.jpg){kind=link}
- Published
- Updated
No alerts can indicate good performance or a broken control. Monitoring teams should examine evidence that a rule operates separately from its alert count. Controlled examples that do not alter real participant data can help make that distinction visible.
Scope of the decision
A control example defines an expected result for a known situation. One example may need to produce an alert while an appropriate counterpart should not. The purpose is not to simulate every clinical risk, but to demonstrate that the basic decision route and required data flow operate visibly.
Worked review
Suppose an overdue-report rule has remained silent for a month. In an appropriate test environment, the team runs an overdue example and a timely counterpart. Neither appears. Investigation shows that a filter accidentally excluded one site group. The possible impact on the production period is then examined separately.
Practical checks
- Define the expected alert and expected silence for each example.
- Record the represented data flow and situations outside the example set.
- Assign review of the production period affected by a failed control.
Acceptance and limits
Acceptance requires expected outcomes and an assessment of the historical impact of a failure. A passing test does not prove that no previous event was missed. Revisit whether the examples remain meaningful after introducing a different source or changing the monitoring rule.
Frequently asked questions
- Should a fake event be added to real data?
- No. Use an authorised testing approach that preserves the integrity of production records.
- Is one positive example sufficient?
- Its coverage is limited. Appropriate negative and boundary examples help explain what the control has demonstrated.