm.nih.gov/mesh/pubtypes.html ). We consider a publication to be related to a clinical trial if it lists any of the following descriptors:

• D016430: Clinical Trial
• D017426: Clinical Trial, Phase I
• D017427: Clinical Trial, Phase II
• D017428: Clinical Trial, Phase III
• D017429: Clinical Trial, Phase IV
• D018848: Controlled Clinical Trial
• D065007: Pragmatic Clinical Trial
• D000076362: Adaptive Clinical Trial
• D000077522: Clinical Trial, Veterinary

In our analysis of the impact of COVID-19 on publications related to clinical trials, we only consider MeSH terms that are associated at least once with a clinical trial publication over the two years. We apply this restriction to filter out MeSH terms that are very unlikely to be relevant for clinical trial types of research.

Open access

We proxy the availability of a journal article to the public, i.e., open access, if it is available from PubMed Central. PubMed Central archives full-text journal articles and provides free access to the public. Note that the copyright license may vary across participating publishers. However, the text of the paper is for all effects and purposes freely available without requiring subscriptions or special affiliation.

We infer if a publication has been funded by checking if it lists any grants. We classify grants as either ‘old’, i.e. existed before 2019, or ‘new’, i.e. first observed afterwards. To do so, we collect all grant IDs for 11,122,017 papers from 2010 on-wards and record their first appearance. This procedure is an indirect inference of the year the grant has been granted. The basic assumption is that if a grant number has not been listed in any publication since 2010, it is very likely a new grant. Specifically, an old grant is a grant listed since 2019 observed at least once from 2010 to 2018.

Note that this procedure is only approximate and has a few shortcomings. Mistyped grant numbers (e.g. ‘1234-M JPN’ and ‘1234-M-JPN’) could appear as new grants, even though they existed before, or new grants might be classified as old grants if they have a common ID (e.g. ‘Grant 1’). Unfortunately, there is no central repository of grant numbers and the associated metadata; however, there are plans to assign DOI numbers to grants to alleviate this problem (See https://gitlab.com/crossref/open_funder_registry for the project).

Impact factor weighted publication numbers (IFWN)

In our analysis, we consider two measures of scientific output. First, we simply count the number of publications by MeSH term. However, since journals vary considerably in terms of impact factor, we also weigh the number of publications by the impact factor of the venue (e.g., journal) where it was published. Specifically, we use the SCImago journal ranking statistics to weigh a paper by the impact factor of the journal it appears in. We use the ‘citation per document in the past two years’ for 45,230 ISSNs. Note that a journal may and often has more than one ISSN, i.e., one for the printed edition and one for the online edition. SCImago applies the same score for a venue across linked ISSNs.

For the impact factor weighted number (IFWN) of publication per MeSH terms, this means that all publications are replaced by the impact score of the journal they appear in and summed up.

COVID-19-relatedness

To measure how closely related to COVID-19 is a MeSH term, we introduce an index of relatedness to COVID-19. First, we identify the focal COVID-19 terms, which appeared in the literature in 2020 (see Table 1 ). Next, for all other pre-existing MeSH terms, we measure how closely related to COVID-19 they end up being.

Our aim is to show that MeSH terms that existed before and are related have experienced a sudden increase in the number of (impact factor weighted) papers.

We define a MeSH term’s COVID-19 relatedness as the conditional probability that, given its appearance on a paper, also one of the focal COVID-19 terms listed in Table 1 are present. In other words, the relatedness of a MeSH term is given by the probability that a COVID-19 MeSH term appears alongside. Since the focal COVID-19 terms did not exist before 2020, we estimate this measure only using papers published since January 2020. Formally, we define COVID-19-relatedness ( σ ) as in Eq (1) , where C is the set of papers listing a COVID-19 MeSH term and M i is the set of papers listing MeSH term i .

Intuitively we can read this measure as: what is the probability in 2020 that a COVID-19 MeSH term is present given that we chose a paper with MeSH term i ? For example, given that in 2020 we choose a paper dealing with “Ageusia” (i.e., Complete or severe loss of the subjective sense of taste), there is a 96% probability that this paper also lists COVID-19, see Table 1 .

Note that a paper listing a related MeSH term does not imply that that paper is doing COVID-19 research, but it implies that one of the MeSH terms listed is often used in COVID-19 research.

In sum, in our analysis, we use the following variables:

• Papers: Number of papers by MeSH term;
• Impact: Impact factor weighted number of papers by MeSH term;
• PMC: Papers listed in PubMed central by MeSH term, as a measure of Open Access publications;
• Trials: number of publications of type “Clinical Trial” by MeSH term;
• Grants: number of papers with at least one grant by MeSH term;
• Old Grants: number of papers listing a grant that has been observed between 2010 and 2018, by MeSH term;

Difference-in-differences

The difference-in-differences (DiD) method is an econometric technique to imitate an experimental research design from observation data, sometimes referred to as a quasi-experimental setup. In a randomized controlled trial, subjects are randomly assigned either to the treated or the control group. Analogously, in this natural experiment, we assume that medical subject headings (MeSH) have been randomly assigned to be either treated (related) or not treated (unrelated) by the pandemic crisis.

Before the COVID, for a future health crisis, the set of potentially impacted medical knowledge was not predictable since it depended on the specifics of the emergency. For instance, ageusia (loss of taste), a medical concept existing since 1991, became known to be a specific symptom of COVID-19 only after the pandemic.

Specifically, we exploit the COVID-19 as an unpredictable and exogenous shock that has deeply affected the publication priorities for biomedical scientific production, as compared to the situation before the pandemic. In this setting, COVID-19 is the treatment, and the identification of this new human coronavirus is the event. We claim that treated MeSH terms, i.e., MeSH terms related to COVID-19, have experienced a sudden increase in terms of scientific production and attention. In contrast, research on untreated MeSH terms, i.e., MeSH terms not related to COVID-19, has been displaced by COVID-19. Our analysis compares the scientific output of COVID-19 related and unrelated MeSH terms before and after January 2020.

Consider the simple regression model in Eq (2) . We have an outcome Y and dummy variable P identifying the period as before the event P = 0 and P = 1 as after the event. Additionally, we have a dummy variable identifying an observation belonging to the treated group ( G = 1) or control ( G = 0) group.

In our case, some of the terms turn out to be related to COVID-19 in 2020, whereas most of the MeSH terms are not closely related to COVID-19.

Thus β 1 identifies the overall effect on the control group after the event, β 2 the difference across treated and control groups before the event (i.e. the first difference in DiD) and finally the effect on the treated group after the event, net of the first difference, β 3 . This last parameter identifies the treatment effect on the treated group netting out the pre-treatment difference.

For the DiD to have a causal interpretation, it must be noted that pre-event, the trends of the two groups should be parallel, i.e., the common trend assumption (CTA) must be satisfied. We will show that the CTA holds in the results section.

To specify the DiD model, we need to define a period before and after the event and assign a treatment status or level of exposure to each term.

Before and after

The pre-treatment period is defined as January 2019 to December 2019. The post-treatment period is defined as the months from January 2020 to December 2020. We argue that the state of biomedical research was similar in those two years, apart from the effect of the pandemic.

Treatment status and exposure

The treatment is determined by the COVID-19 relatedness index σ i introduced earlier. Specifically, this number indicates the likelihood that COVID-19 will be a listed MeSH term, given that we observe the focal MeSH term i . To show that the effect becomes even stronger the closer related the subject is, and for ease of interpretation, we also discretize the relatedness value into three levels of treatment. Namely, we group MeSH terms with a σ between, 0% to 20%, 20% to 80% and 80% to 100%. The choice of alternative grouping strategies does not significantly affect our results. Results for alternative thresholds of relatedness can be computed using the available source code. We complement the dichotomized analysis by using the treatment intensity (relatedness measure σ ) to show that the result persists.

Panel regression

In this work, we estimate a random effects panel regression where the units of analysis are 28 318 biomedical research fields (i.e. MeSH terms) observed over time before and after the COVID-19 pandemic. The time resolution is at the monthly level, meaning that for each MeSH term, we have 24 observations from January 2019 to December 2020.

The basic panel regression with continuous treatment follows a similar setup as Eq (2) but with MeSH term random effects and monthly fixed effects.

The outcome variable Y it identifies the outcome at time t (i.e., month), for MeSH term i . As before, P t identifies the period with P t = 0 if the month is before January 2020 and P t = 1 if it is on or after this date. In (3) , the treatment level is measure by the relatedness to COVID-19 ( σ i ), where again the γ 1 identifies pre-trend (constant) differences and δ 1 the overall effect.

As mentioned before, to highlight that the effect is not linear but increases with relatedness, we split σ into three groups: from 0% to 20%, 20% to 80% and 80% to 100%. In the three-level treatment specification, the number of treatment levels ( G i ) is 3; hence we have two γ parameters. Note that I (⋅) is the indicator function, which is 1 if the argument is true, and 0 otherwise.

In total, we estimate six coefficients. As before, the δ l coefficient identifies the DiD effect.

Verifying the Common Trend Assumption (CTA)

To show that the pre-event trends are parallel and that the effect on publication activity is only visible from January 2020, we estimate a panel regression with each month modelled as a different event. Specifically, we estimate the following model.

We show that the CTA holds for this model by comparing the pre-event trends of the control group to the treated groups (COVID-19 related MeSH terms). Namely, we show that the pre-event trends of the control group are the same as the pre-event trends of the treated group.

Co-occurrence analysis

To investigate if the pandemic has caused a reconfiguration of research priorities, we look at the MeSH term co-occurrence network. Precisely, we extract the co-occurrence network of all 28,318 MeSH terms as they appear in the 3.3 million papers. We considered the co-occurrence networks of 2018, 2019 and 2020. Each node represents a MeSH term in these networks, and a link between them indicates that they have been observed at least once together. The weight of the edge between the MeSH terms is given by the number of times those terms have been jointly observed in the same publications.

Medical language is hugely complicated, and this simple representation does not capture the intricacies, subtle nuances and, in fact, meaning of the terms. Therefore, we do not claim that we can identify how the actual usage of MeSH terms has changed from this object, but rather that it has. Nevertheless, the co-occurrence graph captures rudimentary relations between concepts. We argue that absent a shock to the system, their basic usage patterns, change in importance (within the network) would essentially be the same from year to year. However, if we find that the importance of terms changes more than expected in 2020, it stands to reason that there have been some significant changes.

To show that that MeSH usage has been affected, we compute for each term in the years 2018, 2019 and 2020 their PageRank centrality [ 17 ]. The PageRank centrality tells us how likely a random walker traversing a network would be found at a given node if she follows the weights of the empirical edges (i.e., co-usage probability). Specifically, for the case of the MeSH co-occurrence network, this number represents how often an annotator at the National Library of Medicine would assign that MeSH term following the observed general usage patterns. It is a simplistic measure to capture the complexities of biomedical research. Nevertheless, it captures far-reaching interdependence across MeSH terms as the measure uses the whole network to determine the centrality of every MeSH term. A sudden change in the rankings and thus the position of MeSH terms in this network suggests that a given research subject has risen as it is used more often with other important MeSH terms (or vice versa).

To show that COVID-19-related research has profoundly impacted the way MeSH terms are used, we compute for each MeSH term the change in its PageRank centrality ( p it ).

We then compare the growth for each MeSH i term in g i (2019), i.e. before the the COVID-19 pandemic, with the growth after the event ( g i (2020)).

Publication growth

To estimate growth in scientific output, we compute the year over year growth in the number of the impact weighted number of publications per MeSH. Specifically, we measure the year by year growth as defined below, where m is the impact weighted number of publications at time t .

Changes in output and COVID-19 relatedness

Before we show the regression results, we provide descriptive evidence that publications from 2019 to 2020 have drastically increased. By showing that this growth correlates strongly with a MeSH term’s COVID-19 relatedness ( σ ), we demonstrate that (1) σ captures an essential aspect of the growth dynamics and (2) highlight the meteoric rise of highly related terms.

We look at the year over year growth in the number of the impact weighted number of publications per MeSH term from 2018 to 2019 and 2019 to 2020 as defined in the methods section.

Fig 1 shows the yearly growth of the impact weighted number of publications per MeSH term. By comparing the growth of the number of publications from the years 2018, 2019 and 2020, we find that the impact factor weighted number of publications has increased by up to a factor of 100 compared to the previous year for Betacoronavirus, one of the most closely related to COVID-19 MeSH term.

Each dot represents, a MeSH term. The y axis (growth) is in symmetric log scale. The x axis shows the COVID-19 relatedness, σ . Note that the position of the dots on the x-axis is the same in the two plots. Below: MeSH term importance gain (PageRank) and their COVID-19 relatedness.

Fig 1 , first row, reveals how strongly correlated the growth in the IFWN of publication is to the term’s COVID-19 relatedness. For instance, we see that the term ‘Betacoronavirus’ skyrocketed from 2019 to 2020, which is expected given that SARS-CoV-2 is a species of the genus. Conversely, the term ‘Alphacoronavirus’ has not experienced any growth given that it is twin a genus of the Coronaviridae family, but SARS-CoV-2 is not one of its species. Note also the fast growth in the number of publications dealing with ‘Quarantine’. Moreover, MeSH terms that grew significantly from 2018 to 2019 and were not closely related to COVID-19, like ‘Vaping’, slowed down in 2020. From the graph, the picture emerges that publication growth is correlated with COVID-19 relatedness σ and that the growth for less related terms slowed down.

To show that the usage pattern of MeSH terms has changed following the pandemic, we compute the PageRank centrality using graph-tool [ 18 ] as discussed in the Methods section.

Fig 1 , second row, shows the change in the PageRank centrality of the MeSH terms after the pandemic (2019 to 2020, right plot) and before (2018 to 2019, left plot). If there were no change in the general usage pattern, we would expect the variance in PageRank changes to be narrow across the two periods, see (left plot). However, PageRank scores changed significantly more from 2019 to 2020 than from 2018 to 2019, suggesting that there has been a reconfiguration of the network.

To further support this argument, we carry out a DiD regression analysis.

Common trends assumption

As discussed in the Methods section, we need to show that the CTA assumption holds for the DiD to be defined appropriately. We do this by estimating for each month the number of publications and comparing it across treatment groups. This exercise also serves the purpose of a placebo test. By assuming that each month could have potentially been the event’s timing (i.e., the outbreak), we show that January 2020 is the most likely timing of the event. The regression table, as noted earlier, contains over 70 estimated coefficients, hence for ease of reading, we will only show the predicted outcome per month by group (see Fig 2 ). The full regression table with all coefficients is available in the S1 Table .

The y axis is in log scale. The dashed vertical line identifies January 2020. The dashed horizontal line shows the publications in January 2019 for the 0–20% group before the event. This line highlights that the drop happens after the event. The bands around the lines indicate the 95% confidence interval of the predicted values. The results are the output of the Stata margins command.

Fig 2 shows the predicted number per outcome variable obtained from the panel regression model. These predictions correspond to the predicted value per relatedness group using the regression parameters estimated via the linear panel regression. The bands around the curves are the 95% confidence intervals.

All outcome measures depict a similar trend per month. Before the event (i.e., January 2020), there is a common trend across all groups. In contrast, after the event, we observe a sudden rise for the outcomes of the COVID-19 related treated groups (green and red lines) and a decline in the outcomes for the unrelated group (blue line). Therefore, we can conclude that the CTA assumption holds.

Regression results

Table 3 shows the DiD regression results (see Eq (3) ) for the selected outcome measures: number of publications (Papers), impact factor weighted number of publications (Impact), open access (OA) publications, clinical trial related publications, and publications with existing grants.

t statistics in parentheses, Std. Err. adjusted by MeSH-id. All outcome variables are in natural log.

* p < 0.05,

** p < 0.01,

*** p < 0.001

Table 3 shows results for the discrete treatment level version of the DiD model (see Eq (4) ).

Note that the outcome variable is in natural log scale; hence to get the effect of the independent variable, we need to exponentiate the coefficient. For values close to 0, the effect is well approximated by the percentage change of that magnitude.

In both specifications we see that the least related group, drops in the number of publications between 10% and 13%, respectively (first row of Tables ​ Tables3 3 and ​ and4, 4 , exp(−0.102) ≈ 0.87). In line with our expectations, the increase in the number of papers published by MeSH term is positively affected by the relatedness to COVID-19. In the discrete model (row 2), we note that the number of documents with MeSH terms with a COVID-19 relatedness between 20 and 80% grows by 18% and highly related terms by a factor of approximately 6.6 (exp(1.88)). The same general pattern can be observed for the impact weighted publication number, i.e., Model (2). Note, however, that the drop in the impact factor weighted output is more significant, reaching -19% for COVID-19 unrelated publications, and related publications growing by a factor of 8.7. This difference suggests that there might be a bias to publish papers on COVID-19 related subjects in high impact factor journals.

t statistics in parentheses, Std. Err. adjusted by MeSH-id. All outcome variables are in natural log. σ is the MeSH term relatedness to COVID-19.

By looking at the number of open access publications (PMC), we note that the least related group has not been affected negatively by the pandemic. However, the number of COVID-19 related publications has drastically increased for the most COVID-19 related group by a factor of 6.2. Note that the substantial increase in the number of papers available through open access is in large part due to journal and editorial policies to make preferentially COVID research immediately available to the public.

Regarding the number of clinical trial publications, we note that the least related group has been affected negatively, with the number of publications on clinical trials dropping by a staggering 24%. At the same time, publications on clinical trials for COVID-19-related MeSH have increased by a factor of 2.1. Note, however, that the effect on clinical trials is not significant in the continuous regression. The discrepancy across Tables ​ Tables3 3 and ​ and4 4 highlights that, especially for trials, the effect is not linear, where only the publications on clinical trials closely related to COVID-19 experiencing a boost.

It has been reported [ 19 ] that while the number of clinical trials registered to treat or prevent COVID-19 has surged with 179 new registrations in the second week of April 2020 alone. Only a few of these have led to publishable results in the 12 months since [ 20 ]. On the other hand, we find that clinical trial publications, considering related MeSH (but not COVID-19 directly), have had significant growth from the beginning of the pandemic. These results are not contradictory. Indeed counting the number of clinical trial publications listing the exact COVID-19 MeSH term (D000086382), we find 212 publications. While this might seem like a small number, consider that in 2020 only 8,485 publications were classified as clinical trials; thus, targeted trials still made up 2.5% of all clinical trials in 2020 . So while one might doubt the effectiveness of these research efforts, it is still the case that by sheer number, they represent a significant proportion of all publications on clinical trials in 2020. Moreover, COVID-19 specific Clinical trial publications in 2020, being a delayed signal of the actual trials, are a lower bound estimate on the true number of such clinical trials being conducted. This is because COVID-19 studies could only have commenced in 2020, whereas other studies had a head start. Thus our reported estimates are conservative, meaning that the true effect on actual clinical trials is likely larger, not smaller.

Research funding, as proxied by the number of publications with grants, follows a similar pattern, but notably, COVID-19-related MeSH terms list the same proportion of grants established before 2019 as other unrelated MeSH terms, suggesting that grants which were not designated for COVID-19 research have been used to support COVID-19 related research. Overall, the number of publications listing a grant has dropped. Note that this should be because the number of publications overall in the unrelated group has dropped. However, we note that the drop in publications is 10% while the decline in publications with at least one grant is 15%. This difference suggests that publications listing grants, which should have more funding, are disproportionately COVID-19 related papers. To further investigate this aspect, we look at whether the grant was old (pre-2019) or appeared for the first time in or after 2019. It stands to reason that an old grant (pre-2019) would not have been granted for a project dealing with the pandemic. Hence we would expect that COVID-19 related MeSH terms to have a lower proportion of old grants than the unrelated group. In models (6) in Table 4 we show that the number of old grants for the unrelated group drops by 13%. At the same time, the number of papers listing old grants (i.e., pre-2019) among the most related group increased by a factor of 3.1. Overall, these results suggest that COVID-19 related research has been funded largely by pre-existing grants, even though a specific mandate tied to the grants for this use is unlikely.

The scientific community has swiftly reallocated research efforts to cope with the COVID-19 pandemic, mobilizing knowledge across disciplines to find innovative solutions in record time. We document this both in terms of changing trends in the biomedical scientific output and the usage of MeSH terms by the scientific community. The flip side of this sudden and energetic prioritization of effort to fight COVID-19 has been a sudden contraction of scientific production in other relevant research areas. All in all, we find strong support to the hypotheses that the COVID-19 crisis has induced a sudden increase of research output in COVID-19 related areas of biomedical research. Conversely, research in areas not related to COVID-19 has experienced a significant drop in overall publishing rates and funding.

Our paper contributes to the literature on the impact of COVID-19 on scientific research: we corroborate previous findings about the surge of COVID-19 related publications [ 1 – 3 ], partially displacing research in COVID-19 unrelated fields of research [ 4 , 14 ], particularly research related to clinical trials [ 5 – 7 ]. The drop in trial research might have severe consequences for patients affected by life-threatening diseases since it will delay access to new and better treatments. We also confirm the impact of COVID-19 on open access publication output [ 1 ]; also, this is milder than traditional outlets. On top of this, we provide more robust evidence on the impact weighted effect of COVID-19 and grant financed research, highlighting the strong displacement effect of COVID-19 on the allocation of financial resources [ 15 ]. We document a substantial change in the usage patterns of MeSH terms, suggesting that there has been a reconfiguration in the way research terms are being combined. MeSH terms highly related to COVID-19 were peripheral in the MeSH usage networks before the pandemic but have become central since 2020. We conclude that the usage patterns have changed, with COVID-19 related MeSH terms occupying a much more prominent role in 2020 than they did in the previous years.

We also contribute to the literature by estimating the effect of COVID-19 on biomedical research in a natural experiment framework, isolating the specific effects of the COVID-19 pandemic on the biomedical scientific landscape. This is crucial to identify areas of public intervention to sustain areas of biomedical research which have been neglected during the COVID-19 crisis. Moreover, the exploratory analysis on the changes in usage patterns of MeSH terms, points to an increase in the importance of covid-related topics in the broader biomedical research landscape.

Our results provide compelling evidence that research related to COVID-19 has indeed displaced scientific production in other biomedical fields of research not related to COVID-19, with a significant drop in (impact weighted) scientific output related to non-COVID-19 and a marked reduction of financial support for publications not related to COVID-19 [ 4 , 5 , 16 ]. The displacement effect is persistent to the end of 2020. As vaccination progresses, we highlight the urgent need for science policy to re-balance support for research activity that was put on pause because of the COVID-19 pandemic.

We find that COVID-19 dramatically impacted clinical research. Reactivation of clinical trials activities that have been postponed or suspended for reasons related to COVID-19 is a priority that should be considered in the national vaccination plans. Moreover, since grants have been diverted and financial incentives have been targeted to sustain COVID-19 research leading to an excessive entry in COVID-19-related clinical trials and the ‘covidisation’ of research, there is a need to reorient incentives to basic research and otherwise neglected or temporally abandoned areas of biomedical research. Without dedicated support in the recovery plans for neglected research of the COVID-19 era, there is a risk that more medical needs will be unmet in the future, possibly exacerbating the shortage of scientific research for orphan and neglected diseases, which do not belong to COVID-19-related research areas.

Limitations

Our empirical approach has some limits. First, we proxy MeSH term usage via search terms using the PubMed EUtilities API. This means that the accuracy of the MeSH term we assign to a given paper is not fully validated. More time is needed for the completion of manually annotated MeSH terms. Second, the timing of publication is not the moment the research has been carried out. There is a lead time between inception, analysis, write-up, review, revision, and final publication. This delay varies across disciplines. Nevertheless, given that the surge in publications happens around the alleged event date, January 2020, we are confident that the publication date is a reasonable yet imperfect estimate of the timing of the research. Third, several journals have publicly declared to fast-track COVID-19 research. This discrepancy in the speed of publication of COVID-19 related research and other research could affect our results. Specifically, a surge or displacement could be overestimated due to a lag in the publication of COVID-19 unrelated research. We alleviate this bias by estimating the effect considering a considerable time after the event (January 2020 to December 2020). Forth, on the one hand, clinical Trials may lead to multiple publications. Therefore we might overestimate the impact of COVID-19 on the number of clinical trials. On the other hand, COVID-19 publications on clinical trials lag behind, so the number of papers related COVID-19 trials is likely underestimated. Therefore, we note that the focus of this paper is scientific publications on clinical trials rather than on actual clinical trials. Fifth, regarding grants, unfortunately, there is no unique centralized repository mapping grant numbers to years, so we have to proxy old grants with grants that appeared in publications from 2010 to 2018. Besides, grant numbers are free-form entries, meaning that PubMed has no validation step to disambiguate or verify that the grant number has been entered correctly. This has the effect of classifying a grant as new even though it has appeared under a different name. We mitigate this problem by using a long period to collect grant numbers and catch many spellings of the same grant, thereby reducing the likelihood of miss-identifying a grant as new when it existed before. Still, unless unique identifiers are widely used, there is no way to verify this.

So far, there is no conclusive evidence on whether entry into COVID-19 has been excessive. However, there is a growing consensus that COVID-19 has displaced, at least temporally, scientific research in COVID-19 unrelated biomedical research areas. Even though it is certainly expected that more attention will be devoted to the emergency during a pandemic, the displacement of biomedical research in other fields is concerning. Future research is needed to investigate the long-run structural consequences of the COVID-19 crisis on biomedical research.

Supporting information

Full regression table with all controls and interactions.

Funding Statement

The author(s) received no specific funding for this work.

Data Availability