Fraud will always be an ongoing fight in the insurance sector, but the problem appears to have worsened with digital transformation.
Statistics from the Coalition Against Insurance Fraud indicate that annual fraud increased from $80 billion in 1995 to $308 billion in 2022, paralleling the wider adoption of digital data storage, EHR platforms, automation, and AI. The lesson is not that these tools are inherently unsafe, but that hackers and scam artists have embraced them in creative and unexpected ways insurance industry leaders need to continuously overcome.
How can healthcare payors try to level the playing field? Artificial intelligence does offer a compelling tool to detect fraud rapidly and accurately, but there needs to be a foundation of data hygiene, management, and governance to catch criminals in the act.
Enhancing AI’s Accuracy with Third-Party Data
Are your auditors ready and able to spot fraudulent CT scans and X-rays? Are they trained to identify phony medical records? That’s the new threat on the horizon as criminal gangs or even opportunistic providers learn to better create counterfeit documentation to make their fake claims more convincing.
In the past, this type of fraud was less concerning because the barrier to entry (e.g., in-depth anatomical knowledge, medical diagnoses, exceptional Photoshop skills, etc.) prevented most scammers from using this tactic with any level of reliability. Now, artificial intelligence trained in medical imaging can generate pictures with minimal effort that are almost indistinguishable from real images.
Fortunately, artificial intelligence can also be used to spot this type of fraud. Studies show that AI is already capable of fraudulent images, identifying similarities in biological features and body composition that radiologists and auditors might miss. Generative AI also has the potential to spot discrepancies in claims, evaluating them based on an individual member’s medical claim history and similar claims made to your organization and others.
That’s where third-party data is essential. Combining data from reliable external sources (e.g., federal agencies, non-profits, and other insurance companies) with machine learning algorithms can aid payors as they identify discrepancies and trigger investigations. However, using data matching as a fact-checking strategy requires both data hygiene and governance, practices which require some effort on your part.
Data schemas need clear documentation, and data models and pipelines will need to ingest claims into a secure but easy-to-use AI platform. Unless those underlying data lakes are stocked with comprehensive sources of data, you still might struggle to separate approximate claims documents conceived by generative AI from the real deal.
Upgrading Fraud Detection with Data Integrity
Fraud detection hinges on the ability of auditors to spot irregularities in a sea of reliable and acceptable patterns. If fraud investigators don’t trust the consistency and accuracy of information across appointments, bills, policies, and electronic medical records, they’ll struggle to find a reliable baseline for their decision-making. The same applies to artificial intelligence platforms.
Though AI can detect patterns with greater sophistication and accuracy than human beings, they’re still limited by their own training data. Spotting anomalies requires quality and coherence of the information in your database, meaning garbage in, garbage hurts large language models (LLMs) just as much as your human auditors. With accurate, consistent, and complete data, AI can spot instances when reviews when claims deviate from a physician’s usual prescriptions, or the frequency or types of medical supplies or devices fall outside typical orders.
What about false positives? If your databases have duplicates, outdated information, or inconsistencies, your automated fraud detection systems and auditor team might waste precious time and money on needless investigations. Sure, the speed and adjustability of AI algorithms does trim some waste, but when false alarms happen frequently enough, you might start to miss legitimate instances of fraud.
Even if you’ve done your diligence to implement data hygiene across your organization, there’s still a chance that integrity and accuracy can be undercut – if your cybersecurity practices are weak. There is a rising trend called data poisoning, where hackers compromise LLMs to damage the outcomes and prevent AI from coming to accurate conclusions. From a payor perspective, that might mean sabotaging fraud detection to allow millions to pass undetected.
Establishing data integrity and security in turn enables your organization to establish reliable standards. With this foundation, you can trust the results of machine learning and artificial intelligence platforms to parse large and complex data sets, accurately detecting fraud at a higher success rate.
Improving Your Data Practices to Prevent Fraud
If your data is dirty or you lack the processes to analyze claims with speed and accuracy, the reality is that you likely lack proper data best practices. That’s not uncommon. With the volume of data and new sources entering healthcare payor systems, there needs to be a champion for holistic data governance if your fraud detection is to be a success.
Data hygiene and governance are ongoing processes. If you consolidated and cleaned up your data several years ago, but have not taken any proactive steps since, you will have to repair and revamp your foundation.
Conduct an assessment of your current data practices to identify strengths and opportunities to enhance quality, accuracy, and management. Verify that the current classifications of data offer a complete view for fraud detection and audits, comparing it with third-party sources and standards to create a better blueprint for early scam detection. Most importantly, create a data governance framework that defines how data is stored, shared, and eliminated.
When all this is complete, you’ll be able to empower your algorithms, artificial intelligence strategy, and agents to detect fraud in real time, preventing claims payments from going to uncovered individuals or unauthorized procedures.
Are you looking to create a better data foundation to combat healthcare fraud and abuse? Collaborate with the w3r Consulting team to enhance how you work.
Learn about our health solutions
Related Articles
How Generative AI Boosts Performance, Accuracy, & Retention in Healthcare
3 Use Cases for AI in Insurance to Revamp Your Business (w3r.com)
Unlock Your Artificial Intelligence Strategy with this Foundation (w3r.com)