Healthcare Decision Support
Problem:
Customers lack confidence when selecting the correct medical coverage because of varying plan types, confusing medical terms, and the inability to predict what their future medical needs may be.
On average, only 40% of Americans are very confident they have the correct medical coverage.
Additionally, only 4% can correctly define these top coverage terms: Deductible, Coinsurance, Co-pay, and Out-of-pocket.
(Image) Was a proposal to add interactivity to our questionnaire. The animation was an inclusive exploration for a modern family. Single- You have a motorcycle and a small house. + Spouse- a small sedan and a larger home. Family of 3- upgrade to a minivan. Family 4+- a 2 story home.
What is Decision Support:
The DST is a short questionnaire that provides participants with their best fit medical plan based on the lowest cost. It utilizes artificial intelligence and medical claims data to generate this recommendation.
The DST 1.0 experience consisted of a couple of questions and asked the user to estimate their fixture medical usage for the next 12 months. This was worked out in a spreadsheet and a health plan recommendation was given. We saw varying utilization rates from 8-35%
To enhance this experience and provide the user with better confidence, we purchased 3 years of de-identified medical claims data from Truven. This provided us with 4.88 billion de-identified medical claims to use as reference. Our AI engineers created algorithms to sort through the claims and identify the 7 largest cost drivers that affect insurance premiums! Can you guess what some of those might be?
Below is more detail about our data sets and cost drivers. Throughout the experience, we provide "Did you know" to educate the user as they work through the experience. An example would be; Going to urgent Care vs the ER and using generic prescriptions vs name brands.
Robust data sets:
- 60M unique participants
- 30M families
- 3 years of claims history
Primary cost drivers:
- Demographics
- Prior year medical history
- Key events
- Medications
Competitors:
Some of our competitors in this space are seeing a utilization rate of 87% (enrolling in the recommended plan) when users use the tool end-end. They have robust experiences with a voice-over that explains things along the way and the content is witty and engaging. The downside we have seen is that these tools can be expensive based on configuration and aren’t budget-friendly for some companies. We also saw other offerings with a large mix of AI, Virtual Assistants, multiple language support, and other strategic guidance.
User Flows:
One of the challenges in large UX teams is unifying patterns, color, and 'next best Action'. Even with a mature design system, it is easy for a designer to become siloed within their dev teams. The designs on their own look nice but issues arise when those patterns cause user conflict for cross-enterprise customers.
Image coming soon...
Sample screens from the Decision Support Tool
The layout for the questionnaire is clear and meaningful. We understand that annual enrollment can be stressful, we are reducing cognitive loads by embracing whitespace, keeping content and graphics simple, and minimizing the length of information on pages. Throughout the experience we offer “Did you know” and “Savings tips” to help educate users. In the future, we hope to integrate with chatbots and build out our machine learning.
We introduced a summary experience to allow users to review their answers and make any changes before getting their personalized recommendation. All previous answers are stored so the user will only need to change what they need.
The HSA calculator helps users prepare for future medical expenses by recommending contributions to their HSA maximum, deductible, or a custom amount. It also highlights any contributions from their employers and provides estimated costs per paycheck. This design went through several iterations based on my user testing to get ensure the user understood how to interact with the tool and all relevant data was accounted for. The #1 finding from the research is the user wants to understand what their contribution is per paycheck. I feel too often we need to try and solve for the perfect end-end solution and forget that a large majority of individuals and families live paycheck- paycheck.