What was built, and what it replaced.

DRG Impact Navigator

Static site · precomputed JSON · Python build tooling · Vercel

The problem, in the buyer's words

CMS dropped the FY2026 IPPS final rule. What does it do to us?

Every hospital finance team asks this once a year and most cannot answer it cleanly. The rule is published as a set of tables, not as a number. Turning it into "our inpatient revenue moves by this much" requires holding the caseload constant across two rate years, and almost nobody does that. So the answer that reaches the board is contaminated by volume changes and is not really a policy read at all.

The manual process

For a single hospital: pull the CMS Medicare Inpatient Hospitals by Provider and Service file for the caseload, pull FY2025 and FY2026 IPPS final rule standardized amounts and Table 5 relative weights, pull that provider's row from the CMS impact files for wage index, DSH, teaching, capital and uncompensated care, then compute payment per MS-DRG twice and difference it. Then repeat the whole thing for a market comparison. Roughly two days of analyst time, and that assumes you already know where the files live.

At a loaded rate of $100 an hour that is about $1,600 per hospital in internal cost, against $3,500 to $5,000 if the same question is quoted out as a small project. Across all 2,833 IPPS hospitals it would be roughly $4.5M of analyst time, which is why nobody has ever done it exhaustively. The tool did not save one hospital two days. It made an analysis exist that otherwise would not.

What runs now

A public URL with no login. Type a hospital name and the repriced number renders in under a second, with the FY2025 to FY2026 bridge, the payment components, the MS-DRGs that moved the number most, a scenario panel for volume, Part A deductible and sequestration, and a market table. A board-ready one-page brief generates as a printable PDF.

The same 2024 cases are priced at both rate years, so every dollar of change is CMS policy rather than the hospital's volume. That constraint is the entire point of the tool.

What broke

• Non-IPPS hospitals returned "No matches," which reads as a broken tool. Long-term care, rehabilitation, psychiatric, children's and critical access hospitals are paid under entirely different systems and were never in scope, but the tool said nothing and simply looked wrong. Rebuilt to classify the facility from its CCN and name the payment system that actually covers it. This was the single most valuable fix in the project.
• The lead-capture modal rendered on page load. A hidden attribute was being overridden by display:flex, so the gate appeared before anyone had looked anything up.
• Search failed on spelling variants. "St" against "Saint" split the same facility into misses.
• Retired MS-DRGs cannot be followed one-to-one across rate years, mainly the FY2025 spinal fusion restructuring. Most tools silently drop those cases. This one reports them separately as a crosswalk estimate and keeps them out of the headline number.
• Deployment broke when the repository went private, and nearly committed local Vercel and environment files in the process.

What it says about the build

Three of the five failures above are not about the pricing model. They are about gracefully explaining what the tool does not cover. The math was the straightforward part. Handling the cases outside the math is what separated a working tool from one that looked broken to a third of the people who tried it.