01 · Source files
Raw inputs are stored in data/raw/us/{FY} and are never modified. Each fiscal year is parsed from the authoritative CMS document; TXT counterparts are used only as a second independent source for cross-validation.
| Fiscal Year | Rule | Source file | Sheet / Format |
|---|---|---|---|
| FY2023 | CMS-1771-F IPPS Final Rule | CMS-1771-F Table 5.xlsx | Single sheet |
| FY2024 | CMS-1785-F IPPS Final Rule | CMS-1785-F Table 5.xlsx | Single sheet |
| FY2025 | CMS-1808-F Correction Notice | FY2025 IPPS Final Rule & Correction Notice Table 5.xlsx | Sheet 0 (authoritative) |
| FY2026 | CMS-1833-F IPPS Final Rule | Table 5 addendum | Sheet 0 |
| FY2027 | CMS-1849-F IPPS Final Rule | Table 5 addendum | Sheet 0 |
02 · Methodology decisions
FY2025 uses the Correction Notice, not the Final Rule.
CMS issued a correction to the FY2025 IPPS Final Rule. The Correction Notice supersedes the Final Rule where values differ. Sheet 0 of the combined xlsx file contains the Correction Notice values and is used throughout this pipeline.
Cap-applied weight is the operative reimbursement weight.
CMS introduced a 10% year-over-year weight change cap starting in FY2024. Where a cap was applied, two weight columns exist: Weights — Before Cap and Weights — 10% Cap Applied. This pipeline uses the cap-applied weight, consistent with actual Medicare payment. FY2023 has no cap structure and its single weight column is used as-is.
DRG restructuring is preserved, not masked.
Some DRG codes appear in only one or two of the covered fiscal years due to CMS restructuring (e.g., cardiac defibrillator DRGs 222–227 were consolidated into 275–277 in FY2024). These are flagged with restructured=True in Silver and Gold. Year-over-year deltas are computed only where a DRG exists in both comparison years.
03 · Validation statuses
Every DRG row carries a validation_status:
04 · QA summary
QA is performed by etl/qa_cms_pipeline.py, which reports 74 passed, 0 failed, 0 warnings. That script is one of twelve gates run by etl/run_qa.py; the release invocation --require-raw reports 12 / 12 with zero skips. Any gate failure is a non-zero exit, so a build cannot ship past a red check.
05 · Dollar-impact model
Gold-layer outputs estimate reimbursement impact per DRG at volume tiers from 10 to 500 cases per year. Each estimate multiplies the FY-to-FY weight delta by a national standardized base rate.
06 · Known limitations
- Dollar-impact estimates use a national average base rate. Hospital-specific results will vary based on wage index, DSH adjustment, IME, and local market factors.
- OPPS (outpatient) data is staged but out of scope for this module. See the roadmap.
- This pipeline covers IPPS inpatient DRG weights only — not quality measures, VBP adjustments, or site-neutral payment policies.
- FY2023 and FY2024 are included for trend context; the ranked movers table and every dollar estimate are computed on FY2026 → FY2027.
- Anchor weight assertions cover all five fiscal years, FY2023 through FY2027.
- The FY2027 policy narrative in IPPS Policy Highlights is still extracted from the Proposed Rule (CMS-1849-P) and has not yet been re-extracted against the Final Rule text. Every weight, rate and dollar figure is Final; each policy row carries its own source badge.