How many notes do I need to review?
Audit sample size is not a power calculation. This works out how many records give you a compliance figure precise enough to act on.
Audit sample size is a precision problem, not a power problem
Most sample size calculators answer a different question from the one an audit asks. They size a study to detect a difference between two groups. An audit has no second group: it compares one population against an agreed standard, and what you need is a compliance figure precise enough that people will act on it.
So the inputs are different. You need the number of eligible cases in the period, a guess at compliance, and the margin of error you can live with. If you have no idea what compliance will be, enter 50% — that is the value that requires the largest sample, so it is the safe assumption.
The finite population correction matters here and is missing from most online calculators. Those assume you are sampling from an effectively infinite population. An audit samples from a known, often small, list — every caesarean section last quarter, every discharge summary in March. When your sample is a large fraction of that list, the correction reduces the number you need substantially, sometimes by half.
One caution that no formula can fix: this arithmetic assumes a random sample. Pulling the most recent fifty notes, or the ones that happen to be on the ward, produces a number with no defensible margin of error at all. Number your eligible cases and select with a random sequence.
Frequently asked questions
Why does 50% expected compliance need the biggest sample?
Because the variance of a proportion is largest at 50% and shrinks as you move towards 0 or 100%. If you genuinely do not know what to expect, 50% is the conservative entry: it will never leave you under-sampled.
Is there a minimum sample for an audit?
No universal one, though many departments default to 20 or 50 out of habit. The table above shows what those conventional numbers actually buy you in precision, which is usually less than people assume.
My whole population is only 30 cases. What then?
Review all of them. Below roughly 50 eligible cases, sampling saves little effort and costs you the ability to state a margin of error at all.
Does this apply to the re-audit as well?
Yes, and use the same margin of error both times. If the first cycle had ±10 points and the second ±20, an apparent improvement may be nothing more than the wider interval on the second sample.
The book behind this tool
The Clinical Audit & Quality Improvement Project Workbook
From a problem you noticed on the ward to a signed-off, portfolio-ready project.