Category: QC Basic Concepts | Estimated reading time: 10 minutes
Why 1 Chart Is Never Enough
Most basic QC guides teach how to read a single Levey-Jennings chart. But at the bench, you don't have one chart — you have two, three, sometimes more, one for each control level run.
And the three often do not tell the same story.
Level 1 normal. Level 2 reject. Level 3 normal. What does it mean? Is it a random error? A systematic error? A problem with the Level 2 control material only? Or something more serious?
Those questions cannot be answered by looking at just one chart. The power of multi-level QC actually lies in the comparison between charts — and this article discusses how to do it systematically.
Three Questions for Reading Charts in Parallel
When opening charts from two or three control levels simultaneously, there are three questions that need to be answered in sequence:
1. Are all levels moving in the same direction?
If Level 1, 2, and 3 all move up simultaneously — or all sit on the same side of the mean — that is a different signal from a situation where only one level moves while the other two remain stable.
2. Did the violation occur in the same run?
Timing is key. Violations occurring in different runs (Level 1 rejected yesterday, Level 3 rejected today) are more likely to be isolated. Violations occurring in the same run, across all levels, are far more significant.
3. How many levels are involved?
This forms the basis of the decision rule. One level, two levels, or all levels — the answer leads to different conclusions.
Decision Matrix: Combination of Violations → Initial Interpretation
Below is an initial guide, not a definitive diagnosis. Final interpretation still requires context: incident logs, types of violations, and patterns in the charts.
All levels rejected in the same run, same direction. This pattern is the most consistent with a systematic error. If Level 1, 2, and 3 all exceed the limits in the same direction in one run, it is almost certain that something is affecting the entire analytical process uniformly — a shifted calibration, a problematic reagent lot, or unstable instrument conditions. First step: hold patient results, perform a systemic investigation.
All levels rejected in the same run, but opposite directions. One level goes up significantly, the other goes down significantly. This is the R-4s pattern — an unusual dispersion within one run, and this is more consistent with a significant random error. Step: check the measurement conditions during that run, repeat the run.
Two out of three levels rejected, same direction. This is a gray area. It could be a partial systematic error that is more pronounced at certain concentration ranges — for example, an instrument that begins to lose linearity at low levels but is still okay at high levels. Step: investigate with a focus on the affected concentration range.
Only one level rejected with a single 1-3s. One point is far out while the other two levels are normal. This is more consistent with a localized random error or a problem with that specific control specimen. Step: repeat that control level before drawing conclusions — do not immediately reject the entire run.
Only one level shows a pattern, not a single violation. If one level shows a trend, shift, or 10x while other levels are stable, this is different from a single outlier point. A sustained pattern in just one level can indicate a problem with the control material itself (vial open too long, degradation) or an uneven instrument response at certain concentrations. Step: investigate that control level material, check the incident log.
All levels show a unidirectional pattern, but none rejected yet. This is the easiest scenario to miss because there is no formal rule that applies. But if an upward trend appears in three charts simultaneously, the probability of that happening by chance is very small. This is almost certainly a developing systematic error — calibration starting to drift, reagents nearing expiration, or instrument components slowly deteriorating. Step: act before the official rule is triggered, record the observation, look for the starting point of the pattern.
Rules That Can Only Be Evaluated with 2 Levels
There are two Westgard rules that, by design, require more than one control level in a single run. These rules do not appear in single-level charts, and therefore are often missed in laboratories that read charts level by level separately.
R-4s: Random Error Within One Run
This rule is satisfied when, in one run, the difference between the highest and lowest values of all control levels exceeds 4 SD. If Level 1 falls at +2.5 SD and Level 2 falls at −1.8 SD, the difference is 4.3 SD — the R-4s is met.
This is specifically designed to detect large random errors in one measurement opportunity. Unlike 1-3s which detects one extreme point, R-4s detects unusual dispersion between levels.
A practical way to evaluate it: look at the Z-score (or SD position) of all levels in one run. If there is one above +2 SD and another below −2 SD in the same run, R-4s is likely met and needs to be calculated more carefully.
Important note: R-4s is a reject rule, not a warning rule. If met, the run must be rejected.
2-of-3-2s: Two Out of Three Levels Outside ±2 SD
For laboratories with 3 control levels, this rule marks a situation where 2 of 3 levels in one run are outside ±2 SD in the same direction.
One level outside ±2 SD is a warning (1-2s). But if two levels simultaneously violate ±2 SD in the same direction in one run, the probability of that happening by chance is much smaller — and this falls into the reject category.
This rule is not discussed much in basic QC materials, but it is relevant for laboratories running 3 control levels.
Three Concrete Scenarios
Scenario 1: All Rejected, All in One Direction
Monday morning, you run QC and find: Level 1 at +3.2 SD, Level 2 at +2.8 SD, Level 3 at +2.5 SD. All are above the mean, all exceed ±2 SD, Level 1 exceeds ±3 SD.
This pattern is highly consistent with a systematic error. All levels rise proportionally. None are in the opposite direction. You look at the previous days' charts and find a gradual upward trend over the last 5 days that had not yet met the 7T rule, but today they all exploded simultaneously.
Step: hold patient results, check records — was there a new reagent lot, recalibration, or other change 5 days ago? This is most likely the systemic starting point.
Scenario 2: Only 1 Level Rejected
Level 2 (pathological concentration) is out at −3.4 SD. Level 1 and Level 3 are normal, both within ±1 SD. There is no trend in all three charts.
This is more consistent with a localized random error or a problem with the Level 2 control material itself. Before concluding, check: was the Level 2 vial used today freshly opened? Has it been open for too long? Was the reconstitution different?
Step: repeat the Level 2 control with a new vial or from the same source. If the second result is normal, the problem is likely with the control specimen, not the analytical process.
Scenario 3: Trend Pattern in All Levels, Nothing Rejected Yet
This is the scenario that is most easily missed. No formal rule is violated. All points are still within ±2 SD. But if you look at all three charts together, they all show an upward tendency over the last 6–7 days.
As discussed in the Levey-Jennings Chart Patterns article, trends that are still within limits are still meaningful. If that upward pattern appears in all levels simultaneously, the probability of it happening by chance is very small — this is almost certainly a developing systematic error, not random noise.
Step: increase vigilance, record the observations, find the cause before any official rule is met. This is an example of proactive, not reactive, QC.
Limitations to Keep in Mind
Multi-level is not a guarantee of perfect detection. As discussed in the Target Value & SD article, if the SD used is too wide for all levels, all charts can look calm even though there is a significant error. The number of levels does not compensate for incorrect SDs.
Patterns are not always symmetrical between levels. Instruments with non-linear responses at certain concentrations can show larger errors at Level 1 (low concentration) compared to Level 3 (high concentration) — or vice versa. The interpretation of "only 1 level is affected" needs to consider this.
Control materials with different lots between levels complicate comparisons. If Level 1 just changed lots while Level 2 and 3 have not, a shift in Level 1 could reflect the lot difference, not an analytical problem. This is exactly the scenario discussed in the Changing Control Material Lots article.
Low Sigma requires more than just multi-level. If your method's Sigma Metric is below 3, adding control levels helps but does not replace the need to re-evaluate the method, reagents, or instrument. Multi-level is a reading tool, not a substitute for good SQC design.
Summary
- All levels rejected in the same run in the same direction: almost certainly a systematic error. Perform a systemic investigation, hold patient results.
- Only 1 level rejected with a single 1-3s: more consistent with random error. Repeat that control level before concluding.
- Only 1 level shows a pattern (trend, shift): likely a local problem with the control material or uneven instrument response.
- All levels show a unidirectional pattern even though none are rejected: this is a signal of a developing systematic error. Act before the rule triggers.
- R-4s and 2-of-3-2s are rules that can only be evaluated with 2–3 levels, and are often missed if charts are read one by one.
- The number of levels does not compensate for incorrect SD or low Sigma.
Closing
Reading multi-level charts is not about looking at more numbers — it's about finding patterns that cannot possibly be seen from a single chart. Violations that occur simultaneously at all levels speak differently than isolated violations. Trends that appear concurrently in three charts speak differently than a trend in just one chart.
Medical Laboratory Technologists who are accustomed to reading charts in parallel detect problems earlier, identify their type more accurately, and make better-documented decisions. This is not a skill that takes years — but it does require practice to see the charts as a single system, not three separate images.