Bayesian and Trend-Based Quality Assurance for Binary Urinary Antigen Tests

Objective: Rapid urinary antigen (uAg) tests for Streptococcus pneumoniae and Legionella pneumophila are widely used to support early decision-making in community-acquired pneumonia. These lateral flow assays yield binary results without signal intensity, limiting the use of traditional quality assurance methods. Population-level monitoring of test positivity and Bayesian predictive values may offer complementary approaches for long-term performance evaluation. Methods: A retrospective study of all pneumococcal and Legionella uAg tests (BinaxNOW, Abbott, USA) was performed in Uppsala County, Sweden, from 2007 to 2024. Trends in test positivity and volumes were analyzed using descriptive statistics, logistic regression, and correlation analysis. Annual positive and negative predictive values (PPV, NPV) were calculated using Bayes’ theorem, applying either national incidence data or observed test positivity as proxy for disease prevalence. Quality assurance thresholds were defined using the interquartile range method for outlier detection. Results: A total of 17,356 pneumococcal and 15,280 Legionella uAg tests were analyzed. Pneumococcal positivity displayed significant seasonal variation, with the lowest odds of a positive result in August (OR 0.49; 95% CI 0.35–0.68). Legionella positivity varied primarily by year, with peaks in 2007, 2012, and 2021–2022. Pneumococcal positivity was highest in children and female patients, while Legionella showed no significant demographic trends. Bayesian PPVs based on observed positivity were substantially higher than those based on incidence (median 53.2% vs. <0.2% for pneumococcus; 12.2% vs. <0.04% for Legionella). Quality assurance intervals for test-based PPV were defined as 39.5–67.0% and 4.2–20.2%, respectively. Conclusions: Rapid urinary antigen tests pose challenges for quality assurance due to their binary nature and lack of quantitative signal. Positivity trends and Bayesian analysis suggest that long-term, population-based monitoring can support quality assurance efforts. This approach may help detect analytical drift and improve the interpretability of binary test results over time.