Healthcare Metadata Generator: Generate Data Dictionary, Business Glossary & PHI Governance Report from Any SQL Schema

If you've ever inherited a healthcare data warehouse with zero documentation, you know the pain. Hundreds of columns like mbr_elig_ind, clm_adj_rsn_cd, and prvdr_credntl_sts — and not a single description in sight.

Today we're launching the Healthcare Metadata Generator — a free tool that turns any SQL schema into a complete data dictionary in under one second.

---

What It Does

Paste any healthcare SQL schema — a CREATE TABLE statement, a plain column list, a BigQuery DDL, or even a MongoDB JSON schema — and instantly get:

1. Business Metadata Every column gets a plain-English business description drawn from our 100,000+ ISO-11179 standard healthcare term definitions. Not generic AI guesses — real healthcare definitions written by data engineers for data engineers.

2. HIPAA PHI Classification Every column is automatically classified against all 18 HIPAA Safe Harbor identifiers:

• 🔴 CRITICAL — SSN, MBI, MRN (encrypt immediately)
• 🟠 HIGH — Names, DOB, phone, email (mask in non-production)
• 🟡 MEDIUM — Member IDs, dates, ZIP codes (review required)
• ✅ NONE — Codes, amounts, indicators (standard governance)

3. dbt YAML Ready-to-paste dbt documentation including column descriptions, PHI meta tags, and auto-generated tests (unique, not_null, accepted_values).

4. Governance Report A full Markdown governance report with PHI summary, Snowflake dynamic data masking DDL, and HIPAA compliance notes per column.

5. Data Dictionary Export Download a CSV or Markdown data dictionary with every field a data governance team needs: table name, entity name, column name, ISO standard name, attribute name, business definition, data type, PHI category, masking requirement, and confidence score.

6. ISO-11179 Rename SQL If your schema uses non-standard naming like memberEligibilityIndicator or DATE_OF_BIRTH, the tool generates the ALTER TABLE rename SQL to bring it in line with ISO-11179 healthcare data naming standards.

---

Who Is This For

• Healthcare data engineers documenting new tables before they hit production
• dbt developers who need schema.yml descriptions fast
• Data governance teams auditing PHI exposure across a data warehouse
• New team members trying to understand an undocumented schema
• Healthcare analytics consultants onboarding to a client's data environment

---

How It Works

Step 1 — Paste Your DDL

The tool accepts any format:

CREATE TABLE dim_member (
  mbr_key        INTEGER        NOT NULL,
  mbr_id         VARCHAR(50)    NOT NULL,
  mbr_first_nm   VARCHAR(100),
  mbr_dob        DATE,
  mbr_ssn        VARCHAR(9),
  mbr_elig_ind   BOOLEAN,
  mbr_raf_scr    DECIMAL(10,3),
  load_dt        TIMESTAMP
);

It also handles Snowflake DDL with COMMENT fields, BigQuery DDL with backtick table names, Oracle VARCHAR2 and NUMBER types, MongoDB JSON Schema, plain column lists, and CSV headers.

Step 2 — Resolution Pipeline

For each column the tool runs a 9-step resolution pipeline:

1. Composite term lookup — checks 111 known healthcare composites (dos, raf, mbi, hcc, etc.)
2. Database lookup — queries 100,000+ mdatool term definitions by column abbreviation
3. Entity prefix extraction — identifies mbr, prvdr, clm, rx, fac, and 150+ other healthcare entity prefixes
4. Suffix extraction — maps _ind, _cd, _amt, _dt, _scr and 85+ other data type suffixes
5. Attribute resolution — resolves elig, enrl, raf, dob and 280+ attribute abbreviations
6. PHI detection — applies HIPAA 18-identifier rules with severity classification
7. Confidence scoring — calculates 0-99% confidence using all resolution signals
8. Description generation — builds business descriptions from DB lookup or pattern matching
9. AI fallback — for columns scoring below 40%, Claude generates contextual descriptions using full table context

Step 3 — Get Your Output

Results appear in under 100ms. Average confidence across a standard healthcare schema is 90-99% for ISO-11179 compliant column names.

---

Example Output

For mbr_ssn VARCHAR(9):

Business Description: The nine-digit Social Security Number assigned to a health plan member by the Social Security Administration.

PHI Category: 🔴 CRITICAL HIPAA Identifier: Social Security Numbers (PHI #7) Masking Required: YES Confidence: 99% (DB Lookup)

Snowflake Masking DDL:

SHA2(mbr_ssn || 'SALT_KEY', 256) AS mbr_ssn

dbt YAML:

- name: mbr_ssn
  description: "The nine-digit Social Security Number..."
  meta:
    phi: true
    phi_category: "CRITICAL"
    masking_required: true

---

For mbr_raf_scr DECIMAL(10,3):

Business Description: The Risk Adjustment Factor score calculated for an individual health plan member representing their predicted healthcare cost relative to the average Medicare beneficiary.

PHI Category: ✅ NONE Confidence: 99% (DB Lookup)

---

Supported Platforms

The tool auto-detects your database platform from DDL syntax and normalizes accordingly:

---

What Makes This Different From ChatGPT

You can ask ChatGPT to describe healthcare columns. It will give you something reasonable. But it won't:

• Pull from 100,000+ curated healthcare-specific definitions
• Apply HIPAA 18-identifier PHI rules with masking DDL
• Generate Snowflake column-level security policies
• Produce ISO-11179 compliant rename scripts
• Give you a confidence score so you know what to review
• Export a complete data dictionary CSV with 18 fields per column
• Flag mbr_mbi_id as CRITICAL PHI and explain exactly why

The Healthcare Metadata Generator is purpose-built for healthcare data engineering — not a general-purpose AI that happens to know some medical terms.

---

Supported Schema Types

Beyond SQL DDL, the tool handles:

camelCase (FHIR / MongoDB / Java)

memberId → member_id → "Member identifier"
dateOfBirth → date_of_birth → "Date of birth — PHI"
eligibilityIndicator → elig_ind → "Eligibility indicator"

UPPER_SNAKE (Oracle legacy)

MBR_ELIG_IND → mbr_elig_ind → "Member eligibility indicator"
PRVDR_NPI → prvdr_npi → "National Provider Identifier"

Plain column list (no DDL)

mbr_id
mbr_first_nm
dos
raf
dual

CSV headers

member_id,first_name,date_of_birth,ssn,elig_ind,zip_code

---

The Data Dictionary Export

The CSV export includes every field your data governance team needs:

This CSV opens directly in Excel and can be imported into data catalog tools like Alation, Collibra, or Atlan.

---

What's Coming Next

• Auto-learning: Columns that get AI-generated descriptions are tracked. The most common unknown columns are automatically added to the term database so future users get 99% confidence instead of AI fallback.
• Batch processing: Upload multiple DDL files at once for warehouse-wide documentation
• Atlan / Collibra export: Direct push to your data catalog via API
• dbt package: Run the metadata generator directly from your dbt project

---

Try the Healthcare Metadata Generator

The Healthcare Metadata Generator is completely free — no login required. Paste your schema and get results in under one second.

Try the Healthcare Metadata Generator →

---