Pattern Intelligence: Discovering Revenue Cycle Management Issues in Healthcare

Paper: Pattern Intelligence: Unsupervised, LLM-Interpreted Discovery of Systematic Underpayment and Denial Patterns in Healthcare Revenue Cycle Management

Authors: Harshil Lodhiya, Alex McManus

Executive Summary

Healthcare organizations often face significant financial losses due to issues like underpayments and claim denials from payers. Identifying the root causes of these issues has traditionally been a manual and time-consuming process that struggles to keep pace with the volume of claims hospitals handle. This research introduces a new system, called Pattern Intelligence, designed to automate and enhance the detection of issues in claims processing.

Pattern Intelligence operates by analyzing claims based on their adjudication behavior rather than their identity, which helps to uncover repeating patterns of denial and underpayment that can be addressed. The system uses a sophisticated method of clustering claims into groups based on a 130-dimensional vector that includes various metrics such as expected versus actual payment ratios, denial reasons (captured through claim adjustment reason codes), and other relevant signals. Importantly, it does not factor in the identity of the payer or specific contracts in its initial analysis, as these factors can skew the clustering process and lead to overly simplistic groupings.

The system employs HDBSCAN, a clustering algorithm, and performs additional analysis to refine and validate the groupings. Each grouping is then described in a way that is useful for analysts, with a large language model translating the core characteristics of each cluster into understandable terms, detailing the most common procedures, reasons for adjustments, and types of failures observed.

Research findings indicate that a significant portion of high-value claim denials—estimated between 30% to 50%—exhibit persistent patterns that do not conform to typical clustering, suggesting that these may represent fundamental issues within the claims process rather than flaws in the analytical model itself. This insight is critical for healthcare organizations aiming to improve their revenue cycle management.

The effectiveness of Pattern Intelligence was validated through cluster-coherence analysis and a recoverable-opportunity metric, tested on claims data from two different healthcare providers, which had approximately 10,000 and 157,000 claims, respectively. This evaluation provides a tangible measure of potential recoverable revenue from addressing the identified issues.

The research contributes to the field by providing a new way to approach the problem of claim denials and underpayments. It emphasizes the importance of representation design and empirical findings drawn from actual production environments rather than theoretical models. This could help healthcare providers identify and rectify revenue loss areas more effectively, potentially leading to improved financial performance and operational efficiency.

Academic Abstract

Healthcare providers lose substantial revenue to systematic payer underpayments and claim denials, yet identifying the recurring, correctable patterns behind them remains largely manual and does not scale to hospital claim volumes. We present Pattern Intelligence, a deployed system that discovers such patterns without supervision by clustering claims on their adjudication behaviour rather than their identity. Each claim is embedded in a 130-dimensional vector combining per-line expected-versus-actual payment ratios, claim adjustment reason (CARC) codes, denial and payment signals, and service composition. Payer and contract identity are deliberately excluded from the distance metric - we show they dominate the geometry and collapse clusters into trivial per-payer groups - and are reintroduced only as a post-hoc segmentation (by contract and visit type) that keeps every surfaced pattern coherent and actionable. We cluster with HDBSCAN, recover structure from the noise set via second-pass sub-clustering, and gate low-confidence memberships. A large language model translates each centroid into an analyst-ready description of dominant procedures, adjustment reasons, and failure mode. We report a recurring finding: roughly one-third to one-half of high-dollar denials, depending on tenant and segmentation, form an irreducible long tail that no density-clustering configuration or representation we tried groups - a domain property, we argue, rather than a modeling deficiency. We evaluate validity through cluster-coherence analysis and quantify impact with a recoverable-opportunity ($) metric on two production tenants (approx. 10K and 157K claims) within a six-tenant deployment. Rather than a new algorithm, the contribution is a problem formulation, a representation-design insight, and empirical findings from production.