Entman (1993) defines framing as “to select some aspects of a perceived reality and make them more salient in a communicating text” (p.52), recognizing the selective aspect of framing that helps people to understand the complex world. Certain aspects are focused upon while the rest are disregarded. In so doing, focusing on different aspects of a similarly complicated world leads to differently simplified views of the same world. Then, what would be a selective aspect that journalists focus on when covering the COVID-19 pandemic?

There are multiple studies that identified particular framings in COVID-19 news coverage. In the case of the New York Times coverage of COVID-19, for example, the seven types of framing were examined (i.e., attribution of responsibility, action, risk magnitude, science/medical, human interest, conflict, and economic consequences), whereby attribution of responsibility and action were the salient features (Zhang 2021). The former focuses on holding government or others accountable for an issue’s cause (blame) or solution (credit), while the latter pay attention to actions or attempts to act against the disease (e.g., prevention and potential solutions).

In another study, COVID-19 framing by Italian news media was examined by combining semantic network analysis and topic modeling (Ophir et al. 2021). As a result, a total of 35 topics were extracted, from which three frames were ultimately clustered. The clustered frames were scientific frame paying attention to symptoms and health effects, containment frame focusing on attempts to reduce health risks, and social frame highlighting political and social impact. The empirical studies mentioned above have shown that seemingly different framings of the COVID-19 outbreak have been reported in multiple regions, including the US and Italy.

The variety of framings for the COVID-19 crisis mentioned above can be considered as exemplifying the “fractured paradigm” of framing studies that do not provide reliable knowledge on news framing of an issue (Entman 1993). However, it is also notable that the framings of COVID-19 identified in the US and Italy have something in common. Specifically, attribution of responsibility that focuses on the causes of a problem can be considered a problem diagnosis. Similarly, scientific frames focusing on symptoms and health effects, and social frames highlighting the political and social impact of the pandemic can also be used to diagnose a problem as they are commonly oriented toward identifying the nature and consequences of a health problem. On the other hand, action and containment both seek solutions to health crises. Methods and attempts for the prevention and potential solutions against the disease have been selectively emphasized by news media (Zhang 2021; Ophir et al. 2021). This type of framing is oriented toward what should be done to resolve a problem, unlike earlier framings (i.e., attribution of responsibility, scientific, and social) seeking an understanding of what has happened and what a problem is.

Notably, several studies have examined the very problem-oriented framings. In particular, diagnosis and prognosis framings have been examined as collective action framings in the sociological studies of social movements (Wilson 1973; Benford 1993). “Diagnosis” focuses on identifying a problem and attributing blame or causality, while “prognosis” cares more about an alternative version of reality and what to do in order to achieve the alternative reality (Benford 1993). People engaged in a social movement are expected to pay attention to identifying a problem and attributing responsibilities for the problem on the one hand, while addressing what-to-do aspects in order to achieve an alternative version of reality on the other hand (Wilson 1973; Benford 1993; Benford and Snow 2000; Eilders and Luter 2000; Snow et al. 2007; van der Meer et al. 2014).

An urgent goal for the whole world in the unprecedented pandemic crisis of COVID19 would be controlling the viral spread. This is similar to the social movements for nuclear disarmament (Benford 1993), a just, sustainable and healthy food system (Gaechter and Porter 2018), or socially responsible investment (SRI, Feront and Bertels 2019), in that they all pursued a common goal for fixing a problem. Therefore, it is reasonable to anticipate a societal and collective motivation for coping with a prevalent risk of a society in general, not to mention the pandemic disease crisis. In this context, the problem-oriented framings, diagnosis vs. prognosis, can be expected in the news coverage of COVID-19 as a prevalent risk in a society.

Indeed, news coverage of H1N1, another infectious disease representing a global health risk (Ju 2011), has been examined in terms of diagnosis vs. prognosis. Opinions and editorial pages on climate change as a global environmental risk have been subject to a similar type of framing analysis as well (Hoffman 2011). When it comes to H1N1, Korean newspapers predominantly utilized diagnosis framing to cover the infectious disease that caused 270 mortalities throughout the nation (Ju 2011). However, climate change coverage by US news media varied according to whether news stories conveyed the climate change position of being convinced or skeptical. Here, the stories of convinced were predominantly framed by prognosis while those of skeptical were framed by diagnosis (Hoffman 2011).

In an interventional approach, it would be worthwhile to examine whether diagnosis or prognosis is more salient in news coverage of COVID-19 since the variances of framing saliency can guide public responses to the health crisis differently. Recognizing the potential significance of framing in coping with the health crisis, we examined which framing was salient in news coverage of COVID-19 by establishing the following research questions:

There are so many studies of media framing that it may be challenging to overview all the studies conducted. However, it is also notable that the framing studies can be categorized into two groups. On the one hand, framing research has been conducted to identify frames per se (Antilla 2005; Holton et al. 2014; Muhamad and Yang 2017) or to examine the frame-building process (Scheufele and Tewksbury 2007). On the other hand, the effects of media framing have been examined (Arendt et al. 2018; Gesser-Edelsburg et al. 2015; Ophir et al. 2021) by addressing the dimension of frame-setting in the public mind or society in general (Moy et al. 2016).

However, when it comes to news framing for a health crisis, we posit that a deeper understanding of risk framing can be achieved by further examining message features associated with a particular framing even though they may not constitute a frame itself. Investigating message features differs from identifying a frame, which is required to capture framing devices in the text through which a particular framing is constructed (Pan and Kosicki 1993). Framing devices that have been examined include message structures such as synthetic, script, thematic, and rhetorical structures (Pan and Kosicki 1993) or journalistic elements of reporting (i.e., topics, news sources, and quotations; Jy 2006). Metaphorically, framing devices can be regarded as vessels that contain the content through which framing is conducted. A certain type of information can be placed in a headline, lead, or quotation or constitute a certain topic, rhetoric, script to construct a frame. Notably, the public response to a health crisis is influenced not only by risk messages contained in framing devices, but also by message features that may not directly constitute a risk frame. If both can influence public risk responses, then examining what would be message features accompanying a particular framing of a health crisis would also be worthwhile. This type of investigation can open a new research area that contributes to a deeper and more comprehensive understanding of media framing as well as journalism behavior in covering a health crisis.

The first type of message attribute that are likely to accompany a diagnosis/prognosis framing of the pandemic crisis is affective tones. In Fiske and Pavelchak’s (1986) theoretical perspective on social cognition, affect as a label or tag is likely to be attached to most of cognitive constructs in individuals’ mind. A frame on a risk in this respect is more likely to be attached by certain affect when considering that a risk involves danger and threat. In a similar vein, there are multiple types of theoretical perspectives on affect-driven information processing, such as “affect heuristic” (Slovic et al. 2004; Slovic et al. 2007), and risk-as-feeling hypothesis (Loewenstein et al. 2001). These perspectives recognize the powerful influence of positive or negative affect on information processing and relevant behaviors, including responses to risk.

The diagnosis/prognosis framing of COVID-19 news coverage is examined in this study based on the assumption that framing can play a significant role in guiding public response to the pandemic. If affective tags are likely to be associated with mental constructs including risk frames, it would surely be worthwhile to examine whether an association exists between affective tones and the diagnosis/prognosis framing. In particular, an additional question can be explored regarding whether positive or negative affect is comparatively prevalent in COVID-19 news stories with diagnosis framing compared to stories with prognosis framing or vice versa. Although affective tone is not directly utilized as a type of framing device for diagnosis or prognosis framing, identifying a pattern of association between framing and affective tones as message attribute can lead to a deeper and more comprehensive understanding of framing dynamics and journalism behaviors in covering health crises. Based on this rationale, we examine the following additional research question:

The other message attribute examined in terms of its association with the news framing of a health crisis is efficacy information. In risk communication studies, efficacy information is explained to help people know that something effective exists in reducing risks, and its role in guiding risk response is well known based on empirical evidence (Evensen and Clarke 2012; Song et al. 2023; Lu et al. 2023). Although news coverage of the pandemic does not hold nominal responsibility for assisting the public in responding wisely to a health crisis, it can and should be a criterion for evaluating the quality of health journalism. This means that COVID-19 news coverage with diagnosis or prognosis framing can be examined in light of whether efficacy information is associated with a particular news framing. Therefore, we examine the following additional research question:

In order to examine whether and how the problem-oriented framings are observed globally in news coverage of COVID-19, the multiplicity of the data collection was sought. In particular, under the limit of the feasibility of conducting a content analysis for news coverage worldwide, we tried to cover multiple continents selecting North America, Europe, and Australia. Also, we tried to select news media with different political orientations in order to strengthen the representativeness of news media in each country. As a result, a set of two newspapers and two television news channels in US, UK, Germany, and Australia were chosen for each country, enlisting a total of 16 news media. (Table 1).

Data were collected through the websites of the newspapers and the TV news channels. For TV news stories, the text stories on the websites were used. The term used for searching relevant news stories was “COVID-19,” which was the official name of the pandemic disease. For the German news coverage, English translation provided by each news media was used. Columns and opinions from readers were not considered for coding. Also, if the main topic of a news story did not involve COVID-19, it was excluded from the analysis even if it included the key word, “COVID-19.”

There were so many COVID-19 articles released by each news media that content analysis of the whole news stories by each news media was not feasible. Data were collected in a way that 1,000 articles per each news media, recently published from the time point of data collection, were first selected and random sampling of 100 stories for each data set was proceeded to make a total of 1600 stories for the data set to be analyzed. The time period of the uploaded news stories was from February 4, 2020 to November 10, 2021.

Identifying a frame requires a close look at framing devices used in a news story. According to Pan and Kosicki (1993), framing devices can be found in four types of text structures: syntactic, script, rhetoric, and thematic structures. Syntactic structure can be identified in a typical news story with an inverted pyramid style, a common pattern for news writing. A headline, the most important part in the inverted pyramid, is the strongest tool for framing, while a lead could be the next significant layer (Pan and Kosicki 1993). Script structure is a device used when framing is conducted through independent episodes in a story, while rhetoric structure is the one in which a metaphor, catch phrase, or visual image is used to build a frame. A thematic structure is characterized by its hypothesis-testing. A hypothesis is suggested with some pieces of evidence in a news story with this framing device.

Although the rest of framing devices may or may not be used in a news story depending on a journalist who writes it, the syntactic structure is more likely to be used compared to the others as this type of structure can be identified in a typical news story with an inverted pyramid style, a common pattern for news writing. Therefore, we examined the syntactic structure of each story, focusing on the headline and the lead.

In order to conduct a framing analysis, both framings are operationalized by determining the main contents of the information expected on the headlines and leads that are associated with both framings. The operationalization was conducted by one researcher who read a dozen of articles per each news media and develop a scheme for operationalization by taking the conceptual aspects of both framings into consideration, and utilizing frequently identified subject matters of headlines and leads, which can be considered as representing the aspects of COVID-19 that are selectively focused on in framing the pandemic crisis. The operationalization of both framings is presented in Table 2 with exemplary headlines from the news media in the four countries.

Even though some information in a news story addresses different aspects from an original frame, it does not affect the coding process unless the information is placed in a headline or lead. This is because placing heterogeneous facts at the later parts of an inverted pyramid story should be a part of establishing an original framing, which is to “select, exclude, or highlight a certain aspect” (Entman 1993, p. 52). Here, including a heterogeneous piece of information in the later part may be regarded as highlighting a certain aspect that is placed on the headline or in the lead. Six coders participated in the coding process after having a training session in which the coding instruction was explained with samples. Before the main coding, a preliminary coding was also proceeded using a total of 80 news article that consist of five stories from each news media. Unclear points of coding for each coder were clarified and shared. The inter-coder reliability for framing was assessed using the measure of Fleiss kappa considering that there were six coders, and the magnitude of the measure was .60. It is interpreted that values between .40 and .75 represent fair to good agreement (Fleiss 1971).

Considering that framing and affective tones are categorical variables, we used monthly number of each framing, affective tone, and efficacy information. An independent sample t-test was conducted to examine the significant difference between both framings in an aggregate level. To find the country difference in the news framing, four separate independent sample t-tests were conducted. To identify the association between news framing and affective tone or efficacy information, we tested a multiple regression model. Here, news framing coded with “1” being prognosis and “0” being diagnosis was expected by effective tones and efficacy information. In the model, crisis stage was entered for control purposes because social responses to a crisis, including media framing, can vary according to the crisis stage (Fink 1986). Media type was also considered a control factor, considering the possibility that the media presentation of an issue (i.e., framing, affective tones embedded, and efficacy information provided) may vary according to different types of media.

We examined the subject matter covered in the headlines and leads since this message feature would be the highlighted focus of a news story and thereby constitute the major framing device for both diagnosis and prognosis framing. The most frequent subject matter was public health measures (35.3%), followed by virus prevalence (26.8%), individual coping (14.6%), socioeconomic consequences (12.2%), non-public health measures (4.5%), science of virus (3.4%), and attribution of responsibility (3.3%). When taking a look at the number by countries, the four countries showed a similar pattern in that the first and second most frequent subject matter were public health measures and virus prevalence. Regarding the third and fourth most frequent subject matter, however, US, UK, and Germany were similar in their favoring individual coping over socioeconomic consequences, while Australian media covered more socioeconomic consequences than individual coping (Figure 1).

When the type of news media was considered, both TVs and newspapers covered public health measures the most frequently. The next level of frequency was observed with virus prevalence for both types of news media (Figure 2). However, when it comes to the third level of frequency, both types of media were different in that TV channels favored individual coping, while newspapers covered socioeconomic consequences.

When the news coverage from the four countries was examined in aggregate for RQ 1, considering the global nature of the health crisis, there were 860 news stories focusing on prognosis while there were 740 stories focusing on diagnosis. However, the numeral excess of prognosis framing (M=0.54, SD=1.44) over diagnosis framing (M=0.46, SD=1.28) was not significant when the difference between the monthly numbers was subject to an independent sample t-test, indicating that there was no dominant tendency of favoring either diagnosis or prognosis in an aggregate level across the four countries [ t(3198) = −1.561, p=0.119].

When news coverage of each country was separately examined for RQ1-1, the US news media exceptionally favored prognosis (M=0.62, SD=1.22) over diagnosis (M=0.38, SD=0.72) [ t(798) = −3.46, p < .001]. The pandemic news coverage in the rest three countries did not show any preference for a particular framing. Here, the number of diagnosis and prognosis framing was exactly the same in the case of Germany, and there were even more diagnosis stories in Australia when the statistical significance was not considered. (Figure 3).

RQ2 asks whether a particular affective tone tends to be prevalent among news stories about COVID-19 with a particular frame. When a multiple regression model was tested based on monthly numbers with affective tones entered as a predictor of framing, it was found that a positive tone was associated with more prognostic framing (b = 0.18, t [5, 1593] = 9.38, p < .001). The significant association between positive tone and prognostic framing was consistent across all countries examined when the multiple regression analyses were conducted separately (Table 3). Efficacy information was also prevalent in the prognosis stories. The multiple regression model showed a much stronger association between prognosis framing and efficacy information (b = 0.22, t [5, 1593] = 12.46, p < .001) than with positive tone. A significant association was also observed in each country when the multiple regression analyses were conducted separately (Table 3).