Health Data Visualization Literacy Skills of Young Adults with Down Syndrome and the Barriers to Inference-making
Rachel Wood, Jinjuan Heidi Feng, Jonathan Lazar · 2024 · ACM Transactions on Accessible Computing · doi:10.1145/3648621
Summary
This exploratory study is the first to examine how young adults with Down Syndrome read and interpret health data visualizations (HDVs). The researchers conducted semi-structured interviews with ten participants (ages 16-29) as they viewed six different health visualizations related to weight management: a body fat table, daily steps bar graph, macronutrients stacked bar chart, weekly walking line graph, dual y-axis activity intensity graph, and a food healthiness scatterplot. The study framework examines three progressive stages of graph reading: (1) reading the data—identifying visualization elements like titles, axes, and values; (2) reading between the data—comparing values and identifying patterns; and (3) reading beyond the data—making inferences and connecting information to outside knowledge. The research was co-developed with a self-advocate with Down Syndrome who helped ensure materials and procedures were accessible. Analysis of nearly 700 direct participant quotes revealed a "downward cascade" in performance: participants achieved 79.3% success at identifying elements in stage one, but this dropped to 53.6% when comparing and connecting information, and fell further to 37.9% when making inferences. The researchers identify this pattern as resulting from compounding design-based and task-based barriers that increasingly tax working memory and cognitive flexibility as visualization complexity increases.
Key findings
Participants demonstrated strong visuo-spatial abilities during initial identification tasks, confirming that people with Down Syndrome can effectively perceive and compare visual elements. Six of ten participants identified more than 75% of HDV components. However, performance declined dramatically when tasks required holding information in working memory while making connections or inferences. Several design-based barriers emerged: invisible number lines in ranges (e.g., "18-25" implying 19, 20, 21...) confused participants; abbreviations like "K" for thousand were misinterpreted as kilogram; color coding had no meaning for 30% of participants; and vertical/transposed text was difficult to read. Participants showed a preference for concrete "token" language (e.g., "All the Steps You Walked This Week") over abstract "type" language (e.g., "This Week's Activity"). Icons and images helped when they reinforced familiar concepts with clear labels, but caused confusion when meanings shifted across visualizations or when partial icons (like half a shoe) represented fractional values without explanation. The stacked bar chart achieved 95% accuracy in color mapping because it used layered encoding—icons, colors, and labels all reinforced each other. Participants often relied on pre-existing health knowledge and familiar skills (reading, pattern recognition) when uncertain how to interpret abstract visualization features.
Relevance
This research has significant implications for anyone designing health technologies, dashboards, or data visualizations intended for diverse audiences. The twelve design suggestions provide concrete guidance: use specific/concrete language over abstractions; avoid abbreviations or use multi-letter versions (e.g., "Sat" not "S"); include definitions for unfamiliar terms; layer multiple encoding channels (color + icon + label) to reinforce meaning; provide step-by-step guided walk-throughs; and support question generation with prompts like "What do you want to know?" The findings extend beyond Down Syndrome to benefit anyone who struggles with numeracy, data literacy, or health literacy—roughly one-third of American adults. The study demonstrates that accessibility barriers in data visualization compound over time: early misunderstandings become embedded in mental models and cause larger errors during later inference-making stages. For practitioners, this underscores the importance of testing visualizations with diverse users at multiple stages of comprehension, not just initial recognition. The co-design methodology with a self-advocate also models inclusive research practice that centers the expertise of people with intellectual disabilities.
Tags: Down Syndrome · data visualization · health informatics · graphicacy · cognitive accessibility · intellectual disability · health literacy · inclusive design