CDI and Charge Capture: The Revenue That Never Became a Claim

Spend enough time looking at where revenue goes missing, and a hierarchy emerges, ordered by how hard each failure is to notice.

A denial is loud. It arrives with a reason code, lands in a queue, and somebody owns it. An underpayment is quieter: the claim pays, posts and closes, and you only find it if you independently calculated what the payment should have been.

A missed charge is in a different category entirely. The service happened, and nothing was ever created. No claim, no remittance, no variance, no denial. There is no artifact anywhere in your systems, which means no report you own can find it, because every report you own is built on claims.

That is the failure this article is about, along with its close relative: the charge that did get created but understated what actually happened. If you want the wider measurement argument first, how to improve revenue cycle management covers it, and the 13 steps of the revenue cycle is the plainer orientation.

The Three Failures, Ordered by Visibility

Worth setting out precisely, because the operational response to each is different.

1. Denial. A claim was submitted and rejected. It generates a document, a reason code, and a work queue. Whatever else is wrong with denial management, the failure announces itself.
2. Underpayment. A claim was submitted and paid, at less than the contract entitled you to. Nothing announces it. It is detectable, but only if you have modeled the expected allowed amount and compared, which is the argument in payer contract management and the detection layer described in revenue cycle management software build or buy.
3. Missed charge. No claim was ever created. This one defeats the tooling built for the other two, including variance detection, because variance detection compares an expected amount against a paid amount and here there is neither.

Each step down that list moves further upstream and gets harder to see. And the further upstream you go, the more the failure looks like nothing at all rather than like a problem.

Charge Capture and CDI Solve Different Halves

These two get discussed together and treated as interchangeable, and they are not. The distinction determines which problem a program actually fixes.

Charge capture asks whether the service was recorded and coded at all. Its failure mode is absence: a procedure happened, a supply was used, a service was delivered, and nothing entered the system.

Clinical documentation integrity asks whether the documentation supports the code and reflects the patient’s real complexity and severity. Its failure mode is understatement: something was recorded, but the record does not carry what actually happened, so the reimbursement reflects a simpler patient than the one who was treated.

Both are upstream of billing. Both are documentation disciplines. They fail in opposite directions and they need different interventions, and an organization can be excellent at one while losing steadily to the other.

The practical tell: if your charges reconcile against your schedule but your case mix looks lighter than your clinicians describe, that is a CDI problem. If your documentation is rich but your charge volume does not match your activity, that is a charge capture problem.

Where Charges Actually Go Missing

Not through carelessness. Through structural gaps where nothing owns the handoff.

1. Lag. The longer between service and charge entry, the more gets lost, because reconstruction depends on memory and notes rather than on the event. Common practice targets capturing charges within three to five days, and treats a late-charge rate above roughly 2% of total as a signal rather than noise.
2. Chargemaster mismatches. A service is performed that has no corresponding chargemaster entry, or has one that is outdated. Nothing can be captured for something the system has no code for.
3. Manual entry that depends on someone remembering. Any workflow where the charge is created by a person recalling to create it will lose a predictable percentage, and the percentage rises with workload.
4. Supplies and implants that bypass the workflow. Bill-only and consignment items frequently move through a different process from the procedure they belong to, which is why the orthopedic revenue cycle analysis found device lines routinely omitted while the surgical line went out correctly.
5. Services delivered outside the main documentation flow. Anything performed by someone who is not the person completing the note.

Notice what these have in common: none of them produces an error. They produce an absence, and absences do not raise exceptions.

Reconciliation Is the Only Way to See an Absence

Since a missed charge produces no artifact, the only way to find it is to compare what you billed against an independent record of what you did.

That is what charge reconciliation means: taking captured charges and checking them against other sources that describe the same activity. The schedule. The clinical documentation. The operating room log. The implant log. The supply system. Anywhere the event left a trace that is not a charge.

The reconciliation source has to be independent of the charge process, which is the part organizations most often get wrong. Reconciling charges against the claim file tells you nothing, because both derive from the same upstream step. If the charge was never created, both records agree perfectly, and both are wrong.

Two design choices matter more than the rest.

Retrospective or real-time. Retrospective reconciliation finds what you lost. Real-time reconciliation, run before claims go out, prevents it. The second is harder and worth substantially more, because a charge caught before submission is simply billed, while one caught afterwards is a late charge with its own consequences.

Threshold or exhaustive. Reconciling everything is expensive. Reconciling above a dollar threshold systematically misses the small, high-volume items that add up, and the same trap appears in the timely filing analysis where value-sorted worklists starve small claims of attention.

What CDI Is Actually For, and What It Is Not

CDI is frequently described as a revenue program, which is true but incomplete and leads people to the wrong interventions.

What CDI does is ensure the record reflects the patient who was actually treated. Complexity, severity, comorbidities, and the clinical reasoning behind decisions. When it works, reimbursement matches acuity, quality measures reflect the real case mix, and audits find documentation that supports what was billed.

Two things follow that are worth being clear about.

It is not upcoding, and the distinction is not a technicality. CDI aims at accuracy in both directions. A record that overstates complexity is a compliance exposure, and a good program surfaces those too. Any CDI effort measured solely on reimbursement lift will drift, because it has been given a one-directional target for a two-directional problem.

It cannot be automated the way charge capture can. Whether documentation supports a code is a judgment about clinical narrative. Software can prompt, template, flag missing elements, and route queries to clinicians, and all of that helps. It cannot decide whether the note establishes what it needs to establish. The dermatology analysis in this cluster makes the same point about coverage determinations resting on documentation language.

So the honest split: charge capture is largely a systems and reconciliation problem, and CDI is largely a clinical workflow and query problem with systems support around it.

What a CDI Query Is, and Why Response Rate Decides Everything

I have referenced queries twice now without explaining them, and they are the mechanism the whole discipline runs on.

A CDI query is a question put to a clinician about their own documentation. Not a correction and not a suggestion of what to write. A properly constructed query presents what the record currently says, identifies what is ambiguous or missing, and asks the clinician to clarify, without leading them toward a particular answer. That last constraint is a compliance requirement, not a nicety: a query that suggests the answer is a query that manufactured it.

The industry standard here is the Guidelines for Achieving a Compliant Query Practice, produced jointly by AHIMA and the Association of Clinical Documentation Integrity Specialists and last updated in 2022. It is freely available, it carries a long FAQ section on the edge cases teams actually argue about, and it is worth reading before designing any query workflow rather than after.

Everything about a CDI program’s value passes through this one interaction, which makes response rate the metric that matters most and the one most often left unmeasured.

Three things move it, and none is about query quality:

1. When the query arrives. A question about a patient seen this morning gets a considered answer. The same question three weeks later competes with several hundred intervening patients, and the honest response is often that the clinician cannot now recall.
2. Where it arrives. A query in a separate CDI inbox is a task in a system a clinician opens because they were reminded to. A query surfaced where they already document is part of the work.
3. Whether answering changes anything visible to them. Clinicians who never learn what happened as a result of their answers deprioritize them, entirely reasonably.

The uncomfortable implication for anyone buying CDI software: the product features that get demonstrated are usually detection and analytics, and the variable that determines the outcome is a workflow property of where queries land. Those are not the same purchase.

What a Late Charge Actually Costs

The article has asserted that catching a charge before submission beats catching it afterwards. Worth being specific about why, because the gap is larger than “a bit of rework.”

A charge caught before the claim goes out is simply billed. It costs nothing beyond the reconciliation that found it.

A charge caught after submission is a late charge, and it carries a series of consequences that compound:

It requires a corrected claim or a separate one, which is rework at a per-claim cost.

It restarts adjudication, which extends the time to payment for revenue you had already earned.

1. It can land outside a timely filing window. A charge discovered ninety days after service against a payer allowing ninety days is not a late charge; it is a permanent write-off, for the reasons set out in days in AR and timely filing.
2. It distorts your reporting in both directions. Revenue is attributed to the period it was posted rather than the period the service occurred, which makes period comparisons unreliable exactly when you are trying to measure whether capture is improving.
3. And it attracts attention. A high late-charge rate is a recognized indicator of charge integrity problems, which is why HFMA’s benchmark holds late charges under 2% of total rather than treating them as neutral.

So the sequence matters more than the catch rate. A program that finds 95% of missed charges after submission is worth substantially less than one finding 85% before it, and organizations frequently chase the number rather than the timing.

What Your Systems Have to Do

Five things, and they divide along the split above.

For charge capture, which is the automatable half:

Reconcile against at least one independent source, ideally more than one, and never against the claim file alone.

Run it before submission, not after. The value gap between prevention and discovery is large, and it compounds.

Cover the paths that bypass the main workflow, particularly supplies, implants, and services delivered by someone other than the documenting clinician. These are the highest-yield gaps because nothing else looks at them.

For CDI, which is the assisted half:

Prompt for specificity at the point of documentation, not through a retrospective query weeks later when the clinician has seen four hundred patients since.

Route queries where clinicians already work. A CDI query in a separate inbox competes with everything else in a clinician’s day and loses.

The integration reality is worth stating: the reconciliation sources you need are scattered across systems that were never designed to be compared, and identifiers rarely line up cleanly between a schedule, an operating room log, a supply system, and a charge file. Reconciling identity across those sources is the actual project, and the comparison logic on top of it is straightforward once that is solved.

Anyone quoting this work without asking which systems hold your independent records has not scoped it. Where the rules live downstream is covered in claims processing configuration.

Six Checks Before Your Next Close

In the order I would run them.

1. Ask what your charges are reconciled against today. If the answer is the claim file, or nothing, that is the finding and everything else is secondary.
2. Pick one independent source and compare a single week. The schedule is usually easiest. Count services delivered against charges created.
3. Check the supply, implant, and bill-only path separately. It almost always travels differently, and it is rarely included in whatever reconciliation exists.
4. Measure your charge lag from date of service to charge entry, and compare it against the three-to-five-day benchmark and the 2% late-charge guideline.
5. Ask whether CDI queries reach clinicians in their workflow or in a separate queue. Query response rates follow this more than they follow query quality.
6. Check whether your CDI program is measured on reimbursement lift alone. If so, it has a one-directional target for a two-directional problem, and it will drift.

For where this sits in the wider cycle, medical billing versus revenue cycle management sets out the scope; upstream eligibility problems are covered in insurance eligibility verification, and if the answer points toward handing the operation over rather than building, outsourcing revenue cycle management covers where that line falls.

Worth noting that an outsourced biller can only bill what you capture, so this particular failure stays yours regardless of who submits the claims. A revenue cycle analytics view is where reconciliation output would surface, and denial management handles the loud failure at the other end of the hierarchy.