DDHA 8703 Week 8 Visualization Critique Example

Four Measures, Two Axes, One Flattering Starting Quarter: A Critique of a Quarterly Quality Dashboard Tile and a Rebuild Around the Comparison a Manager Has to Make

Student Name

Walden University

DDHA 8703: Advanced Health Analytics and Data-Driven Decision Making

Instructor Name

Month Day, Year

Four Measures, Two Axes, One Flattering Starting Quarter: A Critique of a Quarterly Quality Dashboard Tile and a Rebuild Around the Comparison a Manager Has to Make

The Chart and the Decision It Should Support

The chart under review is a tile on Northgate Health's quarterly nursing quality dashboard, one tile per inpatient unit. Each tile shows four measures over six quarters: two rates, falls and hospital-acquired pressure injuries, each per thousand patient days, as bars against a left axis, and hand hygiene compliance and the percentage of surveyed patients rating nurse communication at the highest level as lines against a right axis scaled from 60% to 100%. Each measure has its own color, and each quarter's bars are shaded by a fifth color for the year.

The tile was introduced to support one decision: whether a unit's fall rate warrants a formal falls-prevention review this quarter. The system's rule is that a unit whose fall rate is above the system's upper target for two consecutive quarters is reviewed. That is the comparison a manager needs to make: this unit's fall rate, over recent quarters, against the target. Every design choice below is judged by whether it helps or hinders that comparison.

What the Design Does to the Decision

The second axis invites a comparison the data do not support. With bars on the left and lines on the right, the eye reads crossings and parallel movements between, for example, falls and hand hygiene, as if they were related. Where a line crosses a bar has nothing to do with the data; it depends only on how the two axes were scaled. Isenberg et al. (2011) studied charts that combine two scales experimentally, and the design they examined asks readers to hold two unrelated scales in mind at once, which is exactly what a busy manager will not do. On this tile, a manager looking at a quarter where falls rise above the hand hygiene line may conclude that hygiene and falls are linked, which the chart cannot show.

Color separates categories nobody needs separated. Four measure colors and a fifth for the year make five hues on a small tile, and the year shading adds nothing because the quarters are already labeled. Meanwhile, the one thing that matters for the decision, the system's upper target for falls, is not shown at all. The manager has to remember the target and compare it with a bar's height against an axis shared with pressure injuries.

The time span flatters the trend. The six quarters begin with the fourth quarter of 2023, when many units had unusually high fall rates during a winter surge of respiratory admissions. Starting there makes nearly every unit's fall rate appear to decline. A span of eight quarters, or one that starts at a typical quarter, would show several units as flat or rising.

Bars also encode the rates by length from a zero baseline, which is appropriate, but placing two rate measures side by side in each quarter halves the space for each and puts falls next to pressure injuries, a comparison the decision does not need. Cleveland and McGill (1984), testing how accurately people judge graphical elements, found that judging position along a common scale is more accurate than judging length or area across separated elements. The tile asks the manager to compare bar heights across quarters while visually separated by other bars.

The Rebuilt Version

The rebuilt tile shows a single measure, this unit's falls expressed against patient days, drawn as points joined by a line across eight quarters on one axis that starts at zero. A shaded horizontal band marks the range between the system's lower and upper targets, so that a quarter above the band is visible at a glance. A thin gray line shows the system median for comparison. The unit's line is drawn in one dark color; nothing else on the tile has color except the target band, in a light neutral shade.

The ordering follows the decision. The two most recent quarters are labeled with their values, because the review rule depends on them. A short text line under the chart states the rule's result: for example, "Above upper target in two consecutive quarters: review due."

What was removed is as important as what remains. Pressure injuries, hand hygiene and nurse communication move to their own small tiles in a row below, each on its own axis with its own target band, so that each can be read alone. No measure shares an axis with another measured in different units. The year shading is gone. Dowding et al. (2015), reviewing clinical dashboards, noted that dashboards support improvement when they make the information needed for action easy to see; the rebuilt tile is designed so that a manager can make the review decision in a few seconds.

Testing the Rebuild

Before the rebuilt tile replaces the old one, the managers who will rely on it should try it first. Six unit managers will be shown each version for their own unit and asked whether a review is due. If the rebuilt tile is better, more of them should answer correctly and in less time. The test is small, but it judges the design by the decision it supports rather than by anyone's preference.

What the Rebuild Does Not Solve

The rebuild clarifies the comparison but cannot fix the data behind it. Fall rates on small units vary considerably from quarter to quarter by chance, so two quarters above the target may reflect noise. A control chart, with limits based on each unit's own variation, would address that better than a fixed band, and the critique recommends testing one. The rebuild also does not say why falls rose, which belongs to the review the tile triggers.

Conclusion

The original tile placed four measures on two axes, used five colors and began at a quarter that flattered the trend, and none of those choices served the decision it was built for. The rebuilt tile shows one measure against its target on one axis over eight quarters, with the rule's result stated in words, and moves the other measures to tiles of their own.

References

Cleveland, W. S., & McGill, R. (1984). Graphical perception: Theory, experimentation, and application to the development of graphical methods. Journal of the American Statistical Association, 79(387), 531-554. https://doi.org/10.1080/01621459.1984.10478080

Dowding, D., Randell, R., Gardner, P., Fitzpatrick, G., Dykes, P., Favela, J., Hamer, S., Whitewood-Moores, Z., Hardiker, N., Borycki, E., & Currie, L. (2015). Dashboards for improving patient care: Review of the literature. International Journal of Medical Informatics, 84(2), 87-100. https://doi.org/10.1016/j.ijmedinf.2014.10.001

Isenberg, P., Bezerianos, A., Dragicevic, P., & Fekete, J.-D. (2011). A study on dual-scale data charts. IEEE Transactions on Visualization and Computer Graphics, 17(12), 2469-2478. https://doi.org/10.1109/TVCG.2011.160