To investigate media mentions in this observational study, annual counts of articles, excluding advertising, were examined from the archives of both the New York Times and the London Times. Changes in article mentions were compared to obesity prevalence patterns in the US from 1960 (the first year both US obesity prevalence and New York Times media mentions were available) to 2010 (i.e., fifty years) and in the UK from 1993 (the first year both UK obesity prevalence and London Times media mentions were available) to the same year, 2010. Although not fully representative of their countries, the newspapers considered were nevertheless influential and were also the only ones where all issues were fully-indexed online and available over a significant time period. The New York Times had a wide daily circulation (950,000 copies) and won more Pulitzer Prizes than any other newspaper [11]. Including pass-along readership and on-line readership, over 57 % of men and 52 % of women in the US read a daily newspaper [12].
One objective way to assess media coverage is to examine the number of times a particular word or term – like “lettuce” or “potato chips” – was mentioned in the average newspaper story. While this does not capture the manner or valence in which it used, it is an objective measure of usage and exposure. Furthermore, because the total number of articles can vary year-to-year, using the percentage of articles mentioning a term enables a fair comparison to be made across years.
A research assistant identified each term’s media mention by searching the online archives of non-advertising articles provided by the newspapers on their websites. All non-advertising articles mentioning the term counted, regardless of context. The research assistant’s search was validated by calling the New York Times and e-mailing the London Times to ask for counts of a sample of terms (they did not have time to complete the task for all terms) to compare with the counts retrieved using the online search tools provided on these newspaper’s websites. In all cases, they matched. The assistant used quotation marks if two words were separated by a space (e.g., “potato chips”) within the date range of January 1 to December 31 for each of 50 years from 1960 to 2010. This time scope was used because online data from the newspapers were available for this time period. The plural term was used for fruits and vegetables in order to increase the likelihood of an article’s use of the term as food. For example, an article with “oranges” was more likely to be an article about the fruit versus one with “orange,” which may be referring to the color. Terms were searched without considering compounds so that words like “corn” did not tally counts of “popcorn” or vice versa.
Fruit and vegetable mentions were examined as two primary categories of healthy foods, and sweet and salty snack mentions as two primary categories of unhealthy food [13, 14]. To determine the four categories that represent food consumption extremes – fruits, vegetables, sweet snacks, and salty snacks [13] – the researchers relied on reports from government agencies to determine which foods belong to each category. For example, the United States Department of Agriculture’s (USDA) most recent report on agricultural consumption [14] listed the top five fruits consumed in the US—oranges, apples, grapes, bananas, and pineapples—and the top five vegetables—lettuce, corn, onions, carrots, and cucumbers. The United States Department of Commerce (USDC) listed the top five sweet snack food transactions in the US—cookies, chocolate, candy, cake, and ice cream—and the top five salty snack foods—potato chips, tortilla chips, crackers, popcorn, and pretzels [15]. Since items like pretzels and crackers could be perceived as either sweet or salty, analyses were conducted with and without each item; the results did not differ.
A generalized linear model with a logit link was used to determine the association between annual obesity prevalence, as the dependent variable, and the percentage of articles mentioning vegetables, fruits, salty snacks and sweet snacks, as the independent variables. Actual obesity prevalence in the United States came from waves of the National Health and Nutrition Examination Survey published by the National Center for Health Statistics for select years from 1960 to 2010 [16].
The researchers investigated whether media mentions predated or followed obesity prevalence with three-year lagged time-series analyses [17]. Thus, these were analyses just as before, but with three-year lags for each of the independent variables in the model. Obesity prevalence was associated with the percentage of articles mentioning vegetables, fruits, salty snacks, and sweet snacks three years prior, thus testing whether media mentions followed obesity changes. The reverse was also run; that is, the number of articles mentioning each of the food categories was associated with obesity prevalence three years in the past, thus testing whether obesity changes preceded media mentions. Last, all of these models were compared simultaneously using seemingly unrelated regression (SUR) analyses to see which models had the highest R-Square estimates, investigating which models might best explain the relationships: those where changes in obesity follow changes in media mentions, or those where changes in obesity precede changes in media mentions. Models compared in the SUR analyses included the same number of independent variables.
To determine whether this phenomenon was unique to the US, an informative point of comparison would be a change over time relative to that of another English-speaking country, the UK. To accomplish this, similar media mention analyses were conducted for the London Times from 1993 to 2010 using British-equivalent terms (i.e., chips = crisps; candy = sweets; cookies = biscuits, etc.). The time period of 1993 to 2010 was selected because obesity and media mention data prior to this time period were not fully available.
Similar GLM analyses were employed using London Times article mentions and obesity outcomes in the UK, using the top snack items because they differed from the US in name only while top fruits and vegetables differed in actual foods consumed. Analyses include UK obesity prevalence since 1993 – the earliest official records reported on obesity prevalence in the UK.
Actual obesity prevalence in the UK came from the National Health Service Health Survey, Department of Health (2011) [18]. Because of the reduced range of obesity data for the UK, the US data were presented first.
Finally, media mention patterns of the word “obesity” and other comorbidity terms were reported. The researchers included mentions of diabetes, high blood pressure, high cholesterol, and hypertension because these comorbidities were highly associated with obesity [2].