---
license: cc-by-4.0
language:
  - en
task_categories:
  - text-generation
  - table-question-answering
tags:
  - text-to-sql
  - clinical
  - fhir
  - healthcare
  - synthetic
  - duckdb
  - synthea
size_categories:
  - 10K<n<100K
configs:
  - config_name: default
    data_files:
      - split: train
        path: train.jsonl
      - split: heldout_familiar
        path: heldout_familiar.jsonl
      - split: heldout_unseen
        path: heldout_unseen.jsonl
---

# FHIR-to-SQL with Database-Resolved Clinical Terminology

Natural-language hospital questions paired with a **structured JSON query plan and compiled DuckDB
SQL**, over a FHIR-derived clinical schema. Built entirely from **synthetic** Synthea patients — no
real patient data.

Companion to the paper *Plan-Then-Compile: Turning a General-Purpose Coder Model into a FHIR Data
Analyst*.

- **Code & paper:** https://github.com/adelelsayed/fhirsql-reasoning-sql
- **Adapters:** https://huggingface.co/adelelsayed1991/fhirsql-reasoning-sql-adapters

## What makes this dataset different

Gold SQL **never embeds a literal clinical code**. Every query resolves its concept at runtime
through a lookup against a `valuesets` terminology table:

```sql
WITH resolved AS (
    SELECT code, code_system FROM valuesets
    WHERE table_name = 'procedure' AND display ILIKE '%Depression screening (procedure)%'
)
SELECT DISTINCT patient_id
FROM procedure, resolved
WHERE procedure.code = resolved.code AND procedure.system = resolved.code_system
  AND status = 'completed'
```

This turns terminology resolution from a memorization problem into a queryable one, and makes
generalization to clinical concepts absent from training measurable.

## Splits

| Split | Rows | Answerable | Abstention |
|---|---:|---:|---:|
| `train` | 10,696 | 9,680 | 1,016 |
| `heldout_familiar` | 9,300 | 8,284 | 1,016 |
| `heldout_unseen` | 2,156 | 1,140 | 1,016 |

Held-out splits come from a **disjoint 6,383-patient population** (training: 18,999 patients; zero
`patient_id` overlap).

- **`heldout_familiar`** — clinical concepts also used in training, different patients. Isolates
  population generalization. **96.5% of its answerable rows are byte-identical `(question, gold_sql)`
  pairs from `train`**, so it measures whether learned queries stay correct on new data; it cannot
  distinguish memorization from generalization.
- **`heldout_unseen`** — 84 clinical concepts appearing **nowhere** in training (verified by set
  intersection; zero shared question/query pairs). This is the split for measuring transfer.

⚠️ **The 1,016 abstention rows are the same rows in all three splits.** Unanswerable questions
reference no patient data, so an identical set was reused. Abstention metrics computed on the
held-out splits therefore measure *retention* of trained refusal behavior, not held-out refusal
generalization.

⚠️ **Execution match is weakly discriminative on `heldout_unseen`.** Its concepts are rare (1–3
patients each) and gold answers are small integers: 53.9% of within-archetype concept pairs return
identical results, so a wrong concept can go undetected. Text-match metrics are not subject to this.

## Fields

| Field | Description |
|---|---|
| `question` | Natural-language question, phrased for one of 11 hospital roles |
| `target` | Training target: `json.dumps(plan, indent=2)` + fenced ` ```sql ` block |
| `gold_plan` | Structured query plan (entities, joins, constraints, aggregation, abstain) |
| `gold_sql` | Compiled DuckDB SQL, or the literal `UNANSWERABLE` token |
| `archetype_id` | Query archetype (82 total: 64 answerable + 18 unanswerable) |
| `tier` | Difficulty tier 1–4, or 5 for unanswerable |
| `persona` | Hospital role the phrasing targets |
| `concept_display` | The clinical concept instantiated (empty for structural/unanswerable) |
| `schema_ref` | Schema the query is written against |

## Provenance

Every answerable row's gold SQL was **executed against the real database** and kept only if it ran
without error and returned at least one row, then checked for repeated-execution stability.
Questions come from 160 role-appropriate phrasings authored with LLM assistance under the author's
direction, then parameterized by concept.

`schema.sql` is the exact schema injected into every prompt. `valuesets_ddl.sql` is an ablation
fragment that was **not** part of the prompt during the published runs — see the paper, §5.6.

## Limitations

Synthetic data only; one question-generation distribution (templated phrasing, not free-form
clinician language); terminology resolution is substring matching against a corpus-derived
dictionary with one display string per code, so synonymy, hierarchy, and ambiguity are out of scope
(ambiguous concepts are excluded from gold). Some regulatory questions say "dispensed" but are
answered from `MedicationRequest` (prescription orders). Full discussion in the paper's §7.

## Citation

See `CITATION.cff` in the GitHub repository.