# Laboratory Quality Control

> Laboratory quality control for Medical Laboratory Technology: Levey-Jennings charts, Westgard rules, precision vs accuracy, z-scores and EQAS explained.

- Canonical URL: https://prepelephant.com/topics/allied/medical-laboratory-technology/quality-control-lab
- Exam / course: Allied Health · Subject: Medical Laboratory Technology
- Publisher: PrepElephant (https://prepelephant.com) — Prepared and reviewed by the PrepElephant Academic Review Team
- First published: 2026-10-02
- Last updated: 2026-10-02
- How to cite: "Laboratory Quality Control", PrepElephant, https://prepelephant.com/topics/allied/medical-laboratory-technology/quality-control-lab

## Direct answer
Every analytical batch travels with a chaperone: at least two levels of assayed control material whose results must behave before any patient report is released. Those control values are plotted run by run on a Levey-Jennings chart drawn around the assigned mean with lines at one, two and three standard deviations, and Westgard multi-rules convert the pattern into a verdict — 1-2s is a warning, while 1-3s, 2-2s, R-4s, 4-1s and 10x each demand rejection. Internal quality control (IQC) is this daily self-check; external quality assessment (EQAS or proficiency testing) sends blind samples to many laboratories and returns each one a z-score, with values within plus or minus 2 counted satisfactory. Precision (reproducibility, the scatter of random error, summarised by the coefficient of variation) and accuracy (closeness to truth, spoiled by systematic error) are different virtues — a laboratory can be precisely, consistently wrong.

## What you must remember
- Run at least two control levels (normal and pathological) with every batch, exactly as patient samples are processed.
- Levey-Jennings chart: results plotted in sequence against limits of ±1, ±2 and ±3 SD; by the normal distribution, about 95 per cent of points should fall within ±2 SD.
- Westgard rules: 1-2s = warning only; 1-3s = reject (random); 2-2s = two consecutive beyond the same 2 SD (systematic, reject); R-4s = one beyond +2 SD and the next beyond −2 SD (random, reject); 4-1s = four consecutive beyond the same 1 SD (systematic, reject); 10x = ten consecutive on one side of the mean (systematic shift, reject).
- Coefficient of variation CV = (SD ÷ mean) × 100; electrolyte assays demand tighter CVs than enzyme assays — imprecision limits differ analyte by analyte.
- EQAS z-score = (laboratory value − assigned value) ÷ SD of the group; within ±2 satisfactory, 2-3 questionable, beyond 3 unsatisfactory with documented corrective action.
- Calibration is not QC: calibration fixes the signal-concentration relationship with traceable calibrators (after reagent lot change, major maintenance, or drift); QC verifies daily that the calibration still holds.
- Random error scatters results (pipetting inconsistency, bubbles); systematic error shifts them (miscalibration, reagent deterioration, expired standards).
- NABL requires documented IQC plus EQAS/proficiency testing participation for accredited tests — both, not either.

## A 2-2s violation, worked through
Monday morning, glucose control level 2 (the pathological high) reads 2.4 SD below the mean — and yesterday the same control sat beyond −2 SD, two consecutive results past the same limit on one side triggering the 2-2s rule. The run is rejected, and results since the last acceptable QC are held. Troubleshooting proceeds by likelihood: a new reagent lot was opened yesterday — check the lot-to-lot bridge and recalibrate; verify control expiry; inspect the analyser. On recalibration the control recovers to −0.3 SD, and held patient samples are re-run.

The documentation closes the loop: the rejection, the suspected cause, the corrective action and the re-run all enter the QC log, because an assessor reconstructs exactly this sequence. Contrast a lone 1-2s point with no rule violated — proceed watchfully; rejecting every marginal point wastes runs, while ignoring patterns builds silent bias. The Westgard rules exist to make that judgement rule-based rather than temperamental.

## How the exam frames QC
Two questions carry the section. First, "precise but not accurate": picture four shots tightly clustered off the bullseye — an analyser with offset calibration reproducing the same wrong value every run; recalibration against traceable material corrects it, not more repeats. Second, the rule-identification drill: a control at −3.1 SD (1-3s, random error, reject and look for a sporadic event), a control alternating +2.3 then −2.2 SD (R-4s, random), six consecutive results all above the mean (heading for 10x, systematic shift). And the IQC-versus-EQAS confusion: internal control proves today's run behaved; external assessment proves results agree with peers — competence demands both.

## Frequently asked questions
### What is a Levey-Jennings chart and how is it built?
A time-sequence plot of daily control values against the assigned mean with ±1, ±2 and ±3 SD limits, revealing shifts, trends and random scatter at a glance.
### List the Westgard rejection rules and the error type each signals.
1-3s (random), 2-2s (systematic), R-4s (random), 4-1s (systematic) and 10x (systematic); 1-2s serves only as a warning to inspect further.
### Define precision and accuracy with the CV formula.
Precision is reproducibility under repeat conditions; accuracy is closeness to the true value; CV = (SD ÷ mean) × 100 summarises imprecision.
### What is a z-score in EQAS and its interpretation bands?
z = (laboratory value − assigned value) ÷ group SD; within ±2 satisfactory, 2-3 warrants investigation, beyond 3 unsatisfactory with documented corrective action.
### How does calibration differ from quality control?
Calibration establishes the measurement relationship using traceable calibrators at defined intervals; QC monitors every batch to confirm that relationship still holds.
### Why run two levels of control rather than one?
Analyser performance can differ across the analytical range; a normal-level control passing while the high level drifts would be missed by a single level.