Clinical audit in depth: standards, sampling, action and re-audit

• Distinguish a clinical audit from QI, service evaluation and research in practical cases.
• Translate authoritative guidance into explicit, measurable audit criteria and standards.
• Design a defensible sample and data collection plan.
• Interpret audit findings without confusing data quality, case mix and compliance.
• Build action and re-audit into the project so that the audit closes the loop.

Clinical audit asks whether care meets explicit standards and uses the findings to improve care. The defining logic is criterion -> measurement -> action -> re-measurement. An activity that only counts current practice, without explicit standards or a plan to improve and recheck, may be useful review work but is an incomplete audit cycle.

Audit is usually assurance-linked improvement rather than research. It does not normally seek new generalisable knowledge; it asks whether known or agreed standards are being achieved locally. However, local registration, information-governance rules, confidentiality and ethical considerations still apply. HQIP's 2025 ethics guidance explicitly rejects the idea that non-research work is ethically neutral.

The strongest audits choose a topic where better adherence matters to patients, staff or service reliability, rather than selecting whatever is easiest to count.

A guideline recommendation is not automatically an audit criterion. The criterion must say exactly what should happen, for whom and under what circumstances. “Patients should receive VTE assessment” is weaker than “100% of adult inpatients without a documented exclusion should have VTE risk assessment completed within the locally specified timeframe.”

State the source and version of the standard. Where a national standard is not numeric, the project team may need to agree a local target and clearly label it as locally set rather than implying national authority. Avoid invented thresholds.

Criteria should be clinically meaningful, observable in the available data and few enough to drive action. An audit with 40 weak criteria often produces less improvement than one with five carefully chosen criteria.

Most audit errors arise in the denominator. Define the eligible population before collecting data. Exclusions should be clinically justified and applied consistently. If a criterion allows exceptions, record them separately rather than treating every non-compliant case as a failure.

For example, an audit of anticoagulation review might exclude patients receiving treatment for less than a defined minimum period, or distinguish documented contraindications from omissions. The data form should capture why a case was excluded.

Changing the denominator after seeing the results creates bias. If the definition genuinely proves unusable, document the amendment and consider recollecting the affected cases.

A clinical audit sample should be large enough to describe current adherence with reasonable confidence and practical enough to complete. Sampling may be consecutive, random, systematic, stratified or a census of all eligible cases. The appropriate choice depends on case volume, variability and the purpose of the audit.

A convenience sample of readily available records is vulnerable to bias. A single week's activity may not represent the service if staffing or workload is highly seasonal. For rare events, reviewing all cases over a longer period may be more sensible.

Do not use a ritual sample size such as “30” without thought. Record why the chosen period and number are adequate for the decision the team needs to make.

Each field should map to a criterion or a planned explanatory variable. Define exactly how ambiguous records will be coded. Use yes/no/not applicable/unknown where appropriate rather than forcing missing information into “no”.

Pilot the tool on a handful of records. If two reviewers interpret the same field differently, refine the operational definition. A data dictionary is particularly valuable when multiple staff collect data.

Collect only information needed for the audit. Identifiable data should be minimised, stored securely and retained according to local policy. Registration with the relevant clinical governance or audit function is commonly required and should be checked before collection.

Report both numerator and denominator, not percentages alone. A result of 80% means something different when it represents 4/5 cases versus 160/200. Display variation by time, location or relevant subgroup when this helps identify a mechanism or equity issue.

Distinguish non-compliance from undocumented compliance. If care may have been given but not recorded, the clinical and documentation problems are different; both may still matter. Do not simply assume absent documentation means good care occurred.

Avoid league-table thinking when numbers are small or case mix differs. The purpose of local audit is usually to improve a process, not to rank individuals.

An audit report is not the endpoint. Translate findings into actions with named owners, timescales and measures. The action should address the cause, not merely repeat the standard. If 40% of discharge summaries omit escalation plans because the field is hidden in the electronic template, “remind doctors” is weaker than redesigning the workflow and testing the new field.

Prioritise actions by clinical importance, feasibility and expected impact. Some findings require immediate escalation rather than waiting for a committee meeting, particularly where they indicate serious patient-safety risk.

Communicate the findings to the people who can change the system, including patients or service users where appropriate.

Re-audit should be planned at the start, not added if someone remembers. The interval should allow the action to take effect while remaining soon enough to maintain momentum. Use the same criteria and comparable population unless there is a documented reason to change them.

A second measurement that still misses the standard is not necessarily failure. It may show that the action was insufficient, poorly implemented or undermined by a new constraint. That learning should lead to further improvement and another check.

Closed-loop audit means the project can show what changed because of the first measurement and whether care subsequently improved or became reliably compliant.

A community hospital audits patients aged 65 and over with acute change in cognition. Criterion: eligible patients should have a documented delirium assessment using the locally approved approach within the defined timeframe. The first cycle finds 31/52 compliant. Review shows night admissions are most often missed and staff report uncertainty about who should initiate assessment.

The action is not simply an email. The team clarifies responsibility at handover, adds a prompt to the admission checklist and tests the workflow with night staff. Eight weeks later, re-audit finds 47/54 compliant, with the remaining cases reviewed individually for valid exceptions and system failures.

The report documents the denominator, exclusions, action, repeat measure and unresolved risks. This is materially stronger than reporting “compliance improved from 60% to 87%.”

Choose one clinical standard. Write the audit question, source, criterion, target, eligible population, exclusions, sample method, data fields, collection period, governance route and planned re-audit date. Then ask a colleague to decide whether a hypothetical record is compliant using only your definitions.

If they cannot do so reliably, the protocol needs refinement before data collection. The Stage 4 Audit Protocol and Close-the-Loop worksheet is designed for this purpose.

• Auditing a vague recommendation rather than an explicit criterion.
• Picking a sample size by habit rather than rationale.
• Treating missing documentation as automatically equivalent to clinical failure without analysing the issue.
• Producing an action plan with no owner or deadline.
• Calling an audit complete before re-audit or monitoring.

1. Healthcare Quality Improvement Partnership (HQIP): Guide to managing ethical issues in quality improvement or clinical audit projects, updated October 2025. https://www.hqip.org.uk/guidance/guide-to-managing-ethical-issues-in-quality-improvement-or-clinical-audit-projects/
2. HRA Decision Tools: Is my study research? and Defining Research terminology. https://www.hra-decisiontools.org.uk/research/

Educational reference. Verify current local governance requirements before starting a project.