Improving the accuracy of Respiratory Syncytial Virus (RSV) incidence estimates among hospitalised adults in Bristol, UK

Background: The burden of Respiratory Syncytial Virus (RSV) infection in adults is of interest in the context of recently-licensed vaccines. However, burden estimates are affected by test error associated with the testing platform, and number and type of samples tested. Methods: We conducted a prospective cohort study of adults with acute lower respiratory tract disease (aLRTD) hospitalised in Bristol, UK, from April 2022–March 2023. RSV was detected by RT-PCR both by routine standard-of-care (SOC) testing, and by testing of additional nasopharyngeal swabs, saliva and sputum samples from a patient subset. Latent class analysis was used to quantify and adjust for test error rates, including effects of multiple testing. RSV test-positivity rates are reported, and after adjustment for test error, are used to calculate adult population incidence/1000 person-years. Results: 6906/11445 aLRTD cases (60%) were tested and 251 were positive (3.6%; 251/6906). Test-positivity peaked in December (95%CI 7.9–12.7%). Among cases, 43% had pneumonia, 55% had non-pneumonic infection, 59% chronic respiratory disease exacerbations, and 16% heart failure. Test-positivity was highest in 75–84-year-olds, and 30-day mortality was highest in ≥ 75-year-olds (7.1%; 9/127). Due to low positivity-rates and imperfect specificity (0.98–1.00), test-positivity (3.6%) overestimated inferred true prevalence (2.3%). After adjustment for test error, we estimate overall adult population incidence/1000-person-years to be 0.33 (0.21–0.49), and 2.02 (1.10–3.06) in ≥ 75-year-olds. Conclusions: RSV contributes significantly to hospitalised adult aLRTD, particularly among the elderly. The implementation of effective RSV vaccines could reduce morbidity, mortality and associated costs of disease. Adult RSV burden accuracy is improved by adjustment for test characteristics due to the impact of imperfect specificity when positivity-rates are low, and this is particularly important for out-of-season estimates. Multiple samples can improve burden estimation accuracy only when tests have near-perfect specificity.