Web App Design, Healthcare, AI
Designing decision support system that collaborates with Cardiologists?
In United States, there are 5.5 million heart failure patients. When their hearts completely failed them, they have an opportunity to receive a VAD that supports their heart and prolongs their life. However, more than 50% VAD receivers pass away shortly after the surgery.
This is largely due to that the cardiology team was not fully informed about patients' conditions before reaching a decision, allowing unqualified patients receive VADs. CORA is built specifically to help reduce the number of unqualified candidates receiving a VAD
CORA is both the name of the research initiative at CMU as well as the final product we produced. CORA's mission is simple, help cardiology teams make informed decisions about providing end-stage heart-failure patients VAD devices.
The cardiology team makes uninformed decisions because their diagnostic process was inefficient and their human errors. Time was wasted printing out EMR and identifying key clinical data. Furthermore, the team sometimes incorrectly updating patients' information on UPMC database.
How might CORA reduce the time the cardiology team spends identifying key clinical data and reduce their human errors.
Length of the project: 6 months
My role: User research, User experience design, Visual design, Prototyping
Teammates: JeongMin Seo (Product Designer), Saumya Karbanda(Product Designer), Qian Yang (Product Designer), John Zimmerman ( Project Lead)
Collaborating departments: UPMC Cardiology, CMU Bioengineering
Keywords: Healthcare, Web App, Decision Support System
I interviewed a cardiology team at a local Pittsburgh hospital, UPMC. The cardiology team consists of four persons, which is typical of the cardiology team. The team consists of two cardiologists and two nurse practitioners. Every week, they need to decide the life and death of 5-12 patients.
Late stage heart failure
1. Referred to the hospital for VAD by home cardiologists
2. Live by himself, family away
1. Assistant to cardiology surgeon
2.Interacts with patients everyday and has information about patients’ social background
3.Directly interact with and encode patients’ data in hospital’s database
1.The final decision maker about whether patients receive a VAD
2.Very familiar with patients’ medical backgrounds but not personal information
3.Assess patients’ survival rate based on personal intuition
Task flow for preparing meeting about VAD
Identified pain points of VAD teams
Currently, when the cardiology team decides which of their patients should receive an artificial heart, their efficiency in reaching the final diagnosis and the accuracy of the final diagnosis could be greatly improved.
The time the cardiology team spends in identifying the key information that helps them reach the final decisions could be reduced. They currently sift through hundreds of pages of Patients' Medical Records to identify valuable information that are averagely 5% of the total information volume they have to go through.
Human errors often occur in the diagnostic process, and one of the causes is the time gap between when the cardiology team reaches the final diagnosis and when the diagnosis is updated on UPMC website.
Help cardiologists highlight information on patients' EMR that is key for cardiology team's final decisions, and presents the clinical dats in consumable format such as data-visualization. As a result, the cardiology team can glean information from a single interface that allows key clinical data to be effectively analyzed and comprehended.
Help reduce human errors by allowing cardiology team to update their diagnosis in real time.
Features of CORA
1.Highlighting key clinical data
This feature of CORA corresponds to the first goal I intend CORA to achieve. Since the nurse handles patient's data, nurse on the cardiology team primarily interfaces with CORA at this stage. CORA can help the cardiology team select information on each patients' EMR that are most important to the cardiology team's decision making process and display it on the slide. The cardiology team can select more information if it is necessary.
2.Real-time patients’ information update
During the presentation stage, the cardiology team can update patients’ information on CORA’s interface. The change is live updated in UMPC database and for other teams to see.
Interview & Data-collection
We talked to cardiologists and then we understood 3 key factors about them:
They believe that their decision making process is fine
They are not paid to look at any graphs and number. They are paid to be in the operation room.
If we want to build something for them to use, it should be something that save them time on looking at EMR (Electronic Medical Record). Their review process now consists of printing out 12 pages of the patients' EMR for review purposes.
We used the theory of Customer Co-creation in Value Production to guide our design thinking and thus produced the CORA Slide Generator Service to fulfill the need of doctors and embed our need within the slide generator.
Here is the storyboard I drew in to demonstrate how the cardiology team would interface with CORA.
1: Reduce CORA's learning curve
I designed CORA with the intention to reduce the learning curve to the minimum. Because the cardiology team values efficiency, it is important this tool can be learned in little time. I designed visual and interaction experiences familiar to the cardiology team with tools that the cardiologists already familiar with and use often, which is powerpoint.
2: Visualize key clinical data
One insight I discovered from research was that cardiologists understand different types of data in unique processes. Doctors rely on their own expertise to analyze patient’s vitals and computers for other information such as medications. Therefore, I incorporated doctors’ needs in my design. As a result, CORA provides some data to doctors in number format and others as visual manifestations of the data.
One of the most successful data visualizations i created is the visualization of medication data: The cardiology team cares to see which medicine the patient has been taking recently and how long the patient has been taking it. After sketching out several iterations, I decided to use Gant Diagram because it communicates the information hierarchy that fits the cardiology team's need.
Each patient card contains patient’s profile, his/er profile photo and his/er medical ID. Another important function of patient card is to alert doctors whether this patient is in late-stage heart failure
2.Electronic Medical Record design
EMR section has two functions: 1.Display the patients' medical records 2.Allow both CORA and the cardiology team to select patients' medical data that is worth attention. Therefore, it is necessary that CORA's (machine) selection can be differentiated from human selection. I used different colors to represent the varying selections.
3.The presentation space design
Divided the presentation space into left and right column. On the left column are two of the most important information that the cardiology team needs to see: Inclusion and Exclusion criteria (2 & 3)
“Very often, the personal background of the patient instead of the medical record of the patient determines whether or not the patient receives a VAD”
Next step & Reflections
I have applied the visual guidelines of the some other applications created in UPMC to allow the cardiology team to be more familiar with using the product Also, moved the EMR data section close to the patients to allow the cardiology team to visually understand the relationship between patients and data
"If you design something that allows cardiologists to work faster, then we might use it! "
Empathetic of users' needs
The project taught me that regardless of the technologies in the experience/service, I should always consider users’ needs and if these needs are being thoughtfully met. User-centered experiences and services must empower the people it serves.
Incorporating User Feedback
As I designed the application, I interacted with Dr.Kormos and his cardiology team bi-weekly to test our iterations and receive feedback from the them on what works for them, can the design integrate well into their current workflow and if our design is showing the information that they need to see. We will integrate their feedback in our next iteration.
Healthcare is a new domain for me, and I enjoyed reading existing literatures on Clinical Decision Support Tools (the professional term for tools like CORA) to onboard myself.