GHS Prices Update - Annual Process

This document describes the annual process for updating GHS (Groupes Homogènes de Séjours) prices and grouping tables when new versions are published by the French health authorities.

Overview

The French health authorities typically publish new GHS tables around May each year, with retroactive effect to March of the same year. This creates a transition period where we need to carefully manage the update process.

Problem Statement

• Publication Timing: New 20XX tables are published late (around May 20XX)
• Retroactive Application: Tables are retroactive to March 20XX
• Technical Constraint: Setting cutoff to January 1st would cause errors (table not available yet)
• Current Solution: Use June 1st as cutoff to ensure table is published before activation

Annual Update Process

1. Preparation Phase (Before May)

Set Up Next Year's Cutoff Date

In the coding service, the cutoff dates are pre-configured with conservative dates:

// services/coding/src/domain-layer/coding/coding.constants.ts
export const GROUPING_CUTOFF_DATES = [
  { cutoffDate: new Date('2025-01-01'), referenceYear: 2025 },
  { cutoffDate: new Date('2026-06-01'), referenceYear: 2026 }, // Conservative date
  { cutoffDate: new Date('2027-06-01'), referenceYear: 2027 }, // Conservative date
];

The June 1st date provides a buffer period after the expected May publication.

2. Data Collection Phase (May - After Publication)

Obtain the Official GHS Tables

1. Download the official GHS price tables from ATIH (Agence Technique de l'Information sur l'Hospitalisation)
2. Convert the data to the required JSON format
3. Save as ghs-YYYY-prices.json in the valuation service

Required Data Format

[
  {
    "ghsCode": 1234,
    "referenceYear": 2025,
    "ghm": "01M01Z",
    "description": "Description of the GHM",
    "upperDayCountThreshold": 10,
    "lowerDayCountThreshold": 2,
    "upperExtremePrice": 5000.0,
    "lowerExtremePrice": 1000.0,
    "defaultPrice": 3000.0
  }
]

3. Database Update Phase

Step 1: Prepare the Seed File

Place the new prices file in the valuation service:

services/valuation/prisma/seed/seed-data/ghs-YYYY-prices.json

Step 2: Create/Update Population Script

If needed, create a new population script (or update the existing one):

// services/valuation/prisma/scripts/populate-prices-YYYY.ts
export async function populatePricesYYYY(prisma: PrismaClient) {
  const filePath = path.resolve(
    __dirname,
    '../seed/seed-data/ghs-YYYY-prices.json',
  );
  const ghsPrices: IPriceRow[] = JSON.parse(fs.readFileSync(filePath, 'utf8'));

  console.log(`Seeding ${ghsPrices.length} GHS prices...`);
  await runPromisesInParallelBatches<IPriceRow, void>(
    ghsPrices,
    async (price) => {
      await prisma.price.upsert({
        where: {
          ghsCode_referenceYear: {
            ghsCode: price.ghsCode,
            referenceYear: price.referenceYear,
          },
        },
        update: priceData,
        create: priceData,
      });
    },
    50,
  );
}

Step 3: Run the Population Script

1. Update the main script to import and call the new function:

// services/valuation/prisma/scripts/index.ts
import { populatePricesYYYY } from './populate-prices-YYYY';

async function main() {
  console.log('Running Prisma scripts...');
  await populatePricesYYYY(new PrismaClient());
}

2. Execute the script:

cd services/valuation
npx ts-node prisma/scripts/index.ts

4. Cutoff Date Update Phase

Update the Cutoff Date to Retroactive Date

Once the table is successfully loaded in the database, update the cutoff date to the actual retroactive date (typically March):

// services/coding/src/domain-layer/coding/coding.constants.ts
export const GROUPING_CUTOFF_DATES = [
  { cutoffDate: new Date('2025-01-01'), referenceYear: 2025 },
  { cutoffDate: new Date('2026-03-01'), referenceYear: 2026 }, // Updated to retroactive date
  { cutoffDate: new Date('2027-06-01'), referenceYear: 2027 },
];

5. Recomputation Phase

Trigger Recomputation of Existing Groupings

After updating the cutoff date, existing medical stays that fall within the retroactive period need to be recomputed:

1. Identify Affected Records: Find all medical stays with discharge dates between the new cutoff date (e.g., March 1st) and the old cutoff date (e.g., June 1st)

2. Trigger Re-evaluation: For each affected optimization, trigger a re-evaluation to recalculate the GHM and pricing using the new tables

3. Monitor Progress: Track the recomputation process to ensure all affected records are updated

WARNING

The specific recomputation script is to be developed. This process should:

• Query affected medical stays
• Re-trigger grouping calculations
• Update optimization gains
• Emit appropriate events