AI-driven automated medical insurance claim analysis to detect fraud

Neurons Lab delivered a cutting-edge AI-powered tool to support treatment claim authorization processes for one of the world’s largest insurance companies.

Business Challenges

To authorize insurance claims, medical officers manually review each treatment claim submitted by hospitals to determine its legitimacy.

This process involves analyzing PDFs with prescribed treatment details – procedures, medications, and other expenses – and determining their legitimacy for the specific diagnosis.

Challenges include:

• Time-consuming manual review of extensive documentation.
• Potential for human error and inconsistency in applying international guidelines.
• Delays in claim approvals due to the manual process.
• Losing money due to fraudulent claims and overtreatment.

Project Overview

Neurons Lab implemented an AI-powered tool leveraging a large language model (LLM) and a knowledge graph of clinical guidelines.

This tool analyzes treatment claims, compares them with international guidelines, and provides assessments to facilitate rapid decision-making. In addition, it includes a Human-in-the-Loop (HITL) validation mechanism for doctor validation, ensuring continuous model improvement.

• AI-based claim analysis: The AI tool analyzes treatment claims by extracting individual treatment elements – procedures, medications, etc. – and comparing them against relevant international guidelines and patient-specific data. It assesses treatment appropriateness in the context of the patient’s medical history, diagnosis, and other relevant clinical factors.
• Nuanced scoring and explanation: The solution provides a nuanced scoring system that reflects the complexity of clinical judgment. Each treatment element is assigned a clear color code and a score based on adherence to guidelines, potential risks, and alternative options. The solution also generates detailed explanations highlighting specific guideline deviations, potential risks, alternative treatment options, and the assessment’s rationale.
• HITL validation: Medical officers can review the AI-generated assessments and provide feedback through a user-friendly interface. Their feedback is used to refine the AI model’s performance and ensure accuracy continuously.

Solution

Our proprietary, highly customized SaaS AI solution uses generative AI from Anthropic Claude 3 in Amazon Bedrock and a comprehensive knowledge base of treatment procedures and medicines to automate claim analysis. It streamlines the medical necessity review process and provides expert recommendations.

1. Document input

The user, a medical officer, uploads the treatment claim document into the AI tool, or it can be automatically uploaded.

2. AI Analysis

The analysis stage involves:

• Intelligent Document Processing (IDP): The AI tool processes the claims, extracting individual treatment elements – procedures, medications, and other information – row by row.
• Knowledge base enrichment: The extracted information is linked to a comprehensive knowledge base.
• Contextual matching: The AI tool compares the extracted treatment element with the corresponding recommendations in the pre-defined international guidelines databases. It also retrieves the patient’s medical history, diagnoses, and other relevant information from the database.

Based on the level of concurrence, the AI tool gives each element a score corresponding to the relevance of the treatment in the submitted claim.

• Relevant (100%): A full match with guidelines and procedures.
• Relevant—with a warning (99-75%): It matches the guidelines, but alternative options may exist, or further investigation is needed.
• Not relevant (<75%): Either there is no match or significant deviation from guidelines.

The AI-predicted color-coded relevance score is displayed on the processed claim for each element.

For each treatment element suggested in the insurance claim, the LLM generates a detailed natural language explanation for the assigned relevance score based on its concurrence with international treatment guidelines and patient medical data.

The explanation highlights key factors impacting the score and supporting evidence – for example, links to relevant sections in the source guidelines used for comparison.

For more information about IDP… Take a look at our IDP guide for the modern enterprise and how to maximize its ROI.

3. Doctor validation

The medical officer: 

• Reviews the AI-generated assessments for each treatment element.
• Can access the specific international treatment recommendations used for comparison through the link and other relevant details for the claim – such as the patient’s medical history and diagnoses.
• Adds their professional judgment to the input field next to the AI generation relevance score based on the same rules as before.
• May add additional comments to the specific treatment element. Their input is captured through the HITL interface, facilitating continuous model improvement.

4. Decision and rationale:

Once the treatment assessment is finalized, the medical officer decides on the claim. 

The medical officer makes a final decision—approve, deny, or request further information—on each claim based on AI-generated insights and their own clinical judgment.

The medical officer documents the rationale for approving or denying the treatment. The insurance provider’s system records the decision for audit and compliance purposes.

Results

Key benefits include:

• Fraud detection and cost containment: Insurance companies’ primary goal is to ensure that claims are legitimate and adhere to established guidelines. This system, powered by AI, helps flag potential fraud—for example, unnecessary procedures and overtreatment—saving the provider significant costs.
• Efficient claims processing: The automated analysis and clear explanations streamline the review process for medical officers. This leads to faster claim decisions, improved customer satisfaction, and reduced administrative overhead.
• Objective and consistent assessments: The AI consistently applies standardized guidelines, reducing the risk of human error and bias in claim decisions. This ensures fairness and can help defend against potential disputes.
• Data-driven insights: The system can gather data on claim patterns, treatment trends, and guideline adherence. This information can be used to identify areas for improvement, refine insurance policies, and negotiate better rates with healthcare providers.
• Continuous improvement: HITL validation allows the system to learn from medical officer feedback and become more accurate while adapting to changing guidelines and medical practices.

Combining AI with human expertise can empower insurance teams to make informed decisions faster, leading to significant cost savings and improved healthcare outcomes.

About Neurons Lab

Neurons Lab is an AI consultancy that provides end-to-end services – from identifying high-impact AI applications to integrating and scaling the technology. We empower companies to capitalize on AI’s capabilities.

As an AWS Advanced Partner, our global team comprises data scientists, subject matter experts, and cloud specialists supported by an extensive talent pool of 500 experts. We solve the most complex AI challenges, mobilizing and delivering with outstanding speed to support urgent priorities and strategic long-term needs.

Ready to leverage AI for your business? Get in touch with the team here.