22.3 Conducting and Closing a Second-Level Review
Key Takeaways
Document the referral question, source locations, prior query activity, responsible owner and needed action; a workqueue label is not a diagnosis or disposition.
Separate clinical/documentation, reporting and downstream calculation questions before reconciling a working and final DRG.
A physician advisor can provide expertise and collegial assistance; an advisor without direct clinical care in the encounter cannot establish its diagnosis by answering a query.
Address defective drafts, delivered unanswered queries and answered queries according to their actual state. A private invalid flag does not amend the health record.
Define cohort, unit and denominator for review metrics. Targeted-sample defect rates and observed claim outcomes do not establish national benchmarks or proven denial avoidance.
Conducting and Closing a Second-Level Review
A second-level review should produce a supported disposition of a defined question. A workqueue label, senior job title or agreement between departments cannot substitute for the clinical record or applicable reporting rules. Design the process so the reviewer can explain what was examined, what remains uncertain, who must act and how the result was verified.
The sequence below is an example for local workflow design. It can be adapted for concurrent, pre-bill or retrospective work. An automated queue, particular conference schedule or four-pathway structure is not a universal requirement. AHIMA Clinical Validation, 2023 Update; 2026 query guidance, Section XIII
Start with a useful referral
Record the encounter, specific concern, relevant locations and timeline, prior query activity and the action requested. State the uncertainty neutrally. “Please evaluate conflicting documentation about this condition” is more useful than “save the MCC.” Include an appropriate owner and due date under local policy, with access limited to authorized participants.
Distinguish the referral from a provider-facing query. A referral asks another reviewer to evaluate the issue; a query asks the appropriate provider to clarify documentation. The review can determine that no provider query is needed because the record or applicable coding instruction already resolves the question.
Review the record and rules independently
Read the relevant course rather than only the referring specialist's summary. Compare the provider's assessments with dated observations, treatment, monitoring, consultations and subsequent documentation. Include evidence that challenges as well as supports the concern. Determine whether information comes from this encounter, a prior encounter or a copied entry.
Separate three questions:
- Clinical/documentation question: What diagnosis or course did the responsible provider document, and does an unresolved conflict or support concern need clarification?
- Reporting question: What do the applicable coding conventions and guidelines require, including setting and additional-diagnosis criteria?
- Downstream question: Given the final supported reporting, what does the relevant grouper or quality specification calculate?
A medication, procedure or laboratory value contributes context; it is not a universal diagnostic test. Likewise, failure to find a particular treatment does not automatically make a diagnosis invalid. Inpatient additional-diagnosis reporting can be supported by clinical evaluation, treatment, diagnostic procedures, extended stay or increased nursing care/monitoring under the applicable guideline. FY2027 ICD-10-CM guidelines, Sections I.A.19 and III
Assign the right decisions to the right people
| Participant | Appropriate contribution | Limit |
|---|---|---|
| Second-level CDI reviewer | Evaluate the concern, query compliance and documentation needs; record the rationale | Does not independently establish or erase a provider's diagnosis |
| Coding professional or coding leader | Apply classification conventions, guidelines and reporting requirements | Does not replace the provider's diagnosis with a diagnosis inferred from clinical thresholds |
| Physician advisor | Provide expertise, education, process review and appropriate collegial communication | An administrative title alone does not authorize a diagnostic query response |
| Provider involved in care | Clarify the clinical assessment and documentation within the provider's responsibilities | Must exercise independent judgment rather than supply a requested payment result |
| Quality, compliance, HIM or appeals staff | Address their respective measure, compliance, amendment or appeal responsibilities | Their involvement does not turn an internal disposition into a binding external ruling |
The 2026 query brief directs diagnostic queries to providers who provided direct clinical assessment and care in the encounter. An advisor who did not participate in that care cannot establish the encounter diagnosis by answering a query after administrative chart review. When documentation conflicts, the attending provider as defined by policy should clarify the discrepancy. Advisor coaching and peer discussion can help that process without replacing the responsible provider. 2026 query guidance, Section VII
Worked conflict: The attending documents encephalopathy while a consultant explicitly rules it out. An unsent query asks the attending to confirm encephalopathy “as an MCC.” Replace that draft with an appropriate neutral clarification that accurately presents the conflict and relevant evidence. Use the attending route specified by policy; seek advisor assistance if useful. Preserve both original notes. The goal is an accurate clinical statement, not a forced consensus that protects reimbursement.
Select and document a supported disposition
Possible outcomes include:
- No further clarification needed: Identify the existing documentation and rule that resolve the referral.
- Clarification needed: Send a compliant query to an appropriate recipient and track the response under policy.
- Reporting correction needed: Route the supported correction to the authorized coding or reporting owner and verify it.
- Query defect found: Address the draft, delivered query and any existing response through the appropriate process.
- Concern remains unresolved: Record the issue and escalate to the designated clinical, coding, compliance or medical-staff authority according to its nature.
These are useful disposition categories, not a nationally mandated list. “Reviewed” is insufficient if it leaves the actual question unanswered. Record the reason for the decision and the next owner, including when the correct result is no change.
Handle a defective query without rewriting the record
An inappropriate unsent draft can be corrected before delivery. A delivered but unanswered query can require withdrawal or other action under policy. An answered query is different: do not simply mark its clinical response “invalid” in a private workqueue and treat that as an amendment to the health record. Evaluate the defect and clinical documentation separately; involve the responsible provider and HIM's amendment process when appropriate. Preserve the relevant history and required records. Section 24.1 develops remediation in more detail. 2026 query guidance, Sections XIII and V
If a response does not support the reviewer's preferred diagnosis, a second reviewer should not repeat the same question merely to override that judgment. A subsequent query may be appropriate for new or evolving information or genuinely needed specificity. Apply the brief's intent-and-pattern standard and the organization's closure process. 2026 query guidance, Section IX
Reconcile a DRG difference by explaining its cause
A working DRG may differ from the final result because information, diagnoses, procedures or grouping inputs changed. It may also reveal an error. Compare the actual inputs and applicable rules before assuming a conflict between departments. A scheduled conference is one possible method; direct documented discussion may suffice under local policy.
For example, the working abstract includes a secondary diagnosis that the final abstract omits. Determine whether the issue is documentation, a reporting criterion, a code assignment or the grouper configuration. Seek provider clarification for a clinical documentation problem and apply the appropriate coding process for a coding error. A physician advisor or finance leader cannot resolve the matter by selecting the higher-paying DRG. Recalculate downstream results only after the supported reporting decision.
Verify closure and use the findings
Confirm that the provider response is authenticated and incorporated through the approved record process where applicable, the coding/reporting owner completed any required action and the referral disposition matches the result. Address the actual remaining concern before release or further processing under the organization's authorized workflow. An internal review cannot guarantee that a payer will accept the claim or that a later appeal will succeed.
Use metrics that describe the work actually performed. Define the eligible population, unit, numerator, denominator and time window. Completed referrals, time to resolution, documented query defects, unresolved issues and recurrence after intervention can each answer different questions. A high agreement rate does not establish accurate review, and a high exclusion rate is not the purpose of reviewing quality flags.
Worked feedback: In a targeted review of 30 malnutrition queries, reviewers identify 6 with a defined construction defect. The defect proportion in that reviewed group is 6/30 = 20%. It is not automatically the rate for all queries or proof that novice staff caused the problem. Examine the cases, template design, indicator sourcing and review conditions. Provide specific education or tool correction, assign an owner and recheck appropriate later work. Track reporting increases, decreases and no-change outcomes as accuracy results; do not reward a preferred financial direction. AHIMA clinical-validation education guidance; 2026 query governance and auditing guidance
Finally, distinguish observed outcomes from estimates. A claim not yet denied does not prove that review prevented a denial. Report the cohort, follow-up period and limitations of any financial or denial analysis. The review program improves through verified corrections and reduced recurrence, not an invented national agreement, retraction or denial-avoidance target.
The attending documents encephalopathy while a consultant explicitly rules it out. A physician advisor who did not provide direct care is asked to resolve the case. What is the appropriate role and route?
The advisor may answer a diagnostic query solely from administrative chart review because the advisor has a senior medical title.
The advisor may help evaluate and communicate the concern; use the appropriate attending clarification route under policy to resolve the conflicting documentation.
The advisor must select the consultant’s statement as the final diagnosis without involving a provider responsible for the encounter.
The diagnosis can be selected by a departmental vote as long as the majority agree on the final DRG.
A second-level reviewer finds that an already answered query used leading wording. What is the best next action?
Erase the query and clinical response so no record of the defective communication remains.
Mark the response invalid in the private CDI system and assume that this automatically changes the permanent clinical record.
Repeat the same query until the provider withdraws the diagnosis, without considering the evidence or prior response.
Evaluate the query defect and clinical documentation separately, involve appropriate provider/HIM/compliance owners, and follow the authorized correction and retention processes.
Reviewers identify 6 queries with a defined construction defect in a targeted group of 30 malnutrition queries. Which use of the result is most appropriate?
Report a 20% defect proportion in that reviewed group, investigate contributing causes, implement appropriate correction or education and recheck later work.
Report a 6% department-wide defect rate because 6 defects were identified, regardless of how the sample was selected.
Conclude that all novice specialists have a 20% defect rate without examining who generated the queries or how they were reviewed.
Set a required zero-clarification target for malnutrition so that fewer queries will be available to fail review.
Sections you finish are checked off in the contents.