# PDSA Cycle Worksheet: [Change idea being tested]

> Worksheet for one Plan-Do-Study-Act (PDSA) test of change, set inside
> the three questions of the Model for Improvement, with a ramp log for
> linked cycles and a run chart of the outcome measure. The Model for
> Improvement was developed by Associates in Process Improvement and is
> described in Langley GJ et al., *The Improvement Guide* (2nd edition,
> Jossey-Bass, 2009); IHI teaches it as its core improvement method. Run
> chart rules follow Perla, Provost and Murray, BMJ Quality and Safety
> 2011. Replace every `[bracketed]` value and every row marked
> *Example*. This is a quality improvement format, not medical advice;
> follow local policy, clinical judgement and your organisation's
> governance process. Use aggregate, de-identified data only.

<!--
Guidance: How to use this template
- Use one worksheet per test. A test is small on purpose: one clinician,
  one shift, five patients. Scale up only after the test works under
  the conditions you have tried.
- Answer the three questions of the Model for Improvement once for the
  project (section 1), then reuse them for every cycle.
- Write the prediction before you run the test. Comparing the result
  with the prediction is where the learning happens; without a
  prediction, Study becomes a description.
- Plot data over time (weekly or daily) on a run chart, not before and
  after averages. Apply the run chart rules in section 6 before you
  claim an improvement.
- QI or research? Testing a change in local practice is usually
  improvement work, but if you plan to randomise, withhold care or
  publish generalisable findings, ask your governance or ethics team.
- Related templates: Clinical Audit Report (`healthcare/clinical-audit`),
  Quality Improvement Report (`healthcare/quality-improvement-report`)
  for writing up the project with SQUIRE 2.0, Patient Safety Incident
  Report (`healthcare/patient-safety-incident-report`), Medication
  Reconciliation (`healthcare/medication-reconciliation`), which the
  worked example below measures.
-->

## Document control

<!-- Guidance: Give each cycle its own ID and link it to the project ID, so a ramp of cycles can be read in order. -->

| Field | Value |
|---|---|
| Document ID | [PDSA-PROJ-NN-C1] |
| Project | [Project name and ID] |
| Cycle number | [1 / 2 / 3] |
| Team | [Names, roles] |
| Sponsor | [Name, role] |
| Version | [1.0] |
| Status | [Planned / Running / Studied / Closed] |
| Owner | [@qi-lead] |
| Approvers/reviewers | [@clinical-lead], [@qi-team] |
| Last updated | YYYY-MM-DD |
| Classification | [Internal, aggregate data only] |

### Revision history

| Version | Date | Author | Change |
|---|---|---|---|
| 0.1 | YYYY-MM-DD | [Name] | Plan and prediction written |
| 1.0 | YYYY-MM-DD | [Name] | Study and Act completed |

## 1. Model for Improvement

<!-- Guidance: Answer the three questions for the project as a whole. The aim is specific, measurable and time-bound, and names the population. Pitfall: an aim like "improve medicines safety" that no run chart could ever show is met. -->

| Question | Answer |
|---|---|
| What are we trying to accomplish? (aim) | [Measure, from baseline to target, for which population, by when] |
| How will we know that a change is an improvement? (measures) | [Outcome, process and balancing measures, section 2] |
| What change can we make that will result in improvement? (change ideas) | [Change ideas from the driver diagram, section 3] |

*Example:* Aim: on Ward 5 of the fictional Halvermoor District Hospital, raise the
share of adult admissions with medicines reconciled within 24 hours from
a baseline of 46.1% (149 of 323, eight weeks) to 75% by YYYY-MM-DD.

## 2. Measures

<!-- Guidance: Define each measure so two people would count the same number. Include a balancing measure: something that might get worse because of the change. -->

| ID | Type | Definition (numerator / denominator) | Source | Frequency | Owner |
|---|---|---|---|---|---|
| M-01 *(Example)* | Outcome | Admissions with a completed reconciliation within 24 hours of the decision to admit / all adult admissions to Ward 5 | Electronic record report | Weekly | [Pharmacy lead] |
| M-02 *(Example)* | Process | Admissions with a best possible medication history taken from two or more sources / all adult admissions | Weekly audit of 10 records | Weekly | [Ward pharmacist] |
| M-03 *(Example)* | Balancing | Pharmacy technician minutes spent on the ward per day | Time log | Weekly | [Pharmacy manager] |

## 3. Driver diagram and change ideas

<!-- Guidance: Link each change idea to the driver it acts on. Test the idea with the best mix of impact and ease first. -->

Read the diagram from left to right: the aim, the primary drivers that
must change to reach it, the secondary drivers under each, and the
change ideas the team will test. *Example* below.

```mermaid
flowchart LR
    A["Aim: 75% reconciled within 24 h"] --> P1["Primary driver: history is taken early"]
    A --> P2["Primary driver: discrepancies reach the prescriber"]
    P1 --> S1["Secondary: trained staff on the ward"]
    P1 --> S2["Secondary: weekend cover"]
    P2 --> S3["Secondary: standard place to record"]
    S1 --> C1["Change idea CI-01: technician takes history for morning admissions"]
    S2 --> C3["Change idea CI-03: trained nurse rota at weekends"]
    S3 --> C2["Change idea CI-02: reconciliation section in admission note"]
```

## 4. This cycle

<!-- Guidance: Fill Plan before the test, Do during it, Study and Act after it. Keep it short: a worksheet longer than the test is a sign the test is too big. -->

### Plan

| Item | Detail |
|---|---|
| Objective of this cycle | [What you want to learn] |
| Change being tested | [Change idea ID and description] |
| Prediction | [What you expect to happen, with a number] |
| Who | [People involved] |
| Where | [Unit, bay, clinic] |
| When | [Dates, shifts] |
| Data to collect | [Measure IDs, plus notes on problems] |

*Example:* Cycle C1 tests CI-01 on one bay for one week. Prediction:
technician-led histories will raise M-01 above 55% that week, and each
history will take under 20 minutes.

### Do

Record what happened, including problems and observations that were not
part of the plan.

*Example:* The technician took 14 histories. Two patients were too
unwell to be interviewed, so the GP record and the patient's own list
were used. Median time per history was 18 minutes.

### Study

Compare the results with the prediction. What did you learn?

| Measure | Prediction | Result | Matches? |
|---|---|---|---|
| M-01 *(Example)* | Above 55% | 58.5% (24 of 41) | Yes |
| Time per history *(Example)* | Under 20 minutes | Median 18 minutes | Yes |

### Act

Choose one: adopt, adapt or abandon, and define the next cycle.

| Decision | Reason | Next cycle |
|---|---|---|
| *Example:* Adapt | Worked on weekdays, weekend admissions missed | C2: extend to all weekday admissions, C3: test CI-03 at weekends |

## 5. Ramp of cycles

<!-- Guidance: One row per cycle, in order, so a reader sees how the change grew from a small test to standard practice. -->

| Cycle | Weeks | Change | Scale | Decision |
|---|---|---|---|---|
| C1 *(Example)* | 9 | CI-01 technician history | One bay, weekdays | Adapt |
| C2 *(Example)* | 10 to 11 | CI-01 plus CI-02 | Whole ward, weekdays | Adopt |
| C3 *(Example)* | 12 to 16 | CI-03 weekend rota | Whole ward, seven days | Adopt, hand over to business as usual |

## 6. Run chart

<!-- Guidance: Plot the outcome measure over time, mark when each cycle started and draw the baseline median. Apply the four run chart rules before you claim a change. Pitfall: declaring success from two good weeks. -->

```mermaid
xychart-beta
  title "M-01 percent reconciled within 24 h, weekly (Example)"
  x-axis [W1, W2, W3, W4, W5, W6, W7, W8, W9, W10, W11, W12, W13, W14, W15, W16]
  y-axis "Percent" 0 --> 100
  line [45.0, 47.6, 43.6, 51.2, 47.5, 42.1, 46.5, 45.0, 58.5, 65.0, 69.2, 69.0, 70.0, 73.2, 73.8, 75.0]
  line [45.75, 45.75, 45.75, 45.75, 45.75, 45.75, 45.75, 45.75, 45.75, 45.75, 45.75, 45.75, 45.75, 45.75, 45.75, 45.75]
```

| Week | W1 | W2 | W3 | W4 | W5 | W6 | W7 | W8 | W9 | W10 | W11 | W12 | W13 | W14 | W15 | W16 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Reconciled *(Example)* | 18 | 20 | 17 | 21 | 19 | 16 | 20 | 18 | 24 | 26 | 27 | 29 | 28 | 30 | 31 | 30 |
| Admissions *(Example)* | 40 | 42 | 39 | 41 | 40 | 38 | 43 | 40 | 41 | 40 | 39 | 42 | 40 | 41 | 42 | 40 |
| Percent *(Example)* | 45.0 | 47.6 | 43.6 | 51.2 | 47.5 | 42.1 | 46.5 | 45.0 | 58.5 | 65.0 | 69.2 | 69.0 | 70.0 | 73.2 | 73.8 | 75.0 |

Run chart rules (Perla, Provost and Murray 2011), applied to the example:

| Rule | Signal of non-random change | Example result |
|---|---|---|
| Shift | Six or more consecutive points all above or all below the median (points on the median are skipped) | Met: weeks 9 to 16, eight points above the baseline median of 45.75% |
| Trend | Five or more consecutive points all going up or all going down | Met at the limit: weeks 12 to 16 are five consecutive points, each higher than the one before, from 69.0% to 75.0% |
| Runs | Too few or too many runs for the number of points, read from the published table | Not used: the median was frozen at baseline |
| Astronomical point | A point obviously different from all others, agreed by people who know the process | None |

*Example* summary: weeks 1 to 8 reconciled 149 of 323 admissions
(46.1%); weeks 9 to 16 reconciled 225 of 325 (69.2%); the last four
weeks reconciled 119 of 163 (73.0%). The shift shows a non-random
improvement, but the 75% aim is not yet met in a sustained way.

## 7. Learning and next steps

<!-- Guidance: Write what the team now believes about the change and what it will do next. Include what did not work: abandoned ideas are learning too. -->

- What we learned: [Key learning in two or three lines]
- Surprises: [What happened that you did not predict]
- Next cycle or handover: [Next test, or how the change becomes standard work and who owns the measure]
- Write-up: [Link to the quality improvement report, if the project will be reported]

## References

<!-- Guidance: Cite the method you used. If your organisation uses its own QI method, record its name and version here. -->

| Reference | Status | Use |
|---|---|---|
| IHI, How to Improve: Model for Improvement, https://www.ihi.org/library/model-for-improvement | Checked 2026-10-30 | Three questions, PDSA, credit to Associates in Process Improvement |
| Langley GJ, Moen RD, Nolan KM, Nolan TW, Norman CL, Provost LP. The Improvement Guide: A Practical Approach to Enhancing Organizational Performance. 2nd ed. San Francisco: Jossey-Bass; 2009 | Cited from IHI page | Source of the Model for Improvement. Official text: Jossey-Bass (Wiley), purchase or licensed access: https://www.wiley.com/ |
| IHI, Plan-Do-Study-Act (PDSA) Worksheet, https://www.ihi.org/library/tools/plan-do-study-act-pdsa-worksheet | Checked 2026-10-30 | One worksheet per test (free download after a form) |
| Perla RJ, Provost LP, Murray SK. The run chart: a simple analytical tool for learning from variation in healthcare processes. BMJ Qual Saf 2011;20(1):46-51, https://doi.org/10.1136/bmjqs.2009.037895 | Citation checked on PubMed 2026-10-30 | Run chart rules |

## Standard and licence note

Structure aligned with the Model for Improvement and PDSA cycle as
described publicly by IHI (page checked 2026-10-30) and in *The
Improvement Guide* (2nd edition, 2009), with run chart rules from Perla,
Provost and Murray (2011). This template is original text written for
this collection; it does not reproduce the IHI worksheet or the book.
All names, wards and figures in examples are fictional. This is a
quality improvement format, not medical advice; check the current
official pages and follow local governance. Template licence: MIT.