4.1 Quality Control and Assessment Terminology
Key Takeaways
- Accuracy is closeness to the true or assigned value; precision is how tightly repeat measurements agree, and the two are not synonyms.
- Mean value is the average of a control level's results; standard deviation (SD) is the spread around that mean on a Levey-Jennings chart.
- Coefficient of variation (%) = (SD ÷ mean) × 100 and compares precision at different concentrations.
- A shift is a sudden, persistent jump to a new mean; a trend is a gradual drift in one direction even if no single point has crossed ±3 SD.
- Linearity is the concentration range in which results stay proportional; results outside that range are not reported as-is.
CMLA Work Area I.2.A.1 requires you to define the quality-control (QC) and quality-assessment (QA) words the exam will reuse whenever a log, meter, or kit is in the stem: precision, linearity, coefficient of variation (CV), reliability, mean value, standard deviation (SD), shift, accuracy, and trend. These are not abstract statistics. On a waived glucose meter or a waived urinalysis (UA) analyzer, they are how you decide whether today's control is acceptable and whether patient results may leave the bench.
Quick Answer: Accuracy is closeness to the true (target) value. Precision is how tightly repeats agree. Mean is the average of control values; SD is the spread around that average. CV = (SD ÷ mean) × 100. A shift is a sudden, persistent jump to a new mean on a Levey-Jennings chart; a trend is a gradual drift in one direction. Linearity is the concentration range where results stay proportional.
Accuracy versus precision (the target analogy)
Picture a glucose control whose assigned (true) value is 100 mg/dL.
- Accurate and precise: five repeats read 99, 100, 101, 100, and 100 mg/dL — a tight cluster on the bullseye.
- Precise but not accurate: 88, 89, 88, 87, and 88 mg/dL — a tight cluster on the wrong part of the target. A mis-coded strip lot or a failed calibration can do this.
- Accurate but not precise: 92, 108, 97, 105, and 99 mg/dL — the average is near 100, but scatter is wide. Uneven wiping of the meter window, incomplete mixing of a control, or an aging lamp can do this.
- Neither: 80, 115, 95, 130, and 70 mg/dL — off target and not reproducible.
Accuracy answers, "Are we close to the truth?" Precision answers, "Would we get the same number again?" You need both before you report patients. Do not treat the words as synonyms, and do not call a pretty off-center cluster "accurate" just because it is tight.
| Term | Plain meaning | Laboratory-assistant example |
|---|---|---|
| Accuracy | Closeness of a result to the true or assigned value | Glucose control assigned 120 mg/dL reads 119–121 mg/dL |
| Precision | Agreement among repeated measurements (reproducibility) | Five low-level meter controls all fall between 48 and 52 mg/dL |
| Mean value | Arithmetic average of a set of control results | (98 + 100 + 102 + 99 + 101) ÷ 5 = 100 mg/dL |
| Standard deviation | Numeric spread of results around the mean | A tight meter has a small SD; a noisy meter has a large SD |
| Coefficient of variation | SD expressed as a percentage of the mean | CV = (2 ÷ 100) × 100 = 2% |
| Reliability | Consistent accuracy and precision over time, lots, and operators | The same waived chemistry analyzer stays in control week after week |
| Linearity | Range in which measured results stay proportional to concentration | A meter that reads 50, 100, and 200 correctly but under-recovers at 500 is not linear at 500 |
| Shift | Abrupt, sustained change to a new mean | Controls jump from ~100 to ~110 after a reagent-lot change and stay there |
| Trend | Gradual, consistent movement in one direction | Controls climb 1–2 mg/dL each day as a reagent degrades |
Mean value and standard deviation
The mean value is the average of a series of QC results for one control level. Laboratories use the manufacturer's assigned mean or, for some systems, a laboratory-established mean after a qualifying period defined in the standard operating procedure (SOP). You do not invent a new mean because today's number "looks nicer."
Standard deviation describes random scatter around that mean. On a Levey-Jennings chart, the center line is the mean. Horizontal lines mark ±1 SD, ±2 SD, and ±3 SD. Most well-behaved control points fall near the mean; a point beyond ±3 SD is almost never ordinary noise.
You do not need to compute SD by hand on the CMLA, but you must know what a large SD means: poor precision. If yesterday's repeats sat on top of each other and today's repeats spray across the chart, something in the system—reagent, dirty optic, control handling, or operator technique—has changed.
Coefficient of variation
CV puts SD on a percentage scale so you can compare precision at different concentrations.
CV (%) = (SD ÷ mean) × 100
Worked example: a waived glucose control has a mean of 80 mg/dL and an SD of 2.4 mg/dL.
CV = (2.4 ÷ 80) × 100 = 3%
A second control at 200 mg/dL with an SD of 4 mg/dL has CV = 2%. The high-level SD is larger in mg/dL, but relative precision is tighter. That is why CV is the fair comparison. CMLA will not ask you to memorize a universal "good CV" cutoff; manufacturers and the laboratory SOP set acceptability. Know the formula, know that you multiply by 100, and know that a rising CV means precision is getting worse.
Reliability and linearity
Reliability is the method's ability to remain accurate and precise across days, operators, lots, and instruments. A meter that is accurate on Monday and wild on Thursday is not reliable, even if Monday's mean was perfect.
Linearity (analytic measurement range / reportable range in manufacturer language) is the interval of concentrations where the measured result stays proportional to the true concentration. If doubling the analyte doubles the reading, you are in the linear range. If a very high glucose is under-recovered, that result is outside linearity. A Certified Medical Laboratory Assistant (CMLA) does not independently dilute and retest a high-complexity chemistry specimen to "force" linearity. You recognize that a result flagged as over-range cannot be reported as a number, follow the SOP (repeat on the waived device if the instructions for use allow, or refer), and notify the supervisor. Off-label modifications of waived systems belong in Chapter 5.
Shift versus trend on a Levey-Jennings chart
A Levey-Jennings chart plots each day's control value against the mean and SD limits.
- A shift is a sudden jump: several consecutive points cluster around a new mean (for example, all well above the old mean after a reagent-lot change, a new strip vial, or a calibration event).
- A trend is a slow walk: each point is a little higher (or lower) than the last, even if no single point has crossed ±3 SD yet. Aging reagents, a deteriorating lamp, evaporation of opened control material, or a slowly failing temperature-controlled block produce trends.
The line chart below shows both patterns on the same 14-day window. The value series is the assigned mean of 100 mg/dL. The Shift series stays near 100 through Day 7, then jumps and stays near 110. The Trend series climbs a little every day. Either pattern means the system is no longer the system you verified—investigate before reporting patients.
Random scatter that still stays within the laboratory's acceptable limits is not a shift and not a trend. One point just inside ±2 SD is usually in-control noise. Do not "fix" a chart by omitting a point you dislike.
Multi-rule QC as industry practice
Many laboratories evaluate Levey-Jennings points with multi-rule QC (often called Westgard rules in industry literature). Typical industry examples include treating a single point beyond ±3 SD as a run rejection, or treating two consecutive points beyond ±2 SD on the same side as a rejection. American Medical Technologists (AMT) does not publish a numbered Westgard set on the CMLA outline. Learn the idea: more than one rule may be used together so that a single random bounce does not shut the bench down, while a true shift or trend is caught. Your legal instruction is the SOP and the manufacturer's instructions for use (IFU), not a rule set you memorize from a textbook and apply in place of the printed control range.
What the assistant actually looks at
On a waived glucose meter you typically run two levels (low and high). Each level has an acceptable range printed on the control vial, the IFU, or the laboratory QC log. If the low is in and the high is out, the system is not in control—do not test patients. On a waived UA strip analyzer, a negative and a positive control (or the manufacturer's two-level set) play the same role. Internal kit lines (the "C" line on a cassette) confirm that the device flowed; they do not automatically replace external liquid controls when the IFU requires those controls.
Exam traps
- Calling a tight, off-target cluster accurate because it is pretty.
- Mixing up shift (abrupt and sustained) and trend (gradual).
- Writing CV as SD minus mean, or forgetting to multiply by 100.
- Treating linearity as "the machine is on."
- Inventing a Westgard number and applying it when the stem only gives the manufacturer's range.
When CMLA shows a QC word, define it in one sentence, then attach it to a meter log or a Levey-Jennings pattern.
A waived glucose control with an assigned value of 100 mg/dL gives five repeats of 88, 87, 89, 88, and 88 mg/dL. Which statement is correct?
A control level has a mean of 80 mg/dL and a standard deviation of 2.4 mg/dL. What is the coefficient of variation, and what does it describe?
On a Levey-Jennings chart, control values remain near the mean for a week and then jump and stay about 10 mg/dL higher after a new reagent lot is opened. This pattern is: