Tools / free calculator
Clinical audit sample size calculator
How many records do you actually need? Either to measure current compliance to a stated precision, or to show that an improvement at re-audit is real rather than chance. The calculator also writes the sentence you can paste into the audit protocol.
- Samples above 500 records per round are usually a sign the change you are trying to detect is smaller than an audit can practically demonstrate. Consider whether a larger, more meaningful improvement target would serve the service better.
The number matters less than how the records are chosen. Taking the first fifty you can find, or the ones with complete notes, produces a biased picture no sample size will correct. Consecutive or randomly selected records over a defined period is what makes the figure defensible.
How to design the audit →In short
How many records do you need for a clinical audit?
Enough that the figure you report is precise enough to act on. To estimate compliance within ±10 percentage points at 95% confidence you need roughly 60 records; to detect an improvement from 70% to 85% between audit and re-audit you need roughly 120 in each round. Twenty records — the number most audits default to — gives a confidence interval around 35 points wide, which is rarely precise enough to support a conclusion.
Common questions
Answers before you ask
How many records do I need for a clinical audit?
It depends on what the audit has to show. To estimate compliance roughly, a few dozen records may be enough — but 80% compliance in 20 records carries a 95% confidence interval of 58% to 92%, which is too wide to act on. To detect a 10-percentage-point improvement at re-audit you typically need around 250 records in each round.
What is the difference between the two modes?
"Estimate compliance" tells you how many records you need to measure current performance to a stated precision. "Detect a change" tells you how many you need in each round to show that an improvement between audit and re-audit is real rather than chance.
Why does detecting a change need so many more records?
Because you are comparing two estimates, each with its own uncertainty. A single estimate only has to be precise enough to be useful; a comparison has to be precise enough that the two intervals separate.
Should I use the population size field?
Only if you are sampling from a genuinely finite, known group — all 400 admissions in a quarter, say. The finite population correction then reduces the sample needed. Leave it blank if you are sampling from an ongoing stream.
What confidence interval does this use?
The Wilson score interval, which behaves properly at proportions near 0% and 100% where the simple normal approximation does not. It is the interval you should report for audit compliance figures.
Is my data stored?
No. The calculator runs entirely in your browser. Nothing you type is transmitted or stored anywhere, and there is no account.
Keep reading
Related
Clinical audit: a practical guide
The six-stage cycle, setting standards, analysis and closing the loop.
Sample size calculator
For surveys and research studies rather than audit, with the written justification.
Independent statistician
For audits that need statistical input a department does not employ full time.