The explanation of a complex problem: A content analysis of causality in cancer news

Abstract

Understanding causality is a critical part of developing preventive and treatment actions against cancer. Three main causality models—necessary, sufficient-component, and probabilistic causality have been commonly used to explain the causation between causal factors and risks in health science. However, news media do not usually follow a strict protocol to report the causality of health risks. The purpose of this study was to describe and understand how the causation of cancer was articulated on news media. A content analysis of 471 newspaper articles published in the United States during two time-frames (2007–2008 and 2017–2018) was conducted. The analysis showed that probabilistic causality was most frequently used to explain the causal relationship between risk factors and cancer. The findings also uncovered other important details of news framing, including types and characteristics of risk factors, intervention measures, and sources of evidence. The results provided theoretical and practical implications for public understanding and assessment of cancer risks.

Keywords

cancer causality framing news media risk communication

Cancer is a leading cause of death around the world (Siegel et al., 2020). To develop effective treatments against cancer, scientists attempt to understand its causes. Across different areas of health sciences, understanding causality is essential to develop pragmatic and scientific treatments (Russo and Williamson, 2007).

In past centuries, the true nature of causal relations has been the center of debate (Hume, 2000 [1748]; Lewis, 1974; Mackie, 1965; Puddephatt and Prus, 2007; Russell, 1929; Salmon, 1984). As human beings gain more experience and knowledge, the philosophical view has expanded from mechanical to probabilistic causality. Scientists have gradually adopted multiple, probabilistic, and interconnected causations to explain the onset and development of diseases (Broadbent, 2009; Krieger, 1994). While the pursuit of a definitive cause for every truth has never ceased, the paradigmatic transition of causality reflects the intention to practically explain the complexity and uncertainty of human beings and the unknown in nature (Parascandola, 2011).

The adoption of different causality models in scientific reports can influence laypersons’ risk perception through information exchanged in social experience, including news on mass media (Kasperson et al., 1988). When disseminating knowledge about risk events, news media do not always follow technical definitions; rather, they frame risk events by selectively emphasizing only certain aspects of information (Entman, 1993; Kitzinger, 1999). As part of the news framing, causal explanations of risk can influence how audiences perceive the causes and effects of different risks (Clarke and Everest, 2006; Entman, 1993; Escobar and Demeritt, 2014).

However, the presentation of more complex and uncertain causation may improve information accuracy and also cause confusion, anxiety, and distrust among laypeople (Fischhoff, 2012; Frewer et al., 2003). Also, news media has become an alternative space for different concerned voices (Peters and Dunwoody, 2016), which influences how causality is used and perceived. Thus, it is relevant and important to explore how news media frame the causality of cancer risk between the scientific community and laypersons.

1. Causality models

Causality is complicated and has always been the focus of philosophical debate (Mellor, 1998). Aristotle’s notion views the distinctions between objects, organisms, and humans as the foundation of knowledge (Puddephatt and Prus, 2007). In contrast, Hume was skeptical of causality. Holding a reductionist view, he defined a cause as an object “followed by another, and where all the objects, similar to the first, are followed by objects similar to the second” (Hume, 2000 [1748]: VII). While many different views of causality have emerged in history, this study examines three causality models that are commonly used to empirically explain the causes of health risks (Parascandola and Weed, 2001; Williamson, 2009).

Monocausality

The idea of seeking a single cause, held as a classic model of mechanical causation (Bunge, 1979; Elwood, 1992), was pursued by mechanics physicists, such as Galileo and Newton (Bohr, 1950). In a monocausal model, a necessary cause is a condition that must be present for an event to occur; in contrast, a sufficient cause is a factor that guarantees the occurrence of the effect (Parascandola and Weed, 2001). For instance, an influenza virus is a necessary cause of influenza without which influenza will not develop; however, influenza may not occur even with the presence of the virus. Strangulation is sufficient to cause death but death can be caused by other reasons. In a canonical monocausal model, only one causal factor can satisfy both necessary and sufficient conditions of the outcome (Broadbent, 2009). For example, Koch (1933) maintained that the Mycobacterium tuberculosis is a necessary and sufficient cause of tuberculosis.

Sufficient-component causality

Mackie (1965) proposed that a cause can be an Insufficient but Necessary component of a condition that is Unnecessary but Sufficient (INUS) for an effect. Similarly, the sufficient-component cause model was developed by Rothman (1976). Specifically, in a causal statement, such as C causes P, C itself as a cause is not a sufficient and necessary condition for P. Instead, C, along with other necessary causal factors, forms a single sufficient condition E. C and other causal factors are necessary parts of the sufficient condition E, which, as Rothman and Greenland (2005) theorize, are “a set of minimal conditions” (p. S144) with no redundant components (see Figure 1). In health contexts, each necessary cause is one part of a sufficient condition causing the disease; all necessary causes together constitute a full, sufficient cause for the disease onset (Rothman, 1976).

Figure 1.

An illustration of sufficient-component causality with a set of four causal factors.

Sufficient-component causality is sometimes mixed with monocausality. In daily language use, many simple causal statements actually express one component of sufficient condition causation. For example, in a common causal statement “alcohol consumption causes high blood pressure,” drinking alcohol is one INUS cause. Alcohol consumption alone is not a sufficient cause of hypertension. Rather, other INUS factors, such as age and stress, are also necessary to cause hypertension (National Institute on Aging, 2018). A cause (e.g. alcohol consumption) can be a necessary cause in any condition or in only one specific condition, depending on different individuals or contexts (Parascandola and Weed, 2001).

Probabilistic causality

In light of quantum physics, scientists have attempted to avoid the assumption that a cause always exists and will lead to the effect (Bohr, 1950). Also, because existing scientific measurement cannot fully account for uncertain and unknown parts in observation, a single cause in monocausality or many causes in sufficient-component causality cannot be identified and verified (Parascandola and Weed, 2001). Thus, the need for developing probabilistic models arises to explain causation beyond the classic cause-and-effect doctrine (Bohr, 1950; Eells, 1991; Fisher, 1934; Humphreys, 2014; Lewis, 1974; Mellor, 1998; Menzies, 1989; Salmon, 1984; Suppes, 1970).

Many different probabilistic models have been proposed but all maintain the same basic assumption: if C causes P, C either increases or decreases the probability of P, while controlling other causal factors of P (Cox, 1992; Holland, 1986; Williamson, 2009). For example, in probabilistic causality, smoking raises the chance of lung cancer, all else being equal. The overall probability of developing lung cancer is different between the presence versus the absence of smoking.

2. Causality of health risks

Early biologists insisted every disease must have a corresponding necessary-sufficient cause (Bernard, 1957). However, complex human behaviors and biological living processes pose significant challenges to find the etiologic chain of diseases, especially for chronic health risks (Elwood, 1992). Thus, pursuing a single, necessary-sufficient cause can ignore other factors that are equally important to understand the scope and solution of various health problems (Krieger, 2007; Venkatapuram and Marmot, 2009).

In response to those controversies, multifactorial causality emerges as the basic tenet of modern epidemiology (Krieger, 1994). Sufficient-component causality addresses the complexity of causal relations in health research and provides explanations that integrate multiple factors toward each health problem (Parascandola, 2011; Rothman and Greenland, 2005). Nevertheless, relying on sufficient-component causality, researchers still need to find out all the causes making up a sufficient condition that invariably leads to the effect, which may trivialize more important factors in disease development (Parascandola and Weed, 2001).

In response to these problems, a probabilistic model is believed to provide more concrete and practical explanations to the causality of health risks (Hacking, 1990; Parascandola, 2011). Using probabilistic causality, researchers can account for causal factors that change the likelihood of developing chronic diseases, such as cancer. Probabilistic causality indicates the uncertain nature of cancer while acknowledging the limitation of existing knowledge (Parascandola and Weed, 2001).

3. Media and risk causality

Scientific understandings of risk causality are not always translated directly and accurately into lay beliefs of causality (Gifford, 1986; Hunt and Emslie, 2001). In this process, news media play a significant role in connecting experts with laypersons (e.g. Breakwell, 2007; Kasperson et al., 1988). According to the social amplification of risk framework, news media represent and construct objective risks into a mediated reality, which generates a powerful impact on people with no direct experience about the risks (Kasperson et al., 1988). Also, media coverage prioritizes particular risk events that carry more sensation values, controversy, and symbolic meanings among the public (Hughes et al., 2006; Kasperson et al., 1988; Mazur, 1990). Consequently, mediated information can shape public opinions and beliefs about health and illness, including the causality of diseases (Davison et al., 1992; Hunt and Emslie, 2001; Prior, 2003). Public risk perception can thus be amplified or attenuated by media reporting over time (Frewer et al., 2002; Lewis and Tyshenko, 2009).

According to the framing theory, news coverage of risk events does not always reflect the technical definitions of risk but selectively reports some aspects of a risk to promote “a particular definition, causal interpretation, moral evaluation, and/or treatment recommendation” (Entman, 1993: 52). The examination of media frames shows news coverage creates salient perspectives that depart from an objective truth (Entman, 1993). In this sense, risk events can be framed in a certain way to attract attention, to guide public debates on specific aspects of issues, and to influence risk perception (Entman, 1993; Kitzinger, 1999). Previous studies found news stories tend to focus more on the impact of health risks, rather than on prevention or detection (Berry et al., 2007; Slater et al., 2008; Stryker et al., 2005). News is also likely to describe cancer screening as a frightening experience and portray cancer risks to be pervasive (Clarke and Everest, 2006).

Causal explanation plays a significant role in media framing (Entman, 1993). News framing gives attention only to certain causes leading to risk events while neglecting others. Moreover, framing causality of risk is often intended to communicate responsibility and further influence the linkage between science and risk perception (Jang, 2013). In a more complex diffusion process, various actors of cancer research may seek to frame causality differently to influence how the public perceives relevant scientific findings (Peters and Dunwoody, 2016). Thus, an examination of causality models in the news frames will help reveal the forces that shape public understanding and assessment of cancer risks.

Different causality models can bring conflicting implications to the public understanding of health risks through news media. Traditionally, public health professionals often disseminate health claims based on simplified monocausal claims (e.g. smoking causes cancer) to emphasize the harmful effects of a risk factor and hence to motivate behavioral change (Allmark and Tod, 2006; Davison et al., 1991). However, multiple definitive messages could be often contradictory, further influencing lay interpretation and understanding of health risks, especially when the public experience unexpected survival and anomalous deaths (Allmark and Tod, 2006; McConnachie et al., 2001).

However, by expressing the complexity and uncertainty of cancer, probabilistic causality can show the transparency of scientists and their willingness to acknowledge the limitation of knowledge (Johnson and Slovic, 1995). Cancer news can be perceived to be more credible and trustworthy, especially among those who place more trust in science (Jensen, 2008; Johnson and Slovic, 1995). From the perspective of uncertainty management, however, critics argue that increased uncertainty can jeopardize personal control and autonomy of illness and risk (Maxim et al., 2013). Laypersons may experience confusion when coping with uncertain risks (Fischhoff, 2012). Thus, different causality models on the news can result from the balance between pursuing scientific accuracy and the practicality of meeting the needs of multiple interests.

4. The present study

A number of studies have focused on the frames in cancer news (e.g. Hoffman-Goetz and Friedman, 2005), but none of the previous studies, to the best of our knowledge, have specifically examined causality models in the frames of cancer news. While some studies have focused on what causes were linked to cancer (e.g. Jensen et al., 2010; Slater et al., 2008), how news reporting uses different types of causality to frame health risks is notably unknown. Addressing this gap in the literature is the first step to understand causality shown in public information as well as its effects on risk perception.

The purpose of this study is to describe which type of causality is present and how causation of cancer risks is articulated on news media. We focused on cancer as the specific context of this study; the reason for which is twofold. First, cancer remains and will still be a leading cause of death in the foreseeable future around the world (Bray et al., 2018; Siegel et al., 2020). Second, as a chronic disease, cancer could be caused by a mix of factors on different levels (Ezzati et al., 2002; Lim et al., 2012). Actions against cancer are closely related to how causality is presented and processed by scientists, physicians, and the public (Allmark and Tod, 2006; Lagiou et al., 2005; Vineis and Wild, 2014).

We studied newspaper articles only, for several reasons. First, newspapers usually carry more detailed factual content, including background and context information than television news (Eveland et al., 2002). Newspaper news uses thematic frames that are oriented toward broader overriding and structural factors behind social issues (Dudo et al., 2007; Iyengar, 1994). In comparison, television commonly uses episodic content that limits the reporting scope to single events and individuals (Iyengar, 1994). Second, magazines are not an optimal source to sample general cancer-related information. Compared with newspapers, magazines usually appeal to a readership with specialized interests (Greenwood and Jenkins, 2015; Johnson and Meischke, 1993). As readers actively seek the information they want, magazine writing is increasingly focused and narrow (Sumner, 2012). Third, newspapers provide greater depth and breadth in reporting than fully online news sites (Maier, 2010). Through online platforms, newspapers can maintain and expand readership (Chyi et al., 2010).

This study was a content analysis of cancer news in US newspapers. We examined causality in news reporting and the association with other framing characteristics. News articles published within two periods of time (i.e. 2017–2018 and 2007–2008) were sampled to observe temporal changes in framing patterns over the past decade. The following research questions were proposed to guide this exploratory study:

RQ1. What type of causality is used most frequently in cancer news?

RQ2. What type and characteristic of risk factors are used in association with different causality models in cancer news?

RQ3. What are the sources of evidence used in cancer news to support the causality of cancer risk factors?

RQ4. What are intervention measures of risk factors in cancer news?

RQ5. What are the associations between types of causality and characteristics of framing (risk factors, sources of evidence, or intervention measures)?

RQ6. Does news framing of causality models change in the past decade?

5. Method

Overview

We conducted a quantitative content analysis on cancer news published in the United States. Content analysis allows researchers to systematically observe recorded communication and evaluate underlying symbolic meaning (Krippendorff, 2012; Lombard et al., 2002). This approach reduces text data to quantitative forms for replication and comparison of different text contents (Gordon, 2003; Mello and Hornik, 2016). Three coauthors served as the coders of this study. The unit of analysis for this project was each single news article.

Sampling

The newspaper articles were sampled from Nexis Uni, an academic database for accessing archive news and digital information. This third-party database allowed us to impartially sample articles without the influence of the existing interest or agenda of various media outlets. This sampling approach also avoided the discrepancy between different search engine algorithms provided by publishers’ own websites. All kinds of published articles on newspapers except obituaries were sampled. The supplemental material provides the details of the sampling procedure and rationale.

To observe the temporal change in framing, we sampled news articles published in 2017–2018 when the study was conceptualized and also those published a decade ago in 2007–2008. These two time-frames were selected because of significant decreases in cancer incidence rates and revolutionary developments in cancer treatment over this decade (National Cancer Institute, 2020; Topper et al., 2020). The official federal statistics (2020) show that the cancer incidence rate peaked in 2007 but steadily declined to the lowest point in 2017 since 1999. The stark contrast between these two time points suggests that cancer may have become more preventable and treatable (Siegel et al., 2020). Research initiated a decade ago based on emerging paradigms of knowledge has led to major breakthroughs recently (Baylin and Jones, 2011; Bernards et al., 2020). As a result, laypersons’ perceptions, behaviors, and assessments of the disease could be different. The comparison between recent and a-decade-old newspaper articles may reveal how the mediated information contributes to the changes in public understanding of cancer.

Coders and intercoder reliability

The same method was followed to sample a separate pool of posts for coder training. A total of 45 posts, approximately 10% of the main sample, were randomly collected. The first author led the training process. Three coders completed coding independently and convened again to discuss any questions and disagreements. Necessary adjustments to the codebook were also discussed. The training was completed at the time when the articles in the training pool were coded and intercoder reliability was satisfied (Lombard et al., 2002). Cumulative total coding training time was about 40 hours. Intercoder reliability (Krippendorff’s alpha) is included in the supplemental material.

After the completion of the training, the coders worked independently in a supervised environment under the guidance of the first author. To counter the influence of fatigue and loss of attention, the first author continuously checked the quality of a random 5%–10% subset for every 50 articles coded. Questions and disagreements identified in the coding process were addressed immediately before coding the next batch of the main sample.

Coding schemes and variables

The first author developed the initial coding scheme. Adjustments to the coding schemes were finalized throughout the coder training. The following variables were coded: publication year, type of causality, type and characteristic of risk factor, causal explanation, statistical evidence, primary source of evidence, secondary source of evidence, and intervention for risk factor. Cases that could not be independently determined by a coder for a category in a variable were treated as missing values for that variable. Operational definitions and coding instructions are provided in the supplemental material.

6. Results

Descriptive statistics

Excluding duplicate and irrelevant articles, a total of 471 news articles published by 118 media outlets were coded and analyzed. In the final sample, 272 (57.7%) were published in 2007–2008 and 199 (42.3%) in 2017–2018. Excluding two cases of missing data, probabilistic causality was the most frequently used model as found in 253 (53.9%) articles. Monocausality and sufficient-component causality were used in 103 (22%) and 113 (24.1%) articles, respectively.

Descriptive statistics concerning the type and characteristic of risk factors are reported in Table 1. Individual lifestyle, comorbid conditions, and human-related activities were the three most discussed types of risk factor. Among characteristics of risk factor, about 46% were categorized as physiological factors and another 20.6% bad and poisonous.

Table 1.

Types and characteristics of risk factors.

	Risk factor
	N	%
Types of Risk Factors
Individual lifestyle	147	32.5
Individual demographics	43	9.5
Individual health care	9	2
Individual comorbid condition	100	22.1
Genetic causes	67	14.8
Social or systematic issues	1	0.2
Human-related activities	79	17.4
Non human-related causes	6	1.3
Other	0	0
More than one level	1	0.2
Characteristics of Risk Factors
Bad and poisonous	96	20.6
Bad and desirable	66	14.2
Good	34	7.3
Physiological	214	45.9
Neutral/cannot decide	17	3.6
Not applicable	39	8.4

Percentage (%) is calculated as the ratio of the frequency of a particular category and the total number of valid cases (excluding 18 missing values in Types of Risk Factors and 5 in Characteristics of Risk Factors.). The sum of percentage is equal to 100.

Causal explanation and statistical evidence of risk factors are reported in Table 2. About 54.9% of all articles provided causal explanation and 45.5% used statistical evidence. Table 3 lists all sources of evidence to support causality of risk factors. Health care organizations, individual providers, individual scientists were the three most commonly cited primary and secondary sources of evidence.

Table 2.

Causal explanation, statistical evidence, and temporal distance of causality of Risk Factor.

	Risk Factor
	N	%
Causal Explanation
Yes	258	54.9
No	212	45.1
Statistical Evidence
Yes	214	45.5
No	256	54.5

Percentage (%) is calculated as the ratio of the frequency of a particular category and the total number of valid cases (excluding 1 missing value in Causal Explanation and 1 in Statistical Evidence). The sum of percentage is equal to 100.

Table 3.

Primary and secondary source of evidence to support causality of risk factors.

	N	%
Primary Source of Evidence
Individual scientists	62	13.4
Individual health care providers	77	16.6
Health care organizations	132	28.4
Pharmaceutical company	2	0.4
Academic publications	42	9.1
Mass media	2	0.4
Lay non-patient	31	6.7
Individual patients	42	9.1
Other	1	0.2
Not applicable/unknown	43	9.3
Court/non-health-related governmental agency	30	6.5
Secondary Source of Evidence
Individual scientists	52	11.4
Individual health care providers	33	7.2
Health care organizations	72	15.8
Pharmaceutical company	3	0.7
Academic publications	18	3.9
Mass media	10	2.2
Lay non-patient	24	5.3
Individual patients	18	3.9
Other (please specify)	1	0.2
Not applicable/unknown	213	46.7
Court/non-health-related governmental agency	12	2.6

Percentage (%) is calculated as the ratio of the frequency of a particular category and the total number of valid cases (excluding 7 missing values in Primary Source of Evidence and 15 in Secondary Source of Evidence). The sum of percentage is equal to 100.

Types of causality

A set of chi-square analyses were conducted to understand the relationships between causality models and framing characteristics. Of the news articles employing monocausality, 43.1% discussed the individual lifestyle risk factors. By contrast, significantly fewer articles using sufficient-component and probabilistic models (33.6% and 27.5%, respectively) focused on this type of risk factor, χ²(16, N = 453) = 45.31, p < .001, V = .22. Most of the articles on demographics (81.4%), genetics (59.7%), human activities (59.5%) levels of risk factors were found to use probabilistic causality.

Among the articles employing monocausality, 47.6% of bad and poisonous risk factors were attributed to human activities while most of bad and desirable risk factors (63.6%) were about individual lifestyle, χ²(16, N = 101) = 62.79, p < .001, V = .39. Similar proportions were observed in articles using sufficient-component and probabilistic causality. In addition, comorbid illness and genetic factors were two primary types of physiological risk factors in articles employing sufficient-component and probabilistic causality.

In articles employing either monocausal or probabilistic causality, health care organizations were most frequently cited as the primary source of evidence to support causality of risk factors. In contrast, individual providers were cited more frequently in articles using sufficient-component causality, χ²(20, N = 463) = 63.29, p < .001, V = .26. Individual scientists and academic publications were cited as the primary source in 17.2% and 12% of articles employing probabilistic causality, respectively. These two sources were used in less than 10% of news employing necessary and sufficient-component causality. In addition, secondary sources of evidence did not vary with different causality models, χ²(20, N = 455) = 25.64, p = .18, V = .17.

An in-depth examination shows that healthcare organizations were frequently cited as the primary source of evidence to support most types of risk factors. This source was found in one-third of the articles discussing risk factors related to lifestyle (29.9%), demographics (34.9%), comorbid illness (32.3%), and human-related activities (34.6%), χ²(80, N = 447) = 126.57, p < .01, V = .19. However, among articles discussing genetic issues as risk factors, individual scientists were cited more frequently (25.8%) than other sources.

Prevention was a common intervention against cancer risk factors. Prevention was discussed in 67.6% of the articles using monocausality versus approximately 46% using sufficient-component or probabilistic causality, χ²(6, N = 461) = 28.16, p < .001, V = .18. By contrast, detection was found in 21.4% of probabilistic model articles but only in 4.9% and 9.9% of articles using necessary and sufficient-component causality models. About 13.5% of articles employing sufficient-component causality discussed treatment, which was higher than other causality models.

Over half of the articles with causal explanation (57.4%) used statistical information, χ²(1, N = 470) = 32.29, p < .001, V = .26. However, neither causal explanation (χ²(2, N = 469) = 3.96, p = .14, V = .09) nor statistical information (χ²(2, N = 469) = .47, p = .79, V = .03) for risk factors differed by the type of causality.

Temporal variations from 2007–2008 to 2017–2018

The comparison between news articles published one decade ago (2007–2008) and recently (2017–2018) showed additional changes in framing. The overall use of probabilistic causality increased from 41.3% to 71.2% in a decade, χ²(2, N = 469) = 43.70, p < .001, V = .31. In contrast, the monocausal and sufficient-component models decreased from 25.8% to 16.7% and from 32.8% to 12.1%, respectively.

In addition, individual lifestyle was mentioned less frequently, χ²(8, N = 453) = 15.72, p < .05, V = .19. Instead, demographic and genetic risk factors were increasingly discussed in recent articles. Notably, compared to articles published a decade ago (χ²(14, N = 258) = 31.91, p < .01, V = .25), there was no significant relationship between the type of causality and the risk factor level in recent news, χ²(14, N = 195) = 16.54, p = .28, V = .21.

Moreover, the characteristics of risk factors were discussed differently in the past decade. About 26.1% articles of monocausality published 10 years ago focused on bad and poisonous risk factors; the percentage dropped to 9.1% among recent articles, χ²(4, N = 102) = 16.28, p < .01, V = .40. In contrast, an increase from 4.3% to 27.3% of the articles focusing on bad and desirable risk factors was observed. This change was not significant in articles employing other causality models.

The primary source of evidence also changed over time. Recent articles increasingly cited individual providers and health care organizations as the evidence source. Individual scientists and academic publications were less cited, χ²(10, N = 463) = 27.94, p < .01, V = .25. Notably, these changes were only observed in articles employing sufficient-component (χ²(8, N = 113) = 16.14, p < .01, V = .38) and probabilistic causality models (χ²(8, N = 250) = 23.62, p < .01, V = .31).

Finally, prevention as an intervention measure was discussed less frequently over the past decade, χ²(3, N = 462) = 27.22, p < .001, V = .24. Detection and treatment were increasingly framed in articles using sufficient-component (χ²(3, N = 111) = 11.01, p < .05, V = .32) and probabilistic (χ²(3, N = 248) = 10.30, p < .05, V = .20) models, This variation, however, was not observed in articles using monocausality.

7. Discussion

The purpose of this content analysis was to examine how news media represented causality between cancer and risk factors. Probabilistic causality was identified as the most frequently used model in US newspapers, suggesting a long-term shift toward probabilistic uncertainty in news reporting. Along with other framing details concerning the causality of cancer, the findings raised theoretical and practical implications for public perception and understanding of cancer-related risks and uncertainty.

At least two major reasons could explain the shift toward probabilistic causality in news framing. First, probabilistic causality might be intended to enhance the accuracy and credibility of cancer news. As a multidimensional construct, credibility is often associated with expertise and trustworthiness of a communicator (McCroskey and Young, 1981). In cancer news, probabilistic causality can explain more precisely how cancer is related to various risk factors as revealed in scientific discoveries (Parascandola, 2011). This could explain our finding that probabilistic causality was commonly associated with some intangible risk factors, such as demographic, genetic, and comorbid illness. This strength of probabilistic causality shows the expertise of scientists and journalists in news reporting. Also, scientists are seen as more honest and trustworthy by communicating the uncertainty associated with probabilistic causality (Johnson and Slovic, 1995). In contrast, highly certain health claims, which often involve monocausality, are considered less credible after unwarranted cases contradict such information (McConnachie et al., 2001).

Second, probabilistic causality could be more acceptable due to people viewing scientific research differently. Such a changing view could be shaped by a dominant value and culture in Western society that science should constantly explore the unknown and expand human knowledge (Brossard and Nisbet 2007). People expect scientists to stay critical of existing knowledge (Jensen, 2008; Popper, 1961 [1934]). Thus, newspaper readers could be more open to probabilistic causality, especially from scientific sources, as it expresses the tentativeness and limitation in scientists’ knowledge and research findings.

However, one should interpret the prevalence of probabilistic causality with caution because this model can also express skepticism or negative evaluations of research findings. This consideration is critical due to a more complex knowledge diffusion process with various actors involved (Peters and Dunwoody, 2016). For example, probabilistic causality could allow some stakeholders to interpret research findings differently for their own interests (Post and Maier, 2016). News stories may use probabilistic causality to maintain the balance between competing scientists’ claims and to spice up their controversies (Hughes et al., 2006). Journalists might also imply through probabilistic causality that in their opinion some scientific claims are implausible (Peters and Dunwoody, 2016). These potential implications call for future research to further explore the purpose and rationale of employing different causality models in news framing.

In addition, increasingly ambiguous and complex causal information may cause the public to rely heavily on trust, instead of rational evaluation, in risk management (Siegrist, 2021). As a heuristic state, trust allows those with less knowledge to reduce complexity and count on specific expert sources in risk assessment (Siegrist, 2021). Laypeople placing trust in a particular entity usually perceive lower risk and accept recommended measures (Siegrist, 2021). Thus, trust can influence how laypeople process causal explanations supported by different sources. This notion explains frequent citations of scientific sources in probabilistic causal information. Those who have greater trust in science can better tolerate uncertainty in scientific information (Johnson and Slovic, 1995). However, individuals who trust health professionals usually expect security and certainty when seeking medical attention (Simpkin and Schwartzstein, 2016). Thus, clinical evidence from health care sources was often used to support monocausal claims and enhance the perceived trustworthiness of medical sources.

Nevertheless, the importance of trust in evaluating more complex cancer news could increase the public’s vulnerability to the performance of different stakeholders. This trend is particularly worrisome as journalists continue to pursue failure and controversies of science (Peters and Dunwoody, 2016). Future research should further illuminate the role of trust in the perception and assessment of different causality models.

Our findings also raised the importance of uncertainty management related to cancer. Some researchers believed that uncertainty should be eliminated because it would cause confusion and anxiety among laypeople (Fischhoff, 2012; Frewer et al., 2003). Thus, monocausality for lifestyle risk factors might be intended to enhance scientific authority and gain compliance (e.g. Zehr, 2000). However, certainty is not beneficial in all situations (Brashers, 2001). Health suggestions that claim the certainty of different cancer risks are often contradictory, which could only intensify worry about cancer (Brashers, 2001). Also, people who have developed a fatalistic view of cancer tend to avoid highly definitive information (Davison et al., 1992).

According to uncertainty management theory (Brashers, 2001), people may seek to reduce or increase uncertainty, depending on the appraisal of uncertainty as a threat or opportunity, respectively. In this sense, uncertainty through probabilistic causality can be appreciated because it may bring opportunities to manage cancer more effectively (Jensen et al., 2017). For example, probabilistic causality can provide additional knowledge that disconfirms one’s belief about cancer and encourages alternative coping strategies (Brashers, 2001). Probabilistic causality can explain anomalous exceptions of survival or death in personal experiences (Allmark and Tod, 2006; Hunt and Emslie, 2001). For those with fatalistic beliefs, probabilistic causality may introduce another layer of uncertainty that increases their hope for future treatments (Bylund et al., 2012). Thus, future science communication should consider the benefits of uncertainty when evaluating the practical implications of causality models.

Furthermore, prevention information constituted the majority of news coverage concerning the counteracting measures against cancer. Deviating from previous content analyses (Jensen et al., 2010; Slater et al., 2008), this finding nevertheless suggests a positive development, which could have contributed to lowering the cancer incidence rate in a decade (U.S. Cancer Statistics Working Group, 2020). Interestingly, prevention was frequently paired with monocausality. For example, one news article (McCullough, 2018) used monocausality to explain the relationship between textured breast implants and a form of lymphoma. The article further pointed out the preventive measure (i.e., avoiding textured implants) against the occurrence of lymphoma. Thus, preventive actions could be the best solution to the control of newly identified risk factors. Additionally, detection and treatment appeared more frequently as scientists gained knowledge of emerging risk factors over the years.

We acknowledge several limitations that warrant future research. First, newspaper articles were sampled from two time-frames due to significant changes in cancer incidence rates and developments in cancer treatment over this decade. More research is needed to examine media coverage from other time points if additional factors are considered, such as trends of public opinions, political climate, and specific achievement in cancer research. Second, articles were sampled solely from US newspapers. To further understand the public understanding of cancer-related causality, future research should also evaluate the content on emerging media platforms and that from popular cultures (e.g. television shows). Third, content analysis could not infer the effects of causality models on readers’ assessment, attitudes, and behaviors related to cancer risk factors. Other research methods should be employed to examine the antecedents and effects of using different causality models. Finally, the accuracy of causality claims in sampled articles was not assessed. Future researchers could use expert and lay judges to evaluate actual or perceived quality of causality claims in news articles.

8. Summary

This study was among the first to focus on causality as the frame of cancer news. The analysis revealed probabilistic causality was commonly employed to discuss causation between risk factors and cancer. Probabilistic causality may help news media provide the latest scientific findings and also engage the public in preventive actions. The frames of causality, however, differed among the type of risk factors, intervention measures, and sources of evidence. It is our hope that the study findings shed light on how news media frame causality of cancer risks and further inform the direction of future research on causality and risk perception.

Supplemental Material

sj-pdf-1-pus-10.1177_09636625211005249 – Supplemental material for The explanation of a complex problem: A content analysis of causality in cancer news

Supplemental material, sj-pdf-1-pus-10.1177_09636625211005249 for The explanation of a complex problem: A content analysis of causality in cancer news by Wei Peng, Gabriel Alexander de Tuya, Andrea Alexandra Eduardo, Jessica Allison Vishny and Qian Huang in Public Understanding of Science

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Wei Peng

Qian Huang

Supplemental material

Supplemental material for this article is available online.

Author biographies

Wei Peng is an assistant professor in the Edward R. Murrow College of Communication at Washington State University. His research examines health communication, risk analysis, and uncertainty management.

Gabriel Alexander de Tuya recently graduated from the University of Miami with his Masters of Arts in Communication Studies. He currently serves as teacher of Ethics and Director of Debate at Christopher Columbus High School in Miami, Florida. His interests in research include deceptive communication and misinformation.

Andrea Alexandra Eduardo is currently a JD student at the University of Miami School of Law. Her area of concentration is public interest. She looks forward to conducting more research that seeks to address and rectify systemic issues.

Jessica Allison Vishny is currently a JD student in the College of Law at DePaul University. Her academic and professional interest is health law.

Qian Huang is a postdoctoral researcher at the Department of Health Behavior in the Gillings School of Global Public Health at the University of North Carolina at Chapel Hill. Her research interests include health communication and message design.

References

Allmark

Tod

(2006) How should public health professionals engage with lay epidemiology? Journal of Medical Ethics 32(8): 460–463.

Baylin

Jones

(2011) A decade of exploring the cancer epigenome—Biological and translational implications. Nature Reviews Cancer 11(10): 726–734.

Bernard

(1957) An Introduction to the Study of Experimental Medicine. New York, NY: Courier Corporation.

Bernards

Jaffee

Joyce

Lowe

Mardis

Morrison

, et al. (2020) A roadmap for the next decade in cancer research. Nature Cancer 1(1): 12–17.

Berry

Wharf-Higgins

Naylor

(2007) SARS wars: An examination of the quantity and construction of health information in the news media. Health Communication 21(1): 35–44.

Bohr

(1950) On the notions of causality and complementarity. Science 111(2873): 51–54.

Brashers

(2001) Communication and uncertainty management. The Journal of Communication 51(3): 477–497.

Bray

Ferlay

Soerjomataram

Siegel

Torre

Jemal

(2018) Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA: A Cancer Journal for Clinicians 68(6): 394–424.

Breakwell

(2007) The Psychology of Risk. New York, NY: Cambridge University Press.

10.

Broadbent

(2009) Causation and models of disease in epidemiology. Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences 40(4): 302–311.

11.

Brossard

Nisbet

(2007) Deference to scientific authority among a low information public: Understanding US opinion on agricultural biotechnology. International Journal of Public Opinion Research 19(1): 24–52.

12.

Bunge

(1979) Causality and Modern Science, 3rd rev edn. New York, NY: Dover Publications.

13.

Bylund

Fisher

Brashers

Edgerson

Glogowski

Boyar

, et al. (2012) Sources of uncertainty about daughters’ breast cancer risk that emerge during genetic counseling consultations. Journal of Genetic Counseling 21(2): 292–304.

14.

Chyi

Yang

Lewis

Zheng

(2010) Use of and satisfaction with newspaper sites in the local market: Exploring differences between hybrid and online-only users. Journalism & Mass Communication Quarterly 87(1): 62–83.

15.

Clarke

Everest

(2006) Cancer in the mass print media: Fear, uncertainty and the medical model. Social Science & Medicine 62(10): 2591–2600.

16.

Cox

(1992) Causality: Some statistical aspects. Journal of the Royal Statistical Society. Series A 155(2): 291–301.

17.

Davison

Frankel

Smith

(1992) The limits of lifestyle: Re-assessing “fatalism” in the popular culture of illness prevention. Social Science & Medicine 34(6): 675–685.

18.

Davison

Smith

Frankel

(1991) Lay epidemiology and the prevention paradox: The implications of coronary candidacy for health education. Sociology of Health & Illness 13(1): 1–19.

19.

Dudo

Dahlstrom

Brossard

(2007) Reporting a potential pandemic: A risk-related assessment of avian influenza coverage in U.S. newspapers. Science Communication 28(4): 429–454.

20.

Eells

(1991) Probabilistic Causality. New York, NY: Cambridge University Press.

21.

Elwood

(1992) Causal Relationships in Medicine: A Practical System for Critical Appraisal. New York, NY: Oxford University Press.

22.

Entman

(1993) Framing: Toward clarification of a fractured paradigm. The Journal of Communication 43(4): 51–58.

23.

Escobar

Demeritt

(2014) Flooding and the framing of risk in British broadsheets, 1985–2010. Public Understanding of Science 23(4): 454–471.

24.

Eveland

Seo

Marton

(2002) Learning from the news in campaign 2000: An experimental comparison of TV news, newspapers, and online news. Media Psychology 4(4): 353–378.

25.

Ezzati

Lopez

Rodgers

Hoorn

Murray

CJL

(2002) Selected major risk factors and global and regional burden of disease. The Lancet 360(9343): 1347–1360.

26.

Fischhoff

(2012) Communicating uncertainty fulfilling the duty to inform. Issues in Science and Technology 28(4): 63–70.

27.

Fisher

(1934) Indeterminism and natural selection. Philosophy of Science 1(1): 99–117.

28.

Frewer

Hunt

Brennan

Kuznesof

Ness

Ritson

(2003) The views of scientific experts on how the public conceptualize uncertainty. Journal of Risk Research 6(1): 75–85.

29.

Frewer

Miles

Marsh

(2002) The media and genetically modified foods: Evidence in support of social amplification of risk. Risk Analysis 22(4): 701–711.

30.

Gifford

(1986) The meaning of lumps: A case study of the ambiguities of risk. In: Janes

Stall

Gifford

(eds) Anthropology and Epidemiology. Dordrecht: D. Reidel Publishing, pp. 213–246.

31.

Gordon

(2003) Risk communication and foodborne illness: Message sponsorship and attempts to stimulate perceptions of risk. Risk Analysis 23(6): 1287–1296.

32.

Greenwood

Jenkins

(2015) Visual framing of the Syrian conflict in news and public affairs magazines. Journalism Studies 16(2): 207–227.

33.

Hacking

(1990) The Taming of Chance. New York, NY: Cambridge University Press.

34.

Hoffman-Goetz

Friedman

(2005) Disparities in the coverage of cancer information in ethnic minority and mainstream mass print media. Ethnicity & Disease 15(2): 332–340.

35.

Holland

(1986) Statistics and causal inference. Journal of the American Statistical Association 81(396): 945–960.

36.

Hughes

Kitzinger

Murdock

(2006) The media and risk. In: Taylor-Gooby

Zinn

(eds) Risk in Social Science. New York, NY: Oxford University Press, pp. 250–270.

37.

Hume

(2000 [1748]) An Enquiry Concerning Human Understanding: A Critical Edition. Oxford: Clarendon Press.

38.

Humphreys

(2014) The Chances of Explanation: Causal Explanation in the Social, Medical, and Physical Sciences. Princeton, NJ: Princeton University Press.

39.

Hunt

Emslie

(2001) Commentary: The prevention paradox in lay epidemiology—Rose revisited. International Journal of Epidemiology 30: 442–446.

40.

Iyengar

(1994) Is Anyone Responsible? How Television Frames Political Issues. Chicago, IL: University of Chicago Press.

41.

Jang

(2013) Framing responsibility in climate change discourse: Ethnocentric attribution bias, perceived causes, and policy attitudes. Journal of Environmental Psychology 36: 27–36.

42.

Jensen

(2008) Scientific uncertainty in news coverage of cancer research: Effects of hedging on scientists’ and journalists’ credibility. Human Communication Research 34(3): 347–369.

43.

Jensen

Moriarty

Hurley

Stryker

(2010) Making sense of cancer news coverage trends: A comparison of three comprehensive content analyses. Journal of Health Communication 15(2): 136–151.

44.

Jensen

Pokharel

Scherr

King

Brown

Jones

(2017) Communicating uncertain science to the public: How amount and source of uncertainty impact fatalism, backlash, and overload. Risk Analysis 37(1): 40–51.

45.

Johnson

Slovic

(1995) Presenting uncertainty in health risk assessment: Initial studies of its effects on risk perception and trust. Risk Analysis 15(4): 485–494.

46.

Johnson

Meischke

(1993) A comprehensive model of cancer-related information seeking applied to magazines. Human Communication Research 19(3): 343–367.

47.

Kasperson

Renn

Slovic

Brown

Emel

Goble

, et al. (1988) The social amplification of risk: A conceptual framework. Risk Analysis 8(2): 177–187.

48.

Kitzinger

(1999) Researching risk and the media. Health, Risk & Society 1(1): 55–69.

49.

Koch

(1933) The aetiology of tuberculosis. Archives of Internal Medicine 51(5): 818–818.

50.

Krieger

(1994) Epidemiology and the web of causation: Has anyone seen the spider? Social Science & Medicine 39(7): 887–903.

51.

Krieger

(2007) Why epidemiologists cannot afford to ignore poverty. Epidemiology 18(6): 658–663.

52.

Krippendorff

(2012) Content Analysis: An Introduction to Its Methodology. Thousand Oaks, CA: SAGE.

53.

Lagiou

Adami

H-O

Trichopoulos

(2005) Causality in cancer epidemiology. European Journal of Epidemiology 20(7): 565–574.

54.

Lewis

(1974) Causation. The Journal of Philosophy 70(17): 556–567.

55.

Lewis

Tyshenko

(2009) The impact of social amplification and attenuation of risk and the public reaction to mad cow disease in Canada. Risk Analysis 29: 714–728.

56.

Lim

Vos

Flaxman

Danaei

Shibuya

Adair-Rohani

, et al. (2012) A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990-2010: A systematic analysis for the Global Burden of Disease Study 2010. The Lancet 380(9859): 2224–2260.

57.

Lombard

Snyder-Duch

Bracken

(2002) Content analysis in mass communication assessment and reporting of intercoder reliability. Human Communication Research 28(4): 587–604.

58.

McConnachie

Hunt

Emslie

Hart

Watt

(2001) “Unwarranted survivals” and “anomalous deaths” from coronary heart disease: Prospective survey of general population. BMJ 323(7327): 1487–1491.

59.

McCroskey

Young

(1981) Ethos and credibility: The construct and its measurement after three decades. Central States Speech Journal 32(1): 24–34.

60.

McCullough

(2018) Textured breast implants linked to rare cancer pulled from market in Europe. Available at: https://www.inquirer.com/health/breast-implant-lymphoma-allergan-20181219.html

61.

Mackie

(1965) Causes and conditions. American Philosophical Quarterly 2(4): 245–264.

62.

Maier

(2010) All the news fit to post? Comparing news content on the web to newspapers, television, and radio. Journalism & Mass Communication Quarterly 87(3–4): 548–562.

63.

Maxim

Mansier

Grabar

(2013) Public reception of scientific uncertainty in the endocrine disrupter controversy: The case of male fertility. Journal of Risk Research 16(6): 677–695.

64.

Mazur

(1990) Nuclear power, chemical hazards, and the quantity of reporting. Minerva 28(3): 294–323.

65.

Mello

Hornik

(2016) Media coverage of pediatric environmental health risks and its effects on mothers’ protective behaviors. Risk Analysis 36(3): 605–622.

66.

Mellor

(1998) The Facts of Causation. New York, NY: Routledge.

67.

Menzies

(1989) Probabilistic causation and causal processes: A critique of Lewis. Philosophy of Science 56(4): 642–663.

68.

National Cancer Institute (2020) Milestones in cancer research and discovery. Available at: https://www.cancer.gov/research/progress/250-years-milestones (accessed 29 September 2020).

69.

National Institute on Aging (2018) High blood pressure. Available at: https://www.nia.nih.gov/health/publication/high-blood-pressure

70.

Parascandola

(2011) Causes, risks, and probabilities: Probabilistic concepts of causation in chronic disease epidemiology. Preventive Medicine 53(4–5): 232–234.

71.

Parascandola

Weed

(2001) Causation in epidemiology. Journal of Epidemiology and Community Health 55(12): 905–912.

72.

Peters

Dunwoody

(2016) Scientific uncertainty in media content: Introduction to this special issue. Public Understanding of Science 25(8): 893–908.

73.

Popper

(1961 [1934]) The Logic of Scientific Discovery (trans. Popper

). New York, NY: Science Editions.

74.

Post

Maier

(2016) Stakeholders’ rationales for representing uncertainties of biotechnological research. Public Understanding of Science 25(8): 944–960.

75.

Prior

(2003) Belief, knowledge and expertise: The emergence of the lay expert in medical sociology. Sociology of Health & Illness 25: 41–57.

76.

Puddephatt

Prus

(2007) Causality, agency, and reality: Plato and Aristotle meet George Herbert Mead and Herbert Blumer. Sociological Focus 40(3): 265–286.

77.

Rothman

(1976) Causes. American Journal of Epidemiology 104(6): 587–592.

78.

Rothman

Greenland

(2005) Causation and causal inference in epidemiology. American Journal of Public Health 95(Suppl. 1): S144–S150.

79.

Russell

(1929) On the notion of cause. In: Russell

(ed.) Mysticism and Logic. New York, NY: W. W. Norton, pp. 180–205.

80.

Russo

Williamson

(2007) Interpreting causality in the health sciences. International Studies in the Philosophy of Science 21(2): 157–170.

81.

Salmon

(1984) Scientific Explanation and the Causal Structure of the World. Princeton, NJ: Princeton University Press. Available at: https://philpapers.org/rec/SALSEA

82.

Siegel

Miller

Jemal

(2020) Cancer statistics, 2020. CA: A Cancer Journal for Clinicians 70(1): 7–30.

83.

Siegrist

(2021) Trust and risk perception: A critical review of the literature. Risk Analysis 41: 480–490.

84.

Simpkin

Schwartzstein

(2016) Tolerating uncertainty—The next medical revolution? The New England Journal of Medicine 375(18): 1713–1715.

85.

Slater

Long

Bettinghaus

Reineke

(2008) News coverage of cancer in the United States: A national sample of newspapers, television, and magazines. Journal of Health Communication 13(6): 523–537.

86.

Stryker

Solky

Emmons

(2005) A content analysis of news coverage of skin cancer prevention and detection, 1979 to 2003. Archives of Dermatology 141(4): 491–496.

87.

Sumner

(2012) American magazine winners and losers: 2001 to 2010. Journal of Magazine and New Media Research 13(2): 1–11.

88.

Suppes

(1970) A Probabilistic Theory of Causality. Amsterdam: North-Holland Publishing Company.

89.

Topper

Vaz

Marrone

Brahmer

Baylin

(2020) The emerging role of epigenetic therapeutics in immuno-oncology. Nature Reviews Clinical Oncology 17(2): 75–90.

90.

U.S. Cancer Statistics Working Group (2020) U.S. Cancer Statistics data visualizations tool, based on 2019 submission data (1999-2017). Available at: https://www.cdc.gov/cancer/uscs/dataviz/index.htm (accessed 29 September 2020).

91.

Venkatapuram

Marmot

(2009) Epidemiology and social justice in light of social determinants of health research. Bioethics 23(2): 79–89.

92.

Vineis

Wild

(2014) Global cancer patterns: Causes and prevention. The Lancet 383(9916): 549–557.

93.

Williamson

(2009) Probabilistic theories of causality. In: Beebee

Hitchcock

Menzies

(eds) The Oxford Handbook of Causation. New York, NY: Oxford University Press, pp. 185–212.

94.

Zehr

(2000) Public representations of scientific uncertainty about global climate change. Public Understanding of Science 9(2): 85–103.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.14 MB