# ARKA Documentation Concordance — Data request

**Turnaround: under a week from receipt of a clean extract.**  
**Analyst time: under an hour** (Epic Clarity or Oracle Health).  
**Cost: nothing.**

This is the extract we need to run the free Documentation Concordance Report on your imaging orders. An Epic analyst who already pulls radiology throughput extracts can produce it without a call. Exact Clarity / Oracle Health SQL lives in `tools/ire-baseline/README.md`.

## What to send

One flat file (CSV preferred) with **one row per imaging order**. Optional second file for de-identified note text when you want Documentation Concordance Score (DCS) and the missing-element callout.

### Required: orders CSV

| Field | Required? | Notes |
|---|---|---|
| `order_id` | **Required** | Site order id (opaque; not an MRN). |
| `ordered_at` | **Required** | Order timestamp (ISO-8601 preferred). |
| `modality` | **Required** | e.g. CT, MRI, XR, US. |
| `body_region` | **Required** | e.g. chest, abdomen, head. |
| `laterality` | Optional | L / R / blank. |
| `typed_indication` | **Required** | Free-text requisition indication as typed. |
| `ordering_clinician_hash` | **Required** | **Hash only** (SHA-256 or HMAC of provider id). Never a name or NPI plaintext. |
| `service_line` | **Required** | ED, inpatient, ambulatory, … |
| `encounter_class` | **Required** | emergency, inpatient, ambulatory, … |

Window: **90 days** of advanced-imaging orders is enough. Six months is welcome; do not invent a multi-quarter project.

### Optional: notes CSV

| Field | Required? | Notes |
|---|---|---|
| `order_id` | **Required** | Must join the orders file. |
| `note_text` | **Required** | Already de-identified by you. Used only for reconstructed grade + DCS. |

Without notes we still return your typed RI-RADS distribution vs the published inpatient benchmark. With notes we also return DCS, discordance (separately), and the single element most often missing from your requisitions.

## Nothing else

Do **not** send:

- Patient names, MRNs, government IDs, addresses, phones, or emails
- Ordering clinician names or plaintext NPIs (hash at source)
- Day-precision dates of birth
- Anything that is not in the tables above

Column names can be yours — send a one-page mapping. We do not need a code change to onboard you.

## PHI, integration, commitment

| | |
|---|---|
| **PHI** | None required. Hash clinician ids before the file leaves your system. |
| **Integration** | None. No interface, no FHIR scopes, no security review of a live connection. |
| **Commitment** | None. You may run `tools/ire-baseline/` inside your firewall and never send ARKA anything — it makes zero network calls. |
| **Small cells** | Any stratum with n < 11 is suppressed and the suppression is stated. |
| **Cost** | Zero. |

**And if the number isn't there, we will tell you, and we will stop.**

## What you get back

A Documentation Concordance Report (HTML + PDF + JSON):

1. Your RI-RADS distribution, benchmarked against the published inpatient series (n = 762).
2. Your Documentation Concordance Score (when notes are supplied) — completeness gap and discordance **separately**.
3. The single clinical element your requisitions most often omit.

Public worked example (synthetic): download from `/baseline` or `GET /api/ire/sample-concordance-report`.

## Who this door is for

Any organisation that orders imaging — including fee-for-service hospitals, associate CMOs, radiology chairs, and CDI — not only risk-bearing entities with claims authority. The claims-based ledger door (`docs/arka-tcoc/DATA_REQUEST.md`) remains the path for attributed-population TCOC buyers.

## Module 3 connection (A.4)

When you later deploy Accountable Justification (Module 3), a verified write to `ServiceRequest.note` lands on the same requisition this report grades. That shared surface is measured as a secondary finding (RI-RADS / concordance vs ceiling-suppressed controls) only after a live tenant produces real pairs — see `docs/ARKA_Documentation_Concordance_Onepager.md`.

## Appendix B marginal fields — suspected-PE / CTPA (Study 2)

ARKA-IP Protocol Appendix B (suspected-PE / CTPA) is scored from the **same imaging-orders extract**, not a second negotiation. Study 1 (discharge gate) and Study 2 (CTPA) share consent language, DUA/BAA, and pre-registration discipline. The argument for the site is the count below.

**Marginal field count: four.** One recommended for reliable CTPA identification; three optional for pathway and yield. Nothing else. If a column is absent, the report states the limitation and still returns what the existing columns support — it does not invent a multi-week data project.

| Field | Required? | Notes |
|---|---|---|
| `cpt_or_proc_code` | **Recommended** | CPT 71275 or local PE-protocol CT chest code. Without it, CTPA is inferred from modality + body region + study description and the limitation is stated. |
| `pretest_probability_band` | Optional | `low` / `intermediate` / `high` / `not-documented` (Wells, revised Geneva, or site band). Structured only — free-text indication alone does not count. |
| `d_dimer` | Optional | Resulted value (positive / negative / numeric with units) **and** result timestamp. Counts toward pathway completeness only when the timestamp is strictly before `ordered_at`. |
| `ctpa_result_class` | Optional | `positive-pe` / `negative` / `nondiagnostic` / `unknown`. Yield secondary only; unknown is never folded into negative. |

**Four fields.** No names, no MRNs, no note text beyond what you already chose to send for concordance. Claims-door buyers who already pull the TCOC extract may recover D-dimer as a lab line; the structured columns above remain preferred and are labelled `measured`.

## Next step

Reply with a time, or forward this one-pager to whoever owns clinical documentation integrity / radiology informatics. Kit README with Clarity SQL: `tools/ire-baseline/README.md`.