Quick answer
A quality dashboard needs a reproducible definition and a source population, not only a percentage. Specify the counted unit, numerator, denominator, time basis, exclusions, duplicate rules and version. Keep local operational indicators distinct from official CMS measures. Then test whether another analyst can reproduce the result and explain how missing data, unit transfers or changed definitions affect its interpretation.
Identify the decision the number supports
For federally certified nursing facilities, 42 CFR 483.75(c)(3) addresses performance-indicator development and evaluation. The CMS nursing-home quality-measures hub separately provides versioned official materials. A local workflow indicator is not automatically an official CMS measure or a star-rating calculation.
This guide uses an original fictional indicator: clarification requests without an identified owner. Its purpose is to find a coordination problem. It does not measure clinical outcomes or urgency.
Define the measure before counting it
| Dictionary field | Worked definition |
|---|---|
| Unit | Distinct clarification request, not notification or resident |
| Population | Requests opened in the selected week on included units |
| Numerator | Requests without an identifiable current owner at extraction |
| Denominator | All requests in that population, including those missing an owner |
| Duplicate rule | Link confirmed repeated reports to one request; preserve sources |
| Time basis | Opened during the week; ownership status as of the stated extraction |
| Uncertainty | Unresolved identity or unit attribution reported separately |
| Version | Local definition 1; source extract E14; calculation build C2 |
Add the decision owner and reasons for exclusions. A definition that changes whenever an inconvenient record appears cannot support a meaningful trend.
Reproduce the count, then challenge it
Extract E14 contains twelve rows. Two are confirmed repeat reports, leaving ten distinct requests. Eight have identified owners and two do not. The unresolved-owner proportion is 2/10, or 20%. The complementary ownership proportion is 8/10, or 80%. Label which one the chart displays.
If missing-owner rows are excluded, the ownership display becomes 8/8, or 100%, without assigning anyone. If duplicate reports are counted as requests, the denominator represents a different question. Retain the row-to-request mapping so another analyst can reproduce the result.
An uncertain duplicate should not be silently deleted. If its relationship remains unresolved, show that limitation and its possible effect. Data cleanup must not delay any clinical response, required reporting or investigation attached to the original record.
Compare units with the actual denominators
| Unit | Distinct requests | Without owner | Local proportion |
|---|---|---|---|
| North | 20 | 2 | 10% |
| South | 60 | 3 | 5% |
| Combined | 80 | 5 | 6.25% |
South has more unassigned requests; North has the higher proportion. The combined proportion is 5/80, not the unweighted average of 10% and 5%. Which view helps depends on whether the decision concerns total workload or the fraction of requests affected.
Neither proves better care. Differences in capture practices, included work or unresolved attribution can change the comparison. If five additional rows lack a usable unit, disclose them; do not drop them silently or assign them to the unit that makes the chart look favorable.
Time and location are measurement choices
Occurrence time, discovery time and entry time can differ. For an incident review, preserve them and specify which drives local grouping. Do not fabricate an occurrence time from the moment a user opened the form. Official falls or other clinical measures require their own definitions; this local timeline supplies none.
A unit move creates a similar choice. A current-location worklist can legitimately use today’s unit while a historical analysis uses the unit associated with the event. The labels must explain the basis. Reassigning all historical events to the current unit can create a false trend after transfers.
For pressure-injury or other clinical dashboards, inspect source observations, transformation and qualified review separately. Missing data should not become a reassuring default category. This lineage method does not validate staging, risk algorithms or clinical outcomes.
Mark a changed definition as a changed question
Suppose definition 2 includes older requests carried into the week. The denominator grows because the population changed. A continuous line that omits that change encourages readers to attribute the movement to performance.
Keep the original definition and result. If a restated historical comparison is useful, label it and preserve the relationship to the original packet. A current dashboard refresh should not rewrite what a prior committee saw without explanation.
The same discipline applies to official specifications: identify the applicable version and associated updates, and validate detailed logic separately. A link to the CMS hub is not proof that a local formula reproduces CMS output.
Retain an evidence index, not only a screenshot
Keep the dictionary, included/excluded population, source extract identity, calculation, displayed result and review disposition connected through authorized references. A screenshot preserves presentation; it rarely preserves enough to reproduce the computation.
A second analyst should be able to obtain 20%, reproduce 6.25% and explain the effect of the definition change without private notes. If not, classify the gap: unavailable source, ambiguous rule, undocumented adjustment or stale output. That diagnosis tells the team what to repair before the dashboard guides a consequential decision.
Questions & answers
Should a local QAPI indicator be described as a CMS quality measure?
Only when it actually implements the applicable official specification. Otherwise label it as a local indicator with its own purpose and definitions. Similar wording does not establish identical calculation logic.
How should unit percentages be combined?
Use the appropriate shared numerator and denominator when definitions and populations align. In the example, 2/20 and 3/60 combine to 5/80, or 6.25%, rather than an unweighted average of the percentages.
Is a dashboard screenshot enough to reproduce a result?
Usually not. Retain the source population, definition version, extraction context, calculation and review disposition. The screenshot can link to that evidence as the presentation used at the time.
Primary sources
Use the cited source and its applicable version when making a facility-specific decision. These guides explain operational evaluation methods; they are not individualized clinical, coding, or legal advice.