Improving prognosis communication for patients facing complex medical treatment with a user-centered design approach
Myself — Prototyping, Data Collection & Analysis
Dr. Sunyoung Kim — Research Advisor, Data Analysis
Dr. Mark Aakhaus - Critical Revision
Dr. Lisa Mikesell — Critical Revision
Sarah Fadem — Data Collection Support
September 2018 - August 2019
September 2018 - July 2019
Meaningful communication of a patient’s prognosis information is important for making decisions in clinical settings, but it can be very stressful for a patient to understand their diagnosis when it comes to numbers. Therefore, visuals like graphs and charts are widely used to help make decisions since it reduces the amount of mental computation required to interpret this information. This study investigates how patients facing a complex medical treatment perceive and respond to different types of visual representations that present numerical data of projected treatment outcomes.
How do patients perceive and interpret projected data on their survivability via different types of visualizations?
When it comes to electronic health aids, it is essential to understand how different types of visuals for presenting numeric information can influence a patient's understanding of a prognosis. Still, there is not much empirical research on this, especially when the risks may be life-threatening or concern their life expectancy.
For this research project, we parterned up with the Rutgers Cancer Institute of New Jersey, who have fortunately given us the opportunity to work with patients who have experienced Acute myeloid leukemia (AML), a type of blood cancer. For AML, it is a common challenge for a patient to decided whether they should recieve allogeneic hematopoietic stem cell transplant or proceed with chemoterapy. Because survivability predictions for either of these choices do not always positive outcomes, it can be really stressful for patients. We decided to focus on AML as our main context for this research because it deals with complex medical outcomes and communications challenges.
To achieve our objective, we created a tool for patients in the form of a website. We followed a user-centered, iterative design process for its implementaiton.
First, a set of paper-based, low-fidelity prototype sketches was created by Sarah Fadem to visually present treatment outcomes. During this phase, there was a continuous interaction between the research team and the clinician team to facilitate the match between the patient’s requirements and the clinical practices. In brief, three graphical representations with a varying degree of abstraction were shown to participants that essentially predict the percentages of survivability. The three forms were a plain pie chart, natural frequencies using a human icon, and abstract heatmap-colored bar.
Patients from the first-round study rejected all graphical representations of survivability but a pie chart. This first round of evaluation sessions resulted in a strong user preference for pie charts because it is simple, straightforward, and strategically ambiguous. You can read more about this part of the process here.
Based on the findings from the first iteration, we created three interactive prototypes of graph representations with outcome information. To be more specific, these were working demos that run on a web server so that a user can have some real interactive experience. Unlike the first iteration, these prototypes were to be interactive on a website. Because there was a strong preference for pie charts in the last iteration, we decided to focus on presenting more popular and simple kinds of charts: a pie chart, a vertical bar chart, and a horizontal bar chart.
In the first-round study with the paper prototypes, another key finding was that there was a conflicting preference among patients on if they wanted to know the number of survivability or not. Therefore, we created each prototype in two versions: one with a chart's scale and numeric labels of survivability and one without them. On these charts, a projected timeline was shown from Day 0 of the transplant to 4 years past the transplant. Clicking each timeframe refreshes a chart with survivability data for the time frame selected. A variation of each prototype was also created that shows a numeric value of survivability when a mouse hovers over a chart. At the bottom of the chart, a short description of how the values are calculated and instructions about how to interpret the graph was provided.
The first prototype is a pie chart. Two pie carts are juxtaposed horizontally to present a chance of survival with and without receiving a transplant. There are different opinions on using pie charts since they have been found to be very useful in displaying multiple values of information in a single chart, yet have also been found to result in poor comprehension of risk information. Still, we included this version of visualization because our partner clinicians strongly envisioned pie charts as usual for clinic practice. Additionally, they were the preferred visualization in the firt iteration. We created the prototype to be interactive by animating the colored portion of a pie chart to gradually fill up clockwise after the page loads.
Secondly, we have a vertical bar chart. Two vertical bar charts are juxtaposed horizontally to present a chance of survival with and without receiving a transplant. A vertical bar chart is known to be useful to compare groups side by side on the same measure. We created the prototype to be interactive by animating the colored portion of a bar chart to gradually fill up from bottom to top after the page loads.
The third prototype is a horizontal bar chart. Two horizontal bar charts are juxtaposed vertically to present a chance of survival with and without receiving a transplant. Information presented in a horizontal bar chart is known to be more readable than the vertical layout and to be preferred by users. We created the prototype to be interactive by animating the colored portion of a bar chart to gradually extend from left to right after the page loads.
We conducted semi-structured, think-aloud interviews with twelve leukemia survivors to understand their perspectives regarding the graph presentations of survivability likelihoods. Each participant was recruited through a pool of AML patients, and all but one patient received allogeneic hematopoietic stem cell transplants.
Interviews lasted between one and a half and two hours and sought patients’ feedback on the prototypes highlighting their perspectives, experiences, and preferences. Participants were first asked to share their experiences when they were informed about possible outcomes. After we explained the purpose of the OPEN:AML tool, they were asked to freely interact with it and provide feedback reflecting on their experiences. While exploring each prototype, participants were asked to verbalize their thoughts through a think-aloud process.
The results showed a preference for vertical bar charts over horizontal and pie charts while animating the charts to “fill-up” generally conveyed a subtle sense of positivity even when survival projection was low. The value of explicitly indicating numeric values and scale varied but the results suggest that what matters to participants is having control over when such details would be seen. The results also point out that making sense of prognostic information involves balancing the tension between information utility and patient fear and judgments about authenticity and credibility of survivability calculations.