Underpayment Detection and Payer Contracted Rate Audits

Underpayments happen when an insurance company pays less than the contracted rate, but still pays something. That single fact makes them almost invisible. A denial gets flagged and worked. An underpayment gets posted, closed, and forgotten, because as far as the billing system is concerned, the claim is paid and the account balances to zero.

What underpayment costs a practice at current scale

Nobody agrees on the exact size of the problem, but every credible estimate points to significant losses. Industry-wide figures put annual losses to providers in the billions, with underpayments eating 3 to 5% of net revenue that manual review moves too slowly to catch. At the practice level, the number most often cited runs from 1 to 3% of net patient revenue lost every year, with some estimates for physician groups landing harder still, as high as 11% when underpaid and unpaid claims get counted together.

Even at a smaller scale, the number still stings. A $5 million practice running a 2% underpayment rate, conservative by every estimate above, loses $100,000 a year, enough to cover a hire, a lease payment, or a piece of equipment sitting unbought. A $5 million practice running a 2% underpayment rate, conservative by every estimate above, loses $100,000 a year, a real loss rather than a rounding error. That's a hire, a lease payment, a piece of equipment sitting unbought.

None of this is theoretical. In August 2025, a federal judge granted final approval for Blue Cross Blue Shield to pay $2.8 billion in underpayments to hospitals and healthcare organizations, hard proof that underpayment isn't a scattering of individual billing mistakes but something systemic in how payers manage reimbursement at scale.

The margin for error keeps shrinking too. Median health system operating margins held near 1% through 2025, with expenses climbing around 6% a year against revenue growth of only 3%. When the operating margin sits that thin, a 2% revenue leak stops being background noise. It turns a normal year into a bad one.

Why standard billing workflows are structurally blind to underpayments

The mechanism is simple, and the simplicity is exactly why it's so hard to catch. A claim gets paid 5% below the contracted allowable, the payment posts, the balance zeroes out, the transaction closes. No alert fires. Nobody on staff has a reason to look again, because nothing in the workflow ever told them to.

Very few billing systems, and fewer workflows built on top of them, include a step where someone compares the posted payment against what the contract actually obligated the payer to pay. Matt Basham of DoctorsManagement says that comparison is easy conceptually. It's just missing operationally. Layered on top sits a records problem: payer contracts scattered across filing cabinets, old email threads, and whatever a long-tenured office manager happens to remember. With no single place to check a payment against, nobody checks.

Rate increases make the blind spot worse, not better. Payers apply a headline increase, say 3%, unevenly across the fee schedule, tilting it toward low-volume codes while leaving the high-volume E/M codes that actually drive collections flat. Mark Schroeder of Aroris Healthcare estimates practices can end up as much as 15% behind market rates without a single number on paper ever looking wrong. Catching that requires weighting each CPT code's rate change by how often the practice bills it, a step most practices skip.

A psychological layer shapes the operational one. Doral Jacobsen, CEO of Prosper Beyond, calls it the "courage gap," a quiet, defeatist assumption among practice leaders that challenging a payer wastes time better spent elsewhere. Resource math reinforces that assumption. CAQH data shows the average physician group manages around 20 payer contracts. Hospitals staff entire managed care departments to track that kind of load. Physician groups hand the job to an office manager who's also running HR, credentialing, and the front desk schedule, so contracts sit for years without a second look. An MGMA Stat poll found a meaningful share of medical group leaders don't review their contracts on any regular basis.

The three-part audit process that surfaces hidden shortfalls

Every audit worth running has to identify which contracts are costing money right now, and most groups have never asked that question in any structured way.

Inventory and triage comes first. Every active payer contract needs to live in one place before comparison is even possible. Rank payers by revenue contribution, since the top three typically account for 50 to 70% of total collections and deserve the first pass, and flag any contract that hasn't been reviewed in two years or more. Credentialing records deserve a check here too: CAQH attestations, NPIs, provider rosters, effective dates, because a well-negotiated rate that never activates due to a roster gap produces no real gain. It's a paper agreement.

CPT-level payment comparison reveals the silent losses, because contracted rates get loaded into a system that matches them against ERA (Electronic Remittance Advice) data, claim by claim. Contracted rates get loaded into a system that matches them against ERA (Electronic Remittance Advice) data, claim by claim. Spot-checking a sample finds isolated mistakes. Comparing every claim against every contracted rate finds patterns, and the difference matters enormously, since patterns are what get recovered and fixed, not one-off errors. A multispecialty clinic documented by DoctorsManagement recovered $250,000 in underpayments over 18 months doing exactly this, comparing posted payments to contracted rates, nothing more exotic than that. The underpayments had been running for years before anyone looked.

A payer scorecard closes the loop. Reimbursement behavior, denial categories, payment delays, appeal outcomes, and policy changes, tracked by payer over time, turn every denial and delayed payment into contract intelligence instead of a one-off billing headache. That scorecard feeds the recovery work happening now and the renegotiation conversation coming later.

How often this actually happens varies widely, and the variation tells its own story. An MGMA Stat poll, with 193 applicable responses, found 48% of medical group leaders audit payer payments monthly (30%) or quarterly (18%), while the remainder audit far less frequently. The monthly share rose from 26% in a January 2023 poll, suggesting practices are responding to a problem getting worse rather than better. Annual auditing lets an underpayment run for up to 12 months before anyone notices. That's not a small gap to live with.

Where manual audit breaks down and what automation changes

Hospitals now need to model thousands of contract terms against millions of charge-level data points every day, and manual review was never built to sustain that kind of load. Spreadsheet audits catch samples. They miss patterns, and they can't reliably account for modifiers, locality adjustments, and payer-specific carve-outs across every contract at once.

The MGMA Stat poll backs this up directly. Practice leaders cited staffing and bandwidth limits, EHR and practice management system constraints, and the sheer complexity of juggling fee schedules from different payers as the top obstacles to auditing well.

Automation changes the math in a few concrete ways, and here's where the real disagreement should end: manual sampling is not an adequate substitute for full-population review, full stop. Instead of checking a representative slice of claims, an automated system checks every claim, every time. Contracts, fee schedules, and carve-outs get digitized into logic the system applies automatically, so it can flag discrepancies between what a claim paid and what the contract required. Remittances get processed in near real time instead of the days or weeks manual review takes. Vendor-reported figures from Ember AI's guide put manual review error rates around 8 to 12%, against a sub-3% error rate for AI-enabled detection, numbers that should be read as directional rather than settled fact given the source.

The recovery figures actually move a budget. Ember AI's 2026 guide cites $50 to $120 recovered per identified underpaid claim, a return several times over the cost of implementation, along with a 1 to 3% improvement in net collection rates and cash flow gains appearing as early as 90 days after go-live. At enterprise scale, McKinsey modeling cited in an AI Executive Media brief shows that trimming a typical 3.5 to 4.0% cost-to-collect by even one or two percentage points saves a $6 billion health system somewhere between $60 million and $120 million.

Running a structured pilot before full-scale rollout

None of this justifies a blind rollout, and the failure rate on record proves the point. A KLAS Research report found only 23% of health system AI initiatives ever made it past proof-of-concept into full production. A 2025 KLAS Research report found only 23% of health system AI initiatives ever made it past proof-of-concept into full production, and that failure rate is worth planning a rollout around.

Ember AI's 2026 guide lays out a pilot structure: pick one payer, one facility, or one specialty as the controlled test case, not the entire revenue cycle at once. Pick one payer, one facility, or one specialty as the controlled test case rather than the entire revenue cycle at once. Run the AI system and human reviewers side by side for four to twelve weeks, comparing results directly by tracking recovery dollars, the drop in manual review hours, and any shift in the clean-claim rate. Bring RCM, contracting, compliance, and IT into the room together, since data integrity and governance questions arise fast once real claims data enters the picture. Adjust thresholds and workflows based on what the parallel run actually shows, then scale, not before.

Governance is not optional here. Every automated recommendation needs documentation that holds up if a payer disputes it or an auditor asks questions, particularly given the False Claims Act exposure tied to how claims get billed and appealed. HIPAA, SOC 2, and HITRUST alignment need to be settled before any platform touches claims data, not after the fact.

More modern automation approaches can reduce the need for deep technical integration that earlier tools demanded. That moves the real bottleneck in deployment away from technical integration and toward governance and scope decisions, a different problem, and a more solvable one. Organizations running production AI agents for tasks like claim status checks report administrative FTE reductions in the 60 to 80% range and cost-per-claim improvements of 40 to 55%, figures cited in material referencing a Smilist deployment, a useful benchmark for what mature automation looks like once it clears the pilot stage.

Turning audit findings into leverage at the negotiating table

A pilot that recovers money is useful. A pilot that produces a clean, dated record of exactly where a payer has been underpaying is worth more, because that record becomes the foundation of the next contract negotiation.

RevGuard's 2026 framework describes building an evidence packet from 12 to 24 months of internal claims history, broken out by payer, procedure, denial reason, allowed amount, paid amount, and payment timing. Before walking into a negotiation, that packet has to answer a specific set of questions: which services drive volume, which drive margin, where the payer's allowed amount diverges from the contracted schedule, and which denials trace back to actual contract language versus plain administrative error.

Four documents structure the conversation from there. A payer scorecard tracks reimbursement behavior, denial categories, delays, and appeal outcomes over time. An ask sheet lists each requested change, the evidence behind it, the opening position, the target, and the floor below which the practice won't go. A concession log records every movement made during negotiation, along with whatever the payer commits to in return, and an executive brief rounds it out with a summary of the relationship, the financial exposure at stake, the value of network access, the operational burden of the current contract, and the practice's alternatives if talks stall.

The scale and variety of contracts providers manage produces structural complexity, as shown by an analysis cited by RevGuard that found hospitals negotiating a mean of 52 contracts across 33 different payers, and across 524 contracts studied, 59.4% used more than one reimbursement methodology at once. An analysis cited by RevGuard found hospitals negotiating a mean of 52 contracts across 33 different payers, and across 524 contracts studied, 59.4% used more than one reimbursement methodology at once. Walking into that kind of negotiation without a data-backed record of where the current contract is already failing means negotiating blind. Walking in with the audit trail in hand means negotiating from evidence instead of guesswork, which is the reason for running the audit.