Systemic intervention | about 40 minutes | free
Distinguish a measured change from a causal claim and use evaluation to revise both intervention and framing.
A basic mixed-evidence evaluation rehearsal. It is not a complete impact evaluation, statistical inference course or assessment of real practitioner behaviour.
This is original learning material, not a reproduced official course. Mechanical checks have determinate answers within the stated case. Worked comparisons for open judgements show defensible reasoning, not the only possible model. Specialist pedagogical review is not recorded.
Fictional case: after a new triage process, recorded average waiting time falls from 20 to 12 days. During the same period, 30 difficult cases are moved to another team and excluded from the report. Staff report less duplicated work. Three residents describe improved clarity; two describe feeling dismissed. The analyst designed the change and chose the waiting-time measure. There is no comparison group and no information about seasonal demand.
Use paper, your own drawing tool or the notes below. Attempt each step before opening its comparison. Describe a diagram in words when that is more accessible.
Produce: An absolute change, a percentage change and a qualification.
The reported average fell by eight days, or 40% relative to 20. That describes the recorded series. It does not establish an eight-day improvement for all people or prove that triage caused it, because case inclusion changed and other influences are unknown.
Produce: Outcome, process, distribution and unintended-effect evidence.
Outcome: time to usable resolution across both teams. Process: duplicate work and triage experience. Distribution: results by relevant case types and access needs. Unintended effects: dismissal, displaced queues and staff burden elsewhere. Combine trace data with participant accounts and report disagreements rather than averaging them away.
Produce: A reflexive note and one practical safeguard.
The analyst selected both intervention and measure and may favour evidence of improvement. Use an independent challenge or joint review, retain negative cases, document changed definitions and publish the measurement boundary. Reflexivity is not a confession that makes evidence unnecessary.
Produce: A short finding and a next learning action.
'Recorded waiting fell, and some staff and residents report benefits. Changed case boundaries and mixed resident experience prevent a whole-service causal claim. We will trace the transferred cases and review dismissal accounts before expanding.' This supports action without manufacturing certainty.
These checks test the supplied case, not your overall competence. Open answers and model variants still need your judgement.
The evaluator calls the pilot successful because the selected indicator improved, while excluding transferred cases and dissenting accounts.
Restore the end-to-end boundary, explain changes in inclusion and consider alternative explanations. Evaluate effects and experience, not only the measure chosen by the designer.
Transferred cases now show waits of 50 days. What does that tell you, and what else is needed?
It reveals a potentially serious displaced burden. To calculate a combined mean, obtain counts and comparable definitions for both groups; do not average the two means without weights. Examine whether transfer was appropriate and what happened to residents.
A different model can be defensible when its purpose, assumptions and reasoning are explicit. A contradiction, incorrect unit, invented fact or unacknowledged change of purpose needs repair. Keep both your first attempt and revision.
No sign-in or submission is required. Notes are not sent by this practice tool. Use fictional information, not identifiable client or learner data. Saving stores this page's attempt only in this browser; clearing browser data can remove it. Export a copy to keep it elsewhere.
The complete original case above is free. Links below distinguish exercises from explanations, recordings, software and paid study. Reading a page or drawing in a tool is not itself a model check.
Select the named tool, try the associated template and compare with the example. Use the pack's explicit checks where the toolkit provides no answer key.
HTML toolkit inspected, including mapping, stock-flow, theory of change and evaluation sections. Its examples are not automatic validation of a learner's model.
Write a prediction before a change and compare observation with it. The service and evaluation cases here give a small exercise and answers.
The Institute's account of prediction, study and theory revision inspected. It distinguishes PDSA from a simple implementation check.
Use the cases to examine legitimacy, boundaries, agency and learning partnerships. The original case here provides the checkable rehearsal.
Authors' page, free PDF route and paperback links inspected. It is a related social-learning resource, not identical to every method grouped in the bridge exercise.
Follow 'Learn PDSA in DemingNEXT' for the paid learning route, after using the free explanation and this pack's exercise.
The page links to DemingNEXT. The subscription material and current access terms were not tested.
Choose the relevant workshop and check current arrangements. Learning with others is needed beyond a solo worksheet.
Authors' overview describes workshops using participants' live projects. Specific dates, places, prices and availability are not asserted here.
Choose another practice area · Task-only version · Worked-answer version