Quality improvement and clinical audit

Clinical audit is one QI method. HQIP defines it as a quality-improvement cycle involving measurement of healthcare effectiveness against agreed and proven standards, action to bring practice in line and further measurement. An audit criterion states what should happen to whom and under what circumstances. A standard is the target level of compliance. Standards need an authoritative evidence source and justified exceptions; setting every target at 100% can be inappropriate when contraindication, decline or clinical judgement are legitimate.

Choose the project type from the question and activity. Research seeks new generalisable knowledge. Clinical audit compares care with an explicit standard. Service evaluation judges how an existing service performs without determining a new standard of care. QI tests changes in a local system. These categories can overlap in methods, and publication does not itself turn QI into research. Use the organisation’s governance, information and ethics routes before data collection.

Map the current process with people who do the work and receive the service. A simple flow map may reveal that a result reaches the inbox but no one owns recall, or that patients receive separate appointments they cannot attend. Cause-and-effect diagrams and five-whys questioning can generate hypotheses, but they do not prove a root cause. Prioritise modifiable causes supported by data and avoid blaming individuals for a system that makes the safe action difficult.

Write a SMART aim with scope and time: “Increase the proportion of eligible patients with completed, reviewed and communicated monitoring from 54% to 85% within twelve weeks, without increasing urgent appointment waits.” The aim should be ambitious enough to matter and realistic enough to test. Avoid embedding the proposed solution in the aim, because a reminder template may fail while another change succeeds.

The Model for Improvement uses three questions: What are we trying to accomplish? How will we know a change is an improvement? What change can we make that will result in improvement? A driver diagram links the aim to primary system drivers, secondary drivers and candidate changes. It is a theory of change, not proof. Update it when PDSA learning contradicts the original assumptions.

Build a measure family. Outcome measures represent the result that matters, such as harm, control or patient experience. Process measures show whether the intended work occurred. Balancing measures detect displacement or harm, such as increased workload, waiting time or overtreatment. Structure measures describe capacity. Use operational definitions specifying numerator, denominator, inclusion, exclusion, data source, frequency and owner so measurement is reproducible.

PDSA is experimental learning at local scale. Plan the change, prediction, who, where, when and data. Do the smallest useful test and record what departed from plan. Study results against the prediction, including qualitative feedback. Act by adapting, adopting or abandoning and plan the next cycle. A cycle written retrospectively after full rollout misses the method’s protection: learning before exposure of the whole system.

Clinical audit follows a related cycle. Select a priority; define evidence-based criteria and standards; determine population, sample and exclusions; collect reliable data; compare results; agree specific actions with owners and deadlines; implement; and re-audit. Re-audit should use comparable definitions and enough time for change to operate. If performance improves but misses target, investigate mechanism and iterate rather than declaring failure or changing the denominator.

Sampling must answer the project question. A census may be practical for a small register. A consecutive sample reduces cherry-picking. Random sampling can improve representativeness. Convenience samples may be useful for early PDSA learning but cannot support broad claims. Report dates, setting, denominator, exclusions, missingness and data quality. A percentage without counts can exaggerate change in small samples: 50% to 100% could mean one of two to two of two.

Patient involvement changes both aim and design. Patients may identify that the problem is not missed blood tests but repeated travel, unclear messages or no explanation of results. Co-production means sharing influence, not asking for approval after the intervention is fixed. Use accessible engagement and compensate contributions where policy allows. Include experience or burden among measures and communicate findings back to participants.

Sustainability requires incorporation into ordinary work: named ownership, training, induction, electronic configuration, supplies, policy and ongoing measurement. Test under different staff, days and demand before spread. A successful pilot may depend on one enthusiastic clinician. Create a control plan stating what will be monitored, how often, by whom and what threshold triggers action. Share methods and limits honestly so another site can adapt rather than copy blindly.