The aim of external quality assessment (EQA) schemes is to evaluate the analytical performance of laboratories and test systems in a near-to-real-life setting. This monitoring service provides feedback to participant laboratories and serves as a control measure for the epidemiological assessment of the regional incidence of a pathogen, particularly during epidemics. Using data from EQA schemes implemented as a result of the intensive effort to monitor SARS-CoV-2 infections in Austria, we aimed to identify factors that explained the variation in laboratory performance for SARS-CoV-2 detection over the course of the COVID-19 pandemic. For this observational study, we retrospectively analysed 6308 reverse transcriptase quantitative PCR (RT-qPCR) test results reported by 191 laboratories on 71 samples during 14 rounds of three SARS-CoV-2 pathogen detection EQA schemes in Austria between May 18, 2020, and Feb 20, 2023. We calculated the overall rates of false and true-negative, false and true-positive, and inconclusive results. We then assessed laboratory performance by estimating the sensitivity by testing whether significant variation in the odds of obtaining a true-positive result could be explained by virus concentration, laboratory type, or assay format. We also assessed whether laboratory performance changed over time. 4371 (93·7%) of 4663 qPCR test results were true-positive, 241 (5·2%) were false-negative, and 51 (1·1%) were inconclusive. The mean per-sample sensitivity was 99·7% in samples with high virus concentrations (1383 [99·4%] true-positive, three [0·2%] false-negative, and five [0·4%] inconclusive results for 1391 tests in which the sample cycle threshold was ≤32), whereas detection rates were lower in samples with low virus concentrations (mean per-sample sensitivity 92·5%; 2988 [91·3%] true-positive, 238 [7·3%] false-negative, and 46 [1·4%] inconclusive results for 3272 tests in which the cycle threshold was >32). Of the 1645 results expected to be negative, 1561 (94·9%) were correctly reported as negative, 10 (0·6%) were incorrectly reported as positive, and 74 (4·5%) were reported as inconclusive. Notably, the overall performance of the tests did not change significantly over time. The odds of reporting a correct result were 2·94 (95% CI 1·75-4·96) times higher for a medical laboratory than for a non-medical laboratory, and 4·60 (2·91-7·41) times greater for automated test systems than for manual test systems. Automated test systems within medical laboratories had the highest sensitivity when compared with systems requiring manual intervention in both medical and non-medical laboratories. High rates of false-negativity in all PCR analyses evaluated in comprehensive, multiple, and repeated EQA schemes outline a clear path for improvement in the future. The performance of some laboratories (eg, non-medical laboratories or those using non-automated test systems) should receive additional scrutiny-for example, by requiring additional EQA schemes for certification or accreditation-if the aggregated data from EQA rounds suggest lower sensitivity than that recorded by others. This strategy will provide assurances that epidemiological data as a whole are reliable when testing on such a large scale. Although performance did not improve over time, we cannot exclude extenuating circumstances-such as shortages and weakened supply chains-that could have prevented laboratories from seeking alternative methods to improve performance. None.
Read full abstract