QC Failures in a 24-Canister Apartment Batch
Your company's QA plan requires one field duplicate for every 10 field devices, one field blank per batch, and one spiked device per 20 field devices. Acceptance criteria: duplicate relative percent difference (RPD) of 36% or less, or an absolute difference of 0.5 pCi/L or less; field blank at or below 2 times the lab's MDC of 0.3 pCi/L; spike recovery 85–115%. For a 12-unit apartment building you deploy 24 charcoal canisters, two per unit, plus two field duplicates (Units 4 and 9), one field blank, and one spike. Results: the blank reads 0.9 pCi/L; duplicate pair A (Unit 4) reads 2.0 and 2.4 pCi/L; duplicate pair B (Unit 9) reads 1.8 and 5.7 pCi/L; the spike recovers 78%. Unit results otherwise range from 1.1 to 4.6 pCi/L. The lab emails that the batch 'looks fine, most values are low.' The client wants to sign a whole-building mitigation contract next week and needs clearance for the low units so they can be excluded from the contract.
1. Evaluate each QC result against the stated criteria and decide whether the batch is usable.
Read the worked answer
Check each item against the plan, not against the lab's impression.
- Field blank: 0.9 pCi/L exceeds 0.6 pCi/L (2 x MDC). Fail. A blank this high suggests contamination during shipping, storage, or analysis, and it biases all results upward.
- Duplicate A: difference 0.4 pCi/L, RPD = 0.4 / 2.2 = 18%. Pass on either criterion.
- Duplicate B: difference 3.9 pCi/L, RPD = 3.9 / 3.75 = 104%. Fail both criteria. Something varied between two devices at one location: placement, handling, or analysis.
- Spike: 78% recovery is below 85%. Fail. Low recovery means under-response, a downward bias.
Three of four checks fail, and they fail in opposite directions, so you cannot even bound the error. The batch is not usable.
The fact that most values are low is irrelevant. Low values are exactly what a downward-biased batch produces, and the client plans to exclude the low units from mitigation - the decision most vulnerable to that bias.
2. If the batch is not usable, what exactly must be retested and what should be investigated?
Read the worked answer
Because the blank and spike failures are batch-wide, no unit result from this batch can be used to clear a unit or to flag one. The 1.1 pCi/L units are as suspect as the 4.6 pCi/L unit.
Retest all 24 locations with devices from a different lot and, if possible, a different analysis run. Unit 9 requires particular attention given the duplicate spread.
Investigate before redeploying:
- Where canisters were stored - a basement, garage, or area near a known radon source raises blanks.
- Shipping time and temperature.
- Whether the spike was prepared and analyzed in the same run.
- Lab calibration, detector background, and the raw counts.
Ask the lab for the batch QC record and its corrective action. Retesting before finding the cause risks repeating the same failure and doubling the client's cost. Preserve the failed batch data in your QA file; it is the evidence that supports your decision.
3. Write the key points of your message to the client and to the lab.
Read the worked answer
To the client: the batch failed internal QC, so no results from it can support the mitigation contract. State which checks failed and by how much, give the retest schedule and cost, and say explicitly: do not exclude any unit from the contract based on these numbers, and do not treat the low values as clearance.
To the lab: request a written explanation of the blank, spike, and duplicate failures, the batch QC records, and the corrective action taken. Ask whether other clients' batches in the same run were affected and whether the detectors came from the same lot.
Confirm both conversations in writing and retain the failed data. If the retest comes back internally consistent, the client can proceed. If it does not, escalate to a different lab and consider a second measurement method for the discordant units. The credibility of the whole program rests on being willing to throw away a batch that fails.