Graphical Models for Fraud Detection in Healthcare Prescriptions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The healthcare industry faces challenges in detecting fraud and abuse in medical treatment activities, with existing methods struggling to accurately identify fraudulent claims and account for normalization factors, leading to high false positive rates and failure to connect diagnoses with prescriptions effectively.
Innovation Solution
A method utilizing graphical models with latent variables to connect disparate data sources, combining patient care event data and prescription data from independent databases, and employing dynamic programming to generate prescription scores and identify outlier behavior, thereby detecting potential fraud and abuse.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional fraud detection methods are used to identify fraudulent claims, then some fraudulent activities can be detected, but the false positive rate is high and many legitimate claims are incorrectly flagged
Solution Approach 1:
The patent segments the fraud detection process into multiple independent graphical models, each focusing on specific relationships (e.g., physician-patient, pharmacy-patient, diagnosis-prescription). This segmentation allows each model to specialize in detecting particular fraud patterns without being overwhelmed by all possible claim types, thereby improving detection accuracy while reducing false positives through targeted analysis.
Solution Approach 2:
The patent introduces latent variables as intermediary elements that connect observable claim data to unobservable fraud indicators. These latent variables act as mediators that capture hidden relationships and normalization factors, enabling the model to distinguish between legitimate variations in prescribing patterns and actual fraudulent behavior, thus reducing false positives while maintaining high detection reliability.
2Loss of information
If existing methods attempt to connect diagnoses with prescriptions, then some relationships can be identified, but the connection accuracy is insufficient due to lack of normalization factors
Solution Approach 1:
The patent adds a temporal dimension to the diagnosis-prescription connection by incorporating time-dependent latent variables that evolve over time. This allows the model to capture how prescribing patterns change over time in response to diagnoses, accounting for normalization factors such as disease progression, treatment response, and seasonal variations, thereby improving connection accuracy without losing important relational information.
Solution Approach 2:
The patent dynamically adjusts the parameters of the graphical models based on observed data patterns, allowing the strength and nature of diagnosis-prescription connections to vary according to contextual factors. This parameter adaptation enables the model to accurately capture relationships under different conditions while filtering out spurious connections, thus improving measurement precision without losing information about genuine associations.
3Reliability
If multiple data sources are integrated to improve detection accuracy, then more comprehensive fraud identification is achieved, but the system complexity increases
Solution Approach 1:
The patent divides the integration of multiple data sources into separate graphical models, each handling specific data relationships (e.g., one model for physician claims, another for pharmacy claims, another for diagnosis-prescription links). This segmentation reduces system complexity by allowing each model to process a manageable subset of data independently, while the overall detection accuracy improves through the combined insights from multiple specialized models.
Solution Approach 2:
The patent uses latent variables as intermediary layers that simplify the integration of multiple data sources. These latent variables aggregate and synthesize information from various data sources in a standardized form, reducing the complexity of direct multi-source integration while preserving the comprehensive information needed for accurate fraud detection. The intermediaries act as buffers that manage complexity while maintaining detection reliability.
Data Source
AI summary
Software that performs the following steps: (i) receiving data from a first database and data from a second database, (ii) identifying a training subset and a test subset from the received data; (iii) generating a first graphical model using data from the training subset; (iv) generating a second graphical model using data from the training subset; (v) determining respective weights for the first graphical model and the second graphical model by using an expectation maximization method on data from the test subset; (vi) generating a third graphical model by interpolating at least the first graphical model and the second graphical model using their respectively determined weights; and (vii) defining one or more links between the data from the first database and the data from the second database using the third graphical model.