Generated from
cat-harness/sdlc/processes/sme-test-data-review.bpmnbygen-processes-viz.ts— do not edit here. All processes · cat-harness processes
SME Review of Decision Support and Indicator Test Data
Process_SmeTestDataReview · strict · 4 step(s)
SME review of fixed test data for clinical or scientific decision support and indicator definitions. It runs on its caller’s bean (Process_TestDataCuration calls it), so it opens and closes none: every step that touches the plan notes on it.
TWO UNITS OF REVIEW (SmeCaseVerdictKind, $schema: "sme-case-verdict/1.0.0" in schemas/sdlc/test-data.ts):
decision-support: the unit of review is a worked case (case: input state,decisionLogicRef,systemOutput, optionalexpectedGuidance).indicator-definition: the unit of review is a cohort plus an expected value (cohort:indicatorRef,populationSize,inputs,expectedValue,computedValue, optionalperiod,numerator,denominator).
JUDGEMENT-ONLY SME LANE. Lane_SME binds clinical-sme, which is declared judgementOnly: true in scenarios/roles.json. Task_SmeJudge carries no skill reference because no mechanical procedure yields the clinical or scientific answer — that is the whole reason a person is in the lane.
ATTRIBUTABLE REVIEWER AND DATE. Task_RecordCaseVerdict writes the sme-case-verdict/1.0.0 node with the SME’s attributable human identity (reviewer: { kind: "human", provider, handle, role }) and ISO-8601 review date (reviewedAt), because an audited review without an attributable reviewer and date is not audited.
How it connects
- Called by: Test Data Curation — Fixed Sets and Generated Recipes
- Calls: none
- Presented on: no docs page section shows this diagram
- Running instances: none recorded
Lanes — who acts
| lane | role | what it does here |
|---|---|---|
| Authoring agent | authoring-agent | Prepares the review package from the fixed test data set (test-data-fixed/1.0.0): either a worked case with system output for decision support, or a cohort with computed and expected values for an indicator definition. |
| Clinical / scientific SME | clinical-sme | Human subject-matter expert (clinical-sme, judgementOnly: true). Judges whether the decision-support output on a worked case or the indicator value on a cohort is clinically/scientifically right. No procedure yields the answer, which is why a person is in this lane. |
| Tester | tester | Records the SME’s judgement as a validated sme-case-verdict/1.0.0 node with attributable human reviewer identity and reviewedAt date, and updates the fixed data set’s review status when passed. |
Steps
Every one of the 4 step(s) is documented.
| step | lane | skill / sub-process | what it does |
|---|---|---|---|
Prepare the worked case: input, logic, outputTask_PrepareWorkedCase | Authoring agent | test-data-sets | Assemble the DecisionSupportWorkedCase (id, summary, decisionLogicRef, structured input, systemOutput, and optional expectedGuidance) from the fixed data set so the SME can judge clinical/scientific correctness without reading raw schema internals. |
Prepare the cohort: inputs, expected and computedTask_PrepareCohort | Authoring agent | test-data-sets | Assemble the IndicatorCohortCase (id, summary, indicatorRef, populationSize, inputs, expectedValue, computedValue, and optional period, numerator, denominator) from the fixed data set for SME review. |
Judge whether the case or cohort is rightTask_SmeJudge | Clinical / scientific SME | — | Performed by the human SME in Lane_SME (clinical-sme, judgementOnly: true). The SME inspects the worked case (for decision support) or the cohort and expected vs computed value (for an indicator definition) and renders a domain verdict (pass, fail, or needs-revision) with rationale. Deliberately carries no skill reference: no procedure yields the answer, which is the whole reason a person is in this lane. |
Record the SME’s verdict and rationaleTask_RecordCaseVerdict | Tester | test-data-sets | Write the sme-case-verdict/1.0.0 node (SmeCaseVerdictKind in schemas/sdlc/test-data.ts) recording dataSetRef, subjectKind, the worked case or cohort, verdict, rationale, the SME’s attributable human identity (reviewer: { kind: "human", provider, handle, role }), and reviewedAt (ISO-8601 date/timestamp). |
Decisions
Every one of the 2 decision(s) is documented.
| decision | what decides it | branches |
|---|---|---|
Subject kind: decision support or indicator definition?GW_SubjectKind | Selects the unit of review required by SmeCaseVerdictKind (schemas/sdlc/test-data.ts): a worked case for decision-support, or a cohort plus expectedValue for indicator-definition. | decision-support → Prepare the worked case: input, logic, output indicator-definition → Prepare the cohort: inputs, expected and computed |
SME verdict outcome?GW_VerdictOutcome | Routes on the recorded verdict field of the sme-case-verdict/1.0.0 node: pass marks the fixed set reviewed and usable as certification evidence; fail or needs-revision withholds reviewed: true until the logic or case is revised. | pass → Case/verdict pair recorded; fixed data set is audited evidence fail or needs-revision → Discrepancy recorded; logic or case returned for revision |