Writing

Long-form writing on healthcare data engineering, Medicare Advantage Stars methodology, and the places where CMS Technical Notes and production reality diverge.

2026-04-19 The Slowest Way to Learn Web Design Is the Only Way That Worked for Me A personal history of building and rebuilding a website across three eras, from Squarespace drag-and-drop to a pure HTML and CSS rebuild, and what the increasing friction of each step actually taught me.
2026-04-12 How the Star Rating Predictor Works The statistical model behind the interactive Star Rating predictor: ordinal logistic regression calibrated to CMS 2025 weights, running entirely in your browser.
2026-04-05 Building a Browser-Based Star Rating Predictor: Methodology & Evidence Technical documentation and literature review supporting the ordinal logistic regression model used in the interactive Medicare Star Rating predictor on this site.
2026-03-29 HEDIS Measure-Level ETL Patterns: Building the Pipeline from Claims to Measure Rates A practitioner's walkthrough of how HEDIS measure pipelines actually work -- from eligible population builds and numerator event matching to exclusion logic, supplemental data integration, and rate calculation.
2026-03-22 Two Ways to Build a Web App (and What You’re Actually Choosing) Most applications boil down to server-rendered or API-driven designs. The difference isn’t just technical, it shapes how your system evolves.
2026-03-15 The ECDS Shock Index: Modeling Distribution Risk in Medicare Advantage Stars ECDS adoption is not just a data modernization effort. It introduces measurable distribution risk into the Stars ecosystem.
2026-03-08 Concurrency: The Story We Tell vs. the System We Run Concurrency sounds like “do more at once.” In real data systems it’s mostly coordination under constraint, and in healthcare analytics the constraints include trust, auditability, and deadlines.
2026-03-01 Why Pull Requests Changed Everything in Healthcare Data Science How version control and pull requests transformed healthcare analytics from chaotic file management into transparent, collaborative systems that save lives.
2026-02-22 Why You Probably Shouldn't Version Your Fact Table Most teams version their fact tables because they've modeled their dimensions incorrectly. Understanding the true grain of your table—whether it's event-based or state-based—will save you from double-counting metrics, failed audits, and painful refactoring projects in healthcare data warehouses.
2026-02-15 Debugging GitHub Actions: The Heredoc Horror Story A deep dive into debugging a GitHub Actions workflow where heredoc syntax clashed with YAML parsing, leading to five different errors and one elegant solution.
2026-02-08 LLM Inference Is Not Just Bigger Inference Why serving large language models is a fundamentally different systems problem from traditional ML inference, especially in applied healthcare IT.
2026-02-01 What LLM Systems Teach Healthcare IT About Architecture Large language models force healthcare IT teams to rethink batching, memory, routing, and state management at production scale.
2026-01-25 Security Starts With the Problem You’re Trying to Solve A brief description of your blog post (1-2 sentences)
2026-01-18 Why CMS Never Let STAR Ratings Touch Meaningful Use On the surface, it looks like an omission - but CMS understood that capability and outcomes cannot be governed the same way, even when one enables the other. The deliberate separation was probably the right call.
2026-01-11 The Gravity of STAR Ratings How outcomes shape programs they never appear in - STAR Ratings operate less like a measurement program and more like a financial force that influences clinical workflows, data governance, and analytics investments across healthcare.
2026-01-04 America First, Global Health Last A quantitative analysis of documented international health program disruptions, funding cuts, and their human toll during 2025
2025-12-28 The Dismantling of America's Public Health Shield A quantitative analysis of documented workforce reductions, budget cuts, and health outcomes in the U.S. public health system during 2025
2025-12-21 We Were Bending the Cost Curve. Then We Looked Away. For a brief, underappreciated period, Medicare and Medicaid showed real signs of cost stabilization. This post examines what changed, why it mattered, and how we quietly abandoned the lessons when the trajectory shifted.
2025-12-14 From Lambda to Kappa: What Modern Data Architecture Really Means for Healthcare Tech Why the shift away from Lambda architecture matters, what Kappa gets right (and wrong), and how healthcare APIs force us to be more pragmatic than dogmatic.
2025-12-07 Kimball vs. Inmon in Healthcare Tech: Warehouses, Marts, and APIs Walk Into a Clinic A practical, slightly opinionated take on Kimball and Inmon, viewed through healthcare data platforms, APIs, and the realities of shipping software.
2025-11-30 Making PostgreSQL Connections More Resilient with TCP Keepalive Why idle database connections fail, and how TCP keepalive tuning prevents silent disconnects in PostgreSQL.
2025-11-23 I Don’t Actually Care About Making Change Solving the same problem in Haskell, Python, and SQL reveals less about coins and more about how programming languages shape how we think about correctness, process, and structure.
2025-11-16 Deletes Are Data Too Deletes are the quiet failure mode in data pipelines. Here are the main ways to detect them and the tradeoffs you’re actually signing up for.
2025-11-09 Publicly Available but Practically Unavailable Why healthcare executives repeatedly ask for data that technically exists, yet remains inaccessible in usable form—and how misunderstanding 'public data' quietly undermines quality, cost, and governance decisions.
2019-03-07 Weighting Provider Panels Across Primary Care UW Health's innovative approach to patient empanelment using data-driven weighting methodologies that account for age, gender, and insurance factors.

Earlier writing (19 posts, 2009–2011) →

Experiments

Small interactive pages that sit outside the long-form writing.

Life in Weeks A life, 4,680 weeks. Each square is one. Inspired by Tim Urban.