The following explanation of the methodology used to create PHAC’s COVID-19 daily literature scans was kindly provided by the project leader, Dr. Lisa Waddell, on 14 September 2020.
Date: Mon, 14 Sep 2020 19:00:55 +0000
From: "Waddell, Lisa (PHAC/ASPC)" <lisa.waddell@canada.ca>
To: David Earn <earn@math.mcmaster.ca>
The COVID-19 literature daily scan conducted by my team is done in a systematic way as follows.
We initiated this project officially February 4th with the objective of monitoring the COVID-19 literature and as a base for summarizing “what we know” on specific questions that come from a number of offices within our agency.
Monday to Friday a daily scan for new COVID-19 literature is performed by the Emerging Sciences Knowledge Synthesis team at PHAC.
Briefly the search is conducted in the following databases: PubMed, Scopus, bioRxiv, medRxiv, SSRN lancet prepublications, Research Square and arXiv.
In PubMed (for example), the following search uses a combination of keywords and MESH terms: “COVID-19[All Fields] OR SARS-CoV-2[All Fields] OR SARS-Coronavirus-2[All Fields] OR nCov[All Fields] OR “novel CoV”[All Fields] OR (novel[All Fields] AND (“coronavirus”[MeSH Terms] OR “coronavirus”[All Fields])) AND (“2020/02/21”[PDAT] : “2020/12/31”[PDAT])”.
Similar search strategies are employed in Scopus (using the “added in the last 7 days” filter) and BioRxiv and MedRxiv with posted in the last 2 or more days filter. Searches in the other databases are similar and the results are input by hand from Arxiv, research square and SSRN.
COVID-19 literature centers set up by some publishing groups are hand searched for articles omitted by the search (or not indexed yet): Lancet, New England Journal of Medicine, China CDC, Elsevier, British Medical Journal, Nature and Wiley. (Lancet is the only one with a significant delay to being indexed in Pubmed)
When conducting topic reviews, or via cross-checking articles of interest, we may occasionally identify an article that has been omitted by the search and we add it. Initially, when the literature base was small I would also cross-check with WHO and lit COVID when it started. However the WHO literature database is a mess and contains a lot of “reports” that should not be there. And we have what was relevant in lit-COVID, but lit-COVID cast a wider net (more broadly for coronaviruses) So since the end of April, I have not been cross-checking these.
We manage the search results in an Endnote database. And each day two files are produced. 1) new citations captured by the search and 2) Article that were preprints and are now published. These results are put into the Refworks database that people can link to if interested.
The results are also imported into DistillerSR, a systematic review management program. We have an AI classifier set up to classify primary research and synthesis research (e.g. systematic reviews) Our review team is provided primary research articles to summarize and classify into prescribed foci (see below for more info). Reviewer data is double checked by a senior reviewer before the daily report is run. The project lead classifies the non-primary articles. All the data for the day are pulled together into a line list of new articles, prefaced with highlights identified by the review team that day. This is what goes out in the daily emails. The daily scan data is also saved in an excel file to keep a searchable running list of daily findings - that is what is available on the google drive.
Database maintenance: duplicates and preprints that get published are identified on an ongoing basis and refworks, distiller and the excel spreadsheet are updated so that they have only one copy (the most recent one) of each article.
Reviewer tasks: Each reviewer uses DistillerSR to access the citations assigned to them. One at a time, they determine if the link to the paper works (if not they provide a working link) and whether the paper should be excluded or not (selected appropriate exclusion criteria). Included papers are summarised in a short, compact paragraph or couple lines. The researchers then tags the study with at least one focus area outlined below: These foci have definitions - they are not 100% ideal in retrospect, but they were set up very early in the pandemic
That is the process in a nutshell, don’t hesitate to ask for additional clarification.
Cheers,
Lisa
Lisa Waddell BSc. MSc. PhD.
Knowledge Synthesis Team Lead / Chef d'équipe de synthèse des connaissances
National Microbiology Laboratory / Laboratoire national de microbiologie
Public Health Agency of Canada (PHAC) / Agence de la santé publique du Canada (ASPC),
Government of Canada / Gouvernement du Canada
Tel. / Tél.: 226-979-7174
Lisa.Waddell@canada.ca