Graphical literacy: the quality of graphs in a large-circulation journal

We sought to characterize the quantity and quality of graphs in the Journal of American Medicine (JAMA), contrasting articles published in 1999 with those published in 2000 after the addition of a dedicated tables and figures editor. We also sought to compare the quality of graphs in JAMA with the quality of graphs in Annals of Emergency Medicine.

Two reviewers independently assessed all graphs in original research articles from 12 randomly chosen issues of JAMA, 6 each from 1999 and 2000, using a standardized abstraction form. We noted graph type, clarity, and completeness and identified internal discrepancies. We examined the graphs and articles to observe discrepancies with text, to observe efficiency of graph presentation, and to determine whether the graph portrayed unaggregated data. Results were compared with results from a previously published study of graphs from 18 consecutive issues of Annals of Emergency Medicine beginning in January 1998.

The 12 JAMA issues contained 56 research articles, with 64 graphs in the 37 articles that had graphs (28 in 27 1999 articles, 36 in 29 2000 articles). Simple bar or point charts (63%) predominated. We rarely encountered internal errors (8%), contradictions with text (3%), numeric distortion (6%), lack of visual clarity (5%), nonstandard graphing conventions (11%), or extraneous decoration (0%). Graphs generally defined all symbols (98%), but 31% were not self-explanatory; that is, despite knowing the study's design and reading the figure's legend, we could not unambiguously interpret the graph. Fourteen percent contained redundancies. Graphs infrequently portrayed by-subject data (9%) or advanced features (15%) such as pairing, symbolic dimensionality, or small multiples. Forty-eight percent (21/44) of graphs did not illustrate the underlying distribution, 48% (26/54) did not depict important covariates, and 67% (14/21) did not portray pairing inherent in the data. There were no differences between 1999 and 2000 graphs, although we noted more graphs in the 2000 issues. Graph quality was similar in Annals of Emergency Medicine and JAMA, but graphs were more common in the original research articles in JAMA. Although univariate displays predominated in both publications, there were more bivariate displays in Annals of Emergency Medicine but fewer advanced graphic features.

The graphs in JAMA were similar to those in Annals of Emergency Medicine and, although generally clear and without errors, often failed to depict detailed data. Authors and editors could improve data presentations by incorporating graphic formats that depict stratified, detailed data.