Coronavirus misinformation: quantifying sources and themes in the COVID-19 ‘infodemic’ (Preprint)

BACKGROUND The COVID-19 pandemic has unfolded alongside what the Director-General of the World Health Organization has termed an “infodemic” of misinformation . In their coverage of the pandemic, traditional media outlets have reported and sometimes amplified the voices of various actors across the political spectrum who have advocated unproven cures, denied what is known scientifically about the nature and origins of the novel SARS-CoV-2 coronavirus, and proposed conspiracy theories which purport to explain causation and often allege nefarious intent. These competing narratives and explanations have risen and fallen rapidly, behaving almost as viral phenomena themselves. Misinformation about COVID-19 is a serious threat to global public health. If people are misled by unsubstantiated claims about the nature and treatment of the disease, they are less likely to observe official health advice and may thus contribute to the spread of the pandemic and pose a danger to themselves and others. Health protection strategies such as hygiene, sanitation, social distancing, mask wearing, lockdowns, and other measures will be less effective if distrust of public health authorities becomes sufficiently widespread to substantially affect public behavior. Specifically, misinformation about treatments for COVID disease can prompt people to attempt cures that might harm them, while fears and distrust about a possible vaccine could undermine the uptake of any vaccination campaign aiming to immunize the public at a later date. Both misinformation and disinformation center on the dissemination of false information, with the difference being that the former is shared without malice while the latter is spread with the intent to deceive. Though we use the term misinformation in this study, it is clear that some of the nine main topics that emerged do include elements of disinformation in that they appear to have been shared intentionally, primarily to advance political agendas, and others are a combination of misinformation and disinformation. OBJECTIVE It is commonly assumed that misinformation is largely a phenomenon of social media, provoking calls for stricter regulation of the content on platforms such as Facebook and Twitter. However, misinformation also appears in traditional media. Here it typically takes two forms: amplification of false claims through widespread coverage of prominent persons whose views and comments are considered newsworthy; and to a lesser degree, active fact-checking and debunking of false claims and misinformation. In this paper we aim to quantify the extent of the COVID infodemic within traditional media and examine it as a multi-dimensional informational phenomenon. While previous authors have investigated specific types of social media misinformation , including the role of “bots” in its dissemination , to our knowledge our analysis is the first comprehensive survey of the traditional and online media landscape regarding COVID-19 misinformation, encompassing millions of articles published globally within the five-month span that followed the outbreak of the pandemic in January 2020. Ours is not the first media assessment: the Reuters Institute/Oxford Martin School published a factsheet in April 2020 looking at “Types, Sources and Claims of COVID-19 Misinformation,” but this considered a sample of only 225 misinformation examples in the media . By using a quantitative approach examining a comprehensive English-language global media database of tens of millions of articles, we aim to present empirical insights into the nature and impact of the entire infodemic that may better inform response measures taken by public health authorities, media institutions, governmental organizations, academia, and others. METHODS We performed a comprehensive analysis of media coverage of the COVID-19 pandemic using Cision’s Next Generation Communications Cloud platform. This commercial platform aggregates online news (including licensed print and traditional media news content via LexisNexis), blogs, podcasts, TV, and radio, sourced via webcrawlers and third-party content providers. In total, this database encompasses a network of 7 million-plus global sources of print, broadcast, and online news. Cision’s comprehensive coverage and search capabilities make it a potentially powerful tool for the kind of content analysis we perform here. This Next Generation Communications Cloud database aggregates global coverage, with the largest volume of English-language results coming in descending order from the United States, United Kingdom, India, Ireland, Australia, and New Zealand, with African and other Asian nations also represented in the sample. This database was queried using an English-language search string for misinformation topics in the context of COVID-19. The search string included variations on common thematic keywords (“COVID-19”, “coronavirus”, “2019-nCoV”, etc.) and used Boolean operators such as AND, OR, NOT, and proximity terms to sift for relevant content. (For a full reproduction of Boolean operators see Supplementary Information 1.) Media coverage was examined from a sample of articles published between January 1 and May 26, 2020. Misinformation terms were identified by an iterative cycle of reviewing coverage of COVID-19-related misinformation, creating an initial search string, further reviewing coverage, and adding additional terms to improve inclusiveness. Sites known to produce non-news content, such as wordpress.com and livejournal.com, were excluded. Keyword-based, pre-set content filters for press releases, job postings, earnings and stock news, and other irrelevant categories were applied in order to exclude them from results. Specific misinformation topics were identified within media coverage via a similar iterative approach of reviewing sample coverage, search query adjustment, and further review of coverage until it was possible to determine that the leading misinformation narratives over the time period were represented since new topic searches failed to generate a substantial volume of results. Misinformation topics were then searched within the overarching misinformation search, operating as a layering set of context terms. When topic research identified new misinformation keywords, they were added to the master search to further improve comprehensiveness. There is obviously a distinction to be made between misinformation per se (defined as information that is likely to mislead the audience) and information that discusses misinformation topics or the phenomenon of the infodemic with the explicit objective of debunking or correcting factual inaccuracies. We explicitly isolate this fact-checking coverage within the broader misinformation sample by identifying common terms used to identify misinformation as false, such as “fact-check” and “false claim”, as well as the use of terms like “misinformation" and "conspiracy theory" which inherently imply that the narratives they reference are untrue. Coverage falling into the misinformation search was also compared to coverage of COVID-19 generally, which was defined as the misinformation COVID-19 search excluding misinformation context terms. We quantify the extent of misinformation by volume, meaning the number of articles about a topic. To avoid excluding coverage that mentions more than one topic, topics within the report are not mutually exclusive. A notable amount of overlap between certain topics was observed, thus “frequency” is used to ensure accurate representation of each topic. In this report, “frequency” is defined as the volume of a specific topic divided by the total volume for the misinformation conversation. RESULTS From January 1 to May 26, 2020, English-language traditional media outlets published over 1.1 million individual articles (total 1,116,952) mentioning COVID-19 misinformation. This represented just under 3% of the overall COVID-19 conversation (total 38,713,161 articles) during the same timeframe. We identified five different sub-sections within the overall COVID misinformation conversation, summarized in Table 1. (See Supplementary Info for specific search strings that yielded these results.) Specifically: • Misinformation/conspiracies sub-topics: We identified 11 key sub-topics within this conversation, which are shown in Table 2 and profiled in more detail in the discussion section below. • Trump mentions: This topic comprises all mentions of US President Donald Trump within the total misinformation conversation, irrespective of whether other subjects were also referenced in the same news article. This topic is included as a way to quantify the prominence of Trump within the overall COVID “infodemic” without risking double-counting by combining Trump mentions from a number of topics that can be expected to overlap. Any and all mentions of Trump will appear in this category irrespective of whether they also appear elsewhere. • Infodemic coverage: This topic includes articles that mentioned the general term “infodemic” (or related keywords such as “misinformation” or “hoax” combined with mentions of COVID-19) without mentioning a specific additional topic such as 5G or Dr Fauci. • Fact-checking: This topic includes articles that explicitly mentioned conspiracies, misinformation, or factual inaccuracies in a way that aimed to correct misinformation with the audience. Examples of this coverage include articles from established fact-checking sources, such as The Washington Post's Fact Checker, and coverage that mentioned the fact-checking of COVID-19 misinformation. • Trump-only mentions: This topic represents the volume and frequency of articles that mentioned President Trump in the context of misinformation but did not mention a specific other topic at the same time. Examples were articles alleging in general terms that Trump has spread misinformation about COVID-19 or discus