18.1 Departmental Metrics: Definitions, Calculations and Dashboard Use

Departmental Metrics: Definitions, Calculations and Dashboard Use

A dashboard should connect a measured result to a decision: which records need review, which queries need follow-up, or which process needs investigation. A percentage without a defined population, counting unit and time window can mislead even when its arithmetic is correct.

The AHIMA 2022 CDI KPI practice brief recommends explicit definitions in departmental policy. It describes organizational variation and states that there is no industry standard for query or agreement rates. The ACDIS September 2026 query-practice Q&A likewise discusses differing survey results and local expectations. A respondent's practice or survey category is useful context, not a mandatory national target.

Write the metric dictionary before the dashboard

For each measure, define the numerator, denominator, unit, eligible population, time period, source system, exclusions, duplicate handling and responsible owner. Also state when the data become mature enough to compare. Keep the same definitions across periods or identify the change clearly.

An admission-cohort report follows the admissions selected for that cohort; it is not interchangeable with a report of all reviews performed that month. An encounter reviewed three times remains one encounter for a unique-record coverage measure but creates three review activities. Similarly, two queries on one encounter are two queries and one queried record.

Coverage and assigned-review completion

For an explicitly defined eligible admission cohort, a useful coverage measure is:

$\text{Coverage} = \frac{\text{Distinct eligible encounters receiving an initial review}}{\text{Eligible encounters in the cohort}} \times 100\%$

By contrast, AHIMA's review-rate example uses records reviewed divided by records assigned for review. That measures completion of assigned work. If a program assigns only a subset of eligible admissions, a high assignment-completion rate can coexist with lower population coverage. Show both if both questions matter. Do not silently replace the eligible population with the assigned workqueue.

Define what counts as a completed initial review and the local timing expectation. Short stays, staffing, education work, clinical complexity, scope and technology affect capacity. A bed-to-FTE ratio or coverage goal must be justified for that program; it is not established by declaring a universal 18–22-bed staffing standard or an 80–90% coverage requirement.

Re-review: distinguish activity from completion

Subsequent review helps assess evolving documentation and new results. Define which encounters or review opportunities are due under the local clinical workflow. The follow-up schedule can consider clinical change, unresolved issues and pending results rather than a national ICU-versus-floor clock.

A local due-review completion measure could be completed eligible follow-up opportunities divided by all follow-up opportunities due in the period. For example, 72 completed out of 90 due opportunities is 80%. Count opportunities consistently: if two reviews are due for the same encounter, both numerator and denominator must use opportunities.

Total follow-up reviews divided by distinct patients instead measures activity intensity and can exceed 100%. It does not show the percentage of required reviews completed unless each patient represents exactly one due review. Neither a particular follow-up percentage nor high initial coverage proves that complications were found or missed; assess documentation and sampled review quality.

Query rate: records and query counts answer different questions

AHIMA's recommended record-based approach is:

$\text{Record query rate} = \frac{\text{Distinct reviewed records with one or more queries}}{\text{Distinct records reviewed}} \times 100\%$

The brief also recognizes calculations based on actual query counts. If that convention is used, label it clearly. Queries per reviewed record, sometimes expressed per 100 reviewed records, measures query density rather than the proportion of encounters queried. Several questions on one encounter increase density without increasing the number of queried encounters.

Hypothetical example: A cohort contains 500 eligible admissions. CDI reviews 425 distinct encounters. Eighty-five of those encounters have at least one query, with 100 total queries sent.

Do not increase query volume to reach a target. A low rate may reflect strong documentation, narrow review scope or missed opportunities; a high rate may reflect a broader scope, complex cases or unnecessary queries. Examine case mix, query purpose and sampled records before selecting an explanation. Separate concurrent, retrospective and automated/manual queries when that distinction helps interpretation.

Response and agreement require matched cohorts

Response rate uses queries with a response meeting the department's definition divided by queries sent in the same defined cohort. State the response deadline and observation cutoff. Counting every response received this month over every query sent this month can mix older answers with new questions and produce a misleading result.

Define how routing errors, withdrawn queries, “unable to determine” and other dispositions are counted. A new query still within its response window should not be treated as a failed timely response without explaining the reporting method. Show pending work separately when a cohort has not yet matured.

Agreement rate also needs an explicit outcome definition. AHIMA's published equation uses queries meeting the defined agreed outcome divided by queries sent. Some local reports calculate agreement among answered queries instead; those two results are not interchangeable. State whether the unit is a query or a health record, and how an alternative diagnosis, greater specificity, inability to determine or a clinical-validation response is classified.

For a mature hypothetical cohort with 110 queries sent, 108 qualifying responses and 107 locally defined agreed outcomes:

• Response rate: 108 ÷ 110 × 100 = 98.2%.
• Agreement using queries sent: 107 ÷ 110 × 100 = 97.3%.
• Agreement among responses: 107 ÷ 108 × 100 = 99.1%.

These are three different measurements from the same cohort. A provider's supported alternative answer can be a successful clarification even when it does not meet a particular administrative “agreement” definition. The 2026 query brief protects independent provider judgment; it does not require agreement with a suggested diagnosis.

An unusually high agreement trend can justify checking the query content, outcome mapping or provider workflow. AHIMA identifies that possibility, while also noting that 100% can occur and setting no universal agreement standard. The percentage alone establishes neither leading queries nor exceptional clinical quality. A low response rate likewise warrants investigation of access, routing, coverage, workload and definition problems before attributing disengagement or imposing a diagnosis-focused remedy.

Build operational and management views

Aging buckets should reflect local policy and distinguish calendar from business time. Under/over 48 or 72 hours are possible design choices, not national regulatory thresholds. Automation and NLP can assist prioritization, but are not mandatory dashboard architecture and do not replace clinical review or compliant query governance.

Use the actual account status when linking query activity to finance. Under HFMA MAP definitions, DNFB differs from final-billed-not-submitted status; an open post-bill query is not automatically DNFB. A dashboard can reveal a suspected bottleneck, but a notification feature does not guarantee a response, claim release or reduced receivables. Validate the account-level causes, take a specific action and monitor both timeliness and documentation quality.