Medical Provider Fraud requirements

The Medical Provider Fraud content pack requires several types of data to perform the statistical calculations and analysis. Your data must be in this format for IBM® Counter Fraud Management to act on it.

Data Structure

The Medical Provider Fraud detection analyzes seven primary data objects:
Loss Event (event)

The Loss Event defines the incident that results in an insurance claim, such as an automobile accident with

medical claims.
Invoice (transaction)

Claim and Invoice are bills that are sent for insurance payment. Claim and Invoice are often considered to be the same, but they can be separate. For example, a single claim might have multiple invoices.

An invoice might have multiple diagnoses or multiple procedures.

Involved Parties (party)

Parties define the people or organizations that are involved, such as the physician, hospital, patient, or insured.

Policies (account)

The insurance policy defines the coverage that is available.

Diagnosis (event)

The diagnosis defines the reason for the insurance claim, such as a simple office visit, an emergency room visit, or a hospitalization admission. A broken leg is a diagnosis. A single diagnosis might have multiple procedures, such as a cast for the broken leg, physical therapy for the recovery, and medication to combat infection.

Procedure (event)

The medical procedure, treatment, prescription, or other therapy required in response to the diagnosis. X-rays, a cast, and physical therapy are procedures for a broken leg. A single procedure might be the result of several diagnoses, such as an MRI is scheduled due to multiple issues.

Medical Provider Detection Inputs

The Medical Provider Fraud detection analyzes large amounts of data that is gathered from claim forms that are based on data that is usually obtained from Centers for Medicare & Medicaid Services (CMS) CMS1500 or the National Uniform Billing Committee UB-04 specifications. Both claim forms accommodate the National Provider Identifier (NPI) and ICD-10 coding. Different analysis models use different parts of the following data set:
  • Medical providers
  • Patient, insured party, related policies
  • Invoice, diagnosis, procedure, service, treatment, supplies
  • Address and locations
  • Prescriptions and drugs
  • Loss event
  • Reference-able diagnosis and procedure codes

The Medical Provider Fraud content pack includes sample data that fits the required model.. The samples include CSV files and PROPERTIES mapping files that you can use with the Data Import tool.