Medical Provider Fraud pattern analysis

The Medical Provider Fraud detection uses multiple analysis techniques to assess the potential for fraud. These individual analyses roll up into a single assessment as to the risk that a provider is committing fraud.

P1: Suspicious count
This pattern assesses each possible indicator to identify potential abnormalities and generate a scoring curve that is based on peer group profile parameters, such as standard deviation, median, and the behavior of the indicator.
P2: Excessive monetary amount
This pattern assesses each possible indicator that is related to monetary values to identify potential abnormalities and generate a scoring curve individually, based on peer group profile parameters, such as standard deviation, median, and the behavior of the indicator.
P3: Unusual time span
This pattern assesses each possible indicator that is related to time to identify potential outliers and generate a scoring curve individually, based on peer group profile parameters, such as standard deviation, median, and the behavior of the indicator.
P4: Manipulated charge
This pattern applies Benford Law to assesses each possible monetary indicator to identify potential records that might be manipulated or fabricated. It extracts the first and second digits of the monetary indicator and detects the digits that are much higher than expected.
P5: Benford distribution change
This pattern applies Benford Law to assesses each possible monetary indicator to identify potential behavior changes. It extracts the first and second digits of the monetary indicator and detects the distribution change over time.
P6: Potential anomaly group analysis
This pattern selects a set of indicators (either by using the default or through running factor analysis) and applies anomaly and clustering models to obtain anomaly scores.
P7: Provider patient network ring
This pattern analyzes provider and patient relationships to identify groups of providers that use same set of patients for claims (PingPong Association).
P8: Provider peer profile change
This pattern establishes a base providers and peer group level profile. It provides early detection for possible change of behavior over time.
P9: Recency-Frequency-Monetary (RFM) analysis
This pattern characterizes providers based on RFM analysis and provides detection for category changes over time.
P10: Unbundling association
This pattern analyzes the procedure and diagnosis codes that were applied by providers and assigns a higher risk to providers that use a different set of procedures for the same diagnosis.
P11: Provider folios analysis
This pattern analyzes case (folio) information and generates risk scores based on the case context and case state within the selected peer group. The case context refers to various analytics content. For example, a case that is generated by the Medical Provider detection is identified by “mpf_analytics”. Examples of case state include “alert_triage_review”, “alert_closed” or “case_request_info”. Different case states are assigned different weights, and the risk score is calculated based on the count of each case state.
P12: Provider entity analysis
This pattern generates risk scores for each provider based on the entity resolution results. Examples of possible statistics include the count of parties that resolve to the same entity, the count of different zip3 values for the entity, and the count of different states for the entity.
P12: Assessment integration
This pattern integrates all the modeling results and generates an overall risk-based scoring for each provider. The final risk score can be adjusted based on the weights that are associated with each model. The top alerted reasons are also extracted to support any needed review, exploration, or validation.

Parent topic: Medical Provider Fraud detection specification