How I Turned 36 States of Medicaid Data into an Interactive Visualization
Role: Head of Content Marketing, Waymark
Tools: Claude (Anthropic), Tableau Public Desktop, Webflow
Timeline: January – February 2025
Output: Embedded interactive dashboard published to waymarkcare.com
The Idea
Waymark's data team had produced a significant dataset: avoidable acute care usage across 36 states, representing more than 41 million Medicaid beneficiaries and nearly $10 billion in annual spending on hospitalizations and emergency visits that could have been prevented with better primary care access.
The data existed as two spreadsheets — one aggregated at the state level, one broken down to 1,654 counties. The story inside it was compelling. States with the highest per-member costs (Mississippi at $30 PMPM, Louisiana at $28) sat next to states with strikingly lower rates (Montana and Oregon at $15 PMPM), and no obvious geographic logic explained the difference. That gap was the insight.
I proposed turning this data into an interactive visualization rather than a static report. My reasoning was that the data's value was in its explorability: the findings at the state level were interesting, but an interactive map could let readers drill into their own state, toggle between metrics, and draw their own conclusions – which is a fundamentally more persuasive experience for an audience of state Medicaid directors, health plans, and Federally-Qualified Health Centers (FQHCs) who are accustomed to being presented vendor claims rather than invited to interrogate data themselves.
I was also responsible for identifying the tool to build it with, getting it approved for use, and building the deliverable myself.
Getting Tableau Public Approved
I had used Tableau Public as a journalist previously, so I knew what it could do and was confident it was the right tool for this kind of interactive, embeddable visualization work. However, before I could use it in a healthcare context, it needed to go through Waymark's data security review process.
I initiated and drove that evaluation, working with our data security team to get Tableau Public assessed and formally approved as an organizational resource. That process meant the project could move forward on solid ground — and that any future data visualization work at Waymark would have an established, approved tool to work from.
The Process
Starting with a project plan
Before touching either tool, I developed a project plan that defined the goals of the visualization, the target audience, and the specific interactions it needed to support: state and county-level exploration, toggling between three metrics (% avoidable care, total avoidable spend, and avoidable spend PMPM), drill-down capability, and hover tooltips with enough context to be useful without overwhelming the reader.
That project plan became the anchor for every technical decision that followed, and the basis for the step-by-step build I undertook in Tableau Public Desktop.
Data preparation
The state-level dataset had issues that needed resolving before Tableau could use it effectively. State names were inconsistently written and only some numbers were formatted with decimal places or percentage symbols, which would have made tooltips confusing for readers. There was also a typo in several fields and some inconsistency in how spending was formatted across the two files.
I used Claude to clean the data and build out calculated fields – adding % avoidable, total unnecessary spend, and spend PMPM to the county file to match the structure of the state-level data. At each step, I manually reviewed the outputs, checked the calculations against the source data, and verified accuracy before moving forward. The data preparation phase was collaborative in execution but editorially mine: every decision about what to include, how to structure the fields, and what the outputs meant was made and verified by yours truly.
Building the dashboard
With clean, structured data in hand, I built the dashboard in Tableau Public Desktop, a tool I had prior experience with from my journalism background, though I had not previously built an embeddable interactive dashboard of this complexity.
The build involved creating multiple worksheets (a state-level choropleth map, the data previously reviewed and cleaned, and an ROI bar chart), connecting two datasets via a relational join, configuring parameter controls for the metric toggle, setting geographic roles for correct map rendering, and assembling everything into a single dashboard view. I worked through technical sticking points — including the data relationship setup and the calculated field logic for the parameter toggle — and made ongoing design decisions about color scale, tooltip content, and layout throughout.
Stakeholder review and the Webflow challenge
Once the dashboard was built and published to Tableau Public, the review process surfaced a practical problem: the Tableau Public web view displayed the map at a noticeably different scale than both the desktop app and the final embedded context. The continental U.S. appeared small, and the logo placement looked off.
Rather than send stakeholders the raw Tableau Public link, I embedded the dashboard into a Webflow draft blog post and shared that URL for review. This meant stakeholders saw exactly what the published version would look like – correct sizing, surrounding content, full context – which reduced revision cycles and prevented the gap that often appears between approval and launch.
What the Data Showed
The final visualization surfaced several findings of interest to Waymark’s audience:
Across 36 states and 41 million Medicaid beneficiaries, nearly $10 billion is spent annually on avoidable acute care.
Per-member cost ranges from $15 PMPM (Montana, Oregon, Hawaii) to $30 PMPM (Mississippi) – a 2x spread that doesn't follow population size or geography in any simple way.
California accounts for $1.78 billion in total avoidable spending but achieves one of the lowest per-member costs in the dataset ($16 PMPM), while Mississippi spends $30 PMPM despite a far smaller program.
Published research supports the pattern the data suggests: a systematic review of 51 studies found that 72.5% showed a significant inverse relationship between primary care accessibility and avoidable hospitalization rates.
The interactive format makes the wide spread of data, the stark variances in numbers, and the expanse of impacted geography visible in a way that a static chart cannot.
What I Learned
A few things from this project stand out and carry over to every similar effort going forward.
Originating the idea matters as much as executing it. The dataset existed before I got involved. The decision to make it interactive, to choose a tool capable of this type of geographic drill-down, and to frame it for an audience that responds to explorable evidence were editorial and strategic judgments that shaped what the project became.
Tool approval is part of the work. In a healthcare context, getting a new tool evaluated and cleared is a project prerequisite not present in every industry. Doing that groundwork made the project possible and paved the way for future work of the same kind.
The storytelling decisions are harder than the technical ones. Choosing which three metrics to highlight and discuss, how to label tooltip fields, and what the default map view should show require judgment calls that no tool can make automatically. Those decisions determined whether the visualization told a clear story.
The review environment matters. Showing stakeholders the work in its actual production context produced more useful feedback than a standalone link would have.
AI assistance expands what's possible without replacing what's essential. Claude handled data cleaning and field construction; Tableau handled analysis and visualization rendering. What neither tool could do was identify the story worth telling, determine which findings would resonate with a Medicaid director, or understand the strategic context in which Waymark was operating. The judgment layer of what to build, why, for whom, and what it means was purely and only human.
The Outcome
The visualization was embedded into a published blog post on waymarkcare.com and received strong engagement both internally and externally. It was shared on LinkedIn by industry thought leaders, including a former head of the Centers for Medicare and Medicaid Services — an indication that the work reached and resonated with exactly the audience it was built for.
It became a reference point for how Waymark could communicate the scale of the problem it exists to solve, in a format that invited exploration rather than simply asserting a conclusion, and generated interest in producing more data-informed content of this kind.
Tools used: Claude (Anthropic) for data cleaning and calculated field construction, with manual review and verification at each step; Tableau Public Desktop for data analysis, visualization build, and publishing; Webflow for CMS embedding and stakeholder review.