Cohort Atlas
Product Design + Data Viz
Background
One of CareJourney’s main products is its Cohort Atlas dataset. Built off of Medicare Fee-For-Service claims, this data covers geographic data, metrics to understand patients’ overall health, prevalence of health conditions, and cost and utilization metrics. It is used by health systems to assess markets, uncover opportunities for savings, understand patient demographics, identify highest-need populations, and more. In the dataset, patients are grouped into "Cohorts" (ex. Alzheimers patients, Diabetes patients, etc.) to better understand and compare across different health conditions.
The Task
My team was asked to create a front-end solution for users to easily digest this data and integrate it into their workflows. I was tasked with creating the Summary Page where users could get a high level understanding, see the most important insights, and delve deeper into the data if they desired.
My Role
Product Designer
Team
UX Designer
Product Manager
Scope
3 Months
The Result
Users
CareJourney users can be broken out into the following three buckets:
Because we were focusing on data visualizations, we felt our main users of interest are the In and Out's as well as the Eager Beavers. We found these users will spend more time with visual aids to find their insights, whereas Beautiful Minds tend to run their own analysis and create their own visualizations using the raw data.
User Stories
We brainstormed our most important user stories for these personas and defined design requirements based on them.
As a user, I need an easy way to find a cohort and define a region so I can analyze characteristics about a patient population.
As a user, I need to see a breakdown of patient demographics so I can better understand specific segments of a population.
As a user, I need a short list of the most important insights about a cohort so I can quickly identify areas of opportunity.
As a user, I need to I need a way to compare several cohorts at once so I can see trends across many cohorts.
Sample User Stories
Design Requirements
Visually digest key insights quickly
See trends in data easily, with the ability to share learnings easily through screenshots. Find most important metrics immediately.
Delve deeper into data of interest
Explore data points of interest in further detail with quick access to tabular info.
Filter data easily
Provide different cuts of data users are interested in based on geography, patient cohorts, year, and more.
Benchmark the numbers
Understand performance relative to other health systems, geographies, and cohorts.
Most Important Data
With our users and their requirements in mind, we set out to understand what data matters to them the most and can best help them solve for their specific use cases. We interviewed several members of our user support team. They provided us with insights on which numbers bring value to which of our users and why. While our dataset has hundreds of data points, we were able to narrow our list of most important metrics to the following:
Summary of Metrics
-
Total # of patients
-
Total allowed amount ($)
-
Average HCC Score (to help assess patient risk)
Cost Metrics
-
Spend per month ($)
-
Site of service breakdown
Chronic Conditions
-
# of patients in the cohort with the chronic condition​
-
% of patients in the cohort with the chronic condition
Utilization Metrics
-
# of avoidable emergency visits
-
# of readmissions
-
# of unplanned admissions
User Flows
With an understanding of what data matters to our members and why, we created a high-level user flow to get a better idea of how users would navigate through the product to reach their data of interest. Getting this right was important because of the variety of our users' needs; some users only wanted summaries of the numbers while others would want to delve deeper. We went through several iterations of this before landing on a flow we liked that could solve for both types of use cases.
User Flow
Brainstorming
Reviewing our user research, design requirements, and MVP metrics, we began brainstorming solutions. During this process, we thought hard on how to make the visualizations truly meaningful. We found that numbers standing alone from this dataset didn’t mean much; we needed to show comparisons so that a user could interpret their relative meaning.
​
Some questions we asked ourselves during this process of iteration include:
-
How can we build visualizations that will be scalable and deliver meaning for the hundreds of cohorts we have?
-
How can data points be shown through different visual channels (ex. color, size, etc.) to show relationships and paint a fuller picture?
-
How many cuts of the data (and which ones) can we show before we start to lose the user's comprehension in the story?
My Early Sketches
High Fidelity Mocks
.png)
Lessons Learned
While we are still gathering feedback on users' experiences with the tool today, here are some things we learned so far:
-
Scope out scalability during the iteration process: Not all of our visualizations were as scalable as we thought - The donut chart under Chronic Conditions doesn't always convey something valuable, especially if the percentages are small (which they often are). If we had tested our designs with more examples of real data, we might have caught this sooner.
-
Prioritize user feedback: Instead of talking to users directly, we relied on team members who work closely with users to provide insights on what delivers value. While this was certainly useful in discovery, we were still left with some assumptions that needed to be validated in our final designs.