HoloCue
Overview
Providing context at a glance to reduce triage miscommunication
Miscommunication during triage means nurses are more likely to see LEP patients revisit within 72 hrs. We designed an AI tool for indirect translation to support a nurse's workflow for clear and effective care.
My Role
Team
Emily Chan, Andrew Ellyatt, Amanda Guan, Ayomide Ibidapo, Kaitlyn Subcharoen, Giang Tran
Timeline & Results
3 months (May-July 2025)
Final grade of 99%
The Task
Rethinking ER's with a constraint: No phones or websites allowed
We were tasked to "design a technology-based intervention to address one meaningful inefficiency within the Emergency Room (ER) system."
Our professor encouraged us to explore future technologies, explicitly ruling out phones or websites. This constraint pushed us to step outside familiar solutions and experiment with alternative platforms.
Prototype Highlights
HoloCue, a dynamic cue card for triage nurses
After flagging cultural nuances during ER intake, HoloCue provides real time context and guided clarifying questions.
By easing cognitive load and guiding clearer conversations, HoloCue supports nurses' endurance by helping nurses understand patient situations deeper and faster. This in turn improves LEP (low english proficiency) patient mutual understanding, overall reducing repeat visits across the system.
Why a Hologram?
The unique format allowed us to experiment with an intervention that feels present but unobtrusive, enabling shared visibility between nurse and patient. This speculative choice helped us learn about the importance of transparency and collaboration during triage.
The insights uncovered remain relevant to conventional implementations like apps or websites, however the exploration using the hologram was key to revealing these lessons.
The Problem
Nurses see 24% more unplanned revisits from LEP Patients
How communication challenges affect nurses:
Communication challenges significantly impact staffs ability to assess symptoms accurately, delay triage decision-making process, and contribute to misunderstandings about diagnoses or treatment plans.
How communication challenges affects LEP patients:
From the LEP (low English proficiency) patient's perspective, this leads to confusion, emotional distress, and difficulty in participating meaningfully in their care decisions
Research Findings
There is confusion and distrust from language barriers from current methods
Only 20% of LEP triage cases access qualified interpreters
This access is due to limitations of time, logistical burdens and cost.
These factors lead nurses to rely on family members or digital tools.
Translation tools only capture literal translation, not cultural nuance
Culture influences how people discuss health (e.g death.)
Speaking the same language does not mean sharing the same culture. Historical, class, or religious intricacies impact understanding.
HMW… Help triage nurses clarify indirect symptom descriptions from LEP patients without disrupting trust and conversation flow?
Service Blueprint
Understanding the complex layers of triage
By situating HoloCue within the triage service journey, we visualized the complex layers that support the triage nurse and LEP patient’s interaction.
Where HoloCue appears:
During initial symptom assessment, when nurses often encounter culturally indirect language.
Who interacts with it:
Primary User: Triage Nurse
Secondary User: Patient
Data Receiver: Hospital System
What changes backstage:
The AI flags potential cultural miscommunications and surfaces clarifying prompts.
What I would explore if I had more time:
Later feedback from ER professionals mentioned consolidating HoloCue into their native translation software would lead to more valuable next steps.
Combining this with user's desire for history/summaries to create EHR notes within the native system would strengthen communication continuity across the entire triage service.
Research Findings
Triangulating qualitative and quantitative methods
We combined quantitative and qualitative UX methods to better understand user behaviour. These methods helped us see both what users did, why they did it.
We conducted bodystorming, user interviews, and TAM surveys. We analyzed our data with a thematic analysis and affinity mapping.
Low Fidelity Research Highlights
“But I feel like most patients would feel like a lack of professionalism in a doctor like googling things, you know?”
Many of the participants mentioned difficulties in integrating HoloCue smoothly into their workflow.
Participants felt pressured to appear credible by being secretive about their usage— and felt burdened balancing a smooth conversation simutaneously.
Evolving from discreet notification to a large tabletop screen
To shift the perception of HoloCue from a private reference to a shared tool, we moved it's placement from the nurse’s eyeline to the triage counter.
This change reduced one sided assessment through promoting collaborative and transparent usage. aiding mutual understanding between nurse and patient.
Before
After
High Fidelity Research Highlights
Analyzing the changes from High-Fi and Low-Fi
Comparing and contrasting our results from low-fi and high-fi helped us identify what was going well, and which areas needed improvement.
Perceived Usefulness Analysis
5.83
Low-fi
->
6.70
High-fi
Score and tightened score range show users felt:
HoloCue supports understanding of cultural differences
Past design changes further supported triage deliberation
Using AI to help inform nurses when they hear culturally indirect symptom descriptions should continue to be our core functionality.
What affected the PU score: Participants struggled to re-prompt HoloCue
Navigation between listening and prompt providing stages was unclear
Users repeatedly tapped or waited for something to happen
The lack of prompt history forced participants to rely on their memory of past clarifications HoloCue made. Participants often tried to access prompt history, which was not an existing feature. Their repeated attempts demonstrate a clear need for this feature.
Understanding the device’s status and timing is an area we should focus on to improve the PEOU rating.
Perceived Ease of Use Analysis
4.93
Low-fi
->
5.83
High-fi
There is a wide inconsistency in ease of use shown through:
a wide score range (4.83-6.83)
moderate SD (0.82)
Some users are struggling while others are succeeding. We should continue to improve on this area
What affected the PEOU score: Additional ‘Read More’ content is overwhelming
Participants felt overwhelmed by the information overload leading to long pauses
Participants struggled to balance reading and maintaining conversation
To improve the PEOU score it will be key to balance conciseness and depth of information to a triage appropriate situation.
Tweaks to improve PU and PEOU if we had more time:
These changes would minimize awkward pauses and frustration from nurses when trying to multitask understanding the technology, reading a new concept, and conversing with a patient.
Effectively, HoloCue would equip the triage nurse to effectively gain the context they need at a glance to connect with LEP patients.
Reflection
Balancing AI's role in conversation
Our initial idea followed in the steps of successful AI assistants: friendly and warm. It's easy to overly rely on AI as a solution, especially for this project that valued creative and futuristic solutions. Initially our concept involved the AI acting as an intermediary between nurses and patients, doing full translations, and speaking. I raised concerns of treating our conceptual AI as a blackbox, especially since it seemed like a wall between nurse and patient. With the Professor’s guidance we shifted the AI’s role to support triage with quick context and clarifying questions rather than acting as the sole communicator. Keeping the AI assistant neutral and literal communicated the nature of the quick, essential aid we aimed to provide.
Breaking the habit of solving isolated problems
This project taught me it is important for tools to integrate well with existing systems. We learned from an ER professional that staff have their own native apps and well-intentioned tools can add friction if they don’t integrate smoothly into established workflows. It's important to me that my solutions don't create problems in adjacent areas. I'm interested in learning about designing effective solutions for the whole ecosystem. Side effects, tradeoffs, and integration are factors I will consider more consciously.