A Dashboard That Looks Too Good: Query Metrics, a Leading Sepsis Query, and What to Tell the Board

Reassuring on their face: response and agreement rates show engaged physicians.

The warning signs cluster:

• 22% query rate sits well above typical CDI program volumes; many programs run in the single digits to mid-teens. High volume is not automatically wrong, but it demands scrutiny.
• 96% agreement with a 41% DRG change rate is the combination that raises concern. Physicians who almost always accept queries, combined with nearly half of queried cases changing DRG, is more consistent with leading queries than with genuinely ambiguous documentation.
• A CMI jump of 0.12 in one year with half of queries targeting a single high-severity condition (sepsis) is exactly the profile a RAC, MAC, or DOJ auditor looks for.
• The sampled query itself discloses its motive ('to reflect severity') and supplies the answer.

Audit plan: pull a statistically valid stratified sample of 30–50 queries by type (sepsis, respiratory failure, malnutrition, AKI, encephalopathy), and have the physician advisor and coding manager review each for (a) whether the documentation was genuinely ambiguous, (b) whether clinical indicators were cited, (c) whether a real 'no/other/unable to determine' option existed, and (d) whether documentation changes were supported by the clinical record. Then compare CMI and CC/MCC capture against external peer benchmarks (HCUP/MedPAR, O/E mortality) rather than against last year's internal number. Involve compliance early; if queries produced unsupported diagnoses, that is an overpayment analysis, not a training issue.

Original problem: it instructs the physician what to document, discloses a financial rationale, cites three nonspecific findings, and offers no alternative.

Rewrite: 'The record documents hypotension (BP 84/52 on [date/time]), tachycardia (HR 118), temperature 38.6°C, WBC 13.2, and lactate 2.4 mmol/L, with IV fluid resuscitation and blood cultures drawn and antibiotics started in the ED. Please clarify the clinical condition being treated: [ ] Sepsis [ ] Sepsis with septic shock [ ] Systemic inflammatory response due to another etiology (specify): ____ [ ] Bacteremia without sepsis [ ] Other: ____ [ ] Unable to determine.'

What changed and why it matters:

• Add the findings that actually discriminate (temperature, culture and antibiotic timing, fluid resuscitation, shock criteria such as persistent hypotension requiring vasopressors or lactate elevation) rather than a generic triad.
• Offer sepsis, septic shock, and non-sepsis explanations, including a free-text and an 'unable to determine' option.
• Remove 'to reflect severity' and every reference to payment.
• Let the physician supply the diagnosis; the CDI supplies the clinical evidence and the question.

Also verify that the query is linked to the attending responsible for the patient and that authenticated responses are part of the permanent record, not just the CDI software.

Frame it as documentation integrity, not CMI growth, and lead with the finding rather than burying it.

Suggested message: 'Our query volume and acceptance rates are outliers, our case-mix growth is concentrated in one diagnosis, and our query templates show a pattern of directing the answer. Before these numbers are used for any financial projection, we are auditing them and correcting our practice. We would rather find this ourselves.'

Actions:

• Suspend canned and diagnosis-suggestive templates immediately; retrain on compliant query construction.
• Route the internal audit findings to compliance and legal for an overpayment exposure analysis, including whether self-disclosure is warranted. Do not let this be handled as a CDI training issue alone.
• Be explicit that the CDI program's value is accuracy; a CMI increase that cannot survive a records request is a liability, not an achievement.

Metrics to report going forward:

• Query rate by type and by clinical service, with a target range, not a stretch goal.
• Agreement rate paired with an audited accuracy rate - agreement alone rewards leading queries.
• DRG change rate reported alongside overturn and denial rates on external review.
• CC/MCC capture and CMI compared to external peer benchmarks, plus O/E mortality and PSI data to show severity is clinically real.
• Query turnaround, pre-bill versus post-bill query mix, and provider education uptake by physician.

A program that reports only volume and CMI is not measuring documentation quality; it is measuring how persuasively it asks.