Chapter 38 — Quiz

26 questions: multiple choice and short answer, written in the style of the CPC and CCS exams where the material is exam-relevant. The answer key is in the collapsed block at the bottom. All dashboards, engine outputs, and evaluation figures are this book's constructed teaching figures; no benchmark or product performance is asserted.


1. Clinical documentation integrity is best described as the discipline of:

  • A. Selecting the code that produces the correct payment for a documented service
  • B. Making the record describe the patient accurately and completely, in language the classification systems can read
  • C. Reviewing completed claims against a written standard after payment
  • D. Ensuring that every chart is coded within the productivity standard

2. The single structural difference between a CDI review and an audit is:

  • A. Who performs it
  • B. Whether a payer is involved
  • C. When it happens — CDI reads unfinished work that can still be completed by the person who made it
  • D. Whether the result is reported to management

3. Short answer: give the one-sentence rule for which sections of a progress note may carry forward and which may not.

4. A documented negative is:

  • A. A denial reason code indicating a non-covered service
  • B. A sentence recording that something did not happen, which converts silence into evidence
  • C. A negative laboratory result
  • D. A query answered "no"

5. "No imaging guidance used" in the March 14 procedure note is load-bearing because it:

  • A. Establishes medical necessity for the injection
  • B. Supports the modifier 25 argument
  • C. Excludes 20611 on the physician's own statement, confirming 20610
  • D. Documents patient consent

6. An absence of any statement about imaging guidance differs from a documented negative because:

  • A. There is no difference; both support the lower-valued code
  • B. Silence proves nothing and must be asked about; a documented negative decides the code
  • C. Silence supports the higher-valued code
  • D. Only inpatient records require documented negatives

7. A working DRG is:

  • A. The DRG on the submitted claim
  • B. The DRG the record would group to as it stands mid-stay, used as a review tool and never as a bill
  • C. The DRG the payer assigns on adjudication
  • D. The DRG used to compute the case mix index

8. Concurrency matters most because a clarification obtained during the stay:

  • A. Is worth more money
  • B. Is a progress note rather than an addendum dated when it was made
  • C. Does not have to be documented
  • D. Can be entered by the coder

9. Short answer: name the four things a query must never do, per §38.3's markers, in the shortest form you can.

10. A query option list contains five clinically reasonable options, every one of which adds or specifies a condition. The query is:

  • A. Compliant, because five options is more than enough
  • B. Compliant, because each option is clinically reasonable
  • C. Leading, because the record's most likely answer is not on the list with equal weight
  • D. Compliant only in the inpatient setting

11. A verbal query is:

  • A. Prohibited
  • B. Permitted, and must be documented — who asked, whom, when, what was asked, what was answered
  • C. Permitted, and needs no documentation because it is not a written record
  • D. Permitted only in the outpatient setting

12. A physician answers a query and the coder assigns a code from the query form because the physician never documented the answer in the chart. This is:

  • A. Acceptable if the query is retained
  • B. Acceptable if the query is part of the legal health record
  • C. Not acceptable — a code is never assigned from an answer that exists only on the query form
  • D. Acceptable if a second coder concurs

13. Querying the same encounter a second time after receiving an answer the coder did not want is:

  • A. Compliant if the second query is worded differently
  • B. Compliant if new clinical evidence is cited
  • C. Leading, regardless of wording
  • D. Required by professional guidance

14. Short answer: state the hard limit Chapter 36 §36.11 established about what a query can and cannot do.

15. The two CDI metrics that corrupt a program when they become targets are:

  • A. Review coverage and turnaround time
  • B. Query rate and agreement rate
  • C. Response rate and DRG change rate
  • D. Case mix index and financial impact

16. A CDI program can become a leading program without sending a single non-compliant query, by:

  • A. Wording queries suggestively
  • B. Sending only the queries it expects to be answered its way
  • C. Failing to retain queries
  • D. Using verbal queries

17. Of the following, the one that is not defensible is:

  • A. Prioritizing which records get reviewed by financial materiality
  • B. Reviewing a random sample alongside a targeted worklist
  • C. Deciding whether to send a clinically warranted query based on the likely direction of the answer
  • D. Reporting adds and removals as separate lines

18. In the CAC pipeline, the step this chapter calls the most valuable and least discussed is:

  • A. Ingestion
  • B. Concept extraction
  • C. Confidence scoring
  • D. Presenting the code with the source text that produced it, highlighted in place

19. "Mother with colon cancer at 62" produces a coding error through which qualifier?

  • A. Negation
  • B. Experiencer
  • C. Temporality
  • D. Certainty

20. "Probable aspiration pneumonia" documented at discharge is coded as an established diagnosis:

  • A. In every setting
  • B. In no setting
  • C. On an inpatient claim, but not in the outpatient setting
  • D. Only with a query response

21. Short answer: state the structural reason an engine cannot flag the absence of a sentence nobody wrote — and the one thing software can do that looks like it.

22. The four properties that make a domain autonomously codable are:

  • A. Volume, revenue, staffing, and payer mix
  • B. A structured source document, a small code space, low variability, and fast unambiguous feedback
  • C. Accuracy, speed, cost, and vendor support
  • D. Inpatient status, Medicare volume, template use, and coder experience

23. An engine suggests 1,905 codes; 1,712 are supported by the record. The record supported 1,840 codes in total. Precision and recall are:

  • A. 93.0% and 89.9%
  • B. 89.9% and 93.0%
  • C. 89.9% and 89.9%
  • D. 93.0% and 93.0%

24. Between a missed code and an unsupported code that reaches a claim:

  • A. The missed code is worse, because it always costs more money
  • B. They are equivalent errors with equivalent consequences
  • C. Both are real errors, but only the unsupported code creates False Claims Act exposure that compounds with volume and consistency
  • D. Neither is an error if the engine assigned it

25. The driver of model drift that almost nobody monitors is:

  • A. The annual code set updates
  • B. A new documentation template
  • C. A change in what counts as "supported" — a revised payer policy, coverage article, guideline, or edit — with no change to any note
  • D. Turnover among coders

26. Short answer: name the four things this chapter says the human in the loop is actually for, and add the fifth that belongs to this chapter specifically.


Answer key 1. **B** — accuracy and completeness in codable language, corrected only by the clinician who owns the statement. 2. **C** — timing. CDI reads unfinished work; an audit reads finished work against a written standard (Chapter 37). 3. **The sections that record what happened today do not carry forward** — assessment, plan, examination findings for the problem being treated, any time statement, any decision-to-proceed statement. Stable reference content (medications, allergies, surgical history) may. 4. **B** — a sentence whose only purpose is to record that something did not happen. 5. **C** — it excludes 20611 (with ultrasound guidance and a permanent recording and report) on the physician's own statement, so 20610 stands (Chapter 17 §17.7). 6. **B** — silence is an absence and has to be asked about; only the documented negative decides the code. 7. **B** — a mid-stay review tool on incomplete documentation. Never a bill. 8. **B** — a progress note rather than an addendum, which Chapter 4 §4.5 requires to be dated, attributed, and flagged. 9. **Supply the answer · name the code, classification, or level · state or imply the money · offer no real alternative (and no way to decline).** The fifth marker — signaling the preferred response, including re-querying after a "no" — is the one people forget. 10. **C** — the test is not how many options there are; it is whether the record's most likely answer, including "the assessment as written is what I intend," is on the list with equal weight. 11. **B** — legitimate and the daily reality of concurrent review, with the documentation obligation attached. 12. **C** — the response must land in the medical record itself. Otherwise the query answered your question, not the record's. 13. **C** — leading regardless of wording. The record of three queries and one changed answer tells its own story. 14. **A query cannot manufacture an encounter.** If a condition was not addressed at the encounter, no answer to any question makes it reportable for that encounter. 15. **B** — query rate and agreement rate. 16. **B** — the selection is the leading act, and no review of query text can detect it. 17. **C** — that is selection, not triage. (A) is defensible; attention is finite. 18. **D** — the source-text audit trail, which no manual process has ever had. 19. **B** — experiencer. A family history is not this patient's diagnosis. 20. **C** — the inpatient uncertain-diagnosis rule (Chapter 9 §9.5); the outpatient setting is the opposite. Same four words, two opposite answers. 21. **Natural language processing operates on text; where there is no text there is nothing to process.** What software *can* do is flag a **clinical indicator pattern with no corresponding diagnosis** — a present pattern the organization has told it to associate with a usually accompanying sentence. That is how CDI worklists are built, and it carries §38.4's problem into the rule set if every rule fires in one direction. 22. **B.** 23. **B** — precision 1,712 ÷ 1,905 = 89.9%; recall 1,712 ÷ 1,840 = 93.0%. Precision is measured against what the engine *suggested*; recall against what the record *supported*. 24. **C** — both are real errors and downcoding is not the conservative option (Chapter 5 §5.8), but only one of them is a statement to a payer that the record does not support. 25. **C** — the notes read the same, the engine reads them the same, and the answer is now different. Chapter 28's Case Study 1 is the same mechanism: correct when written, decayed, with no internal event to signal it. 26. **The judgment that is an argument rather than a pattern · the rules that are not in the note · the absence · the accountability**, because a certification requires a certifier and a confidence score is not a path. **The fifth: the query** — only a person may ask a clinician a question about a patient.