Accuracy, Precision & Monthly Evaluation: Measuring Lab Performance Quantitatively

Category: Performance Metrics | Estimated reading time: 8 minutes

Questions Beyond Just "Accept or Reject"

So far, we have discussed how to set control limits, visualize QC data, and make accept/reject decisions using Westgard rules. But there is a bigger question that remains unanswered:

"How good is our laboratory's performance really?"

It is not just about whether QC is accepted or rejected today — but overall, in one month, does our measurement system produce accurate and precise results? Is the total error we produce still within clinically acceptable limits? This article answers that quantitatively, not just visually.

Accuracy and Precision: Two Different Dimensions

Before diving into the formulas, it is important to build an intuition about these two concepts separately — because they are often misunderstood as the same thing.

• Accuracy is how close our average result is to the true value. It answers: "Are we measuring the right thing?"
• Precision is how consistent our results are from one measurement to the next. It answers: "Are we measuring the same way every time?"

The bullseye target analogy clearly illustrates the four possible combinations:

CombinationError Profile DescriptionPrecision ✓, Accuracy ✗Shots grouped tightly, but far from the center. Systematic error: small CV, large d%Precision ✓, Accuracy ✓Shots grouped tightly in the center. Ideal condition: small CV, small d%, low total errorPrecision ✗, Accuracy ✗Shots scattered randomly and far from the center. Worst condition: large CV, large d%Precision ✗, Accuracy ✓Shots scattered randomly, but the average is close to the center. Significant random error, without meaningful systematic error

What needs to be internalized: these two dimensions must be evaluated together. A laboratory that focuses only on precision (CV) without checking accuracy (d%) can consistently produce incorrect results without realizing it — because the QC is always within range, even though the range itself has shifted from the true value.

Formulas to Master

D% (Bias / Measure of Accuracy)

d% = ((Result − TV) / TV) × 100%

Where Result is the average QC value obtained during the evaluation period (usually one month), and TV is the Target Value established from preliminary testing.

A positive d% means the average result is higher than the TV (positive bias). A negative d% means the average result is lower than the TV (negative bias). In the TE formula, we use the absolute value |d%| — because both positive and negative bias contribute equally to total error.

CV (Coefficient of Variation / Measure of Precision)

CV% = (SD / Mean) × 100%

SD and Mean here are calculated from all QC data in the evaluation period — unlike the CV from preliminary testing, which represents the best stable conditions. The monthly evaluation CV represents precision in real operational conditions: including inter-day variation, inter-operator variation, and all conditions that occur during a working month.

Total Error: Integrating Accuracy and Precision

From d% and CV, we can calculate the Total Error (TE) — a comprehensive measure of how far our results can deviate from the true value:

TE = |d%| + 2CV

The factor 2 in this formula represents 2 standard deviations — providing a confidence interval of approximately 95% for the maximum error estimate that might occur in individual measurements.

Its two components:

• |Bias%| derived from d% — a measure of systematic error.
• 2CV derived from the monthly evaluation CV — a measure of random error multiplied by 2SD.

TEa: Comparison Standards to Know

The calculated TE is meaningless if it is not compared to something. The comparator is the Total Allowable Error (TEa) — the maximum error limit still acceptable without significantly affecting clinical decisions. There are three sources of TEa commonly used internationally:

CLIA (Clinical Laboratory Improvement Amendments) — United States regulations that establish proficiency testing criteria for hundreds of analytes. CLIA TEa values are very frequently used as an international reference because they are based on solid clinical considerations, cover almost all clinical chemistry, hematology, and urinalysis parameters, and are easily accessible and frequently updated. Example: Glucose TEa = 10%, Total Cholesterol TEa = 10%, Hemoglobin TEa = 7%.

RiliBÄK (Richtlinie der Bundesärztekammer) — guidelines from the German Medical Association. RiliBÄK is the German standard, used primarily in European laboratories and laboratories using German standard systems. Some Indonesian laboratories using equipment or systems from Germany also refer to this standard. RiliBÄK TEa values are generally stricter than CLIA for some parameters, reflecting the higher precision standards demanded in the German health system.

Laboratory-defined TEa — some laboratories, especially tertiary hospitals or reference laboratories, set their own TEa based on the specific clinical needs of their patient population. This is the most sophisticated approach, but also the most demanding, as it requires a deep understanding of how laboratory result variation affects clinical decisions for each condition served.

Decisions from TE vs TEa Comparison

Once TE and TEa are known, the decision is straightforward:

• TE < TEa → Acceptable. Laboratory performance during the evaluation period is still within clinically acceptable limits. The QC system can be continued without changes — although it is still necessary to monitor for any developing trends.
• TE ≥ TEa → Investigation. The total error produced has exceeded the allowed limit. There is something in the system that needs to be fixed — either accuracy (d% is too large → calibration or method problem), precision (CV is too large → operational problem), or both.

This evaluation should ideally be performed every month — not just when there is an obvious problem. Systematic error often develops slowly until it is large enough to be seen, and monthly evaluation is the mechanism to catch it before it becomes a bigger problem.

Relationship with Sigma Metric

After d% and CV are evaluated against TEa, there is one more step that integrates all this information into one of the most informative figures: Sigma Metric.

Sigma = (TEa% − |Bias%|) / CV%

The Sigma metric uses the three values we have calculated here — TEa, d%, and CV — and integrates them into a single comprehensive measure of quality. This is why a solid understanding of accuracy, precision, and TE is a prerequisite for understanding the Sigma Metric.

(Read the full article on how to calculate and interpret the Sigma Metric.)

Conclusion

Accuracy and precision are two different dimensions of measurement quality — and both must be evaluated together. d% measures accuracy. CV measures precision. TE integrates both. And the comparison of TE with TEa provides the answer to the most important question: is our laboratory performance still within clinically acceptable limits?

Monthly evaluation of TE vs TEa is not a formality. It is a mechanism that ensures the daily QC system we run truly protects patients — rather than just generating data.