Revenue Management Reexamined
Flatiron Health
Unpacking and simplifying a convoluted internal process for managing cancer care reimbursement
Flatiron Health is a healthcare technology company with a portfolio of products and services to help oncology practices manage care for their cancer patients.
Soon after joining Flatiron Health, I co-led a project to better understand and streamline a key internal process in the Revenue Cycle Management (RCM) business unit, a process which was both costly and causing employee burnout. The core team consisted of myself and my product manager (PM), while the extended team included an RCM manager, 3 RCM associates, 5 engineers, and two business analysts.
We spent 3 months on discovery and then 6 months on ideation and design before beginning to ship our first changes in November 2022.
- Design strategy
- Design thinking
- User research
- Interaction design
- Miro
- Figma
- Google Sheets
- Google Forms
- Lead Designer (myself)
- Product Manager
- RCM Manager
- Engineering Manager
- 4 Engineers
- 2 Business Analysts
- 3 RCM Associates
Healthcare reimbursement is an endless cycle
At a typical healthcare practice, treatment is performed first and paid for later by the patient’s insurance. It often takes months for insurance claims to be paid, if they’re paid at all.
To stay financially afloat and to avoid unnecessary stress to patients, practices must diligently manage each step in their revenue cycle.
Flatiron Health’s Revenue Cycle Management (RCM) service enabled client cancer clinics to outsource many of their clinic’s revenue cycle activities to Flatiron, whose team of experienced RCM associates would oversee those activities and report back to the practice with updates and recommendations.
The existing process was burning out the team
One of Flatiron RCM’s most valuable services was managing “denials,” the insurance claims where the insurance company had either partially or entirely denied payment to the healthcare practice.
Under the RCM service, all revenue cycle tasks in blue were Flatiron’s responsibility. This project focused on denials management (highlighted in green), which comes toward the end of the cycle after the insurance company has already processed the claim.
However, by the end of 2021, RCM associates needed to spend a tremendous amount of time each week managing denials, leading to high staffing costs and widespread burnout.
Going into 2022, our team’s goal was to reduce the amount of time spent on denials, which would require RCM fewer associates per client and give RCM associates more time to focus on other tasks.
What did we already know about the problem space?
To find the gaps in our understanding of the users and the denials process, I led a Miro whiteboarding exercise with my product manager and a representative from the RCM team to document …
what we already knew
what we assumed we knew
what we were confused about
The high-level denials management process with responsibilities color-coded by user group (BPO users vs Flatiron RCM users). Pink sticky notes represent some of the questions that we flagged in our inventory.
The biggest takeaway from this exercise? The denials process actually involved two distinct user groups.
The business process outsourcing (BPO) team contracted by Flatiron: Responsible for addressing each individual denial.
Flatiron’s internal RCM team: Responsible for reviewing the BPO team’s work, analyzing the data, and reporting to the client practices.
After this exercise, I began writing a research plan with 3 key research objectives:
Define the current denials workflow for each user type
Define the main user types & their core behaviors, needs, and frustrations
Identify top RCM pain points & desires and their relationship to BPO behaviors
Putting on my detective hat
Based on our Miro board, I drafted lines of inquiry for both user groups that would get us a comprehensive understanding of each stage of the denials process.
But when I shared my draft research plan, my PM and my direct manager had the same concern. The team had interviewed these users several times before—
How would this round of research tell us anything new?
After reflecting on the goals of denials management and the relationships between the user groups, I realized we could focus our research by working backwards through the denials process.
Since the goal of denials management is to help the client practices, we could start by learning about pain points related to client deliverables. Then like detectives following each clue to the root of the issue, we could ask about the information and processes feeding into the client deliverables, and so on.
My manager and my PM were both enthusiastic about this new approach!
My revised approach to user research, in green, was to start by learning about the most downstream step of the denials process, and then work upstream based on what we learned.
The denials process was a tangled web of data
After interviewing 9 RCM team members and 8 BPO team members, there was an overwhelming amount of information to sift through. I worked with my PM to iteratively group observations by theme and identify cause-effect relationships between themes, trying different groupings to best make sense of our learnings.
To synthesize our research findings, we went through multiple rounds of clustering sticky notes, labeling them, and connecting clusters with arrows.
Eventually I boiled down our learnings into five key findings. I walked through each finding plus our high-level recommendations with leadership from the RCM team, Product Management, and Product Design.
Our analysis of the pain points in the denials process, into five thematic areas (numbered with the purple dots) became our key findings for our broader shareout.
Of the five findings, Finding #2 was critical.
Not only was denials management data spread across many tools and systems, which was frustrating in itself, but it also hindered RCM when it came to:
future denial prevention (Finding #3)
data analysis (Finding #4)
data integrity (Finding #5)
Our research findings resonated deeply with our stakeholders and we received the green light to move forward with solutioning!
Brainstorming as a team
Next, I had to figure out how to approach such wildly different recommendation areas. With our team’s limited capacity, tackling everything simultaneously wasn’t an option. I worked with my PM to define three ideation workstreams that:
we could manage in parallel, and
would help stabilize the denials process before we turned to more aspirational improvements.
For Workstream #2, I led a set of workshops with cross-functional participants to explore ways of improving the tools and systems in the denials process (tackling Research Finding #2!).
Using Crazy 8s brainstorming, I had participants brainstorm individually before we remixed ideas as a group. We refined our top ideas into 14 solution concepts, which we invited Engineering and RCM stakeholders to score by impact and effort.
In the ideation workshops I facilitated for Workstream #2, we brainstormed and refined our concepts in Miro, but then conducted voting with our broader group of cross-functional participants in Google Forms, making the voting process more accessible to those not involved in brainstorming.
After analyzing the scores, we selected four concepts to refine and implement:
Updating the Denials Group Inbox
To address concepts #3 and #4, I began exploring design changes to the in-house Denials Group Inbox (DGI) tool.
At the time, BPO users were entering lots of semi-structured information into each denial’s “notes” textbox. The “smart denials data entry” concept proposed that we turn each row of the notes into a separate structured field, which would allow RCM to export and analyze the data.
In the “Notes” for each denial, BPO users filled out a text template of extra information—totally separate from the structured fields above.
However, since entering data this way would be slightly more work for the BPO users, we planned to also invest in some basic improvements to DGI to speed up their workflow. I put together a simple prototype that showcased several features in addition to the structured notes:
Main prototype features
Organizing fields into named sections
Using radio buttons instead of a dropdown for fields with fewer than 4 options
Conditionally displaying fields to better guide the user (phone number if method=phone). Instead of always showing the phone number field even when it wasn’t relevant.
A “copy to clipboard” feature to allow BPO users to copy notes in the way they were accustomed to, with a live preview of what will be copied.
Fields appearing conditionally (Feature #3) in the prototype based on user selections.
Slowing rollout to boost user success
I brought these changes and updates to bulk editing (not discussed here) for usability testing with BPO users. Although they felt that most of the changes were user-friendly, they were deeply worried about how much the changes to the Notes section would slow down their work.
“For the operator, it will be time consuming to select all the dropdowns instead of filling out one template of notes.”
I hypothesized that once BPO users had the functionality they’d been requesting, such as searchable dropdowns, they would be more open to structured data entry. Therefore we decided to split the changes into two waves:
Wave 1: performance enhancements, making it easier to enter data in DGI
Wave 2: introducing structured data entry for the “notes”
I handed off annotated Wave 1 wireframes to our engineers who began development in October.
Wrapping up the denials project
By the end of the year, we had built up significant momentum on this project. Cross-functional leadership came away from our research report with a clearer understanding of:
The underlying problems with the denials process, an area that had frustrated the RCM team for some time
The resourcing needed from each function to address these problems
“Thank you for all the work you did getting into our denials workflows, and the high-level mature recommendations you developed [...]. I am very excited to light a fire and start majorly improving those opportunities, and that clarity comes from your work.”
Our collaborators and users on the RCM team were genuinely excited about the solutions we were working on and inspired by our vision for the future state of the denials process. This was a clear change from the severe employee burnout that required investing in this problem in the first place.
While Flatiron Health ultimately made the business decision in November 2022 to phase out the RCM service, our work successfully de-risked the denials workflow if Flatiron had chosen to continue with the RCM business.
If Flatiron had made a different business decision, our parallel workstreams and staggered waves of features would have put us in a good position to keep working toward the goals for this project and tackling new complications as they appeared!
Engage with your cross-functional partners
My PM and I deliberately included a RCM partner in our planning throughout discovery and design, creating two-way communication between us and RCM. We invited RCM team members to participate in our brainstorming workshops and provide feedback on prototypes.
By engaging early and often, we built close relationships with many members of the RCM team, which enabled us to build momentum and enthusiasm for our work!