10.26.26
Major payers now adjust some claims by algorithm, and the shortfalls arrive as payments with no denial code attached. AI agents can check every remit for them. Collecting the money still takes people who know how payers work.
A denial arrives with a reason code and lands in your work queue. An underpayment arrives as a payment. The payer accepts the claim, pays less than your contract requires, and the difference often posts as a CO-45 contractual adjustment. The account closes at a zero balance, and the same shortfall repeats on every claim with that payer, code, and contract term until someone checks.
Industry estimates commonly put commercial underpayment losses at 1% to 3% of net revenue. Compare that with hospital margins. Kaufman Hall data put the median hospital operating margin at 1.4% for the year through July 2026. At that margin, a 1% to 3% revenue loss from underpayments can equal or exceed the median hospital's operating income.
Sources of underpayments
Three changes from the past year belong on your audit list.
Automatic downcoding. Cigna, Aetna, Anthem, and Blue Cross Blue Shield of Massachusetts have launched or expanded programs that adjust higher-complexity E/M claims by algorithm, Becker's ASC reported in May. Cigna's policy took effect Oct. 1, 2025, and covers codes 99204-99205, 99214-99215, and 99244-99245. Aetna expanded its claim and code review program from 12 pilot states to every commercial state except Louisiana. Aetna sends no separate notice when it downcodes a claim, so you see the change only on your remit, according to the Indiana State Medical Association. When your documentation supports the level you billed, a downcode is an underpayment.
Inpatient stays are paid at observation rates. Jefferson Health's April 2026 lawsuit accuses Aetna of paying some Medicare Advantage inpatient stays of one to four midnights at observation rates despite physician documentation that supports admission. The health system argues the policy violates the CMS Two-Midnight Rule and its contract with Aetna.
Fee schedule changes. CMS raised the 2026 physician fee schedule conversion factor 3.26%, to $33.40, for clinicians outside qualifying alternative payment models. It also cut work RVUs 2.5% on about 7,700 non-time-based codes. If your commercial contracts pay a percentage of Medicare, each payer has to load new rates for thousands of codes by the date your contract sets. Every rate update gives a payer's system another chance to pay the wrong amount. CMS has proposed a 1.68% cut to that conversion factor for 2027, so you face the same risk in the new year.
Why your current process misses them
Your claims system sends denials to a work queue. Underpayments need someone to look for them, and three gaps keep that from happening.
One rate per code. If your practice management system stores a single expected rate per CPT code, it cannot apply multiple-procedure reductions, modifier rules, carve-outs, or escalators. Your variance report then flags expected reductions alongside underpayments, and your staff has to sort them by hand.
Sampling. If you audit a sample of remits each month, you catch a pattern only when it appears in your sample. A shortfall of a few dollars on a high-volume code can run for months before that happens.
Deadlines. Each payer contract or provider manual sets an appeal window. An underpayment you find after that window closes becomes a write-off.
How AI agents change detection
In a February 2026 HFMA survey of 95 healthcare finance professionals, 27% said their organizations deploy AI at scale across multiple functions, and 53% said they run pilots in select areas. In the same report, McKinsey's Michael Peterson named underpayment management as one of the back-end functions where many hospitals first apply AI. He pointed to the staff hours the work takes and the fixed rules it follows.
Automated detection starts with a model of each contract: rates, fee schedules, multiple-procedure logic, modifier rules, carve-outs, and amendment dates. The software compares every line on every remit to the payment the contract requires. AI agents add three capabilities.
Plain-English questions. Your analysts can ask, "Which payers paid 99214 below contract since Oct. 1?" and get the answer, with the claims behind it, in seconds. They skip the report request and the wait.
Scheduled monitoring. An agent can review each new batch of remits on a schedule. It flags new payer behavior, such as a payer paying more level 5 visits at level 4, in the first batch where it shows up.
Root cause and priority. An agent can group variances by payer, code, location, provider, and modifier, then rank the groups by dollar value. Your team starts with the patterns worth the most.
Check the math before you trust a flag. Ask any vendor to show the contract term and calculation behind each variance. Your team needs that detail to confirm a finding, and every appeal you send needs it too.
Why recovery still takes people
Finding a variance does not collect it. Recovery means reading contract language, building an appeal in the format each payer requires, following up by phone and portal, and escalating when a payer stops responding.
Appeals also change how payers treat you. The Indiana State Medical Association reported in 2025 that practices could seek early removal from Aetna's downcoding review by winning appeals on 75% of their downcoded claims. Regulators can act too. Maryland's Insurance Administration fined Cigna $80,000 in March 2026 and ordered it to stop automatic downcoding in the state. Cigna must now dispute claims it questions there and request documentation. Your state insurance regulator gives you one more escalation path.
Send repeat findings to your managed care team. A payer that underpays the same code across locations gives you evidence for your next contract negotiation. If your staff lacks the hours for this work, outside recovery specialists can work the backlog while your team handles current claims.
Five steps to take before your next remit cycle
Load every contract, amendment, and fee schedule into one system that calculates the expected payment for each claim line.
Compare 100% of remits to contract terms, including zero-balance accounts and CO-45 adjustments.
Compare billed and paid E/M levels by payer each month. A rise in level 4 and 5 visits that payers pay one level lower points to downcoding.
If your contracts pay a percentage of Medicare, compare 2026 allowed amounts for your top 20 codes to the contracted percentage of 2026 Medicare rates. Repeat the check in January 2027.
Sort open variances by dollars and appeal deadline, and send repeat patterns to managed care.
Pair detection with recovery
Each new payer policy and fee schedule update can create a new underpayment pattern. You protect your margin when software checks every payment against your contracts and specialists pursue every valid variance.
MD Clarity built RevFind, its agentic revenue optimization software, for the first job. RevFind's AI agent answers plain-English questions about your remit and contract data, monitors new remits in the background, and flags underpayment and denial patterns as they appear. Your team can verify the contract terms behind every expected rate.
For the second job, MD Clarity's Revenue Recovery Services team works from RevFind's findings. Its payer reimbursement specialists verify underpayments, generate appeals, follow up with payers, and trace root causes, with a fee contingent on what you recover. Together, RevFind and Revenue Recovery Services help your organization get paid fully and fairly.