PDSA Cycles That Actually Finish

Every department has a folder of half-finished improvement projects. Someone measured antibiotic timing, presented a bar chart at the audit meeting, recommended that everyone try harder, and rotated. Two years later someone else measures antibiotic timing and finds it unchanged.

This is not a failure of effort. It is a failure of method — specifically, of treating a single measurement as a project. The Plan-Do-Study-Act cycle exists to prevent exactly this, and it fails in a small number of predictable ways. This post covers what the method actually asks for, why surgical projects stall, and how to run one that reaches a second cycle.

It sits within our governance series — start at What Clinical Governance Actually Is for how improvement fits alongside audit, incidents and appraisal.

Audit and improvement are not the same thing

This confusion causes more wasted work than any other, so it is worth being blunt about.

Audit tells you whether there is a gap. Improvement is how you close it. A project that measures a gap, recommends that colleagues do better, and stops has done the first half of an audit and none of an improvement project. The re-measurement is not an optional extra — it is the entire point.

The three questions before the cycle

PDSA is the engine, but it does not work without the frame around it. The Model for Improvement, developed by Langley, Nolan and colleagues and popularised by the Institute for Healthcare Improvement, puts three questions first:

1. What are we trying to accomplish?
2. How will we know that a change is an improvement?
3. What change can we make that will result in improvement?

Skipping question two is the most common single mistake, and it is fatal, because a project without a defined measure cannot ever demonstrate that it worked.

A good aim statement is specific, numerical and time-bound. Not "improve VTE prophylaxis" but "increase the proportion of elective arthroplasty patients with VTE risk assessment completed before leaving pre-assessment from 62% to 95% by the end of March."

The measures — and the one everyone forgets

You need three kinds, and the third is the mark of a serious project.

Outcome measure — what actually matters to the patient. Symptomatic VTE rate. Surgical site infection rate. These are the point, but they are often too rare or too slow to guide a short project.

Process measure — whether the change is actually happening. Percentage of patients risk-assessed. Percentage of checklists fully completed. This is what you track week to week, because it moves quickly and tells you whether your intervention is even in place.

Balancing measure — what might get worse because of your change. This is the one that gets skipped, and skipping it is how improvement projects cause harm. Speed up time-to-theatre for hip fracture and you may push other urgent cases later. Add a mandatory field to a form and you may lengthen clinic time or push people to enter something meaningless to get past it. Introduce a new prophylaxis protocol and bleeding is the balancing measure.

If you cannot name what your change might break, you have not thought about it hard enough to test it safely.

Make the first test absurdly small

The single most useful discipline in improvement work: your first PDSA cycle should be small enough to feel trivial.

One clinician. One clinic. One afternoon. Five patients. The instinct — particularly among surgeons, who are decisive by training and temperament — is to design the correct system, roll it out across the department, and measure in three months. That approach fails in a specific way: when it does not work, you have no idea which part was wrong, you have spent your colleagues' goodwill, and there is no appetite for a second attempt.

A ramp looks like this:

Each cycle is a genuine test with a prediction attached, not a rollout stage. If cycle 1 fails, you have lost an afternoon and learned something specific.

What each letter actually requires

Plan. State the change, who does what, and — critically — write down your prediction. What do you expect to happen, numerically? A prediction converts the cycle from an activity into an experiment, and a wrong prediction is the most informative result you can get.

Do. Run it. Record what actually happened, including everything that went wrong and every deviation from the plan. The deviations are usually the most valuable data, because they tell you where the system resists.

Study. Compare the result to the prediction. This is the step that is routinely collapsed into "it went fine". Ask specifically: did it match? If not, why not? What did we learn about the system that we did not know?

Act. Choose one of three: adopt (it worked, standardise it), adapt (partly worked, modify and test again), or abandon (it did not work, try a different change). Every cycle must end with one of these named explicitly. A cycle that ends with "we'll keep an eye on it" has not ended.

Why surgical projects stall

In roughly the order they occur:

1. No measure defined at the outset. The project cannot conclude anything because there is no baseline.
2. The first test was too big. It failed for one of six reasons and nobody can tell which.
3. The change relied on people remembering. Education and exhortation are the weakest interventions available. They decay within weeks and vanish at the next intake.
4. The trainee rotated. No handover, no owner, no continuation.
5. Data collection was manual and painful. If measuring takes more effort than the change, measurement stops first.
6. Nobody with authority was involved. The change needed a decision the project team could not make.

The fix for the third is structural thinking about interventions. Ranked from weakest to strongest:

Surgical improvement projects overwhelmingly reach for the weakest tier because it is the easiest to implement. If your change is a teaching session, expect the effect to decay — and plan the stronger intervention as cycle 2.

Choosing a project that will survive

Good orthopaedic candidates share three properties: the data already exists somewhere, the change is within your team's control, and someone else already cares.

Reliable examples:

• Time to theatre for hip fracture — already collected, already scrutinised, nationally benchmarked
• Prophylactic antibiotic timing relative to knife-to-skin
• WHO checklist completion quality — not compliance percentage, which is always near 100%, but whether the pause actually happens
• VTE risk assessment completion before leaving pre-assessment
• Discharge summary turnaround and whether the primary care team can act on it
• Fracture clinic non-attendance rates and the booking process behind them
• Consent documentation quality against the standard set out in Informed Consent After Montgomery

Avoid projects that require new data collection by busy people, that depend on a department other than yours changing, or that need money you have not secured — that last one is a business case, not a PDSA cycle.

Show it as a trend, not two bars

A before-and-after bar chart is the visual signature of a weak project. Two data points cannot distinguish a real improvement from normal variation, and surgical processes are noisy.

Plot the measure over time as a run chart — weekly or monthly points, with a median line and the intervention marked. This shows whether the change actually shifted the process and whether the shift held. It is also far more persuasive at an audit meeting or an interview than any pair of bars. See Run Charts and SPC for Surgeons for how to build and read one.

Writing it up

Improvement work has its own reporting standard: SQUIRE 2.0 (Standards for Quality Improvement Reporting Excellence). If you intend to publish, follow it from the start rather than reconstructing it afterwards — it asks for the rationale, the context, the specific changes tested, and how you distinguished the effect of your intervention from other things happening at the same time. See CONSORT, PRISMA, STROBE for how the reporting checklists fit together.

For portfolio and interview purposes, what makes a project credible is unglamorous: a defined aim with a number, a baseline, at least two cycles, a balancing measure, and honesty about what did not work. A project that failed and was analysed properly interviews better than a project that succeeded and cannot explain why.

The short version

Pick something small and already measured. Write the aim with a number in it. Name your outcome, process and balancing measures before you start. Make the first test almost embarrassingly small and write down what you predict will happen. Compare the result to the prediction honestly. Decide explicitly whether to adopt, adapt or abandon — then run the next cycle.

The projects that finish are almost never the most ambitious ones. They are the ones whose first cycle was small enough that a second cycle was still possible.