Levey-Jennings charts and Westgard rules, in plain English.

A Levey-Jennings chart plots quality-control material results against their established mean and standard-deviation limits over time. Westgard rules are the statistical tests read off that chart — patterns in the control points that signal when a measuring process may be drifting, shifting or scattering more than it should. Together they are the working core of internal QC in laboratories and point-of-care testing alike.

One boundary is worth stating before the detail: everything on this page concerns control material — manufactured samples with known target values — never patient results. Westgard rules are statistical data-quality signals about the measuring process. What any signal means, and what to do about it, is a human judgement.

What a Levey-Jennings chart plots

For each device, control level and analyte, the clinic first establishes a mean and standard deviation (SD) for the control material. The Levey-Jennings chart then plots every subsequent control run as a point in time order, against horizontal lines at the mean and at ±1, ±2 and ±3 SD. Read across the chart and the behaviour of the measuring process becomes visible at a glance: points hugging the mean (stable), points stepping to one side (a shift), points sloping away (drift), or points swinging wide (increasing scatter). The chart does not say why — a new reagent lot, a calibration, a temperature problem — but it shows that, early and objectively.

The Westgard rules, one line each

Westgard multi-rule QC formalises how to read the chart. The common rules, each as a plain-English statistical signal:

• 1_2s — one control result beyond ±2 SD. A warning, not a failure: it prompts a closer look at the other rules before anything is decided.
• 1_3s — one control result beyond ±3 SD. A strong signal of random error in that run.
• 2_2s — two consecutive control results beyond the same ±2 SD limit. A signal of systematic error creeping into the process.
• R_4s — two control results within one run at least 4 SD apart. A signal of widening random error — the spread, not the centre.
• 4_1s — four consecutive control results beyond the same ±1 SD limit. A signal of gradual systematic drift.
• 10_x — ten consecutive control results on the same side of the mean. A signal of sustained bias, even when no single point looks alarming.

Which rules a clinic applies, and to which analytes, is a policy decision — typically guided by the device manufacturer's instructions for use and the clinic's own quality goals. The rules detect statistical patterns; a trained person decides whether testing pauses, a control is repeated, or an instrument is recalibrated.

QC review discipline in a POCT setting

Point-of-care testing puts instruments in corridors and consulting rooms, run by clinical staff rather than laboratory scientists — which makes the discipline around QC more important, not less. In practice, a sound routine looks like this:

• Run controls to schedule — per device, level and analyte, following the manufacturer's instructions for use and the clinic's policy.
• Review before testing continues. A flagged control is looked at when it happens, not at the end of the month.
• Respond and record. Repeat the control, recalibrate, change the lot, or escalate — and write down which, by whom, and why. An unrecorded response did not happen, as far as any assessor is concerned.
• Review cumulatively. A named reviewer looks at the charts over weeks, where drift and shift live, not just at today's points.
• Close the loop. Recurring patterns feed nonconformity handling, with the records an ISO 15189 assessment will ask to see.

How Catenix automates the charts

None of the above requires software — laboratories ran Levey-Jennings charts on graph paper for decades. What software removes is the friction that makes the discipline decay. Catenix builds the charts automatically for every connected device, per control level and analyte, as control results arrive through the middleware layer; evaluates each new point against the clinic's configured Westgard multi-rule set; can hold testing on a device through a lockout workflow until a failed control is resolved; and records the review trail — who looked, what they decided, when — in a tamper-evident audit log. The result is statistical data-quality monitoring with the paperwork already done. Humans own every decision: Catenix does not interpret patient results, does not judge clinical acceptability, and provides no clinical decision support. See the quality module for the full QC, competency and CAPA workflow, and which analysers feed it.

← Back to resources