Brainard. (2020, May 13). Scientists are drowning in COVID-19 papers. Can new tools keep them afloat? Science. https://doi.org/10.1126/science.abc7839

University of Michigan - Ann Arbor

Google

- As fo May 13, 2020, more than 4,000 COVID-19 related papers were came out during the previous week.
- A collaborate efforts supported by large tech firms and the White House, by collecting free digital papers by building platforms to search relevant scientific papers so that the researchers can quickly find the information.
- It is estimated that the COVID-19 related literature published since January 2020 reached more than 23,000 papers, and the literature is double in every 20 days.
- The researchers just have no time to read the whole articles and distill "value-added" (contributions), "bottom-line" (conclusions), and "limitations".
- Despite the efforts and 20% of new papers are accessible only to paying users, which can grow up to 50%.
- "Not particularly user-friendly."
- The users (researchers) skepticism about the tools' abilities to accurately tell them "what they really want to know.".  For example: How can the algorithms judge "the quality of work"?  There are researches already out there that are over-sell.  It has data, which does not support their conclusions.
- Two strategies used by the teams who want to help the researchers: 
					(1)  Create easily accessible paper collections, including a few carefully curated collections designed to highlight strong papers
			(2) building automated search tools that use artificial intelligence (AI) technologies to cut through the noise
            - While Dr. Colavizza characterizes CORD019 as "amazing work", he also identified some shortcomings: 
	- More than 60% of the papers in CORD-19 don’t mention the search terms used by the collection’s creators—such as “coronavirus” and “SARS-CoV", an indication that such papers are only tangibly related to COVID-19.
- "Only about 40,000 papers in the collection had full test", which is insufficient for more comprehensive data mining.
- This relates to about 20% of subscribers-only journals, which cannot be included in the collection of CORD-19 with a forecasted trend of 50% increase.
- A curated list of COVID-19 related papers with summaries by Johns Hospkins University.
- A target audience includes health care workers, policymakers, and researchers.  It is curated list that aims to filter out not well conducted studies.
- The major criticism that preprint servers receive with that the papers that appear in preprint servers are not peer reviewed, compromising the quality of the articles in the servers.  However, approximately 80% out of more than 11,000 COVID-19 manuscripts had appeared in journals.
- Potential cause of the 80% ultimately appearing to the journals:
	- Publishers' efforts to accelerate publicaition schedule. 
    - 14 medical journals that published most COVID-19 papers reduced the average time taken from submission to publication to about 60 days, although concerns about the quality of research by prioritizing the speed still remains.  
- A caution: The number of citations and the status of "retraction" does not necessarily mean,  at least for now for the COVID-19 papers, that the quality of such papers are good/poor.  
 - But what has been resulting in "impact" for a given paper is: social media mentions.  
 - Some researchers actively takes their papers to social media to share while others use the platform to actively challenge studies that they see flaws, an unofficial form of quality control.
 - SciSight: A highly visuzlized more tool that focuses on usability.
 - The researchers' way of doing science has to be altered for them to start even trying them.  It is harder when the time is so pressed as it has been right now.
 		(1) Word of mouth by trusted colleagued including Tweets
        (2) Reading scientific societies bulletin boards and leading journals.

Learn Before

Related