PNS12 WHY ADVANCING DATA VISUALIZATION DOGMA IS OUR BEST TOOL IN DISAMBIGUATING RESEARCH FINDINGS

Abstract

A well-done data visualization (DV) can be worth more than a thousand words, and suboptimal visualizations can confuse or mislead viewers. Despite much published research on DV principles, defiance persists in the health economics and medical communities. This study curated principles of DV and assessed adherence to these tenants in recently published and highly circulated medical literature. DV experts were selected based on frequency of references to their works, regularity of contributions, and the publication of books or manuscripts on the topic. Tenants of DV repeated across experts comprised a master list: “The Dogma.” Literature published in 2017 by four top medical research journals (Nature, New England Journal of Medicine (NEJM), The Journal of the American Medical Association (JAMA), and The Lancet) were reviewed based on Altmetric (a measure of dissemination and influence) ranking by two researchers. DV techniques used in these articles were cross-checked with The Dogma. Modifications were developed for a sample of visualizations that were nonadherent. DV experts included Scott Berinato, Alberto Cairo, Stephen Few, Matthew Kay, Andy Kirk, Robert Kosara, Elijah Meeks, Cole Nussbaumer, Naomi Robbins, and Edward Tufte. The Dogma was structured into four categories: clarity, engagement, integrity, and messaging. Key concepts within these classifications included clutter, variety, color choice, scaling, chart selection, and annotation. Most manuscripts within the top 50 ranked by Altmetric had limited data visualizations, critical DV flaws, or no visualizations. Four manuscripts with similar but contrarily executed DV techniques were selected to evaluate in depth. Widely disseminated research is currently lacking an emphasis on DV. Adherence to DV techniques could improve the communication and exploration of relevant facts and new findings in medical research. The Dogma is not exhaustive, visualizing data – like health economics research – is a subjective art, and we did not incorporate methods of causal inference.