Multimodal feedback: color, icon and name in diagnosis

You submit your guess in the diagnosis game. The screen lights up a green rectangle. No text. No icon. Just the color. In less than a second you have to decide: did I get the diagnosis right, am I in the same group, or only in the same chapter? If your vision confuses green and red, or if the screen is in high contrast mode that desaturates colors, the rectangle becomes an opaque block. The clinical reasoning was underway. The feedback demanded color discrimination. This is the territory of multimodal proximity feedback: color, icon and written name together, so that diagnostic proximity reaches whoever is playing without depending on a single sensory channel.

Multimodal feedback, in this text, is the deliberate combination of visual channels (and, in passing, of what a screen reader needs to hear) to communicate the same proximity level. In games and clinical teaching platforms, proximity usually maps how close the guess came to the answer in the diagnostic hierarchy. If this information travels only by color, part of the class is left out. The focus here is the design of the immediate return after the attempt. It is not a general checklist of medical software standards, nor a guide to visual assistive technology products, nor the full case flow with a screen reader. It also does not exhaust the long pedagogical explanation after the case closes.

Key points

• Diagnostic proximity that depends only on color excludes people with color vision deficiency, people using desaturated high contrast and people reading the feedback under time pressure or high cognitive load.
• Redundant signaling (color + icon + written name) is a classic principle of accessible design: meaning does not live in a single channel.
• In medical education, color vision deficiency is prevalent enough to stop being an anecdote; tasks that ask only for hue generate more errors.
• Working memory and reading speed call for short, stable, legible labels, not just a colored flash.
• In NEXO, the guess card combines color, icon and the name of the proximity level; the accessibility statement describes independence from color in proximity feedback. No claim of effectiveness.

What multimodal proximity feedback is

Diagnostic proximity is the return that says how close your guess came to the target answer. In systems based on a classification hierarchy, the typical levels range from the exact match to the distant one, passing through category, group and chapter (the naming varies by product). The student uses this signal to calibrate the next attempt: stay on the same axis, go up a level or abandon the diagnostic family.

Multimodal feedback, here, means that the same level appears in more than one channel at the same time. Color (bar, background, badge). An icon with a distinct shape per level. A short written name ("exact", "category", "group", "chapter", "distant" or translated equivalents). Optionally, accessible text for assistive technologies. The golden rule: if you remove the color, the level must still be identifiable. If you remove the icon, the written name is still enough. If the name disappears, color and shape together still communicate to those who see.

This differs from useful feedback in clinical cases, which deals with the content of the explanation (right/wrong, criterion, self-explanation). Here the object is the immediate channel of the proximity level, before or alongside the long explanation. It also differs from dual encoding in studying with image and text: that line is about learning content with two representation systems; this one is about communicating attempt status with redundant channels in the interface.

Not to be confused with the axis of the post about digital accessibility in medical software (general criteria for clinical software) nor with the one about assistive technology for physicians with visual impairment (screen readers, magnifiers, products). The problem in this text is specific: the proximity signal that flashes after the guess. If that signal is only chromatic, the game measures color discrimination, not just reasoning.

For the student, the practical question is: "did I read the level or did I guess the color?". For whoever designs the playable case, the question is: "does each proximity level have its own shape, color and label, stable across screens?". The rest of this text translates that into barriers, remedies and evidence, with links to differential diagnosis with clues and to case-based learning.

Why color alone fails

Color vision deficiency (CVD) is common enough to enter the design of any teaching app. In European Caucasian populations, inherited red-green deficiency affects about 8% of men and about 0.4% of women (Birch, 2012); in men of Chinese and Japanese ethnicity the range cited in the same survey is between about 4% and 6.5%. In medical school classes this is not rare: in any mixed cohort, several colleagues may be reading your "correct green" as a tone close to "error red".

In clinical practice and in study, the literature describes difficulties when the task requires hue as the only criterion. Physicians with CVD report problems observing signs that depend on color; reviews and surveys call for screening and counseling in training (Spalding, 1999; Spalding, 1999, students). In histopathology, professionals with CVD made more errors identifying slides, with a correlation to the severity of the deficiency (Poole et al., 1997). In students, color-dependent clinical and laboratory images led to more errors; recommendations include awareness and accommodations (Dhingra et al., 2017). A recent narrative review in health professions education reinforces gaps in screening and institutional support and points to universal design, grayscale images and alternative teaching strategies (2026 review, The Clinical Teacher).

The parallel for the diagnosis game is direct. If "exact" is only green and "distant" is only red, the student with CVD may invert or merge the levels. If the intermediate levels are yellows and close oranges, the proximity scale becomes an unreadable gradient. Operating system high contrast, screen filters, a washed-out projector and extreme ambient light worsen the same problem for those without CVD: color stops being reliable as the only channel.

There is also the criterion for the use of color in web accessibility: color should not be the only visual means of conveying information, indicating an action, prompting a response or distinguishing a visual element (W3C, Use of Color). This post is not a compliance tutorial; the principle is enough: proximity shown only in color fails the conceptual test before any audit.

Signs that the feedback is only in color: a colored badge with no textual label for the level; a scale legend that exists only in the tutorial and disappears in the game; two levels with identical shapes and close hues; an animation that changes only the background tone for fractions of a second; dark mode or high contrast that equalizes the colors without compensating with text or an icon.

When this happens, the student "gets wrong" the reading of the feedback and adjusts the next hypothesis poorly. The post about a knowledge or reasoning gap helps separate failures; here the additional hypothesis is a channel failure: the proximity did not arrive.

Three channels: color, icon and written name

Redundant encoding is the practice of mapping the same category to more than one visual property: color and shape, color and texture, color and label. In scientific visualization, classic columns call for CVD-safe palettes and alternatives to color alone (Wong, 2011, Color blindness; Wong, 2011, Avoiding color). Perceptually uniform, CVD-friendly color maps reduce misuse of color in scientific communication (Crameri et al., 2020). Packages that add patterns beyond color in scatter plots show legibility gains under CVD simulation (ScatterHatch / Guha et al., 2022).

On the guess card of a diagnosis game, the practical translation is the triad:

• Color: side bar, badge background or outline, with distinct hues that also differ in luminance (not just in hue).
• Icon: a different shape per level (for example check, target, filled circle, empty circle, X), for those who see shapes better than colors.
• Written name: a short, stable word for the level, in the interface language, next to the icon.

Why three and not two? Color + icon without text fails for those who zoom in a lot and lose the small icon, or for those who describe the state out loud ("I'm in group"). Color + text without an icon still works well, but the icon speeds up visual scanning when several guesses are stacked. Icon + text without color still works in monochrome high contrast. The triad covers the three most common failure modes.

Good design practices for the triad:

• One exclusive icon per level; do not repeat the same shape in another color.
• Short name (one word or two); avoid internal jargon in the badge.
• Color with a difference in luminance, not just in hue (test in grayscale).
• Stable order: icon + name in the same corner on every guess.
• Permanent accessible legend (not only in the first onboarding).

In types of games for medical students and in gamification in medical education, immediate feedback is a frequent feature. Illegible immediate feedback is not feedback: it is chromatic noise. The desirable difficulty should be in the differential, not in decoding the badge (see also desirable difficulties).

Working memory and reading speed

Even with typical color vision, feedback competes with working memory. After submitting the hypothesis, the student is still holding the list of differentials, the last clue and the doubt about the next guess. A color flash without a label forces the scale legend to be recalled from memory ("was green category or exact?"). This is extraneous load: processing that does not teach Medicine.

Multimedia learning principles applied to medical education call for reducing extraneous processing, managing essential processing and supporting generative processing (Mayer, 2010). Applying multimedia design principles to slides improved short-term retention in medical students in one trial (Issa et al., 2011). The parallel for the proximity badge: words and figure (icon) aligned, close together and without useless decoration speed up reading of the status.

Feedback reading speed matters under time pressure. If the written name is a long sentence ("you are in the same taxonomic group as the correct answer"), the student under pressure reads only the color. Prefer a short label + optional expandable explanation. If the only channel is a 300 ms toast, those who read more slowly (second language, dyslexia, extreme zoom) lose the signal. Keep the guess card persistent in the list; do not rely on an ephemeral message.

Quick load questions for proximity feedback: does the level fit in one visible word without hover? Is the scale legend one tap away, not three menus away? Do several guesses on screen use the same visual pattern? Is the level name in the same language as the rest of the interface? Is there a clear difference between "not submitted yet" and "distant level"?

Calibrating confidence after the case also depends on having read the right level. The post on calibration of confidence in diagnosis deals with judgment; from the multimodal angle, the precondition is that the channel delivered the status without ambiguity.

Contrast, high contrast and legibility

High contrast in the system and dark themes are allies for text legibility, but they can destroy palettes designed only in hue. A "success" green and an "error" red with similar luminance become two similar grays. That is why the triad matters: when color collapses, icon and name survive.

Practical guidance without turning this text into a contrast audit: test the scale in grayscale and in CVD simulation (protanopia and deuteranopia); ensure sufficient contrast between the name and the badge background; do not use only red and green as a single pair of extremes, add shape and label; avoid conveying severity only through saturation without text; keep the icon with a stroke thick enough for small screens.

People who enlarge the interface (partial low vision) may see only a piece of the badge. A short written name and an adjacent icon increase the chance that the level fits in the viewport. Magnification and screen readers are covered in more depth in visual assistive technology texts; here the point is only this: multimodal proximity reduces dependence on seeing the screen "whole and in color".

Post-case metacognition ("did I read the right level?") becomes a habit. The post on metacognition in Medicine goes deeper into monitoring; from the multimodal angle, include the question "was the feedback legible or did I guess the color?".

People who use a screen reader do not "see" the green bar. The written name of the level must exist in the interface as text (or as an equivalent accessible label), not only as colored pixels. Decorative icons can be hidden from the reader if the name already states the level; if the icon is the only signal, it needs alternative text. This passage does not replace a full case flow with a screen reader: that post goes clue by clue. The minimum here is: multimodal proximity includes an explicit verbal channel, not only a visual one. In practice, the visible label of the level ("group") must be the same (or semantically equivalent) as what the reader announces. Color alone is invisible to the reader; a written name is the structural remedy.

Evidence in design and clinical training

On the design and visualization side, the message converges: do not rely on hue alone; use luminance, shape, pattern and label; test under CVD (Wong, 2011; Crameri et al., 2020; Guha et al., 2022). On the health training side, CVD affects color-based tasks and the institutional response is still uneven (Spalding, 1999; Dhingra et al., 2017; 2026 review). On the educational feedback side, clear and well-calibrated feedback improves the use of information; meta-analyses and reviews in education reinforce the power of feedback when it is understandable (Wisniewski et al., 2019; definition and role of clinical feedback in van de Ridder et al., 2008). In formative multiple-choice tests, going beyond right/wrong improves the use of the feedback (Ryan et al., 2020).

Putting the threads together: multimodal proximity is inclusive design and also pedagogical design. An illegible level wastes the attempt. A legible level allows the next step in reasoning: narrow the differential, change chapter or confirm the correct answer. In ICD-10 and the right code, the hierarchy matters for searching; in the game, the same hierarchy becomes proximity feedback and must be read without guessing the color.

Universal guidance in medical education for students with CVD includes avoiding color as the only marker in slides and exams, using arrows and numbers, and offering grayscale versions when structure (not hue) is what matters (practical synthesis in Meeks and colleagues on universal design and CVD in medical education; see also Rubin et al. on grayscale histology cited in that line). The game's proximity card is the digital equivalent of that care.

Practical table of channels

Use the table when reviewing a diagnostic game, a simulator or a case bank with proximity levels.

Channel	What it communicates	It fails if...	Remedy
Color	Quick scan of the level	CVD, high contrast, washed-out projector	Add icon and name; vary luminance
Icon / shape	Level without relying on hue	Identical icons; stroke too thin	Exclusive shape per level; thick stroke
Written name	Legible and announceable level	Color only; ephemeral toast; jargon	Persistent short label; same language as the UI
Accessible text	Level for screen reader	Pixels only; generic alt	Level name in the accessibility tree
Scale legend	Convention memory	Onboarding only	Legend always one tap away

If one channel fails and the others cover the meaning, the design is truly multimodal. If all channels collapse together (for example, icon and text hidden behind a single colored square), go back to the drawing board.

How NEXO fits in

In NEXO, the case of the day flow presents fictional cases with a real clinical presentation and up to six progressive clues. After each guess, the attempt card shows the proximity level with color, icon and written name (per-level setting in the attempt component). The Accessibility page describes color independence in proximity feedback: levels with color and text labels for people with color vision deficiency. No sentence here claims effectiveness; it describes verifiable interface features.

Ecosystem resources linked to the study: NEXO Researcher to go deeper into topics; NEXO Closer when the focus is closing gaps; usage limits in Economy. The text what NEXO is places the game on the clinical teaching map.

How to use the case of the day with a multimodal feedback mindset (no promise of results):

• When you receive the level, read the written name before looking only at the color.
• Compare two attempts in the list: icon and label should be enough in a mental grayscale.
• If you train with the system CVD filter or a high-contrast theme, check whether the level is still obvious.
• In the case explanation, separate "I read the feedback" from "I understood the clinical criterion".
• If you adapt cases for a class, copy the color + icon + name triad into your own proximity materials.

Honest comparisons: NEXO shows proximity with redundant channels on the attempt card; banks that mark only with green/red demand more color discrimination. The choice of environment depends on the goal of the day. The principle travels: measure reasoning, not the ability to tell hues apart under pressure.

Final checklist before publishing a new feedback mode in any clinical teaching app:

• Does each level have its own color, icon and name?
• In grayscale, do the levels still separate?
• Does the name persist in the attempts list?
• Does the screen reader announce the level?
• Is the scale legend available without a treasure hunt?
• Did you avoid conveying meaning only through red versus green?

Frequently asked questions

Is multimodal feedback the same thing as dual encoding?

No. Dual encoding in studying combines image and text to learn content. Multimodal proximity feedback combines channels (color, icon, name) to communicate the status of an attempt. The two can coexist in the same app, with different roles.

Do I need to drop color in feedback?

No. Color remains useful for quick scanning. The request is not to leave color alone. Keep color and add an icon and a written name.

Does this only matter for people with color blindness?

No. High contrast, projectors, ambient light, haste and high cognitive load also degrade reading by color alone for people with typical vision. CVD makes the failure more predictable, but it is not the only scenario.

Doesn't the written name make the interface "cluttered"?

Short, stable labels reduce mental rework. One word on the badge usually costs less attention than forcing the student to remember the scale legend. Clutter is decoration without function; a level name is function.

How does this relate to the explanation after the diagnosis?

They are different layers. Multimodal proximity delivers the immediate level. The pedagogical explanation after closing deepens the clinical criterion. This post does not exhaust the second layer.

Can I use NEXO just to train reading proximity?

As a rehearsal environment with limited clues and feedback after the guess. Use it to practice the habit of reading name and icon; not as proof of competence nor as treatment.

Play the case of the day.

References

1. Birch J. Worldwide prevalence of red-green color deficiency. J Opt Soc Am A. 2012. PMID 22472762.
2. Wong B. Points of view: Color blindness. Nat Methods. 2011. PMID 21774112.
3. Wong B. Points of view: Avoiding color. Nat Methods. 2011. PMID 21850730.
4. Crameri F, Shephard GE, Heron PJ. The misuse of colour in science communication. Nat Commun. 2020. PMID 33116149.
5. Guha T, Fertig EJ, Deshpande A. Generating colorblind-friendly scatter plots for single-cell data. Elife. 2022. PMID 36524718.
6. Spalding JA. Colour vision deficiency in the medical profession. Br J Gen Pract. 1999. PMID 10562750.
7. Spalding JA. Medical students and congenital colour vision deficiency: unnoticed problems and the case for screening. Occup Med (Lond). 1999. PMID 10474917.
8. Poole CJ, Hill DJ, Christie JL, Birch J. Deficient colour vision and interpretation of histopathology slides: cross sectional study. BMJ. 1997. PMID 9390055.
9. Dhingra R, Rohatgi J, Dhaliwal U. Preparing medical students with congenital colour vision deficiency for safe practice. Natl Med J India. 2017. PMID 28731005.
10. Colour Vision Deficiency in Health Professions Education: A Narrative Literature Review. Clin Teach. 2026. PMID 41866652.
11. Mayer RE. Applying the science of learning to medical education. Med Educ. 2010. PMID 20604850.
12. Issa N, et al. Applying multimedia design principles enhances learning in medical education. Med Educ. 2011. PMID 21752078.
13. van de Ridder JM, et al. What is feedback in clinical education? Med Educ. 2008. PMID 18230092.
14. Wisniewski B, Zierer K, Hattie J. The Power of Feedback Revisited: A Meta-Analysis of Educational Feedback Research. Front Psychol. 2019. PMID 32038429.
15. Ryan A, et al. Beyond right or wrong: More effective feedback for formative multiple-choice tests. Perspect Med Educ. 2020. PMID 32789664.
16. W3C. Understanding SC 1.4.1 Use of Color (WCAG 2.2). w3.org.