Q & A: We don't run QC with a time machine

A question came in about running QC and looking at the initial 20 data points used for validation.

Full Question: "I want to ask: when running in parallel, the first 20 data points don't have Westgard rules applied, and the data is within the manufacturer's 2SD limit.

"However, once the laboratory mean/SD is calculated and the laboratory mean is applied, data points 19 and 20 inadvertently combine with point 21, causing a 3-1s error.

"Is this considered a violation of the rules? Or is the violation only when data points 22 and 23 continue to exceed 1SD?"

A more thorough reply comes from our colleague, Hassan Bayat:

"The first data we collect is in fact 'method evaluation' to find our SD and bias. Then we use this data to calculate Sigma and plan QC program. QC starts them and looks at upcoming data, not look back at previous data. If, according to Sigma, looking at more than one QC result is needed (e.g., 1:3s), the N>1 QC results should be collected in the same run. Even in case of applying look-back approach, it must be applied to results from the start of QC, not to the results collected during 'method evaluation' period."

In QC, we do apply look-back rules, but those only apply to the QC data, not the method validation data.