CDEO common mistakes
CDEO · Understand the traps and how to avoid them.
Scoring an office/outpatient E/M visit by tallying history and exam bullet points under the 1995/1997 guidelines when the practice has adopted the current AMA MDM-or-time framework (or the reverse).
Candidates memorize one rule set during study and default to it under time pressure, especially when the note contains a rich HPI and exam that 'looks' like a high-level visit.
What to do instead
Before scoring any E/M, state out loud which rule set and which year's code set applies to the date of service, then apply only that framework; practice with a two-column worksheet that forces you to record problems/data/risk or total time explicitly and discard the history/exam columns when MDM-or-time governs.
Counting copy-pasted or carried-forward text (an unchanged assessment paragraph, a verbatim ROS from a prior visit) as documented evidence for the current encounter.
The text is present in the note and reads as complete, and reviewers who read for content rather than for provenance do not notice it is identical to the prior note.
What to do instead
When reviewing, pull the prior note side by side for any patient with more than one visit in the record and mark text that is unchanged as non-contributory; if the status of a condition matters to code selection, treat it as undocumented and issue a query rather than crediting the cloned language.
Assigning a time-based E/M level without confirming the note states total time on the date of the encounter and that the stated minutes match what the code descriptor includes (for example counting only face-to-face minutes when the code counts total time).
Candidates see a number in the note ('45 minutes') and assume it is the billable time, without checking whether it refers to counseling, the procedure, or total encounter time.
What to do instead
For every time-based level, verify three things in writing: the number, the phrase tying it to the date of service, and that it represents the time component the code actually measures; if any of the three is missing, score on MDM instead and flag the documentation gap.
Appending modifier 25 to an E/M service on the same day as a procedure based only on the presence of a separately reported procedure code, without identifying a distinct problem with work beyond the pre-, intra-, and post-procedure care.
The modifier resolves the edit and the claim pays, so it feels correct; candidates conflate 'a problem was mentioned' with 'a significant, separately identifiable service was documented.'
What to do instead
Require the note to contain a distinct chief complaint or problem, its own assessment, and management that is not the decision to do the procedure; write a one-sentence justification in the review notes for each modifier 25 you support, and deny the modifier when that sentence cannot be written from the documentation.
Submitting claim lines whose ICD-10-CM codes do not support the service rendered, or defaulting to an unspecified code when the note contains the detail (laterality, encounter type, seventh character, stage) needed for a specific code.
Unspecified codes are always payable and never rejected at the clearinghouse, so there is no feedback signal; and candidates stop at the first code that 'sounds right' rather than reading the full tabular list.
What to do instead
For each CPT line, verify at least one diagnosis code that the provider explicitly documented as the reason for the service; then re-read each chosen code's instructional notes and the 'unspecified' sibling to confirm the note genuinely lacks the specificity before using it, and flag over-use of unspecified codes as a provider education item.
Applying the inpatient 'principal diagnosis' selection logic to outpatient records, or vice versa, when determining the first-listed code.
The two concepts are taught together and both are described as 'the main reason for the encounter,' so the distinction blurs under exam pressure.
What to do instead
Anchor the rule to the setting: for outpatient, ask 'what is the condition, symptom, or reason chiefly responsible for today's visit as documented in the provider's own words,' and check the official guideline section for the setting before finalizing the first-listed code.
Capturing a hierarchical condition category from the problem list, medication list, or a prior year's diagnosis without finding MEAT (monitor, evaluate, assess, treat) language for that condition in the current encounter.
The condition is genuinely present and clinically obvious, and the chart 'feels' complete because the history section lists it; candidates credit clinical truth instead of current documentation.
What to do instead
Run each candidate HCC through a MEAT check line by line, recording the exact sentence that supports it; if no sentence exists, do not report the HCC and instead draft a compliant query to obtain the status and management for this encounter.
Flagging a quality measure care gap or a measure numerator hit without first checking the measure's denominator exclusions, exceptions, and required documentation timeframes.
Candidates read the numerator criteria and stop, treating the measure as a simple 'was the test done' checklist, when exclusions frequently remove the patient from the denominator entirely.
What to do instead
For every measure, walk denominator, then exclusions/exceptions, then numerator; document each step on the abstraction sheet and cite the specific exclusion criterion (and the documentation that supports it) whenever you exclude a patient.
Writing a provider query that names a diagnosis in the question, embeds two unrelated questions in one query, or offers only a single option.
The reviewer already knows what the answer should be and wants a fast, clean response; leading phrasing feels efficient because it produces the desired code.
What to do instead
Use a standard template: state the clinical indicators found in the record, ask an open-ended question about the condition's status or cause, and list clinically plausible, non-leading options plus a request for the provider's own words; split unrelated issues into separate queries so each answer is auditable.
Delivering audit feedback as a bare list of coder corrections, with no linkage to the documented evidence, no expected behavior change, and no scheduled re-audit or acknowledgment.
The reviewer's goal is accuracy of the claim, not behavior change, and a correction list takes far less time than a targeted education session; the result is repeat errors on the next audit cycle.
What to do instead
Convert each finding into an education item that quotes the note, names the rule, states the corrected action, and quantifies the impact; deliver it as a short focused session or one-pager tied to the specialty, obtain documented acknowledgment, and re-audit a sample of that provider's charts within a defined interval to confirm the fix.