Written by the MediCodio AI editorial team. Published 30 September 2026. Last reviewed 30 September 2026.
You have three options on a thin note: code to the level of certainty the record supports, query the provider, or guess. The guidelines endorse the first two. Which one you pick is a routing decision, and it should be made before the chart reaches a coder.
| Option on a thin chart | What the ICD-10-CM guidelines say | What it costs you |
|---|---|---|
| Code to the level of certainty the record supports | Unspecified codes have "acceptable, even necessary, uses" (Section I.B.18) | Specificity you may have been entitled to |
| Query the provider | Directed where the record conflicts or the relationship is unclear (Sections I.A.19, I.B.16) | Turnaround, plus a conversation with the client |
| Assign a specific code the record does not support | "Inappropriate" (Section I.B.18) | Audit exposure on every chart coded that way |
Quotations come from the CMS and NCHS ICD-10-CM Official Guidelines for Coding and Reporting, FY 2027 .
What "incomplete documentation" means in coding terms
A chart is incomplete for coding when the record does not carry the statement a code needs. The diagnosis is implied by the labs but never written. The operative note names the procedure without the approach. Two clinicians in the same chart say different things.
None of that means the care was poor. It means the code you would assign has no anchor, and the guidelines are direct about what follows: "The importance of consistent, complete documentation in the medical record cannot be overemphasized. Without such documentation accurate coding cannot be achieved."
Why a billing company has fewer options than a hospital
A hospital coding manager who finds a pattern of thin notes can walk to the service line and raise it. The physicians are colleagues. The documentation template belongs to the same organization.
You hold the service level agreement and no control over the pen. The note was written by a physician your client employs, in a system your client licensed, under habits that predate your contract. The chart still has to go out this week.
The cost lands on you twice. Coding at the level the record supports puts a weaker claim on the wire and invites a question later about why the coding looks cautious. Querying protects the code and spends the turnaround you sold. Between those, a coder under queue pressure at 4pm makes a judgment call that surfaces again only in an audit.
Remove that judgment call by making the decision earlier, once, and the same way for every chart that looks like it.
What the guidelines actually allow
Code assignment follows the provider’s statement, not the clinical picture. Section I.A.19 is explicit: "The assignment of a diagnosis code is based on the provider’s diagnostic statement that the condition exists ... Code assignment is not based on clinical criteria used by the provider to establish the diagnosis. If there is conflicting medical record documentation, query the provider." A sepsis-shaped set of vitals is not a documented sepsis.
Unspecified is a legitimate answer, and often the correct one. Section I.B.18 states that sign, symptom and unspecified codes "have acceptable, even necessary, uses," and that "each healthcare encounter should be coded to the level of certainty known for that encounter." The same section rules out the alternative: selecting a specific code "that is not supported by the medical record documentation" is inappropriate.
Some details can come from someone other than the treating provider. Section I.B.14 lists the exceptions, including body mass index, depth of non-pressure chronic ulcers, pressure ulcer stage, coma scale, NIH stroke scale, social determinants of health, laterality, blood alcohol level, underimmunization status and firearm injury intent. The associated diagnosis still has to come from the provider, and conflicting entries send you back to a query.
A borderline diagnosis is coded as confirmed. Section I.B.17 treats a documented borderline condition as confirmed unless the classification carries a specific entry, and encourages a query where the documentation is unclear. Coders who downgrade it on instinct are creating variance.
Unspecified coding, undercoding and a query are three different things
Teams use the three terms interchangeably in the same meeting. Each one has a different fix.
Unspecified coding means the record supports a code but not a more detailed one within the same category. Bacterial pneumonia, organism unstated. The guidelines permit it, and where the record genuinely carries no more, it is the accurate answer.
Undercoding means something documented did not reach the claim at all, or reached it at a lower level than the record supports. A procedure described in the note that no coder picked up. A visit level reported below what the documentation carries. The record was sufficient and the coding was not.
A query is what you send when the record contradicts itself or leaves a relationship undefined, which Section I.B.16 directs for complications of care. It is a request for documentation, not a coding decision.
Reading unspecified coding as undercoding sends teams chasing specificity the chart cannot support. Reading undercoding as unspecified coding hides a capture problem inside a documentation problem, and a coding audit is how you tell which one you have.
How to decide before the chart reaches a coder
Write the rule per client, not per coder. Decide in advance which conditions are worth a query for that client and which get coded to certainty. A rule written once is auditable. A rule each coder carries in their head is not.
Agree the query route before go-live. Name who receives a query at each practice, the turnaround you expect, and what happens when it goes unanswered. Doing this at contract time costs one conversation; doing it chart by chart costs the same conversation every week.
Separate the two failure reports. Track charts held for documentation apart from charts coded to certainty. One is a client conversation about the note. The other is a coding conversation about the code. Reporting them as a single queue hides both.
Give the client the pattern rather than the anecdote. Ten charts from one surgeon missing the same element point to a template that needs fixing. One chart reads as a complaint. Clinical documentation work belongs upstream, which is what CDI support is for.
How MediCodio AI routes a thin chart
Per-chart routing is the answer to this pain. Complete charts flow through AutoPilot for fully autonomous coding. Charts where the documentation is thin or contradictory stay in CoPilot, where certified coders review and finalize the output. A thin chart does not get coded on bad input by a queue running late.
Where the documentation does not support what is codeable, CODIO AI flags the gap for follow-up instead of working around it, and Provider Query Dispatch sends the documentation question to the provider and tracks the response against the chart it belongs to. That is the part of the loop most billing companies run on email and a spreadsheet. The chart holds, the question goes out, and the answer comes back attached to the case rather than to somebody’s inbox.
Every code on a submitted claim carries its support with it. Open any assigned code in the Prediction Trail and you see two things: the passage of clinical documentation the code was assigned from, and the coding guideline or payer rule that justified it. That is a 100% code-level audit trail, and it is what turns "why was this coded to certainty?" into a lookup rather than a reconstruction six months later.
MediCodio AI reports 98%+ coding accuracy. The basis is first-pass accuracy across deployments since 2023, measured at the code level on production charts before any correction, against a reference review by AAPC- and AHIMA-credentialed coders applying official guidelines and payer policy to the same documentation. Turnaround runs at <24 hr, measured at enterprise volume and inclusive of certified-coder review, so the charts that need a coder are not queued behind the ones that do not. Nothing runs autonomous on day one. Everything starts in review, and the autonomous lane widens as accuracy is proven on your charts, audited by your team.
See RCM medical coding for the multi-client version of this, and computer-assisted coding software for how routing differs from a list of suggestions handed to a coder who still has to decide.
What this does not fix
The note still has to say it. No coding platform invents a diagnostic statement that a physician did not write, and a tool that appeared to do so would be a compliance problem rather than a feature.
Queries still need an answer. Dispatch and tracking are the platform’s half of the loop. A held chart becomes a coded chart when a provider replies, and that timeline belongs to your client.
And the guidelines still govern the choice. Section I.B.18 asks for the level of certainty known for the encounter, no more and no less, whoever or whatever assigns the code. See security and compliance for how the coding record behind each decision is retained.
See what this looks like on your own charts
Book a Demo and we will run an accuracy check against a sample of your own charts, so you can see which of your clients’ notes route straight through and which ones hold for a coder.
Related: RCM medical coding · Coding audit services · RCM and billing companies .
Sources
- CMS and National Center for Health Statistics, ICD-10-CM Official Guidelines for Coding and Reporting, FY 2027, updated 1 October 2026. Sections I.A.19, I.B.14, I.B.16, I.B.17, I.B.18. cms.gov