The Laboratory Quality Briefing · III
The sodium control had passed after calibration. About 20 minutes later, the moving average of patient results warned that sodium was drifting upwards. The laboratory traced the change to an instrument malfunction. Its routine sodium control was scheduled every four hours, so the patient results gave the team an earlier reason to look.
What a moving average can see
Patient-based real-time quality control uses a running summary of eligible patient results to look for changes in routine testing. A moving average recalculates the mean as new results enter the window. If the patients being tested are broadly comparable from one window to the next, a sustained shift can make a change in the measurement system visible between checks with control material. A different mix of patients can move the same average, which makes the way the laboratory investigates an alert as important as the way it sets the limits.
Adam McShane, Nancy Werts and William Légaré describe how the Cleveland Clinic core chemistry laboratory used this approach for 17 analytes. The team tracked selected adult outpatient results from July 2018 to December 2025. They excluded extreme results and several settings, including dialysis, emergency care and cancer centres. A warning held results from automatic release and prompted checks with control material and comparison of at least five patient specimens on another analyser. This gave staff an independent line of evidence when the patient average moved.
Investigations classified 160 of 623 alerts as actionable (25.7%). Calcium accounted for 31 actionable alerts out of 77, carbon dioxide for 34 out of 182, and creatinine for one out of 15. The raw counts matter. Carbon dioxide generated slightly more actionable alerts than calcium and more than twice as many investigations. The paper also describes falling carbon dioxide results when open sample tubes spent too long on the system, an example of a specimen-handling problem that the control material would have missed.
The 25.7% describes the yield among alerts that the laboratory investigated. The authors did not measure how often a genuine failure went undetected, so the study cannot estimate sensitivity or tell us how much extra protection the system provided across all runs. It also came from one high-volume laboratory. An outpatient-only rule provides little coverage when few eligible specimens arrive, and an alert can take much longer to accumulate on a low-volume assay. The authors reported no known competing interests.
If you are considering a local pilot, start with one assay and a failure you want to catch. Use historical results to see how the running average behaves across busy and quiet periods and when the patient mix changes. Simulate a shift large enough to matter and count how many results and how much time the rule takes to flag it. Run the rule silently before letting it affect release, and agree how staff will investigate an alert, what they will record and who decides when routine reporting can resume. Keep the existing internal quality control (IQC) in place while you establish what the new signal adds.
From the website
A reagent lot change poses a related question. When a control shifts at the point a new lot is introduced, what evidence shows whether patient results shifted too? Our free guide follows the timing of a control shift and uses paired patient specimens to examine whether the change is also visible in patient results. The pairing holds the specimen constant, which a comparison of separate groups of patients cannot do.
Read the guide on QC shifts after a reagent lot change
Further learning
The free IQC email course covers what control materials monitor, how to read a sequence of results and what a repeat control adds to an investigation. It runs over ten days in five short emails.
For a team discussion, take a recent QC incident and ask what would have been visible in the patient results, when it could have been seen and what independent evidence would have confirmed the cause. The answers can help you decide whether a patient-based rule would add anything to the safeguards you already use.