Skip to content

Systems and records

Medical billing payment posting dataset: tracing each posting correction

By SourceX Editorial · Reviewed by Noah Loul ·

10 sections

Short answer

Payment posting errors become traceable administrative training cases when each one is captured as a correction episode: the remittance line as received, the posting as entered, the error and how it was detected, the correction, and a later reconciliation that proves the ledger now matches the remittance. Keep the case administrative, strip clinical detail and identifiers, and accept an outcome only from ledger evidence.

Value drivers

Medical billing payment

This page covers medical billing payment posting dataset: tracing each posting correction.Illustrative

Key takeaways

  • The unit of a useful posting dataset is the correction episode, not the payment line or the claim.
  • Each case needs the remittance as received, the posting as entered and the corrected posting side by side, so the difference is visible.
  • Accept an outcome only when a later reconciliation shows the ledger matching the remittance; a closed status or a note that says fixed is not proof.
  • Record the payer companion guide version and the code list in force on the remittance date, because code usage changes over time.
  • Drop diagnoses and other clinical detail, replace identifiers with keys, and have a qualified privacy reviewer confirm the de-identification method.

How can payment posting errors become traceable administrative training cases?

They become traceable when every error is stored as one correction episode that links four things: what the payer sent, what the poster entered, what was changed, and the reconciliation that later proved the change right. A payment line on its own only shows money arriving. The episode shows a decision, which is what an AI system meant to assist posters or reviewers has to learn.

Most billing offices already hold the pieces. The electronic remittance advice sits in the clearinghouse archive, postings and adjustments sit in the practice management or billing system, corrections sit in an adjustment log or a work queue, and month-end reconciliation reports sit with the finance team. What is missing is the join between them, with dates and reasons.

This page covers the administrative side only: remittance, posting, adjustments and reconciliation. It does not cover coding, medical necessity or any clinical judgment, and nothing here is advice about patient care.

Where posting errors enter between the remittance and the ledger

Errors enter wherever a person or rule translates the remittance into ledger entries, and the 835 structure tells you which translations matter. A claim adjustment in the remittance carries a group code that assigns responsibility for the amount: CO for contractual obligation, PR for patient responsibility, OA for other adjustments and PI for payer-initiated reductions, as X12 defines them. A claim adjustment reason code explains why the line was paid differently than billed, and a remittance advice remark code adds explanation to that adjustment or, when prefaced with Alert, conveys general information.

Above the claim level sit provider-level adjustments. CMS's remittance guidance explains that group codes, reason codes and remark codes describe adjustments at the claim or line level, while PLB reason codes describe adjustments at the provider level, and that the check amount equals the sum of claim-level payments less provider-level adjustments. A recoupment or interest amount in the PLB segment therefore never appears on the claim it relates to, which makes it a common source of misapplied balances.

Before any of that, the payment itself has to be matched to its remittance. The CAQH CORE Phase III reassociation rule (CORE 370) says the X12 EFT trace number in the TRN segment is the only trace number used to match the EFT to its ERA, and that it is not the same number as the ACH trace number. Offices that match on the wrong number create unapplied cash that later has to be traced and moved.

The posting correction case record

The table below is the asset: the fields one de-identified correction case should carry, where each comes from and how it is treated before the case leaves the office. Field names are illustrative and form an editorial design, not an X12 or CAQH CORE structure.

The posting correction case record
FieldSourceTreatmentWhy the case needs it
case_idGenerated at extractionRandom key, no link to account numbersJoins all evidence for one episode
remit_key and payer_guide_version835 trace number; payer companion guide on fileTrace number replaced by a keyed hash; guide version keptShows which remittance and which payer rules applied
line_as_received835 claim and line adjustmentsGroup code, reason code, remark code and amount fields kept as symbolic or scaled valuesThe ground truth the posting should match
posting_as_enteredBilling system transaction historyTransaction type, adjustment code and target bucket kept; user names replaced by roleShows the decision that was wrong
error_typeReviewer label from the taxonomyControlled valueLets cases be grouped and sampled
detection_signalReconciliation report, denial follow-up or patient inquirySignal type and days from postingShows how the error surfaced
correction_entriesAdjustment log or reversal and repostBefore and after entries, role of corrector, approval stepThe corrective action to learn from
routingWork queue historyQueue names generalized; timestamps converted to elapsed daysShows who decided and who approved
reconciliation_evidenceMonth-end or batch reconciliationMatched or unmatched flag with the reconciliation periodProves the outcome
exclusions_appliedPreparation logList of fields dropped or generalizedLets a privacy reviewer check the case

Which administrative evidence the case needs, and which clinical detail stays out

Keep only what is needed to show the remittance, the posting and the correction; drop everything that describes the patient's condition or care. The HIPAA minimum necessary standard asks covered entities to make reasonable efforts to limit protected health information to what the intended purpose requires, and a posting correction case is a narrow purpose.

In practice, diagnosis codes, clinical notes, attachments and anything from the medical record have no place in a posting case. Procedure codes are a closer call: the poster needs them to match a payment to the right line, but the case usually does not. Replace them with a line sequence key, and keep a coarse service category only if a reviewer agrees it is needed to explain the error.

Free-text fields deserve the most care. Posting notes and work queue comments often mention patient names, phone calls or symptoms. Either drop them or replace them with a coded reason chosen by a reviewer; do not rely on automated redaction alone for notes that will train a model.

Six posting errors and the evidence that proves each one fixed

Label every case with one error type, and accept it only if the proving evidence in the last column exists. These categories are editorial and should be adapted to the office's own adjustment codes.

Six posting errors and the evidence that proves each one fixed
Error typeTypical detection signalCorrect postingEvidence that proves the fix
Patient responsibility written off as contractualPatient balance totals below the PR totals on remittances for the periodReverse the write-off and transfer the amount to patient responsibilityNext reconciliation shows PR totals matching; patient balance created
Contractual amount billed to the patientPatient inquiry or statement reviewReverse the patient transfer and post the CO adjustmentPatient balance cleared; contractual adjustment equals the remittance
Provider-level recoupment not tied to the original claimCheck amount differs from the sum of posted claim paymentsApply the PLB amount to the claim it recoversBatch total equals the check amount; claim balance updated
Payment posted to the wrong claim or lineCredit balance on one account, open balance on anotherReverse and repost to the matching claim and lineBoth accounts reconcile to their remittance lines
Unmatched EFTDeposit with no applied remittanceReassociate using the TRN trace number, then postUnapplied cash cleared in the bank reconciliation
Duplicate posting of a reprocessed claimPaid amount exceeds the latest remittance for the claimReverse the duplicate, keep the reprocessed postingClaim paid total equals the latest remittance

Judging a posting by the payer rules in force on the remittance date

A posting is correct only relative to the payer's usage of codes on the date the remittance was issued, so every case has to carry that context. X12 maintains the claim adjustment reason code list and revises it, and the list page shows when it was last modified. How an individual payer applies those codes is its own policy, usually set out in a companion guide, so a case labelled wrong under today's guide may have been right when it was posted.

Store three things with each case: the version or effective date of the payer companion guide the office had on file, the code list effective date that applied to the remittance, and any contract amendment that changed allowed amounts during the period. When the office cannot tell which version applied, label the case unverified rather than guessing.

Routing belongs in the same record. Note whether the poster, a team lead or a payment variance analyst made the correction, whether a second person approved it, and how many days passed between posting and detection. Those details turn a case from a before and after pair into a decision trail, which is what a model that drafts corrections for human approval needs to learn. They also let a reviewer see whether a correction followed the office's own approval rules or skipped them, which is itself a reason to hold the case back.

Worked example: a coinsurance amount written off as contractual

This case is fictional. Amounts are shown as letters and every count is illustrative.

Starting state: a remittance line shows an allowed amount A, a payer payment P and a patient coinsurance adjustment R reported under group code PR with a coinsurance reason code. The poster, working a large batch, keys R as a contractual write-off. The claim shows a zero balance, the work queue item closes and nobody is billed.

Detection: at month end, the reconciliation compares patient responsibility reported on remittances with patient balances created in the ledger. That period shows a gap of 14 lines across 3 payers. Linked evidence: for this line, the case record holds the remittance line with group PR, the posting entry with a contractual adjustment code, and the reconciliation row that flagged the gap.

Decision and correction: a posting supervisor approves reversal of the write-off and a transfer of R to patient responsibility. The case stores both entries, the supervisor role and the approval step. Outcome check: the next reconciliation period shows the line matched, and the patient statement cycle includes the balance. Only that reconciliation row, not the closed queue item, marks the case as resolved.

Counterexample: a second case has a work queue item marked corrected and a note reading fixed per payer call, but no reversal or repost entries and no later reconciliation row. It may well have been fixed. It is still excluded, because nothing in the record proves what changed or that the ledger now matches the remittance.

What to strip before correction cases leave the billing office

Strip direct identifiers, clinical content and exact dates tied to a patient, and have a qualified privacy reviewer confirm which HIPAA de-identification method the case set relies on. HHS guidance describes two methods: Safe Harbor, which removes 18 types of identifiers and requires no actual knowledge that the rest could identify a person, and Expert Determination, in which a qualified expert concludes the risk of re-identification is very small.

Posting data carries several identifiers that are easy to miss: account and claim numbers, member identifiers, check and trace numbers, service dates and payment dates. Replace numbers with keyed hashes held only by the supplier. Whether elapsed day counts derived from service dates are acceptable is a question for the privacy reviewer or the expert, not a default.

A billing company working for providers also has to check what its business associate agreements allow before any reuse. De-identification changes the privacy analysis, but it does not by itself give a billing company the right to license records that belong to its clients.

Which correction cases to prepare first, and which to leave out

Start with error types that are frequent, clearly labelled and closed by a reconciliation, because those produce the most trustworthy cases. Patient responsibility misclassifications and unmatched EFTs usually meet that bar. Wrong-claim postings often do too, once both accounts have reconciled.

Leave out cases still under payer dispute, cases whose remittance file is missing from the archive, and cases where the correction depended on a clinical determination such as a coding change. Those either have no proven outcome or fall outside an administrative task. Record each exclusion and its reason, so the final set can show how it was drawn.

Offering correction cases for a licensing review

For billing offices, SourceX handles rights review, contracting, data preparation, delivery and payment when selected business records are licensed to AI developers. The case record and error table above let an office describe what its posting history contains without sending any of it.

Scope and terms are approved and the agreement is signed before anything is delivered, and de-identification is carried out by SourceX through approved processing routes. A short company assessment, with no records uploaded, shows whether the company qualifies.

Frequently asked questions

What is a medical billing payment posting dataset?

It is a set of records drawn from remittance files, ledger postings and reconciliations. For AI work, the most useful version captures posting errors and their corrections as linked cases, so a model can learn what a correct posting looks like and how errors are found. It should be de-identified and limited to administrative fields.

Do posting cases need diagnosis or procedure codes?

Diagnosis codes are not needed for posting correction cases and should be dropped. Procedure codes help match a payment to a line inside the billing system, but the case can usually use a line key instead. Keep a coarse service category only when a reviewer agrees it explains the error.

Why is a closed work queue item not enough to prove a correction?

A status field records that someone closed the item, not that the ledger now matches the remittance. A reconciliation row from a later period shows the result directly. Cases without it should be excluded or labelled unresolved.

Why record the payer companion guide version?

Payers apply adjustment and remark codes under their own policies, usually described in companion guides, and X12 revises the code lists over time. Recording the guide version and code list date in force on the remittance date lets a reviewer judge a posting by the rules that applied then.

Can a billing company license posting data it processes for clients?

Only if its client agreements and business associate agreements allow it, and a qualified privacy reviewer and counsel confirm the specific facts. De-identification alone does not create that right. Many billing companies start by asking clients for explicit permission.

Sources

  • X12 claim adjustment group codes CO, PR, OA and PI assign responsibility for adjustment amounts. Source
  • Claim adjustment reason codes explain why a claim or service line was paid differently than billed, and X12 maintains and revises the list. Source
  • Remittance advice remark codes add explanation to an adjustment described by a reason code, and codes prefaced with Alert convey general remittance information. Source
  • CMS explains that PLB reason codes describe provider-level adjustments and that the check amount equals claim-level payments less provider-level adjustments. Source
  • The CAQH CORE Phase III (CORE 370) reassociation rule uses the X12 EFT trace number in the TRN segment as the only number for matching an EFT to its ERA. Source
  • HHS describes HIPAA de-identification by Safe Harbor removal of 18 identifier types or by Expert Determination. Source
  • The HIPAA minimum necessary standard requires reasonable efforts to limit protected health information to what the intended purpose requires. Source

See if your company qualifies

A short company assessment. No data uploads are needed.

See if your company qualifies