How NilesAI's 16 Scan Engines Catch What Manual Review Misses

Medical billing errors cost the U.S. healthcare system an estimated $935 billion over the past decade Source. Yet the vast majority of bills are never audited - fewer than 1% receive any scrutiny at all. For the professionals who do review bills - attorneys handling personal injury liens, insurance adjusters processing claims, or compliance officers auditing charges - the process is slow, expensive, and inconsistent.

NilesAI was built to change that. Instead of relying on human stamina or probabilistic AI guesses, NilesAI runs every bill through 16 deterministic scan engines backed by over 2.6 million billing edits. Here’s how they work and why they matter. If you want a high-level overview before diving into the technical details, our Medical Bill Review Guide explains the full review process in plain language.

The manual review problem

A trained medical billing auditor reviewing a complex hospital bill typically spends 2 to 8 hours per case. At professional billing review rates, that translates to $150 to $1,800 per review Source. Scale that across hundreds or thousands of claims and the economics quickly break down.

But cost isn’t the only issue. Human reviewers face:

• Fatigue - accuracy drops after hours of line-by-line code checking
• Inconsistency - two reviewers examining the same bill may flag different issues
• Knowledge gaps - keeping up with CMS NCCI quarterly updates, MPFS rate changes, and new HCPCS codes is a full-time job
• Limited cross-referencing - mentally comparing an EOB against an itemized bill across dozens of line items is error-prone

As we detailed in our whitepaper on medical billing errors, up to 80% of medical bills contain at least one error Source. The sheer volume of potential mistakes makes exhaustive manual review impractical for most organizations.

Rules-based validation vs. AI guessing

Many emerging tools use large language models (LLMs) or retrieval-augmented generation (RAG) to “guess” whether a bill looks right. While AI has its place, probabilistic approaches introduce a critical weakness: they can’t cite their sources.

NilesAI takes a fundamentally different approach. Every scan engine is deterministic and rules-based. When NilesAI flags an NCCI bundling violation, it doesn’t say “this looks like it might be bundled.” It says: “CPT 99213 and CPT 99214 are an NCCI edit pair per CMS Column 1/Column 2 table, effective Q1 2024, with modifier indicator 1.”

Every finding is citable, reproducible, and auditable. The same bill run through NilesAI today will produce the same findings tomorrow - because the rules don’t hallucinate.

The 16 scan engines

NilesAI’s analysis pipeline runs every bill through 16 specialized validation engines, organized into five categories. Each engine checks for a specific class of billing error using authoritative reference data from CMS, the AMA, and other regulatory bodies.

Code Validation (3 engines)

These engines verify that every code on the bill is real, current, and properly modified.

• CPT Validity - Checks every CPT and HCPCS code against the active code set. Catches typos, fabricated codes, and codes that were never valid.
• Code Currency - Flags codes that were valid in a prior year but have since been deleted or replaced. A code that expired in 2022 shouldn’t appear on a 2024 bill.
• Modifier Validation - Verifies that modifiers (like -25, -59, or -LT) are appropriate for the base code and setting. Incorrect modifiers are one of the most common billing errors.

Bundling & Compliance (2 engines)

These engines enforce CMS packaging and bundling rules.

• NCCI Bundling - Cross-references every code pair against the CMS National Correct Coding Initiative edit tables - 2.6 million edit pairs in total. If two procedures should be bundled into one payment, this engine flags the unbundled charge.
• Global Period Violation - Detects when follow-up services are billed separately despite falling within a procedure’s global surgical period (0, 10, or 90 days). A post-op visit on day 5 after a 90-day global procedure shouldn’t generate a separate charge.

Fee & Charge Analysis (3 engines)

These engines compare what was billed against what should have been charged.

• Fee Schedule Variance - Compares every line item against 7,700+ Medicare Physician Fee Schedule rates from the CMS MPFS. Flags charges that exceed the Medicare allowable by a configurable threshold.
• Charge Outlier Detection - Identifies charges that exceed 5x the MPFS rate - a strong indicator of balance billing, upcoding, or data entry errors.
• Charge Mismatch - When both an EOB and an itemized bill are uploaded, this engine compares charges line by line to find discrepancies between what the provider billed and what the insurer processed. (Learn how to read your EOB to understand these documents.)

Duplicate & Pattern Detection (4 engines)

These engines look for suspicious patterns across line items.

• Duplicate Detection - Finds identical services billed on the same date by the same provider. Exact and near-exact matching catches both obvious and subtle duplicates.
• Upcoding Pattern - Flags when a provider consistently bills at the highest E/M level (e.g., 99215 instead of 99213). The OIG has identified upcoding as a top fraud indicator.
• Frequency Abuse - Detects when the same service is billed at an unusually high frequency - for example, daily physical therapy sessions extending far beyond clinical norms.
• Modifier 25 Abuse - Modifier -25 (significant, separately identifiable E/M service) is one of the most overused modifiers in medical billing. This engine flags suspicious patterns of -25 usage, especially when attached to every E/M code on a claim.

Specialty Scans (2 engines)

These engines handle procedure-specific rules.

• Bilateral Procedure - Verifies correct billing of bilateral procedures using modifier -50 vs. separate line items with -LT/-RT modifiers. Incorrect bilateral billing can double the charge.
• Anesthesia Time Validation - Checks anesthesia time units against procedure duration to ensure time-based billing is accurate and within expected ranges.

Context-Aware Scans (2 engines)

These engines validate billing against clinical and setting context.

• Place of Service Mismatch - Confirms that the place of service code matches the services billed. An office-based procedure code billed with a hospital outpatient POS, for example, may indicate an error - or an attempt to bill at the higher facility rate.
• Units Validation - Checks billed units against 10,500+ Medically Unlikely Edits from the CMS MUE tables. If a procedure has an MUE limit of 1 and a provider bills 3 units, this engine flags it immediately.

The reference data powering every scan

NilesAI’s engines are only as good as their reference data. Every quarter, our team updates the rule sets to reflect CMS’s latest publications. Here’s the scale of data behind every analysis:

This isn’t a static snapshot. NCCI edits alone change every quarter - codes get added, removed, or have their modifier indicators updated. NilesAI’s reference data stays current so your analyses reflect the latest CMS guidance.

Real-world impact: a synthetic case study

Consider a hypothetical hospital bill for Jane Doe, who underwent an outpatient knee arthroscopy. Her itemized bill totals $14,200 across 22 line items. Here’s what NilesAI’s 16 engines might find:

Engine	Finding	Potential Overcharge
NCCI Bundling	CPT 29881 and 29880 are an NCCI edit pair - the arthroscopy with meniscectomy bundles into the more complete code	$1,850
Fee Schedule Variance	Physical therapy evaluation (CPT 97161) billed at $480 vs. Medicare allowable of $112 - a 329% markup	$368
Duplicate Detection	CPT 99213 (office visit) billed twice on the same date of service	$125
Units Validation	Drug administration (CPT 96374) billed for 4 units; MUE limit is 2	$340
Modifier 25 Abuse	Modifier -25 appended to every E/M code across three dates of service	$275

Total potential overcharges flagged: $2,958 - on a single bill, in under two minutes.

For attorneys: cited findings that strengthen demand letters

Personal injury attorneys reviewing medical liens need more than a gut feeling that a bill is inflated. They need specificity - the exact CMS rule, the exact code pair, the exact dollar variance.

NilesAI’s scan engines deliver exactly that. Every finding includes:

• The specific rule or edit that was violated (e.g., NCCI edit pair 29881/29880)
• The regulatory source (CMS NCCI Q1 2024, MPFS CY2024)
• The dollar impact - the difference between what was billed and what the rule allows
• A severity rating to help prioritize which findings to include in demand letters

This level of citation transforms a demand letter from “we believe this bill is inflated” to “this bill contains $3,200 in charges that violate specific CMS billing rules, as documented in the attached NilesAI analysis.” Learn more about common billing errors that appear in litigation.

For insurance and claims operations: consistency at scale

Claims adjusters and SIU teams reviewing hundreds of bills per month need consistency. When two adjusters review the same bill, they should reach the same conclusions. NilesAI guarantees this - because the rules don’t vary by reviewer.

Key advantages for claims operations:

• Audit trail - every finding is logged with the rule, the data, and the timestamp
• Volume processing - analyze bills in minutes instead of hours, per the AAPC’s benchmarks for manual audit time
• Prioritization - severity ratings let adjusters focus on high-impact findings first
• Defensibility - when a provider disputes a finding, you can point to the exact CMS rule

Whether you’re processing 10 bills a month or 10,000, every analysis runs through the same 16 engines with the same reference data. No reviewer fatigue. No knowledge gaps. No inconsistency.

Get started

NilesAI’s 16 scan engines are available today. Upload an itemized bill - or an itemized bill paired with an EOB - and get a complete analysis in minutes. Every finding is cited, every dollar is quantified, and every rule is traceable to its source. Not sure where to start? Our bill diagnostic tool can help you determine which scan engines are most relevant to your specific bill.

Try NilesAI free and see what your bills are really hiding.

You can scan a bill for free now to see what NilesAI finds.