Health plans are witnessing tremendous increase in volumes and complexity of claims stemming from value-based contracting, complex coverages, evolving regulations, and innovative vendor partnerships. While there is an uptake of technology across most processes, healthcare claims audit within health plans continues to rely on traditional statistical sampling and manual research. Manual auditing, though effective, involves longer cycle times to identify the root causes and skilled resources to retrospectively analyze and adjust claims. Limited ability of manual claims audit to handle incremental volume and complexity is preventing health plans from taking any further leaps in claims Fraud, Waste and Abuse and payment recovery.
Artificial Intelligence (AI) and Machine Learning (ML) are well positioned to transform claims auditing by improving the sampling, detection, investigation, and adjustment processes.
AI enabled systems can methodically identify and correct errors, while avoiding unnecessary or ineffective interventions. Intelligent solutions can leverage Artificial Intelligence for information extraction, data interpretation and reporting capabilities to enable healthcare payers to analyze a larger number of claims in a much shorter timeframe than normally possible with a traditional manual review. In addition, the continuous learning ability of AI will make it possible for auditors to train Intelligent Systems to work better and smarter.
This creates new opportunities for Auditors to better utilize and optimize their time, enabling them to focus their abilities on analyzing a broader and deeper set of data and documents that cannot be processed by machines. With the advancement in technology, the AI enabled audit is not restricted to after completion of the transaction, it can be implemented at the pre adjudication stage.
AI algorithms, when introduced in a pre-adjudication environment, can be used for all transaction surveillance to proactively flag potential audit failures and trigger actions that can avoid any downstream manual adjustment to core transactional systems.
This not only reduces the requirement of manual effort but enables a continuous learning loop across the payment lifecycle.
Implementing an AI Audit solution can be accomplished in five easy steps:
- Compiling and preprocessing claims data. This will involve diligent cleansing and transformation of data that cognitive algorithms will later draw on; completeness and consistency are essential.
- Analyzing statistical models with data. Analyze patients, providers, contracts, benefits, diagnoses, and claims to establish a correlation between various data sets.
- Developing a valid model for identifying claims anomalies. The test data is then used to train the cognitive algorithms. By feeding in additional insurance data and external information (for example, regional distribution of providers) the model is gradually enhanced until it eventually starts to independently learn new data and case patterns.
- Developing benchmarking metrics. Metrics should be used to benchmark and select from several algorithms to build a pipeline that can most reliably predict the likelihood that a claim can be audited successfully.
- Prototyping with claims audit sample. Piloting with production sample phase serves to audit new claims received in real-world conditions and refine the algorithm further.
Payment inaccuracies cost the health insurance industry anywhere between $60-$250 billion dollars annually, with less than one percent chance of recovery. Artificial Intelligence enabled claims audit not only simplifies and accelerates the overall claims auditing experience, it also enhances its quality, reducing cost incurred for redundant audits, eliminates rejection processes, and fosters better provider collaboration.