README.md

---
license: cc-by-nc-4.0
task_categories:
- tabular-classification
- tabular-regression
language:
- en
tags:
- synthetic
- healthcare
- insurance
- payer-operations
- claims-adjudication
- prior-authorization
- fraud-detection
- ahip
- cms
- nhcaa
- ncqa
- hcup
- caqh
- icd-10
- cpt
- drg
- hcc
- raf
- hedis
- value-based-care
- shared-savings
- capitation
- bundled-payment
- denial-management
- carc
- cob
- coordination-of-benefits
- mlr
- 30-day-readmission
- siu
- special-investigations-unit
pretty_name: HLT-015 Synthetic Insurance Medical Claims Dataset — Payer Operations (Sample Preview)
size_categories:
- 1K<n<10K
---

# HLT-015 — Synthetic Insurance Medical Claims Dataset — Payer Operations (Sample Preview)

**A free, schema-identical preview of the full HLT-015 commercial product from [XpertSystems.ai](https://xpertsystems.ai).**

A **fully synthetic** enterprise-grade payer-side healthcare claims dataset combining **member eligibility / enrollment**, **prior authorization workflows**, **end-to-end claim adjudication** (Professional / Institutional / Pharmacy / Dental / Vision / Behavioral Health), **fraud detection with 12 NHCAA typologies**, **value-based care contract attribution** (5 HHS LAN APM categories), and **HEDIS quality measure tracking** — calibrated to AHIP / CMS / NHCAA / NCQA / HCUP / CAQH benchmarks.

> ⚠️ **PRIVACY & SYNTHETIC NATURE**
> Every record in this dataset is **100% synthetic**. **No real patient data, no PHI, no real provider NPIs, no real claim records.** Population-level distributions match published AHIP / CMS / NHCAA / NCQA / HCUP / CAQH / FAIR Health references but the claims, members, and authorizations are computationally generated.

---

## How does this differ from HLT-008 Healthcare Claims?

Both products cover synthetic healthcare claims, but from different operational perspectives:

| Aspect | HLT-008 (Provider/X12 Billing) | HLT-015 (Payer Operations) |
|---|---|---|
| **Buyer persona** | Provider revenue cycle, X12 EDI integrators | Payer ops, prior auth admin, fraud SIU |
| **Schema focus** | X12 837/835 EDI structure, HCC risk adjustment, CMS CCW chronic conditions | Claim adjudication workflow, prior auth, fraud typology, value-based care attribution |
| **Member view** | 21 CCW chronic conditions, CMS HCC risk score | 15 chronic conditions, RAF + expected annual cost |
| **Fraud taxonomy** | 5 NHCAA patterns (Upcoding/Phantom/Unbundling/Duplicate/Identity_Theft) | 12 NHCAA typologies (full taxonomy including DME, Kickback, Provider Impersonation, etc.) |
| **Quality measures** | HEDIS 5 measures (BCS, COL, CDC, AWV, Depression) | NCQA HEDIS extended (SPC, MPT, TRC, CDC, PCR) |
| **Workflow coverage** | Claim adjudication outcome | Full lifecycle: eligibility → PA → submission → adjudication → fraud SIU → VBC attribution |
| **Adjudication engines** | n/a | Tracked: which payer engine adjudicated |
| **VBC attribution** | n/a | 5 contract types (FFS/Shared Savings/Capitation/Bundled/P4P) |
| **Prior authorization** | n/a | Separate workflow table with peer-to-peer + appeals |
| **Schema width** | 54 cols medical claims | 52 cols claims + 33 cols members + 16 cols prior auth |

**Use HLT-008** for X12 EDI pipeline development, provider revenue cycle, and HCC risk adjustment ML.
**Use HLT-015** for payer operations analytics, prior auth optimization, fraud SIU investigation, and value-based care contract performance modeling.

---

## What's in this sample

**5,000 claims × 300 members × 12-month observation window** linked by `member_id`.

| File | Rows × Cols | Description |
|---|---|---|
| `members.csv` | 300 × 33 | Member master — demographics, coverage tier, deductible/OOP/copay, RAF risk score, 15 chronic conditions, expected annual cost |
| `claims.csv` | 5,000 × 52 | Full claim lifecycle — submission_channel, adjudication_engine, ICD-10 + CPT + DRG, financials, COB, denial categories, 12 fraud typologies, HEDIS, VBC |
| `prior_auths.csv` | ~250 × 16 | Prior authorization workflow — turnaround days, peer-to-peer, appeals, clinical criteria evaluation |

**Total:** ~1.9 MB across 3 CSVs + scorecard JSON.

---

## Schema highlights

### `members.csv` (33 columns)

**Identity & demographics:** `member_id`, `date_of_birth`, `age`, `sex`

**Coverage:** `payer_type` (Commercial / Medicare_Advantage / Medicaid_MCO / ACA_Marketplace), `plan_type`, `coverage_tier`, `enrollment_start_date`, `enrollment_end_date`

**Cost-sharing:** `deductible_individual`, `deductible_family`, `oop_maximum_individual`, `copay_primary_care`, `copay_specialist`, `coinsurance_rate`

**Risk:** `risk_score_raf` (CMS HCC normalized mean=1.0), `expected_annual_cost`

**15 chronic conditions (binary flags):** `cc_diabetes`, `cc_hypertension`, `cc_hyperlipidemia`, `cc_ischemic_heart_disease`, `cc_heart_failure`, `cc_atrial_fibrillation`, `cc_copd`, `cc_asthma`, `cc_ckd`, `cc_depression`, `cc_anxiety`, `cc_obesity`, `cc_osteoporosis`, `cc_stroke`, `cc_cancer`, plus `chronic_condition_count`

### `claims.csv` (52 columns)

**Identity & workflow:** `claim_id`, `member_id`, `claim_type` (6 types: Medical_Professional, Medical_Institutional, Pharmacy, Dental, Vision, Behavioral_Health), `claim_status` (Paid/Denied/Pended/Adjusted), `submission_channel` (Electronic_EDI / Provider_Portal / Paper), `submission_date`, `service_date_from`, `service_date_to`, `adjudication_date`, `adjudication_engine`, `adjudication_turnaround_days`

**Provider:** `rendering_provider_npi`, `billing_provider_npi`, `provider_taxonomy_code`, `provider_specialty`, `place_of_service_code`, `network_status` (In_Network / Out_of_Network / Emergency)

**Clinical coding:** `primary_diagnosis_code` (ICD-10-CM), `secondary_diagnosis_codes`, `procedure_code` (CPT/HCPCS), `drg_code`, `formulary_tier`

**Member context:** `payer_type`, `plan_type`, `member_age`, `member_sex`, `risk_score_raf`

**Denial:** `denial_code`, `denial_category` (8 categories: Medical_Necessity / Timely_Filing / Duplicate / COB / Network / Eligibility / Authorization / Coding_Error)

**Financials:** `billed_amount`, `allowed_amount`, `paid_amount`, `member_deductible_applied`, `member_copay_applied`, `member_coinsurance_applied`, `contractual_adjustment`, `cob_primary_paid`, `cob_secondary_paid`

**Fraud (NHCAA):** `fraud_label`, `fraud_typology` (12 types), `fraud_risk_score`, `anomaly_flags`, `siu_referral`, `provider_fraud_risk_tier`

**Quality & cost:** `high_cost_claimant_flag`, `readmission_30d_flag`, `preventable_admission_flag`, `mlr_contribution`, `leakage_flag`, `generic_substitution_flag`, `hedis_measure_triggered`, `value_based_contract_type` (5 categories per HHS LAN)

### `prior_auths.csv` (16 columns)

`auth_id`, `member_id`, `auth_request_date`, `auth_decision_date`, `auth_turnaround_days`, `auth_procedure_category`, `auth_urgency`, `auth_decision`, `auth_denial_reason`, `auth_units_requested`, `auth_units_approved`, `peer_to_peer_requested`, `appeal_filed`, `appeal_outcome`, `clinical_criteria_met`, `payer_type`

---

## Calibration source story

The full HLT-015 generator anchors all distributions to authoritative payer industry references:

- **AHIP 2023** (America's Health Insurance Plans) — Claim denial rates, in-network shares
- **CMS HCC v28** — Hierarchical Condition Categories risk adjustment
- **NHCAA** (National Health Care Anti-Fraud Association) — 12-typology fraud taxonomy
- **X12 835 (HIPAA EDI)** — CARC denial codes
- **CMS HRRP** (Hospital Readmissions Reduction Program) — 30-day readmission benchmarks
- **CAQH CORE 2023** — EDI 837 adoption rates
- **HHS LAN APM Framework** — Value-based care contract categories
- **NCQA HEDIS** — Healthcare Effectiveness Data and Information Set quality measures
- **CMS CCW** — Chronic Conditions Warehouse
- **FAIR Health** — Independent medical procedure pricing reference

### Sample-scale validation scorecard

| Metric | Observed | Target | Status | Source |
|---|---|---|---|---|
| Fraud prevalence | 3.32% | 3% ± 1.5% | ✅ PASS | NHCAA |
| Claim denial rate | 17.4% | 18.5% ± 4% | ✅ PASS | AHIP 2023 |
| RAF score mean | 1.0000 | 1.0 ± 0.05 | ✅ PASS | CMS HCC v28 |
| In-network share | 87.3% | 87% ± 8% | ✅ PASS | AHIP |
| EDI submission share | 85.0% | 85% ± 8% | ✅ PASS | CAQH CORE 2023 |
| 30-day readmission | 14.9% | 15% ± 5% | ✅ PASS | CMS HRRP |
| Fraud typology count | 12 | 12 (NHCAA) | ✅ PASS | NHCAA taxonomy |
| Denial category count | 8 | 8 (X12) | ✅ PASS | X12 835 / CARC |
| VBC contract type count | 5 | 5 (HHS LAN) | ✅ PASS | HHS LAN APM |
| Chronic condition count | 15 | 15 (CCW) | ✅ PASS | CMS CCW |

**Grade: A+ (100/100) — verified across 6 random seeds (42, 7, 123, 2024, 99, 1).**

---

## Loading examples

### Pandas — explore claim mix

```python
import pandas as pd

members = pd.read_csv("members.csv")
claims = pd.read_csv("claims.csv")
prior_auths = pd.read_csv("prior_auths.csv")

# Claim type & status breakdown
print(pd.crosstab(claims["claim_type"], claims["claim_status"], normalize="index").round(3))

# Payer-stratified denial rates
print(claims.groupby("payer_type").apply(
    lambda d: (d["claim_status"] == "Denied").mean()
).round(3))

# Fraud typology counts
print(claims.loc[claims["fraud_label"] == 1, "fraud_typology"].value_counts())
```

### Fraud SIU referral targeting

```python
import pandas as pd

claims = pd.read_csv("claims.csv")

# High-risk claims for SIU review
siu_candidates = claims[
    (claims["fraud_risk_score"] > 0.5) |
    (claims["siu_referral"] == 1) |
    (claims["provider_fraud_risk_tier"] == "High")
]
print(f"SIU candidates: {len(siu_candidates)}")
print(siu_candidates[["claim_id", "fraud_typology", "fraud_risk_score",
                     "billed_amount", "anomaly_flags"]].head(10))
```

### Prior auth turnaround analysis

```python
import pandas as pd

pa = pd.read_csv("prior_auths.csv")

# Turnaround by urgency
print(pa.groupby("auth_urgency")["auth_turnaround_days"].agg(["mean", "median", "max"]).round(2))

# Approval rates
print(pa["auth_decision"].value_counts(normalize=True).round(3))

# Peer-to-peer conversion (denial → P2P → final outcome)
denials = pa[pa["auth_decision"] == "Denied"]
print(f"\nDenials: {len(denials)}")
print(f"Peer-to-peer requested: {denials['peer_to_peer_requested'].sum()}")
print(f"Appeals filed: {denials['appeal_filed'].sum()}")
print(f"Appeal outcomes: {denials.loc[denials['appeal_filed'] == 1, 'appeal_outcome'].value_counts().to_dict()}")
```

### Value-based care contract performance

```python
import pandas as pd

claims = pd.read_csv("claims.csv")

# Claim outcomes by VBC contract type
print("Mean paid / billed by VBC type:")
print(claims.groupby("value_based_contract_type")[["billed_amount", "paid_amount"]].mean().round(2))

# MLR contribution by VBC type
print("\nMLR contribution sum by VBC type:")
print(claims.groupby("value_based_contract_type")["mlr_contribution"].sum().round(2))
```

### Hugging Face Datasets

```python
from datasets import load_dataset

ds = load_dataset("xpertsystems/hlt015-sample", data_files={
    "members":     "members.csv",
    "claims":      "claims.csv",
    "prior_auths": "prior_auths.csv",
})
print(ds)
```

---

## Suggested use cases

- **Fraud SIU referral classifier** — train on `siu_referral` × claim features; identify provider fraud risk tiers
- **NHCAA fraud typology multi-class classification** — predict `fraud_typology` (12 classes) from claim + provider features
- **Claim denial prediction** — predict `claim_status` (Paid/Denied/Pended/Adjusted) and `denial_category` from submission features
- **Prior auth optimization** — predict authorization decision from request features; reduce unnecessary P2P escalations
- **Appeal outcome prediction** — predict appeal success from denial features
- **Value-based care attribution analytics** — analyze claim economics by VBC contract type
- **HEDIS quality gap identification** — identify members not meeting `hedis_measure_triggered`
- **30-day readmission prediction** — `readmission_30d_flag` ML from baseline claim features
- **High-cost claimant identification** — predict `high_cost_claimant_flag` from early-period features
- **MLR (Medical Loss Ratio) forecasting** — analyze `mlr_contribution` trajectories
- **Out-of-network leakage analysis** — `leakage_flag` patterns by member segment
- **Generic substitution opportunity modeling** — `generic_substitution_flag` rate improvement targeting
- **EDI pipeline testing** — schema-compliant synthetic data for X12 837/835 EDI integration
- **Payer analytics platform development** — claims warehouse, BI dashboards, BI reporting
- **Healthcare AI pretraining** — pretrain payer-side claim models before fine-tuning on real claims (Optum, Truven Marketscan, IBM Watson)
- **Educational use** — actuarial science, health insurance management, healthcare analytics coursework

---

## Sample vs. full product

| Aspect | This sample | Full HLT-015 product |
|---|---|---|
| Claims | 5,000 | 500,000+ (default) up to 50M |
| Members | 300 | 50,000+ (default) up to 5M |
| Observation window | 12 months | 36+ months (multi-year configurable) |
| Schema | identical | identical |
| Calibration | identical | identical |
| License | CC-BY-NC-4.0 | Commercial license |

The full product unlocks:
- **Up to 50M claims** for production-grade payer ML training
- **5M+ member populations** for representative cohort analytics
- **Multi-year longitudinal windows** for trend analysis and intervention impact studies
- **Custom fraud prevalence injection** — control class balance for SIU referral classifiers
- **Multi-LOB (Line of Business) splits** — separate Commercial / MA / Medicaid model training
- Commercial use rights

**Contact us for the full product.**

---

## Limitations & honest disclosures

- **Sample is preview-only.** 5K claims × 300 members × 12 months is enough to demonstrate schema and calibration, but is **not statistically sufficient** for production-grade fraud classifier training (only ~166 fraud-labeled claims at sample scale, across 12 typology classes = ~14 per class). Use the full product for serious work.
- **Sample uses 12-month observation, not 36-month default.** The full product's default 36-month window enables 30-day readmission tracking across full episodes-of-care and multi-year HCC risk adjustment trajectories.
- **Fraud labels are statistically planted, not adjudicated.** When the generator marks a claim fraud=1, it manipulates billing features (excessive amounts, unbundling patterns, etc.) to look fraud-like. Real fraud labels come from SIU investigation outcomes — use the synthetic labels for ML pipeline development, not pathophysiological inference.
- **Provider NPIs are synthetic 10-digit strings**, not real CMS NPPES numbers. Provider taxonomy codes are realistic placeholder strings.
- **Member IDs are synthetic UUIDs**, not real payer member ID formats.
- **No real ICD-10/CPT/DRG-specific payment rates.** `paid_amount` is calibrated to overall paid/allowed ratios, not specific Medicare IPPS/OPPS or commercial fee schedules.
- **Fraud risk scores follow realistic distributions but are not derived from explainable ML.** Use the field for downstream ML, not for fraud explainability research.
- **No coordination of benefits (COB) cascade simulation.** The `cob_primary_paid` and `cob_secondary_paid` fields are calibrated to realistic split ratios but do not simulate multi-payer claim handoff workflows.
- **HEDIS measures are name-only references.** The full HEDIS denominator / numerator / exclusion logic is not enforced — the `hedis_measure_triggered` field flags claims that would trigger a measure but does not validate eligibility populations.
- **Synthetic, not derived from real payer data.** Distributions match published AHIP/CMS/NHCAA/NCQA references but do NOT reflect any specific real payer (UnitedHealth, Anthem, Aetna, Humana, etc.).

---

## Ethical use guidance

This dataset is designed for:
- Payer-side fraud detection methodology development
- Claims adjudication ML pipeline testing
- Prior authorization optimization research
- Value-based care contract analytics methodology
- HEDIS quality measure pipeline development
- Healthcare AI pretraining for payer-side prediction tasks
- Educational use in actuarial science, health insurance management, and healthcare analytics

This dataset is **not appropriate for**:
- Making payment decisions about real claims
- Real fraud accusations against real providers
- Discriminatory analyses targeting protected demographic groups or provider taxonomy
- Insurance underwriting or premium-setting for real members
- Real provider network configuration without validation on real claim data

---

## Companion datasets in the Healthcare vertical

- [HLT-001](https://huggingface.co/datasets/xpertsystems/hlt001-sample) — Synthetic Patient Population
- [HLT-002](https://huggingface.co/datasets/xpertsystems/hlt002-sample) — Synthetic EHR
- [HLT-003](https://huggingface.co/datasets/xpertsystems/hlt003-sample) — Synthetic Clinical Trial
- [HLT-004](https://huggingface.co/datasets/xpertsystems/hlt004-sample) — Synthetic Disease Progression
- [HLT-005](https://huggingface.co/datasets/xpertsystems/hlt005-sample) — Synthetic Hospital Admission
- [HLT-006](https://huggingface.co/datasets/xpertsystems/hlt006-sample) — Synthetic Medical Imaging
- [HLT-007](https://huggingface.co/datasets/xpertsystems/hlt007-sample) — Synthetic Drug Response
- [HLT-008](https://huggingface.co/datasets/xpertsystems/hlt008-sample) — Synthetic Healthcare Claims (X12 / Provider Billing)
- [HLT-009](https://huggingface.co/datasets/xpertsystems/hlt009-sample) — Synthetic ICU Vital Sign Monitoring
- [HLT-010](https://huggingface.co/datasets/xpertsystems/hlt010-sample) — Synthetic Hospital Resource Usage
- [HLT-011](https://huggingface.co/datasets/xpertsystems/hlt011-sample) — Synthetic Rare Disease + Trial Eligibility
- [HLT-012](https://huggingface.co/datasets/xpertsystems/hlt012-sample) — Synthetic Pandemic Spread
- [HLT-013](https://huggingface.co/datasets/xpertsystems/hlt013-sample) — Synthetic Multi-Modal Genomics
- [HLT-014](https://huggingface.co/datasets/xpertsystems/hlt014-sample) — Synthetic Consumer Wearable Health
- **HLT-015** — Synthetic Insurance Medical Claims (Payer Operations) (you are here)

Use **HLT-001 through HLT-015 together** for the full healthcare data stack. HLT-015 specifically completes the **payer-side analytics axis** that HLT-008 began on the provider side — together the two SKUs provide a full bilateral view of healthcare claims (provider billing + payer adjudication).

---

## Citation

If you use this dataset, please cite:

```bibtex
@dataset{xpertsystems_hlt015_sample_2026,
  author       = {XpertSystems.ai},
  title        = {HLT-015 Synthetic Insurance Medical Claims Dataset (Payer Operations) (Sample Preview)},
  year         = 2026,
  publisher    = {Hugging Face},
  url          = {https://huggingface.co/datasets/xpertsystems/hlt015-sample}
}
```

---

## Contact

- **Web:** [https://xpertsystems.ai](https://xpertsystems.ai)
- **Email:** [pradeep@xpertsystems.ai](mailto:pradeep@xpertsystems.ai)
- **Full product catalog:** Cybersecurity, Insurance & Risk, Materials & Energy, Oil & Gas, Healthcare, and more

**Sample License:** CC-BY-NC-4.0 (Creative Commons Attribution-NonCommercial 4.0)
**Full product License:** Commercial — please contact for pricing.