The Institute for Health Metrics and Evaluation (IHME) creates visualizations of global health that inform global response, philanthropy, and public health policy. Their Local Burden of Disease tool is a highly detailed interactive visualization of health trends in Africa. Unfortunately, global health workers using the tool are left with unanswered questions about the validity of sources going into the visualizations.
A page displaying the source data that makes up the visualizations, filterable by country and health indicator, directly connected to the functionality of the toolWatch Demo Video (2 Minutes)
2 of my peers and I partnered with The Institute for Health Metrics and Evaluation (IHME) as part of our HCDE Capstone Project (the final course in our degree). IHME is a research center that measures health problems around the world and creates visualizations of global health data that policymakers can use to inform decisions and target treatment.
When we met them at their office for the first time, IHME presented their problem space: The Local Burden of Disease Tool helps users explore health trends at a very precise level (5-kilometer radius). It's a powerful interactive visualization, but it was designed as a research project. IHME had not considered how the design of the LBD tool could effectively support the goals of their users, which included global health organizations, directors at research foundations, students and policymakers. They wanted to increase uptake by better supporting global health workers at the Washington Global Health Alliance (WGHA) and the BMGF (Bill and Melinda Gates Foundation)
We zoomed out to focus on how potential users of this tool currently use health data. During each of 2 focus groups, we had participants recall specific experiences they had using global health data in their work. We asked them to map out these experiences step-by-step, and then discuss these experiences as a group.
We analyzed themes that came from these conversations, and used those to make Personas that captured 2 distinct types of users, and Journey Maps that helped us make sense of their process.
As we grouped our findings (see visualizations below), we realized we were dealing with two types of people who each used data in different ways; we designated these use cases as data as process, and data as output.
Data as Process - Using data and visualizations as part of a larger research process to answer questions and pose new ones. The LBD visualization would be in the middle of this process, rather than the final step.
Data as Output - Using data and data visualizations primarily to tell a story to an audience or illustrate a problem area.
We organized comments from the focus group in terms of where they fell in the process of using health data for some purpose. We were able to separated quotes and concepts into distinct stages.
Data as Process
"Where do I make an intervention?"
"I doubt the validity of this data, and [my client would have] questions about uncertainty interval"
Compare two countries or districts to see if there has been a change
Data as Output
"How can I tell stories about how things are changing?"
"Is the data a triangulation of sources?"
Talk with a Minister of Health about piecing together a new data system
"Think, Feel, Do" framework helped us unpack the underlying motivations behind the tangible worklflows they described
Understand what data the LBD visualizations are representing, so they can confidently use insights in a grant report, presentation, or
Use data from the LBD Tool as part of a broader research process, which includes validating data sources and using insights from visualizations to generate more research questions
Easily find and start using the LBD Tool on the IHME website
Communicate information learned from LBD visualizations to stakeholders around the globe, including NGOs, a range of nonprofit organizations, research foundations, and the general public
There were 2 ways we thought we could make IHME more useful for global health workers
Direction 1: Source Data Transparency - Increase transparency about source data used in LBD visualizations, making it easier to contextualize the visualization and validate the underlying sources.
Direction 2: Narrative Behind Data - Reframe the options for manipulating the visualizations to emphasize the narrative behind the data
We briefly brainstormed design changes for each of the above options, focusing on how we could address the needs of our personas (see Analysis). After spending about an hour on each, it was clear to us that Direction 1 could make a larger impact because validating source data was a key step in each of our personas' workflows.
The local burden of disease landing page (leftmost box) could link directly to information about the source data. A filterable source data page would let global health workers find sources related to specific locations.
The LBD has filters that change the state of the map. We could give example questions that could be answered within the tool, and let users jump into a visualization with certain filters pre-selected.
We set out to make source data behind the LBD Visualization more transparent. We decided we needed to embed the option to explore sources within user's current process of exploring the visualization.
We individually created wireframes of the LBD landing page, a new "Source Data" page, and the map page. These helped us solidify the overall solution we wanted to propose to IHME.
In critique sessions, we discussed how well our designs addressed the needs of our personas. Iterating individually gave us a breadth of ideas to discuss in sections, so we could narrow in on the most effective elements of our wireframes and mockups.
To test our decisions, we had 5 global health professionals from various organizations/foundations use our interactive prototype within a task-based scenario (pulling use cases from our focus groups)
We learned a lot from our usability tests. We refined our designs based on feedback we got (See 'design decisions' below).
Potential Users of the LBD Tool wanted to know how source visualizations are created. We designed a workflow (shown below) that brought pathways to source information closer to the surface of the tool to make it more accessible and more transparent. We proposed changes that IHME could embed within their current Website.
A dynamic link from the 'Map Page' to a 'Source Data' Page populated with sources used to create estimates within a given country
Rationale - Each of our personas needed to be able to access the underlying data behind LBD visualizations; when they used the visualizations to recommend treatment plans to stakeholders like ministers of health, or to frame a problem in a grant or report, they always fielded questions about the validity of the visualizations. (E.g. "Does this visualization include our district's most recent survey?")
Participants wanted access to Source Data while exploring the map, but the Map Page was already pretty crowded with other functionality. We decided to design a new page designated to source data. This gave us more control over how we could display sources.
Rationale - All our participants jumped into the actual LBD Visualizations without looking for info about Source Data. As they explored the tool, they were concerned with source data associated with specific locations they were interested in (e.g. "we have a hospital there"). A link that updates with the state of the map gives users power to explore Source Data information at any point during their research process
Rationale - Users struggled to find the Local Burden of Disease Visualizations on the site and were confused about the relationship between the Tool and the Visualizations.
Rationale - Participants in usability testing said these were most critical. Provider indicates validity, and year of data collection tells them how relevant the data is for current work.
Rationale - Most participants in usability tests either said they were confused by a term or made an incorrect assumption about a term that effected their interpretation of the data (e.g. "I'll just exclude sources labeled "Neonatal" in my research since those aren't being used to visualize "Under-5 Mortality")
Our persona, Taylor, is using the LBD Tool as one step in a larger research process. She would be looking to dig deeper into the data sources so she could validate them on her own. She would want to link to the original source of the data. One Usability Study participant, a data scientist, was most interested in downloading raw datasets, which IHME makes available on the GHDx.
Since the potential user defined in our persona needs to communicate sources to a variety of stakeholders, an automatically generated citation for sources would streamline their workflow. Some participants told us that they have their own citation methods specific to their organization or report, but for more casual usage, this would help.
Working with a real organization is hard. Working with IHME, who is relatively new to applying a user-centered process, definitely helped me improve how I explain the value of defining your target user and how research informs design decisions.
The focus groups were not what we expected. We thought our participants had experience with Local Burden of Disease tool prior to the meeting. It turned out that almost none of them had experience with the tool. But this was okay; it helped us learn more about potential users who don't currently find the tool useful.
Also, it takes a surprising amount of time and energy to keep a client looped into your process! Keeping IHME updated on our progress was a project in and of itself, and we struggled with this in the initial stages. Luckily, our sponsor was able to gently nudge us when they wanted information, so we found a rhythm of keeping them updated about the most important information.