CCS (AHIMA) common mistakes
CCS · Understand the traps and how to avoid them.
Selects the principal diagnosis from the top line of the EHR problem list, or the condition that consumed the most resources, and builds the DRG from that choice.
The problem list looks like a ranked list and time pressure rewards grabbing the first plausible diagnosis, but the principal diagnosis is the condition established after study to be chiefly responsible for occasioning the inpatient admission. Admitting, working, and discharge diagnoses frequently differ.
What to do instead
Before assigning anything, state in one sentence why the patient was admitted, then confirm the discharge summary and pertinent findings support that condition as chiefly responsible. When the admitting and discharge diagnoses differ, sequence by the definition, not by problem-list order or resource use.
Defaults to the unspecified code for a condition and codes each documented condition on its own line, skipping Excludes1 notes, combination codes, and 'code also' or 'use additional code' instructions at the code, category, and chapter level.
Encoder prompts and the alphabetic index make the unspecified option the path of least resistance, and instructional notes sit above the code title where they are easy to scroll past during a fast read.
What to do instead
Read every note at the code, category, subcategory, and chapter level before finalizing. When a combination code exists, that single code replaces the multiple-code construction; an Excludes1 note means the excluded code cannot be reported with the code under which the note appears; 'code also' conditions must be reported when present. Confirm documentation actually supports the more specific option before selecting unspecified.
Captures diagnoses from nursing assessments, the medication administration record, the problem list, prior encounters, or copy-forwarded text as though they were provider-documented for the current encounter.
These sources are convenient and often clinically accurate, and cloned or carried-forward text makes a condition appear current when it was never reassessed during this encounter. Coders may not know the facility's policy on which document types support code assignment.
What to do instead
Assign codes only from documentation by a provider for the current encounter, per facility policy. When a condition is clinically important but appears only in non-provider sources, do not code from them; generate a provider query instead. Treat copy-forwarded text with no current assessment as unsupported.
Codes from the history and physical and the discharge summary only, skipping the operative report, pathology, imaging, addenda, consults, and late entries, then closes the case as complete.
Record volume and productivity targets push coders toward the familiar summary documents. A distinct procedure, a specimen result that changes the diagnosis, or a revised diagnosis in an addendum often lives only in those other documents.
What to do instead
Work from a fixed document checklist (H&P, consults, operative/procedure reports, pathology, imaging, discharge summary, all addenda) and reconcile that checklist against the encounter's actual document list before finalizing the codes.
Assigns an unspecified code, codes a 'possible' or 'probable' condition as confirmed, or silently drops the condition when the documentation is ambiguous, conflicting, or shows clinical indicators with no linked diagnosis, instead of generating a query.
Querying feels like it stalls the case and can feel like admitting uncertainty; the low-specificity code is faster and less confrontational. Many coders also are not sure which way the uncertain-diagnosis rule cuts.
What to do instead
Treat ambiguity, internal conflict, and unexplained clinical indicators (elevated labs, antibiotics, imaging findings) as mandatory query triggers. On inpatient records code 'probable/suspected/questionable' as if established; on outpatient records do not code it. Query in either setting rather than guessing or omitting.
Writes a query that names or telegraphs the answer ('Would you agree the patient has sepsis? Please document sepsis so we can code it') or a multiple-choice query where only one option is clinically plausible, and keeps the exchange in a personal file or email rather than the health record.
Templated queries circulate as productivity shortcuts and get copied without review; the underlying goal becomes a faster, higher-paying code rather than an accurate one. Query retention is often not covered in onboarding, so coders store responses wherever is convenient.
What to do instead
Write an open, non-leading query that lists the clinical indicators and asks the provider to document a diagnosis in their own words, offering choices only when clinically reasonable and including 'unable to determine' where appropriate. Never introduce a diagnosis the provider has not documented. File the query and the authenticated response in the health record per policy.
Appends modifier 25 to every E/M reported with a procedure, or modifier 59 to push a code pair past an NCCI procedure-to-procedure edit, without documentation of a separately identifiable service or a distinct site, session, or encounter.
The modifier clears the edit and the claim pays, so it becomes the routine fix for denials. The documentation threshold behind each modifier is rarely re-read once the habit forms.
What to do instead
Support 25 with documentation of a significant, separately identifiable E/M service beyond the usual work of the procedure. Support 59 or an X{EPSU} modifier with a distinct anatomic site, separate session, or separate encounter that is documented. If the documentation doesn't meet the threshold, remove the code rather than override the edit.
Reports every condition named anywhere in the record as a secondary diagnosis to lift severity, and assigns present-on-admission indicators by assumption, usually 'Y', rather than from the admission documentation.
More secondary diagnoses mean higher severity and reimbursement, so there is financial pull toward inclusion. POA flags are widely treated as clerical data entry, and the admission H&P is often not reviewed for when a condition first appeared.
What to do instead
Report a secondary diagnosis only when it was clinically evaluated, therapeutically treated, diagnostically worked up, or extended the stay or increased nursing care and monitoring. Assign POA from documentation of the condition at or after inpatient admission, distinguishing present-on-admission from hospital-acquired conditions, and query when timing is unclear.
Accepts the computer-assisted coding suggestion or the encoder's top-ranked code and the autogrouped DRG without independently confirming them against the alphabetic index, tabular notes, and the documentation.
Confidence scores look authoritative and natural language processing surfaces diagnoses from text that is not valid support. Productivity targets reward accepting suggestions quickly, and the DRG output looks like a finished answer.
What to do instead
Treat CAC output as a candidate list only. Verify each code against the index and the tabular instructional notes and confirm the documentation support before accepting, then review the final DRG for clinical coherence with the record rather than accepting the first grouped result.
Releases claims with automated edits still unresolved, overrides NCCI PTP or medically unlikely edits without a documented rationale, or codes in an encoder still carrying the prior fiscal year's code set or grouper.
Edit queues are large and volume pressure makes the one-click override attractive. Code set and grouper versions are updated in the background, so coders assume the system is current and never check.
What to do instead
Resolve or document the clinical rationale for every edit override before release, and escalate edits you cannot support rather than overriding them. Verify the loaded code set and encoder version against the current fiscal year, and confirm the grouper version used for the DRG, before coding and grouping.