| Popis: |
Background: Communicating data and research to relevant stakeholders is essential to ensure impact in practice. This is especially true for global ‘wicked’ healthcare challenges, such as antimicrobial resistance (AMR). Data visualizations can transform the increasingly growing data into more comprehensible information for exploration and communication purposes. In general, there is little guidance on selecting and designing data visualizations. This study aims to lay the ground for improving visual communication of data and research on AMR and other wicked healthcare challenge by creating an overview of the visual dictionary (incl. vocabulary and design space) for AMR research. Methods: Data visualizations in published research articles on antimicrobial stewardship, infection control and institutional surveillance (identified by a prior science mapping study) were studied. One visualization per article was randomly extracted of a random sample of 180 articles. The extracted data visualizations were assessed on content (i.e. vocabulary) and design space (based on Munzner's nomenclature and categorization). Additionally, visualization errors, chart junk, and quality were assessed. Two researchers (CL, JK) independently assessed half of the visualizations. To calculate the interrater reliability 10% was analysed in duplicate, and answers for the remaining 150 visualizations were double-checked. By combining the vocabulary and design space, an AMR research visualization dictionary was created including dos and don’ts in the use and design of data visualization. Findings: The three most used attribute combinations (representing the vocabulary) were time and antimicrobial consumption (n=21), time and incidence (n=18), and antimicrobial consumption and antimicrobials (n=12). Regarding the visual design space, colour and shape channels were frequently used, and time was always visualised with lines. Bar charts (n=54, 36.0%) and line charts (n=42, 28.1% n=42, 28.1%) were the most commonly used of the fourteen visualization types identified. 55.3% of visualizations were interpretable without additional text and visualization quality was rated 3.6 on average (scale 1=poor to 5=good , SD: 1.2). Dos: • Clarify what is shown. E.g. use titles, legends, axes names, labels, annotations or captions, and explain abbreviations. • Prefer colours over shape for grouping/stratifying (keeping in mind colour-blindness and black-white printing) and ensure that groups are distinguishable. • Prefer length over area/volume to compare sizes. • Show even and equal axes for readability and comparison (for a single visualization and across combined or faceted visualizations). Don’ts: • Colour scheme mismatches. E.g. using categorical colours for ordered attributes and vice versa, and using non-intuitive colour schemes. • Use double y-axis. • Hide data points through overlaps. • Use chart junk (e.g. unnecessary 3D, shadows, or colours). • Overcrowd (e.g. channel or attribute overflow). Discussion: This study provides an overview of the use and design of data visualization in the field of antimicrobial stewardship, infection control and institutional surveillance, including visualization dos and don'ts. The field of AMR is under constant change and heavily influenced by new data-driven technologies increasing the need for translating these data into comprehensible information supported by high-quality visualizations. Results of this study can serve as the basis to optimize future communication in AMR research and practice matching visualizations with the visual dictionary of the target group. |
