Insurance Claim Anomaly Curing via Multi-Engine Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for classifying insurance claims rely on single classification engines, which can lead to inefficiencies and inaccuracies, particularly in identifying invalid claims due to insufficient high-quality training data and the inability to handle complex medical code combinations effectively.
Innovation Solution
The implementation of multiple classification engines operating in parallel, including a Rules Engine and an AI/ML Engine, which use a weighted selection procedure based on past performance and generate synthetic training data to improve classification accuracy, along with a Decision Module to interpret results and correct anomalies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single classification engine is used to classify insurance claims, then the system is simple to operate and maintain, but the classification accuracy and reliability are insufficient
Solution Approach 1:
The patent combines multiple classification engines (Rules Engine, AI/ML Engine, and synthesized data) into a unified system that processes insurance claims together. The engines work in parallel and their results are integrated through a decision module that uses weighted selection based on past performance, achieving higher reliability than any single engine could provide alone.
Solution Approach 2:
The patent introduces a decision module as an intermediary that receives classification results from multiple engines and selects the most reliable classification based on weighted performance metrics. This mediator resolves conflicts between different engine outputs and ensures the most accurate classification is used, improving overall system reliability without requiring direct integration of all engine components.
2Reliability
If multiple classification engines operate in parallel, then classification accuracy improves, but the device complexity increases
Solution Approach 1:
The system implements self-service through automated performance tracking and weighted selection. The decision module automatically monitors past performance of each classification engine and dynamically weights their outputs without requiring manual intervention. This automation maintains ease of operation while leveraging multiple engines for improved accuracy.
Solution Approach 2:
The patent makes the system dynamic by allowing the weights assigned to different classification engines to change based on their past performance. The decision module continuously adapts the contribution of each engine based on observed accuracy, making the system flexible and responsive to changing conditions while maintaining operational simplicity through automated adjustment.
3Reliability
If extensive training data is used to train classification engines, then classification accuracy improves, but the loss of time and resources for data preparation increases
Solution Approach 1:
The patent applies preliminary action by pre-generating synthesized training data that mimics real insurance claims data. This synthesized data is created in advance using rules-based methods and historical patterns, allowing classification engines to be trained without requiring extensive manual collection and preparation of real training data, thereby reducing time and resource investment.
Solution Approach 2:
The system creates copies of real training data through synthesis techniques that generate artificial but realistic insurance claims. These synthesized copies preserve the statistical properties and patterns of real data while avoiding the need to collect, store, and process large volumes of actual sensitive data, reducing preparation time and resource requirements.
4Quantity of substance
If synthetic training data is generated, then the need for extensive real training data is reduced, but the manufacturing precision of the synthetic data must be maintained
Solution Approach 1:
The patent employs parameter changes by adjusting the synthesis process to match statistical parameters of real insurance claims data. The synthesized data is generated with controlled variations in key parameters such as claim amounts, diagnosis codes, and procedure codes to ensure they reflect real-world distributions, maintaining manufacturing precision while reducing dependency on extensive real data.
Data Source
AI summary
Disclosed embodiments include determining whether the insurance claim comprises an anomaly based on whether the claim violates a rule in a Rules Engine or is scored by an Artificial Intelligence/Machine Learning engine below a threshold. When the claim comprises an anomaly, (i) for each feature of the claim, (a) determining a sensitivity value of the feature relative to each other feature of the claim, and (b) determining a saliency value of the feature relative categories relevant to the claim, and (ii) estimating a value of the claim based on a multi-variate linear regression model. And based on the sensitivity values, saliency values, and estimated value, recommending one or more ways to cure the anomaly based on (i) a Local Interpretable Model-Agnostic Explanations based recommendation and/or (ii) a Global Surrogate Model based recommendation.