Metric results

A metric result is the calculated value for one approved metric definition, one cohort and one period window. It keeps the numerator, the denominator, the value, the dimension key it was split by and the number of source rows behind it. It also points at the operation run that produced it. When a value looks wrong, the team can open the run, see the steps and check the inputs without recalculating from scratch.

Observations are recorded first. Each metric observation carries the metric name, the period start and end, the numerator, denominator, unit and value, and it points at both the cohort and the source snapshot it came from. Results are then persisted once per approved definition and window with a status of Calculated, Suppressed or Invalid. A suppressed or invalid result keeps the reason, for example a denominator of zero or a cell too small to report.

• No value from a zero denominator. The calculation must suppress the result with a reason instead of writing a rate.
• The original result is kept. A recalculation adds a new result tied to its own run and leaves the earlier one in place.
• Calculation under the analyst’s control. The Outcomes Analyst runs calculation and reconciliation and can see each stored result with its reason.

The Healthcare Analytics Overview page reads these results to show KPI health, reviewed results by unit and the observed readmission-rate trend.

Agree the period windows you report, such as calendar month or rolling quarter, and how a window is named in the dimension key. Agree the minimum denominator or cell size below which a result is suppressed, and the reason text to record. Decide whether the calculation workflow is activated in the pilot or results are entered manually first. Agree who may mark a result invalid and what evidence they attach.

Calculate a monthly readmission rate for a validated cohort. Confirm that the result shows the approved definition version, the numerator and denominator, the source row count and the run. Then calculate a window where the denominator is zero. The result should be stored as Suppressed with a reason. Confirm that the overview trend does not plot it as a zero rate.

Run the same calculation again with the same operation key. The existing run should be returned and no second result written. Ask the Metric Steward to confirm, from the “Suppressed and invalid results” view, that each suppression has a reason a reviewer could accept.

• Metric result windows — every result with its metric, cohort, period start and end, dimension, numerator, denominator, value and status.
• Suppressed and invalid results — results that were not reported, with the recorded reason and the run that produced them.
• Published outcome cohorts — cohorts whose results have been published, with revision and exposure.

• Cohort construction to validation: Build a cohort revision, load its facts and accept or reject each one.
• Run reconciliation and correction: Resolve failed runs and correct a published revision without losing the original.

Review the implementation plan and the data model before preparing your import files.