How to Read an EOB and ERA: A Line-by-Line Guide for Treatment Centers

The remittance is where money quietly disappears. Not in a dramatic, obvious way. A flat denial with a dollar amount of zero is easy to see and easy to work. The leak is subtler. A 90837 paid at the 90834 rate. A contractual adjustment that's $12 larger than your contract says it should be. An allowed amount that drifted down 4% when the payer "reprocessed" a batch. Multiply those small line-level gaps across a treatment center's claim volume and you're looking at five or six figures a year that never shows up as a denial at all.

The Explanation of Benefits (EOB) and its electronic cousin, the Electronic Remittance Advice (ERA), record every one of those decisions: payment, adjustment, denial, downcode, patient responsibility. They arrive as a dense grid of codes and dollar amounts that most billing teams scan rather than read. That habit is expensive. The adjudication detail on a remittance is the single richest source of truth you have about how each payer actually treats your claims. "eob medical abbreviation" is one of the most-searched billing questions for a reason: the codes are not self-explanatory.

This guide walks line by line through both documents. It defines the EOB, ERA, and underlying 835. It maps the anatomy of an EOB field by field, decodes the CARC and RARC codes you'll actually see, and shows how to spot underpayments and silent downcoding. It builds a reconciliation workflow for claim volume rather than a solo caseload, and draws the line between what's worth appealing and what's worth writing off. The goal is a billing team that reads remittances instead of skimming them, and recovers the revenue skimming leaves on the table.

EOB vs ERA vs the 835: what each one actually is

These three terms get used interchangeably, and they shouldn't be. They describe related but distinct things.

An EOB (Explanation of Benefits) is the human-readable statement a payer issues explaining how a claim was adjudicated. Historically it arrived on paper, attached to a paper check. It's written for a person to read, whether patient, provider, or biller, and shows the billed amount, the allowed amount, what the plan paid, what was adjusted, and what the patient owes. The patient gets a version too, often labeled "This is not a bill." Yours, the provider remittance, carries the operational detail.

An ERA (Electronic Remittance Advice) is the electronic equivalent. It delivers the same adjudication information as a structured data file your practice management system can ingest automatically and post against open claims. No re-keying, no scanning. The ERA is what lets a treatment center auto-post hundreds of line items instead of typing them in one at a time.

The 835 is the actual technical standard underneath the ERA. Formally the ASC X12N 835 Health Care Claim Payment/Advice, it's the HIPAA-mandated electronic transaction format for remittance and payment data: the EDI counterpart to the 837 you send claims in. When someone says "we get ERAs from this payer," they mean the payer transmits 835 files. Under HIPAA's electronic transaction rules, payers are required to support the 835 standard for electronic remittance.

Sources: CMS - Electronic Billing & EDI Transactions (835/837) · CMS - Electronic Remittance Advice (ERA) overview

One more pairing to know: every 835 references the ERA/EFT relationship. The EFT (Electronic Funds Transfer) is the actual money moving into your bank account. The 835 is the explanation of what that deposit covers. They're transmitted separately and reassociated using a trace number (the TRN segment / EFT trace / check or EFT number). When a deposit hits your bank and you can't tell which claims it covers, you're missing the 835-to-EFT reassociation, a common and fixable reconciliation gap.

Anatomy of an EOB: the fields that matter

Every payer formats its EOB a little differently, but the underlying data elements are standardized by the 835. Once you can name them, you can read any payer's layout. Here are the fields to find on every remittance, roughly in the order money flows.

Header / payer-provider block. Payer name and payer ID, your provider/group NPI and tax ID, the check or EFT number, the TRN trace number, and the payment date. This is your reassociation key: the link between this remittance and the deposit in your bank.

Patient and claim identifiers. Patient name, member/subscriber ID, your patient account number, the payer claim control number (CLM / ICN / DCN) the payer assigned, and the claim status. The claim control number is what you cite when you call about this specific claim. Keep it.

Dates of service and place of service. The DOS for each line, and the POS (e.g., 11 office, 02 telehealth away from home, 10 telehealth in home). POS mismatches are a frequent adjustment trigger in behavioral health telehealth.

Now the money columns, line by line. These are the six numbers that have to reconcile on every service line:

Sources: CMS - Remittance Advice Resources & Adjustment Codes · X12 - 835 Health Care Claim Payment/Advice

The arithmetic that has to hold on every line: Billed = Paid + Patient Responsibility + Adjustments. When it doesn't balance, something was coded wrong or recorded wrong, and that line needs a human. The single most useful habit you can build into a billing team is checking that this equation closes on every posted line. Most auto-posting tools enforce it. Plenty of manual posting doesn't.

Glossary / footer. Most EOBs print the CARC and RARC code definitions used on that statement at the bottom. Read them. Payers update which codes they use, and the footer is the fastest way to learn a new one without leaving the page.

The adjustment and denial codes you'll actually see: CARC and RARC

Two code sets do almost all the explaining on a remittance, and they work as a pair.

CARCs (Claim Adjustment Reason Codes) explain why an amount was adjusted: paid less than billed, or not paid at all. Each CARC is paired with a Group Code that tells you who is responsible for the adjusted amount. The group code is as important as the number:

• CO (Contractual Obligation): the provider eats it. A contractual write-off you can't bill anyone for.
• PR (Patient Responsibility): bill the patient. Deductible, coinsurance, copay, non-covered.
• OA (Other Adjustment): neither cleanly. Often used when another group code doesn't fit, frequently with COB.
• PI (Payer Initiated Reduction): the payer's own reduction, not contractually agreed and not the patient's.

A "CO-45" is really a Group Code (CO) plus a CARC (45). Reading the group code first tells you immediately whether a line is a write-off, a patient bill, or something to investigate.

RARCs (Remittance Advice Remark Codes) add detail the CARC alone can't carry. They come in two flavors: alert codes (start with "N" or "MA"/"M") that provide supplemental information, and codes that point to a specific missing or invalid element. A CARC tells you the claim was denied for missing information. The paired RARC tells you which information. You almost always need both to act.

Here are the codes a behavioral health treatment center sees most, with plain-English meaning and the usual next step. Verify any code's current definition against the official X12 list before you build a workflow rule on it. Definitions are maintained and revised on a published schedule.

Sources: X12 - Claim Adjustment Reason Codes (CARC) · X12 - Remittance Advice Remark Codes (RARC) · CMS - Washington Publishing Company code lists / RA code maintenance

Two practical notes on this table. First, CO-16 is not a denial you appeal. It's a correction you resubmit. Treating CO-16 lines as denials and routing them to an appeals queue wastes the most expensive resource you have. Read the paired RARC (e.g., N290 for a missing/invalid rendering provider NPI, MA27 for a missing/invalid patient name) and fix the field. Second, the group code changes the whole meaning of a number. A 204 under PR may be billable to the patient. The same 204 under CO is your problem. Always read the group code first.

Spotting underpayments and soft denials: silent downcoding

Flat denials are visible. The revenue that actually goes missing on a treatment center's remittances is the revenue that was paid, just at the wrong amount. This is the underpayment problem, and its most common form in behavioral health is silent downcoding: the payer pays you, but for a cheaper code than the one you billed, with no rejection and often no obvious flag.

The classic example: you bill 90837 (the 60-minute psychotherapy code), and the remittance pays you at the 90834 (45-minute) rate. No CARC screams "we downcoded you" in plain language. The line just shows a lower allowed amount, sometimes with a vague RARC, sometimes with nothing. Because it posts as a payment, auto-posting tools accept it and the variance never reaches a human. As we documented in our behavioral health denials guide, soft denials like this are common and routinely untracked. A 90837 paid at the 90834 rate, at a $30–$50 gap on 8–10% of claims, runs into thousands of dollars per 1,000 claims that no one ever appeals.

The only reliable way to catch this is expected-versus-actual reconciliation by CPT code. You maintain a per-payer expected allowed amount for each code, and you compare every paid line against it. The variance report is where soft denials surface.

Sources: HFMA - Reducing Underpayments and Improving Yield · Supahealth aggregate data from 200+ behavioral health practices.

The IOP line is a visible denial. It shows zero paid and a CO-197, so your team will work it. The first three lines are the dangerous ones: each paid, each posts cleanly, and each is short. Without an expected-versus-actual check by code, none of them gets flagged. The discipline that catches them is unglamorous: a monthly (or, at volume, weekly) variance report grouped by payer and CPT, with a threshold that routes any line paying below contract to a human.

A note on why this is getting harder. Payers increasingly adjudicate with automation at machine speed. The well-documented case is Cigna's internal "PxDx" system, which a ProPublica investigation reported let its doctors reject more than 300,000 claims over two months at about 1.2 seconds each, without opening patient files. Automated adjudication isn't inherently wrong. It speeds legitimate payments too. But the asymmetry is real: when one side reduces at algorithmic scale, a provider checking remittances by hand will miss the patterns. The durable response isn't more manual scrutiny. It's matching automated adjudication with automated reconciliation.

The reconciliation workflow at a treatment center's volume

A solo therapist can eyeball a handful of EOBs a week. A treatment center processing hundreds or thousands of claims across IOP, PHP, outpatient, and multiple payers cannot. The workflow has to be built for that volume, not improvised per remittance. Here's the sequence that holds up at scale.

1. Ingest the 835 and auto-post the clean lines. Pull ERAs electronically from every payer that supports them (most do, under the HIPAA 835 mandate). Auto-post lines that balance (billed = paid + patient responsibility + adjustments) and that pay at or above the expected allowed amount. These need no human.

2. Reassociate 835 to EFT. Match each remittance to its deposit using the TRN trace number. Deposits you can't tie to an 835 are a reconciliation hole. Chase the missing remittance rather than posting a lump sum blind.

3. Route the exceptions. Anything that doesn't auto-post cleanly goes to a worklist, bucketed by type: true denials (zero paid), underpayments (paid below expected), patient-responsibility lines that need statementing, and CO-16-style correctable rejections. Each bucket has a different owner and a different action. Don't dump them into one undifferentiated "denials" pile.

4. Run the variance report. Group paid lines by payer and CPT, compare against expected allowed, and surface every line below threshold. This is the step most practices skip and the one that recovers soft-denial revenue.

5. Work by dollars and by deadline, not by date received. Prioritize the worklist by recoverable dollars and by each payer's appeal/timely-filing window, so high-value and time-sensitive lines get worked before they age out.

6. Feed the patterns back upstream. When a payer consistently downcodes a code or denies for a recurring missing element, that's not just an appeal. It's a signal to fix the front end (the benefits check, the coding rule, the documentation prompt) so the next claim doesn't repeat it.

The reason most centers leak revenue here isn't laziness. It's volume. There is simply too much line-level detail, across too many payers and codes, for a human team to hold in working memory and check by hand on every remittance. The work is rules-based and high-volume, which is exactly the kind of work that gets done inconsistently when it's manual.

When to appeal versus write off

Not every adjustment is recoverable, and chasing the ones that aren't burns money. The decision comes down to three questions: is the adjustment correct, is it appealable, and is it worth appealing?

Write it off when the adjustment is correct. A CO-45 that matches your contracted rate is not a loss. It's the discount you agreed to. PR-1/PR-2/PR-3 lines aren't write-offs at all. They're patient-billable, and writing them off instead of statementing the patient is a quiet, common revenue leak. A CO-29 timely-filing denial with no proof of timely submission is generally dead.

Appeal when the adjustment is wrong or the missing element is fixable. A CO-45 larger than your contract allows is an underpayment, so appeal with your fee schedule. A CO-197 where an auth actually exists is an administrative error, so appeal with the auth number. A CO-50 medical-necessity denial is appealable with documentation that establishes necessity. A CO-16 isn't an appeal at all: read the RARC, fix the field, resubmit.

Then weigh the economics. Reworking a denied claim has a real cost. Industry estimates put rework in the rough range of $25 to upwards of $100 per claim depending on complexity and whether it reaches a formal appeal. If the recoverable amount on a $40 line is less than the cost to chase it, selective write-off is the rational call. The exception is pattern denials: a single $40 underpaid line may not be worth a manual appeal, but the same $40 downcode appearing on 300 claims is a batch recovery, and a front-end fix, worth real effort.

Sources: HFMA - Cost-to-collect and rework benchmarks · KFF - Claims Denials and Appeals in ACA Marketplace Plans in 2024

Worth remembering: most denials that could be overturned never get appealed at all. In Medicare Advantage, KFF found only a small share of denied requests are appealed even though the large majority of appeals succeed. Denials stick mostly because nobody fights them, not because they were right. At a treatment center's volume, the appeal-versus-write-off decision shouldn't be a per-claim judgment call made under time pressure. It should be a rule, applied consistently, with dollar and deadline thresholds set in advance.

AI and agentic systems in remittance reconciliation

AI in revenue cycle has moved past "better OCR" on scanned EOBs. What's emerging now is agentic systems that handle the remittance workflow end-to-end: ingesting the 835, reconciling line by line, and surfacing the exceptions a human should actually look at, with people reviewing at the edges rather than typing in the middle. For reading and reconciling EOBs and ERAs specifically, two agents in the Supabill stack do the relevant work: the claims agent and the denials process management agent.

Here's what that looks like in practice. The claims agent holds payer-by-payer and state-by-state guidelines, including expected allowed amounts by CPT, in its core database. When an 835 comes in, it doesn't just auto-post the balanced lines. It checks every paid line against the expected reimbursement for that code and payer. So when a 90837 pays at the 90834 rate, the agent flags the variance as a soft denial in real time instead of letting it post silently as a clean payment. That's exactly the high-volume, rules-based comparison a human team can't sustain across every code and payer by hand.

Lines that are true denials flow to the denials process management agent, which reads the CARC/RARC pair, routes by code and payer, and drafts the appropriate response: a corrected resubmission for a CO-16 with a missing-NPI RARC, an auth-attached appeal for a CO-197 where an auth exists on file, a medical-necessity appeal for a CO-50 built from the matched clinical documentation. It tracks resubmission cycles so nothing ages past a timely-filing window, and it learns which appeal patterns succeed with which payer. Because these agents share state, a downcode pattern the claims agent keeps flagging becomes a documentation or coding signal that feeds back upstream, so the next claim is built to avoid the adjustment rather than to appeal it after the fact.

Honest limits. An agent can't make a payer pay above its own policy. If a plan genuinely doesn't cover a service, no reconciliation logic changes that. It can't manufacture an authorization that was never obtained, and it can't turn a clinically thin session into a billable one. What it does is close the gap that volume creates: the line-by-line checking, the expected-versus-actual comparison, and the deadline tracking that humans get wrong simply because there's too much of it. The judgment calls (which pattern denials to fight, when to escalate a contract dispute) stay with your team.

If you're interested, book a demo here to learn more.

Quick wins

Things a treatment center's billing team can start this week:

1. Switch every payer that supports it to ERA/835. If you're still keying paper EOBs for any volume payer, electronic remittance plus auto-posting is the fastest reduction in manual effort and posting errors available to you.

2. Build a per-payer expected-allowed table for your top 10 CPT codes. This single table is what makes underpayment detection possible. Without it, you can't tell a downcode from a normal payment.

3. Run a variance report on last month's paid claims, grouped by payer and CPT. Look specifically for 90837 paying at the 90834 rate, and any allowed amount below contract. You will almost certainly find recoverable money.

4. Stop writing off PR lines. Audit a sample of recent remittances for PR-1/PR-2/PR-3 amounts that got written off instead of billed to the patient. That's collectible revenue mislabeled as a loss.

5. Reassociate every deposit to an 835. Find the deposits in your bank that you can't tie to a remittance via the TRN trace number. Each one is a reconciliation hole hiding posting errors.

6. Make the group code the first thing your team reads on any adjustment. CO versus PR versus OA changes whether a line is a write-off, a patient bill, or an investigation. Reading the number before the group code causes mis-posts.

FAQ

Q: What does "EOB" stand for, and how is it different from an ERA?

A: EOB stands for Explanation of Benefits, the human-readable statement of how a claim was adjudicated (billed, allowed, paid, adjusted, patient responsibility). An ERA (Electronic Remittance Advice) is the same information delivered as a structured electronic file your system can auto-post. The EOB is the document. The ERA is the data feed. Both convey the same adjudication; only the ERA can be posted automatically.

Q: What is the 835, and do I need to understand it to read remittances?

A: The 835 is the HIPAA-mandated EDI standard (formally the ASC X12N 835) that underlies the ERA. It's the actual file format payers transmit. You don't need to read raw 835 segments to do your job, because your practice management system translates them into readable lines. But knowing the term matters when you're talking to a payer or clearinghouse about why an ERA isn't posting correctly.

Q: My deposit hit the bank but I can't tell which claims it covers. What happened?

A: You're missing the 835-to-EFT reassociation. The EFT (the money) and the 835 (the explanation) transmit separately and are linked by a trace number: the TRN segment, sometimes called the EFT trace or reassociation trace number. If the 835 didn't arrive or didn't match, you have a deposit with no detail. Contact the payer or your clearinghouse to retrieve the missing 835 rather than posting the deposit as a lump sum.

Q: What's the difference between a CARC and a RARC?

A: A CARC (Claim Adjustment Reason Code) explains why an amount was adjusted and pairs with a group code showing who's responsible. A RARC (Remittance Advice Remark Code) adds supplemental detail the CARC can't carry, often the specific missing or invalid element. A CO-16 CARC says "claim lacks information"; the paired RARC tells you exactly which field. You usually need both to act.

Q: What does the group code (CO, PR, OA, PI) actually change?

A: Everything about how you handle the line. CO (Contractual Obligation) is a write-off you can't bill anyone for. PR (Patient Responsibility) is billable to the patient. OA (Other Adjustment) is usually neither and often appears with coordination of benefits. PI (Payer Initiated) is a payer reduction that isn't contractual and isn't the patient's. The same reason-code number means different things under different group codes, so read the group code first.

Q: Is a CO-45 always a write-off I should accept?

A: No. CO-45 means the charge exceeded the allowed amount, but it only validates as a clean write-off if the allowed amount matches your contracted rate. If the contractual adjustment is larger than (billed minus your contracted rate), the payer underpaid you and the line is appealable with your fee schedule. The mistake is auto-writing-off every CO-45 without validating the allowed amount against the contract.

Q: What exactly is a "soft denial" or silent downcode?

A: It's a claim that was paid, but for a cheaper code than you billed (most commonly a 90837 paid at the 90834 rate) with no formal rejection. Because it posts as a payment, it doesn't appear in a denials report and auto-posting accepts it. The only way to catch it is expected-versus-actual reconciliation by CPT: comparing each paid line against the rate you expected for that code and payer.

Q: How do I tell a downcode from a normal contractual adjustment?

A: You need a per-payer, per-CPT expected allowed amount to compare against. A normal CO-45 brings billed down to your contracted rate. A downcode pays you the contracted rate for a different, cheaper code. Without the expected-allowed table, both just look like "paid less than billed," which is why building that table is the prerequisite for catching underpayments at all.

Q: CO-16 keeps showing up. Should my team appeal these?

A: Usually no. CO-16 is a correctable rejection, not a true denial. It means the claim is missing information or has a submission error, and it's almost always paired with a RARC naming the problem (a missing rendering NPI, an invalid patient identifier, etc.). The fix is to read the RARC, correct the field, and resubmit. That's much cheaper and faster than routing it to a formal appeals queue.

Q: How should a treatment center decide what to appeal versus write off?

A: Make it a rule, not a per-claim judgment. Write off adjustments that are correct (a CO-45 matching contract, a timely-filing CO-29 with no proof of submission). Appeal adjustments that are wrong or fixable (an underpaid CO-45, a CO-197 with an auth on file, a CO-50 you can document). Then weigh economics: a lone $40 underpayment may not be worth a manual appeal, but the same downcode across 300 claims is a batch recovery and a front-end fix worth pursuing.

Q: Why are underpayments getting harder to catch?

A: Payers increasingly adjudicate with automation at high speed and scale. ProPublica's reporting on Cigna's PxDx system described automated review of hundreds of thousands of claims at roughly a second each. Automation isn't inherently the problem; the asymmetry is. When adjudication runs at machine scale and your reconciliation runs by hand, the patterns slip past. Matching automated adjudication with automated reconciliation is the durable answer.

Q: How often should we run remittance reconciliation and variance reports?

A: Auto-post and reassociate continuously as ERAs arrive. Run the expected-versus-actual variance report at least monthly, and weekly if your claim volume is high or you've spotted a payer with a downcoding pattern. The sooner you catch a systematic underpayment, the more of those claims are still inside the appeal and timely-filing windows.

Q: Do all payers have to send ERAs?

A: Under HIPAA's electronic transaction rules, payers are required to support the 835 standard for electronic remittance, and most do. Some smaller or out-of-network payers still issue paper EOBs only. For any payer with meaningful volume, enrolling in ERA/EFT is worth the setup effort. It's the difference between auto-posting and re-keying.

References

• American Medical Association. (2024). Prior authorization survey and reform resources. https://www.ama-assn.org/practice-management/prior-authorization
• Centers for Medicare & Medicaid Services. (2024). Electronic Billing & EDI Transactions (835/837). https://www.cms.gov/medicare/billing/electronic-billing-edi-transactions
• Centers for Medicare & Medicaid Services. (2024). Electronic Data Interchange (EDI) - Coordination of Benefits & Recovery. https://www.cms.gov/medicare/coordination-benefits-recovery/electronic-data-interchange
• Centers for Medicare & Medicaid Services. (2024). Remittance Advice resources and adjustment/remark code maintenance. https://www.cms.gov/medicare/coordination-benefits-recovery/electronic-data-interchange/remittance-advice
• HFMA. (2025). Denials management resources. https://www.hfma.org/revenue-cycle/denials-management/
• HFMA. (2025). Revenue cycle: improving yield and reducing underpayments. https://www.hfma.org/revenue-cycle/
• HFMA. (2024). Claim Integrity Task Force: Standardizing Denial Metrics. https://www.hfma.org/guidance/standardizing-denial-metrics-revenue-cycle-benchmarking-process-improvement/
• ICANotes. (2025). Behavioral Health Billing Metrics & KPIs 2025. https://www.icanotes.com/2025/09/24/behavioral-health-billing-metrics-kpis/
• KFF. (2025). Claims Denials and Appeals in ACA Marketplace Plans in 2024. https://www.kff.org/patient-consumer-protections/claims-denials-and-appeals-in-aca-marketplace-plans-in-2024/
• KFF. (2025). Medicare Advantage Insurers Made Nearly 53 Million Prior Authorization Determinations in 2024. https://www.kff.org/medicare/medicare-advantage-insurers-made-nearly-53-million-prior-authorization-determinations-in-2024/
• ProPublica. (2023). How Cigna Saves Millions by Having Its Doctors Reject Claims Without Reading Them. https://www.propublica.org/article/cigna-pxdx-medical-health-insurance-rejection-claims
• SimiTree. (2025). How to Calculate and Improve Your Behavioral Health Clean Claims Rate. https://simitreehc.com/simitree-blog/how-to-calculate-and-improve-your-behavioral-health-clean-claims-rate/
• Washington Publishing Company / X12. (2025). Claim Adjustment Reason Codes (CARC). https://x12.org/codes/claim-adjustment-reason-codes
• Washington Publishing Company / X12. (2025). Remittance Advice Remark Codes (RARC). https://x12.org/codes/remittance-advice-remark-codes
• X12. (2025). 835 Health Care Claim Payment/Advice - transaction sets. https://x12.org/products/transaction-sets
• Supahealth. (2026). Aggregate billing, remittance, and denials data from 200+ behavioral health practices. Internal dataset.

For more on behavioral health revenue cycle, see our guides on Why Behavioral Health Claims Get Denied and Behavioral Health Billing for Treatment Centers.