Westgard rules explained for POCT teams

The chart first

A Levey-Jennings chart plots each control result in time order against the mean and standard deviation established for that control material on that analyser. Lines at plus and minus 1, 2 and 3 SD turn the chart into a set of zones. A Westgard rule is a statement about where points fall in those zones: how far, how many in a row, in which direction. The rules are named N-L, meaning N control observations exceeding the limit L, so 1-3s reads as one observation beyond 3 SD.

The mean and SD must come from your own data on that instrument, typically at least 20 runs over 20 days after the material is introduced, not from the manufacturer's insert. A chart built on the wrong SD will either fire constantly or never.

The rules, one at a time

Which rule catches which error

Analytical error comes in two kinds. Random error is scatter: results are imprecise but centred on the truth. Systematic error is bias: results are consistently high or low, as a shift (a step) or a trend (a slope). The rules divide along the same line:

• Random error: 1-3s and R-4s. Look for a one-off cause in the run: sample handling, clot, bubble, short sample, a control vial that was not mixed.
• Systematic error: 2-2s, 4-1s, 10x, 7T. Look for something that changed and stayed changed: a new reagent or control lot, a calibration, a maintenance event, a temperature change, a firmware update.

This is why the rule that fired is worth recording. It points the investigation.

Multirules and why 1-2s is a warning

A single 1-2s rejection rule has an unacceptable false-rejection rate: with two controls per run, roughly one run in ten would be rejected with nothing wrong. Westgard's insight was to combine rules so that the false-rejection rate stays around 1% while the probability of detecting a medically important error stays high. The classic multirule uses 1-2s as a warning that triggers inspection of 1-3s, 2-2s, R-4s, 4-1s and 10x; if none of those is violated, the run is accepted.

Rules can be applied within a run (across levels) and across runs (the same level over time). A 2-2s across the two levels in one run points at the run; a 2-2s on the same level across two runs points at the analyser or the lot.

How many rules do you need? Ask sigma

Not every analyte needs the full multirule. The sigma metric, (TEa minus absolute bias) divided by CV, measures how much room a method has before its errors become clinically significant, and Westgard Sigma Rules translate it into a QC design:

The practical consequence for POCT: a high-sigma analyte such as sodium on a good blood-gas analyser can be controlled with 1-3s and two controls, and applying 2-2s and 4-1s to it just generates false rejections on a ward at 3 a.m. A low-sigma analyte needs the multirule and the investigations that come with it.

What to do when a rule fires

1. Stop

   Do not report patient results from that analyser for the affected analyte until the rejection is resolved. If the device supports lockout, this is what lockout is for.

2. Do not just repeat

   Repeating the control until it passes is the most common wrong response. It converts a rejection into a false acceptance and leaves the cause in place. One repeat to rule out a handling error is reasonable; a passing repeat does not close the investigation on its own.

3. Read the rule

   1-3s or R-4s: look inside the run. 2-2s, 4-1s or 10x: look for what changed. Check lots, calibration, maintenance, temperature, the control vial's opened date and storage, the operator.

4. Fix, then re-run

   Correct the cause (new control vial, recalibrate, service, retrain) and run controls again. The repeat is evidence only when the cause has been addressed.

5. Assess patient impact

   If the error was systematic, decide with the medical lead whether results since the last acceptable QC need review or repeat.

6. Document

   The run, the rule, the cause, the action, the repeat and who did what. This record, not the chart, is what the assessor asks for; it is also what feeds a non-conformance if one is needed.

Special cases in POCT

• Electronic QC and internal checks on some devices verify the electronics, not the reagent. They do not replace liquid QC; they change how often liquid QC is needed. Follow the manufacturer and your IQCP where you have one.
• Unit-use devices (strips, cartridges) have QC on each lot and each new box or shipment rather than each run, and the rules are applied over those events.
• Many operators, one device means the operator is a variable. A pattern of rejections that follows a person rather than a device is a training finding, not an analyser finding.
• Monthly review: ISO 15189:2022 expects QC data to be reviewed at a defined frequency and acted on. A monthly summary of runs, rules fired, failures and reviews, signed by the reviewer, is the usual evidence.

Where software helps

Everything above can be done on paper and a spreadsheet, and many services do. The parts software does better are the ones that depend on being there when the control was run: capturing the value from the analyser rather than a transcription, applying the rules at that moment, holding the device until someone has looked, attaching the review to the run, and turning the month into a summary. Catenix quality control does that, records the rule that fired, links the failure review to a non-conformance when one is warranted, and, through the Statistics Workbench, calculates sigma per analyte so the rule set can be chosen deliberately.

Questions people ask

This article is general guidance for laboratory and point-of-care professionals. It is not a substitute for the standards themselves, your accreditation body's requirements or the analyser manufacturer's instructions for use. Catenix does not interpret clinical results and provides no clinical decision support.