Visualizing Multimorbidity in Chronically Ill Populations Using UpSet Plots

To the editor: We read with great interest the recent article in this journal by Dreisbach et al. in which they analyzed and clustered unrelieved symptoms in patients with one or more chronic conditions.¹ In particular, Figure 1, Panel A, which illustrates the overlap of chronic conditions, caught our eye. We would like to use this as an opportunity to point out a novel and innovative approach to visualize intersecting sets which we find particularly useful for illustrating multimorbidity in chronically ill populations.

Figure 1. Recreation of a Venn Diagram as UpSet Plot.

Panel A: Figure “Chronic condition diagnoses in study participant cohort (N=14,127)” reproduced from the Dreisbach et al. Journal of Pain and Symptom Management, Volume 64, Issue 6, P555–566 (permission pending). Panel B: Recreation of the Venn diagram from Dreisbach et al. in the form of an UpSet plot. Abbreviations: COPD, Chronic obstructive pulmonary disease.

As the authors pointed out, an increasing number of the ageing American population is affected by multiple chronic conditions. The most common approach to visualize the intersection of multiple conditions in a population, which was also employed by the authors (Figure 1, Panel A), is using Venn diagrams. Originally proposed by the eponymous John Venn in 1880 in the context of logical reasoning, Venn diagrams have found widespread uptake in all sciences.² However, they are exceedingly difficult to interpret when visualizing more than three intersecting sets. Furthermore, Venn diagrams are commonly constructed without scaling the plot area in proportion to the size of the intersecting sets, which can lead to misrepresentation.

To overcome these shortcomings, in 2014 Lex et al. proposed the use of UpSet plots which visualize set intersections in matrix layout providing bars and columns to allow readers to visually and intuitively grasp the relative size of set intersections.³ To illustrate this alternative visualization tool, we recreated the Venn diagram by Dreisbach et al. as UpSet plot (Figure 1, Panel B). We find this UpSet plot to be more useful in that it immediately displays which intersecting chronic conditions are most prevalent in the sample. In this case, cancer and type 2 diabetes make up the bulk of the trial population. While the Venn diagram spatially overrepresents the intersecting sets in the center, the UpSet plot unmistakably visualizes that most of the study population only experienced a single chronic condition. In addition, UpSet plots can have many more sophisticated features e.g., displaying the attributes of intersecting sets. UpSet plots can be easily constructed using the UpSetR package for R software (R Foundation for Statistical Computing, Vienna, Austria), or using an online application.^4,5

In summary, we believe that UpSet plots can be a valuable addition to the graphical toolbox of clinical researchers who study multimorbid populations in various fields, including internal medicine, geriatrics and palliative care.