Category: Basic QC Concepts | Estimated reading time: 7 minutes
A Misleading First Impression
Once you start delving into the statistics behind QC — Mean, SD, CV, Bias%, Sigma metric, Westgard rules — it is natural for a feeling to emerge: "In that case, QC is very complicated."
Quite the contrary. QC, at its core, is a very simple system. What makes it feel complicated is not the concepts, but our habit of approaching it as a procedure to be followed — rather than logic to be understood.
There are four foundational points that, if understood correctly, are all interconnected and make QC feel much more sensible.
1. Every Measuring Instrument Has Variation — and That Is Normal
There is no perfect measuring instrument. There never has been, and there never will be.
When you measure a control material with the same value 20 times in a row under identical conditions, you will not get 20 exactly identical numbers. You will get 20 numbers that are almost the same — spread around the mean value, with a pattern that can be predicted statistically.
This is not a defect. It is the fundamental nature of every measurement system — from a grocery scale to the most sophisticated spectrophotometer.
What distinguishes a quality laboratory from one that is not, is not the presence or absence of variation — because variation is always present. What distinguishes them is whether they know the limits of acceptable variation, and whether they take action when those limits are exceeded.
This is the core question of the entire QC system: which variations can we accept, and which variations should force us to stop and re-examine?
2. There Are Two Types of Variation — Different Causes, Different Handling
This is the concept most often overlooked in daily QC practice. When QC is rejected, many Medical Laboratory Technologists (MLTs) immediately repeat the test without first diagnosing the type of error that occurred. In fact, knowing the type of error is the key to an efficient investigation.
Westgard (2016) distinguishes between two types of errors that have very different characteristics, causes, and handling:
Random Error — variation without a pattern, the direction of which cannot be predicted. One measurement is high, the next is low, then it is high again. There is no directional consistency. In metrology, this is called imprecision, measured by CV. The larger the CV, the more inconsistent the results produced by our instrument.
Systematic Error — a consistent deviation with a pattern in one direction. Every measurement is always higher, or always lower, than it should be. Something systematic is causing this bias. In metrology, this is called inaccuracy, measured by bias%. The larger the bias%, the further our mean result is from the true value.
This difference is not just academic — it has a direct impact on how to investigate:
- Random error → investigate towards current operational conditions: pipetting technique, sample conditions, temperature fluctuations, or operator variation.
- Systematic error → investigate towards systemic causes: shifted calibration, new reagent lot, contamination, or issues with the instrument itself.
Investigating a systematic error with a random error approach — or vice versa — will only waste time without solving the problem.
3. QC Is How We Know
With the two premises above — variation is always present, and there are two different types of variation — the function of QC becomes very clear:
QC is a system that helps us distinguish normal variation from problematic variation.
More concretely, QC answers three questions that must be addressed every time we are about to release a result:
- When can we trust the instrument? — When QC data shows the analytical process is in a stable condition, within established control limits.
- When should we be suspicious? — When there is an unusual pattern in the QC data, even if it has not officially crossed the rejection limit.
- When must we stop and investigate? — When there is a violation of QC rules that indicates a clinically significant error.
Westgard & Westgard (2016) emphasize that effective QC is not just about "pass or fail" — but about the ability to read signals from the data generated every day. A good MLT does not just look at whether the QC point falls within range — they read patterns, recognize shifts, and act before the problem becomes greater.
4. Without QC — We Do Not Know When to Trust
This is the logical consequence of the three previous points, and perhaps the most important one to truly internalize.
Without QC, we have no mechanism to distinguish the day when the instrument is working well from the day when something is wrong. All results look the same on the surface — numbers come out, reports are printed, doctors read them.
But behind those numbers lies unmeasured uncertainty. And unmeasured uncertainty is uncertainty that cannot be managed.
Plebani (2016), in his discussion on the new paradigm in medical laboratories, emphasizes that the reliability of laboratory results is an absolute prerequisite for every clinical decision that depends on them. Without properly functioning QC, we cannot provide that guarantee of reliability — and the one who bears the risk most is the patient.
Not because we don't care. But because we don't know.
Closing
QC is not a complicated procedure. It is the logical answer to one simple reality: measuring instruments have variation, variation has two types, and we need to know which one is occurring — every day, every run.
These four points are not for memorization. This is a way of thinking that, once embedded, will change the way you approach QC — from a morning ritual into a living decision system.