Making Sense of Cancer News Coverage Trends: A Comparison of Three Comprehensive Content Analyses

Abstract

Cancer stories (N = 5,327) in the top 50 U.S. newspapers were analyzed by a team of four coders and the results were compared with the earliest analyses of this type (from 1977 and 1980). Using cancer incidence rates as a comparison, three cancers were found to be consistently underreported (male reproductive, lymphatic/Hodgkin's, and thyroid) and four cancers were found to be consistently overreported (breast, blood/Leukemia, pancreatic, and bone/muscle). In addition, cancer news coverage consistently has focused on treatment rather than on other aspects of the cancer continuum (e.g., prevention), portrayed lifestyle choices (e.g., diet, smoking) as the most common cancer risk factor, and rarely reported incidence or mortality data. Finally, the data were compatible with the idea that personalization bias (e.g., celebrity profiles, event coverage) may explain some news coverage distortions.

This research was conducted with funding from the National Cancer Institute, CA98437-01A1.

Notes

¹Within the cancer research community, cancer is often discussed in terms of a “cancer continuum.” The continuum represents the cancer process from beginning to end, starting with prevention and moving forward to detection, treatment, survivorship, and end-of life care (e.g., National Cancer Institute, 2007). The purpose of the cancer continuum is to help researchers organize their efforts and to draw attention to the fact that cancer care is often disproportionately treatment focused.

²The search incorporated measures of recall and precision to estimate reliability and validity. Stated briefly, researchers constructed an initial Lexis-Nexis search term and then modified it based a series of trial runs. During the runs, researchers coded a small sample of articles retrieved by versions of the search term to calculate how many were “valid” stories. This data were used to improve the term until an optimal search term had been constructed (for a further description of the search term, see Stryker, Wray, Hornik, & Yanovitzky, 2006). The full search term as well as a list of the newspapers included in the study are available at the lead author's website: http://web.ics.purdue.edu/~jdjensen/

Note: N = the number of valid news stories. Coverage data were not available for all cancers in 1980. Specific cancers were counted in this analysis if they were mentioned at all (i.e., they did not have to be the major theme or topic). A story could mention more than one type of cancer.

Coverage rank = A hierarchical score conveying frequency of mention in the news, from most frequent (1) to least frequent (15).

Incidence rank = A hierarchical score conveying the number of new cases each year, from most new cases (1) to least new cases (15).

Difference Score = A score reflecting the difference between coverage and incidence rank; positive scores mean that a cancer was depicted more frequently in the news than it actually occurred.

∗Some cancers were combined in 1977, but separated in 2003. For example, in 2003, bladder cancer was eleventh in coverage and sixth in incidence, whereas kidney cancer was twelfth in coverage and nineth in incidence.

³The comparison was not perfect, as different data reporting practices and content analytic schemes made the datasets somewhat inconsistent. The 1977 data included an “other cancers” category that altered the coverage rankings (and hence the discrepancy scores). The published 1980 data was incomplete (i.e., the fourth and sixth most covered cancers were not identified) and only included information for five cancers. The 2003 data separated some cancers that previously were grouped together. To make the datasets more comparable, some cancer categories were cut from 1977 (i.e., the other cancer category) and a few cancers were combined in 2003 (e.g., bladder/kidney). Even with all these differences, several patterns were identified upon comparison.

⁴Breast cancer consistently has been one of the most covered cancers; for example, roughly one in four cancer news articles in 2003 mentioned breast cancer. Of course, breast cancer is also a very common form of cancer, so the breast cancer distortion has been relatively small over time. This should not mask the reality that breast cancer is substantially more likely to be reported than other types of cancer.

⁵Bone/muscle cancer coverage is somewhat difficult to interpret. In 1977, bone/muscle cancer had a difference score of plus nine. In 2003, the difference score was considerably smaller; however, readers should interpret the second number with some caution. Bone/muscle cancer is very rare and it can be difficult to meaningfully distinguish the incidence rates of rare cancers (i.e., because of a floor effect). We chose to label bone/muscle cancer as fifteenth in incidence (because that is the total number of cancers we coded for), but it actually is one of several cancers that occurred with extremely limited frequency in 2003. The news coverage rank is different. Bone/muscle cancer was, in fact, the fourteenth most reported cancer in 2003. So, for bone/muscle cancer, the distance between coverage and incidence rank is somewhat subjective. A case could be made that bone/muscle cancer was, for example, twenty-fourth in incidence. But that number would significantly inflate the difference score, disguising the fact that cancers ranking fifteenth to twenty-fourth in incidence are all very rare. Regardless of how one calculates bone/muscle cancer incidence, that cancer site consistently has been overreported in the news.

Note: Cancer continuum components had to be a major topic/theme of the article to be included in this count.

∗In 1977/1980, researchers coded for “coping.” Coping is now thought of as an aspect of several parts of the cancer continuum, including survivorship and end of life care. The present study aggregated survivorship and end of life to form a “coping” variable. Survivorship and end of life are also reported separately to allow readers to interpret the data as they see best.

⁶Some readers may be interested in the percentage of all stories (n = 5,327) that included a specific cancer risk factor (to help visualize the data): lifestyle (11.7%), demographics (10.0%), genetic/heredity (5.7%), environmental/occupational (5.6%), and medical (4.8%).

Note: Cancers are listed vertically by incidence rates (most frequent first). N _f  = the number of news stories about famous people with cancer. N _p  = the number of news stories about nonfamous people with cancer. N _e  = the number of stories that were about cancer fundraisers or events. Specific cancers were counted in this analysis if they were mentioned at all (i.e., they did not have to be a major topic/theme). A story could mention more than one type of cancer.