University of Michigan · 2019-2020
Plain Language Medical Dictionary
Empowering patients to understand their health by translating complex medical jargon into accessible literature.
- Healthcare UX
- Product Strategy
- Information Accessibility
- Sustainable Design
Try the live widget
Project overview
My role
UI/UX Designer · Project Manager · User Researcher
Timeline
October 2019 · 4 months
Partner
Taubman Health Sciences Library
Tech
React.js · Figma
Team
- Xiaoshan He — Co-designer / Developer
- Carol Shannon — Mentor
- Mark Chaffee — Information Services Specialist
Methods
Comparative analysis · Contextual interviews · Affinity mapping · Usability testing (3 rounds)
The problem
Medical terminology and acronyms make it incredibly difficult for patients to understand their own health situations. Doctors often speak in plain language during visits, but after-visit summaries are filled with complex jargon—so patients cannot connect documentation back to the conversation they had in the clinic.
Many people fall back on Google, yet results skew either overly academic or toward unreliable sources—exactly when clarity and trust matter most.
The challenge
The original Plain Language Medical Dictionary launched in 2011 across three surfaces—a web widget, iOS, and Android—but had grown outdated and no longer carried enough depth in its term library.
How might we redesign PLMD into a sustainable product that helps people decode health-related documents across realistic scenarios?
Understanding the landscape
PLMD's legacy strength was refreshingly straightforward and ad-free, but it lacked the search behaviors users needed when confronting messy, real-world documents—not isolated vocabulary flashcards.
We conducted comparative analysis across CDC, WebMD, Merriam-Webster, and MedicineNet—benchmarking how trusted references organize terms, explain nuance, and balance credibility with readability.
Benchmarking findings
- Navigational Load: Long lists slow down retrieval; lacks efficient indexing.
- Authority Perception: Institutional branding fosters immediate user trust.
- Cognitive Load: Ad-free, sparse layouts maximize focus and absorption.
- Semantic Hierarchy: Tiered visual layers differentiate nuance from prose.
Strategic Scoping · Why We Sunset the Apps
Our initial charter was to refresh every legacy surface—the widget, iOS, and Android. However, a critical audit of historical performance and user behavior reframed the investment:
- Low Adoption Data: Legacy app performance showed fewer than 100 total downloads between 2011–2015. The maintenance cost of native apps far outweighed their actual utility.
- High Friction for Episodic Use: As an episodic tool rather than a daily-use app, requiring users to download and constantly update a native application created significant hurdles at the exact moment of need.
- Contextual Alignment: We observed that most healthcare institutions operate via web-based patient portals. To maximize reach, we needed to be where the caregivers and patients already are.
Pivot Takeaway
Guided by historical adoption data and friction analysis, we abandoned low-impact native apps to build a flexible, embeddable widget that seamlessly integrates into the global healthcare web ecosystem.
User Research
We ran hour-long contextual interviews with five participants spanning clinicians, plain-language specialists, and everyday readers navigating their own records.
- Memory gaps. Patients rarely remember exact terminology from visits, so classic search bars tuned for precise spelling break down fast.
- The translation gap. Plain-language conversations in the clinic clash with jargon-heavy documentation afterwards—users experience two different vocabularies for the same episode of care.
- Content is king. If readers cannot find the word they need, they abandon the product—no amount of visual polish compensates for missing coverage or weak lookup paths.
Design iteration
We moved from whiteboard sketches through mid-fidelity wireframes, testing iteratively across three rounds of usability sessions before locking visual polish.
Key features shipped
Fuzzy & alphabetical search
When users only partially remember a term—or struggle with spelling—we layered fuzzy matching plus an alphabetical browse mode so exploration still succeeds without precision typing.
Paragraph search (contextual translation)
Users rarely encounter jargon one word at a time—they read it in paragraphs copied from portals or printed visit summaries. PLMD parses pasted text, highlights recognized medical language, and surfaces plain-language definitions alongside the original snippet for quick cross-checking.
Crowdsourced sustainability (request & report)
To grow the glossary without a massive budget, we paired widget UX with lightweight crowdsourcing pathways that turned reader friction into structured feedback loops for library partners. We shipped Request Term and Report Error flows that funnel reader frustration into actionable tickets so librarians can triage updates asynchronously.
Impact & future vision
Repository scale
1,100 → 1,800
Expanded active definitions while modernizing React architecture for faster iteration.
Embed-ready delivery
We packaged a lightweight embed snippet so hospital sites, library portals, or partner teams could drop the widget into existing patient journeys without bespoke engineering spikes.
<iframe src="https://mlibrary.github.io/medical-dictionary/" style="margin: 2em 1%; height: 650px; width: 98%; border: 2px solid #EEE;"></iframe>COVID lockdowns interrupted our final planned enhancement—importing anatomy imagery from the library collection—but we documented the path forward so partners could resume once assets became accessible again.
For the narrative behind the fellowship—including organizational context and storytelling beats beyond this portfolio outline—read the published UM Library piece linked below.
Reflection · AI and human-centered design
As international students navigating the US healthcare system in a second language, this project was deeply personal. Today, large language models can ingest dense clinical PDFs and return conversational summaries in seconds—yet the exploratory research we ran before that maturity still anchors every decision we made.
PLMD was never only about translating words. It was about mapping the emotional cognitive load carried by vulnerable patients, designing for trust, and building sustainable information systems libraries can maintain. It reinforced that the core of UX is naming the human problem long before reaching for the technological shortcut.