Inaccuracy in Health Research News: A Typology and Predictions of Scientists' Perceptions of the Accuracy of Research News

Abstract

This article introduces an integrated inaccuracy typology to explore the prevalence of inaccurate news coverage of health research. This typology suggests that errors, omissions, and misinterpretations are three common types of inaccuracy; errors and omissions are objective, whereas misinterpretations are subjective. Objective inaccuracy involves errors and omissions in describing the background or substantive information about the research, such as how, when, where, and on whom research was conducted. Subjective inaccuracy entails misinterpretations as a result of a lack of expertise among journalists (e.g., misstating facts, errors in inferences, offering speculations as facts) or media's interest in profits (e.g., overemphasis on unique findings, overgeneralizations of findings, shifting emphases). For this study, coders analyzed objective inaccuracy, while scientists rated subjective inaccuracy. In turn, it identifies what can account for the variance in scientists' perceptions of inaccuracy in news articles citing their research. Objective and subjective inaccuracy offer significant predictors. Of the different types of objective inaccuracy, omissions of research methods represent a significant factor, whereas of the types of subjective inaccuracy, errors in inferences, overemphasis on uniqueness, and overgeneralizations of findings are all significant predictors.

Notes

¹For example, readers may regard findings from a study conducted by Harvard researchers or published in Science or Nature as more credible.

²Analyzing archived news is especially important in the modern era, when more people search online for health information (Cline & Haynes, 2001). More recent health news research similarly has used databases as sources for analyzed news (e.g., Moyer et al., 1995).

³Most of the research (95.5%) cited in the news reports was published in international journals.

⁴Subcategories include research on nutrition (e.g., dietary choices, nutrition and health, human nutrition needs, food sensitivities or allergies, and functional foods, the relation between food and health), addictions (e.g., alcohol, smoking), and so on. Within the disease/treatments topic, subcategories include research on diseases or medical treatment. Last, food safety research comprises subcategories such as food-borne illnesses, food ingredients (e.g., normal ingredients, chemical additives, pesticides) and health, food–disease links, and food additive safety.

⁵The procedures include the following steps: (1) defining each variable (see Table 1); (2) identifying levels and subcategories of each variable that are mutually exclusive; (3) developing coding schemes and coding forms; (4) training coders; (5) establishing pilot reliability using news articles not from the sampled articles; (6) having coders code one-fifth of the sample and calculate intercoder reliability (satisfactory, with Holsti's [1969] agreement formula ranging from .96 to 1); and (7) having coders split up and each code two fifths of the remaining sample.

⁶One week (August 13–15) after the packets were sent, a follow-up e-mail sent to these same scientists mentioned the mailed packet and provided another, electronic version, in case they preferred electronic files. On August 28, 2012, a first reminder was sent. A second reminder was sent 1 month later.

⁷A score of 0 indicated the absence of all four problems; a score of 4 indicated the presence of all four problems for each category.