Medical Insurance Fraud Detection

This dataset contains detailed synthetic records of medical insurance claims, including patient demographics, provider information, claim amounts, service dates, and labeled indicators of fraudulent activity. Designed for machine learning and analytics, it enables robust research and development of fraud detection models in healthcare and insurance. The dataset supports granular analysis of claim patterns, provider behaviors, and patient demographics to identify and prevent fraudulent claims.

  • opened 29 times
  • last updated 22 Jul 2025
  • by GoMask
The brief that made it

Training and benchmarking machine learning models for insurance fraud detection

Sample rows

preview · 8 of 200 rows · all 20 columns
claim_idstringclaim_statusstringclaim_amountfloatis_fraudbooleanhospital_idstringpatient_idstringprovider_idstringclaim_datedateadmission_datedatedischarge_datedatediagnosis_codestringprocedure_codestringapproved_amountfloatservice_start_datedateservice_end_datedatepatient_ageintegerpatient_genderstringprovider_specialtystringbilling_codestringnum_servicesinteger
CLM00001approved185falseblankPT001PVD012023-02-15blankblankE11.999213166.52023-02-152023-02-1535femaleFamily PracticeB992131
CLM00002approved130falseblankPT002PVD022023-02-16blankblankF32.0908341172023-02-152023-02-1527malePsychiatryB908341
CLM00003approved2700falseHOSP001PT003PVD032023-02-172023-02-152023-02-17I109922322502023-02-152023-02-1776femaleCardiologyB992232
CLM00004approved75falseblankPT004PVD042023-02-18blankblankM54.597110752023-02-182023-02-1844malePhysiotherapyB971101
CLM00005denied110falseblankPT005PVD052023-02-19blankblankE66.99940102023-02-192023-02-1953femaleInternal Medicineblank1
CLM00006approved210falseblankPT006PVD062023-02-20blankblankJ06.9992141892023-02-202023-02-2060maleFamily PracticeB992141
CLM00007approved18500falseHOSP002PT007PVD072023-02-212023-02-192023-02-21C34.177427120002023-02-192023-02-2184femaleOncologyB774272
CLM00008pending120falseblankPT008PVD082023-02-22blankblankE78.5800531002023-02-212023-02-2141maleEndocrinologyB800531

What the 200 rows show

from the 200-row sample

Denied (claim status) stands out: 12 of its 23 rows have is_fraud = true, against 5 of 177 for the rest.

  • 9%is_fraud = true
  • 128.5median claim_amount
  • 4patient genders
  • 18provider specialties
  • 25providers
  • 51billing codes
Is fraud rate by claim_statusis_fraud = true
0%50%100%1%approved1 of 14052%denied12 of 2311%pending4 of 37

Denied (claim status)'s mean claim_amount is 24,760, against 2,592 for approved and 6,707 for pending.

claim_amount200 rows, in bands of 20k
0951901831310201080k140kclaim_amount →

Median 128.5, from 48.0 to 128,000.

hospital_id56 rows with a value · 144 left blank
  1. HOSP0024
  2. HOSP0054
  3. HOSP0064
  4. HOSP0094
  5. HOSP0104
  6. HOSP0013
  7. HOSP0033
  8. HOSP0043
  9. HOSP0073
  10. HOSP0083
20 columns by typefrom the column list below
  • string 10
  • integer 2
  • float 2
  • date 5
  • boolean 1

Columns

20 columns in four groups
blueprint · 20 columns
columntypedescriptionexample
Text 10 columns
claim_idstringUnique identifier for each medical insurance claimuniqueCLM00001
patient_idstringUnique identifier for the patient associated with the claimPT001
provider_idstringUnique identifier for the healthcare provider submitting the claimPVD01
diagnosis_codestringPrimary diagnosis code (e.g., ICD-10) associated with the claimE11.9
procedure_codestringPrimary procedure code (e.g., CPT/HCPCS) billed in the claim99213
claim_statusstringStatus of the claim (e.g., submitted, approved, denied, pending)submitted · approved · denied · pendingapproved
patient_genderstringGender of the patientmale · female · other · unknownfemale
provider_specialtystringMedical specialty of the provider submitting the claimoptionalFamily Practice
hospital_idstringUnique identifier for the hospital or facility where services were renderedoptionalHOSP001
billing_codestringBilling code used for the claim (may be different from procedure code)optionalB99213
Numbers 4 columns
claim_amountfloatTotal monetary amount claimed for reimbursement0 or more185
approved_amountfloatAmount approved by the insurance company for payment0 or more · optional166.5
patient_ageintegerAge of the patient at the time of claim submission0 to 12035
num_servicesintegerNumber of services or procedures billed in the claim1 or more1
Dates and times 5 columns
claim_datedateDate when the claim was submitted2023-02-15
admission_datedateDate when the patient was admitted for treatment (if applicable)optional2023-02-15
discharge_datedateDate when the patient was discharged from treatment (if applicable)optional2023-02-17
service_start_datedateDate when the medical service began2023-02-15
service_end_datedateDate when the medical service ended2023-02-15
True or false 1 column
is_fraudbooleanLabel indicating whether the claim is fraudulent (true) or legitimate (false)false

Use it for

  • is fraud9%17 of 200 rowsmean claim amount by …2.6kapprov…24.8kdenied6.7kpending

    An insurance dashboard

    The is_fraud rate, claim_amount by claim_status and a breakdown of hospital_id. Excel, Power BI or Tableau.

  • Why do 17 of 200 rows have is_fraud = true?

    A root-cause class exercise

    Hand out the rows and one question. The answer is in the data, not in the brief.

  • A software demo

    Believable claims with patient_id, provider_id and claim_date to fill a screen in front of a buyer.

Not quite right?

Make it yours.

Same 20 columns, your size and your rules. See 20 rows before you pay.

Preview 20 rows free

10,000 rows of yours: $12.99One-time. No subscription. All prices

This dataset200 rows20 columns
Yours10,000 rows20 columns

blueprint · medical-insurance-fraud-detection

Behind this dataset

Same schema. As many rows as you need.

These 200 rows came out of a blueprint — 20 columns with generation rules behind each one. Open it in Data Factory to retune a column, add your own, wire in foreign keys, and run it at the size you actually need.

Rules it was built with
  • Claims include varying procedures, costs, and patient demographics
  • Fraudulent claims follow patterns like duplicate billing or upcoding
  • Legitimate and fraudulent claims distributed 90/10
  • Temporal features mimic real-world claim submission timelines
Rows
Open the blueprint in Data Factory

1 credit per row. New accounts start with 25 free credits.

Exports
CSV, JSON, JSONL, Parquet, SQL, Excel, TSV, XML
Licence
yours to use, including commercially
API slug
medical-insurance-fraud-detection

More in Insurance

Browse all
Also in InsuranceMedical Claims Fraud DetectionInsurance Policyholder Churn InsightsClaim Fraud Detection DatasetAll 1,947 datasets
In collectionsClassification Datasets for Machine LearningFraud Detection Datasets

What should your data show?

Preview 20 rows free
No signup. No card.