1. Choose one quantity and comparable records
This worked chart monitors control recovery in percent. Define the control material, calculation, unit, method and settings version, acceptable record status, time order, and context fields before collecting the sequence.
Use independent routine runs from one interpretable population. Preserve invalid, repeated, or excluded runs and document whether they appear on the display or contribute to the baseline calculation.
2. Approve and version the baseline
Select a stable representative period and record every included run. Calculate the center and dispersion by the approved method, inspect assumptions and special causes, then approve the baseline version and effective date.
The worked baseline has mean 100% recovery and SD 4 percentage points. Its lines are 104% and 96% at ±1 SD, 108% and 92% at ±2 SD, and 112% and 88% at ±3 SD. These are illustrative statistical lines, not assay acceptance limits.
3. Plot each result against the retained lines
Plot runs in execution order and retain the exact value, run identity, baseline version, method version, and active filters behind every point. Do not round the stored value merely to place it on the display.
In the worked sequence, Run 10 is 109% recovery. It is above the +2 SD line of 108% and below the +3 SD line of 112%; that exact position is the observation to interpret.
Run 10 is above +2 SD, so the defined response begins
- Baseline
- Mean 100%; SD 4 percentage points; +2 SD 108%; +3 SD 112%.
- Observed signal
- Run 10 is 109%: above +2 SD and below +3 SD. That observation is not automatically a failed assay.
View the exact values plotted
| Run | Control recovery | Position |
|---|---|---|
| 1 | 99% | Within ±2 SD |
| 2 | 102% | Within ±2 SD |
| 3 | 98% | Within ±2 SD |
| 4 | 101% | Within ±2 SD |
| 5 | 104% | Within ±2 SD |
| 6 | 97% | Within ±2 SD |
| 7 | 100% | Within ±2 SD |
| 8 | 103% | Within ±2 SD |
| 9 | 96% | Within ±2 SD |
| 10 | 109% | Above +2 SD |
| 11 | 105% | Within ±2 SD |
| 12 | 107% | Within ±2 SD |
4. Apply the predefined rule without changing disposition
If the procedure treats one point beyond ±2 SD as a warning, Run 10 creates that warning. If the approved rule set does not use that pattern, the same point does not acquire a new rule after it is observed.
Store the rule, involved point, observed value, baseline and limits, time, and outcome. A longitudinal signal is not automatically a failed run, and a point within the bands does not prove that run-level suitability passed.
5. Investigate with the source-run context
Open Run 10 and compare control and reagent lots, analyst, instrument and maintenance, plate position, timing, method and software versions, suitability, exclusions, deviations, and neighboring results. The chart suggests where to look; it does not identify the cause.
Record the investigation even when no assignable cause is confirmed. Never remove the point or recalculate the baseline simply to clear the visible warning.
6. Retain the disposition and follow-up
Record the reviewer, conclusion, affected records, disposition, corrective or preventive action, baseline decision, approval, and effectiveness check. Link that record back to the signal and source analysis.
If evidence supports rebaselining, approve a new population and effective point while retaining the original chart and limits. Historical points remain interpreted under the versions active at the time.
Limits and optional rule sets
A Levey–Jennings display does not require one universal number of runs, estimator, distribution, or signaling rule. Select and validate the approach for the monitoring question and data behavior.
Westgard-style multirules, CUSUM, EWMA, and other methods are separate choices. Current Provenarium trending supplies one bounded explainable signal, not a general customer-configurable Levey–Jennings or multirule engine.
Frequently asked questions
How many runs are needed to establish limits?
There is no universal count. It depends on stability, independence, distribution, method variation, and intended use. Use statistical expertise and document the selected baseline.
Should acceptance limits and control limits be the same?
Not necessarily. Acceptance limits define a method or product decision; control limits summarize expected process behavior. Display and label both when both matter.