Designing effective EHR tools to decrease the amount of cognitive load experienced throughout the physician’s daily journey.

Our generative research allowed us to narrow our focus to ways in which we might be able to reduce the amount of cognitive load generated by use of digital systems (specifically, the EHR).

Understanding that we had only a few months to translate insights into design solutions, we followed the GV design sprint framework, pushing us to continually test and iterate based on user feedback.

Sprint 1
Narrowing Scope
Concept Testing
Our insights formed three overarching categories: Information Processing, Workflow Friction, and Psychological Stress. We spent early brainstorming activities ideating design solutions that would target the most common pain points observed.

We created eight conceptual solutions - ranging from EHR-integrated data visualization to digital ‘thank-you’ letters for physicians - which were intended to address key pain points uncovered through journey mapping.

We conducted five rounds of ‘speed-dating’ sessions with physicians working across the U.S. that we recruited via Respondent. Our goal was to understand which of these concepts might make the most impact on physicians and why. Two concepts stood out in particular: Adaptive User Interfaces and Data Visualization. Physicians felt that these types of tools could speed up their workflow and help surface information about their patients.
Sprint 2
Iterating Content
Card Sorting
With feedback from our physicians indicating Adaptive User Interfaces to be a promising concept, we faced a learning curve in terms of our medical knowledge. It was clear that every patient was different, and data that was critical for one patient might not be so relevant to another. But how to know what physicians searched for based on different ailments? Well, we went right to the source.

Our second round of user testing - again with a set of five physicians who participated for an hour each - was centered around understanding what EHR data would be relevant for different medical conditions. We created three scenarios, each with unique patients exhibiting different chief complaints. After going through testing rounds with each physician to understand what data they would look for given these three scenarios, we synthesized this feedback and were able to create rough correlations between data searched for and presenting illness. This not only helped us generate content within adaptive screens, but corroborated our overarching principle that value of information is context-dependent.
Sprint 3
Defining Features: Data Display & Capture
User Testing
Moving forward with an exploration of Adaptive User Interfaces, we created initial prototypes to get a sense of general usability. We realized that there was a range in terms of level of adaptability and predictability, and focused initial usability testing on understanding what physicians might find the most useful.

We also added in another key feature (data capture), positioning gathehr as an overall data display and data capture ecosystem rather than as a stand-alone product.
buildable
adaptive
continuous

Data Capture: How might a note assistance tool have the most impact?

Sprint 4
Evaluating Usability
Task-Based Usability Testing, SUS Evaluation
EHRs have a SUS score of 45.
Gathehr received a SUS score of 80.

After refining our prototypes based on initial usability testing - and more clearly defining tasks asked of physicians - we saw pretty dramatic results.

Findings

Sprint 1

Goal: Scope potential solutions based on value to physicians

Effective data visualizations and adaptive UIs can minimize information foraging.

Sprint 2

Goal: Understand what data needs to be captured and displayed at what points in the physician workflow and in what form

Tracking patient data over time helps notice trends and patterns.

There is distinction between subjective and objective information in the progress note.

Sprint 3

Goal: Testing what level of adaptivity best benefits physiciansTest hypothesis that voice-to-text will decrease charting time

Goal: Test hypothesis that voice-to-text will decrease charting time

Tracking patient data over time helps notice trends and patterns.

There is distinction between subjective and objective information in the progress note

Sprint 4

Goal: Test hypothesis that adaptive UIs and note assistance will lower cognitive load

Goal: Identify and address usability issues

Usability from the EHR was improved by 78%. Current EHRs have a SUS score of 45 and gathehr had a SUS score of 80.

Adapting by problem in problem list was the ideal navigation.

CMU MHCI 2022 CAPSTONE PROJECT

We are a team of five master's students in Carnegie Mellon's Human-Computer Interaction Institute (HCII). Gathehr was created as part of our Capstone (MHCI 2022) Project, working with Optum/ UnitedHealth Group.