AMA-SHLI MJAF "Data is Power" Session
Talk/Presentation · July 2025
Allyson Belton, MPH, Director, Education and Training at SHLI and the AMA-SHLI Medical Justice in Advocacy Fellowship team invited me to lead a learning session on data as an advocacy tool, timed deliberately: fellows had just come off an intersession assignment defining their research questions, anticipated data collection needs, and methodology for their capstone projects, and this session was meant to sharpen that thinking before they locked in a direction.
QE Rationale
Counts toward Theory and Practice — directly demonstrates the three mapped learning objectives: recognizing how data/analytics affect policy and system outcomes (Cobb/Fulton case study), translating public data into advocacy assets (audience-specific reframing), and identifying credible health-equity data sources (HET's CDC, CMS, CAWP, and Bureau of Justice Statistics partners, plus the red-flags framework).
Why This Session, Why Then
By mid-program, the Medical Justice in Advocacy Fellows had just finished an intersession assignment defining their research questions and anticipated data needs for their capstone health equity projects. I was brought in at exactly that point — while they were still shaping what their projects would look like — to walk through the Health Equity Tracker (HET), the tool I work on as Product Manager and UX Engineer at the Satcher Health Leadership Institute, and how to use data like it responsibly in advocacy work.
Why We Built the Tracker
HET is free and open source by design — the codebase is public, so anyone can flag a number that looks wrong or point us to a source we're missing. Three principles drive the build: proportional rates over raw counts, because fair comparisons across populations require it; community-informed development, because we build with communities rather than for them; and spotlighting missing data, because gaps can hide real health problems just as much as the numbers that are present.
The tracker exists because of a specific failure: when COVID-19 hit in 2020, it exposed how little centralized health equity data existed. Communities were dying at two to three times the rate of others, but the data behind that was scattered, incomplete, or missing entirely, and reported inconsistently across regions. HET was built to standardize that picture — tracking equity across health conditions instead of working in the disease-specific silos most trackers use, and pulling from vetted public partners including the CDC, CMS, CAWP, and the Bureau of Justice Statistics.
The Cobb County vs. Fulton County Case Study
I used two neighboring Georgia counties with very different health outcomes to make the historical argument concrete. Fulton County absorbed Milton and Campbell counties during the Great Depression, becoming far larger than Cobb County. The 1935 Social Security Act excluded Black workers from building generational wealth. 1938 redlining maps rated the neighborhood that now includes Morehouse's campus as "hazardous" — land values held for white residents and dropped for Black residents in the same area, for no reason but the rating itself. The Hill-Burton Act then funded hospitals in the white neighborhoods of these counties without scaling to match Fulton's much larger population.
The result shows up in HET's current data: Black residents in Fulton County face poverty rates 53.9% above the county average, while white residents sit 52.6% below it. HIV prevalence tells the same story — 71.8% among Black/African American residents versus a 40% national average, compared to 15.3% among white residents versus a 28.6% national average. The point I wanted fellows to sit with: data isn't neutral. It reflects who decided what got counted, and decades-old policy decisions still determine, in a very literal sense, who lives and who dies. Before collecting their own project data, I asked them to hold three questions: who benefits from this data collection, are the communities involved in the decision, and could this data harm the people it's about.
Making Data Meaningful for Different Audiences
The same statistic lands differently depending on who's hearing it, and I walked through how to reframe HET data for three audiences: for community members, translate a raw percentage into something concrete — "1 in 4 families in our neighborhood" instead of "25% prevalence"; for policymakers, frame it in terms of budget impact — "every dollar spent on prevention saves four dollars in ER costs"; for providers, frame it in terms of clinical action — "a two-minute screening catches 30% more cases."
I also flagged the red flags fellows should watch for in any data source, their own included: no stated methodology, emotional or subjective language, correlation presented as causation ("poverty causes diabetes" instead of "poverty correlates with diabetes through multiple pathways"), and counts presented without denominators (a county having the "most" cases means little without knowing the rate per 100,000 people).
Tour of the Tracker
I closed with a live walkthrough of HET itself: over 40 health topics at the national, state, county, and territory level, disaggregated by race, ethnicity, sex, and age, with insurance status, education, income, and urbanicity (metro vs. rural) as newer additions. I showed the different visualization types we support — choropleth maps, trend lines, population-vs-distribution maps, and maps that spotlight where data is missing rather than hiding the gap — since people process data differently and one chart type doesn't serve everyone. I also previewed what's coming: AI-generated plain-language summaries of each chart, a mobile app, and community data upload features, plus recent additions like gun violence and homicide data and maternal mortality.
Q&A
The strongest question was about how HET holds up as public health agencies face funding and data-collection cuts. My answer: our methodology and sourcing stay documented and current, missing data shows up as missing rather than papered over, and because HET is open source and free to host — built originally with support from Google.org and the CDC Foundation, plus a Google.org fellowship that helped staff the early build — it isn't dependent on any single funder's continued support to stay online. Allyson added context I hadn't fully covered myself: in its early development, the tracker brought in professionals across medicine, public health, and policy to shape what it would include, so the tool was designed with input from the start rather than retrofitted with community voice later.
Reflection
Presenting to doctors and experienced medical professionals felt genuinely intimidating going in — the AMA MJAF's application process is competitive enough that I told Allyson beforehand I didn't feel fully qualified for the room. The reception afterward was warm and motivating, which helped. But I measure the real impact differently: no one reached out afterward for help integrating HET into their capstone projects, and until that happens — whether from this talk or a repeat of it — I don't consider the job done. If the pattern holds across multiple sessions with no one taking HET into their own work, that's a signal the tracker's reach into rooms like this one isn't quite there yet.
Outcomes
- Recognize how data and analytics are useful or harmful in impacting policy and system outcomes. → the Cobb/Fulton case study and the "who benefits, who's harmed" framing
- Demonstrate ability to access, interpret, and translate publicly available data sources as advocacy assets. → the audience-specific reframing section (community / policymaker / provider)
- Identify credible sources of data and information relative to health equity. → HET's data partners (CDC, CMS, CAWP, Bureau of Justice Statistics) and the red-flags-to-watch-for section
Related work
Building a Zero-Cost Grant Monitoring Microservice
With our three team roles confirmed funded through FY27, I began transitioning into a grants-facing capacity alongside my UX engineering responsibilities, taking on the early infrastructure work that a dedicated grants professional would otherwise own.
Rather than monitoring dozens of foundation websites manually or investing in a grant database subscription that tracked historical awards rather than live open calls, I scoped and built Grant Watchdog: a lightweight automation tool that watches the live web pages of curated funder targets for changes, flags potential new opportunities, and generates a plain-language report the whole team can read.
Social Responsibilities Round Table Talk
From Data to Action: How Open-Source Health Data Platforms Democratize Research and Empower Community Advocacy
Open-source platforms like HET democratize research, transforming health equity from an academic concept to a community tool
Claim What Is Yours: A Trauma-Informed Content System for a High-Stakes Writing Task
Claim What Is Yours was a personal initiative I built to address a gap I'd lived through directly: veterans navigating VA disability claims — particularly for conditions tied to military sexual trauma — face a writing task that is procedurally demanding and, for many, close to impossible to approach directly. I designed a four-part content system (a free introductory guide plus a core guide with two companion modules) to walk survivors from "I don't know where to start" to a completed claim package, using AI tools as a structured writing partner along the way.
Description The core design problem wasn't information delivery — the VA's process is documented; it's the access gap between knowing the steps and being able to sit down and do them. I addressed that in a few specific ways: Staged disclosure architecture. Rather than asking a reader to write their most difficult account first, I sequenced the material so the hardest writing task (the formal MST stressor statement) comes after three lower-stakes "shadow work" exercises — third-person narration, a before/after contrast, and a behavioral-markers inventory — each one building material the reader can carry forward into the harder document, so no one starts at the hardest point. Structured AI-prompting as a writing scaffold. I wrote reusable, fill-in-the-blank prompt templates for every major document in the process (personal statements, nexus letter requests, buddy letter asks), each engineered to produce output in the reader's own voice rather than a generic or clinical one, and each paired with explicit guidance on what to check in the output before using it. Explicit safety and data-literacy guidance. Because the audience would be using AI tools for the first time, often around highly sensitive material, I included direct instruction on what not to disclose to an AI platform, how to evaluate a tool's data practices, and clear crisis-line signposting throughout — treating tool literacy as part of the content's job, not an afterthought. A tiered information structure. Free → core guide → bonus modules, so a reader in crisis could access something immediately useful at zero cost, with a deeper system available for those ready to go further.
Let's connect