A method for robust estimation of seasonal onset and intensity in respiratory surveillance data: evaluated using data from 21 European countries

BackgroundSeasonal respiratory pathogens cause hospital admissions and strain health services. Commonly used methods, Moving Epidemic Method (mem), WHO Average Curve Method (WHO-ACM) and Mean Standard Deviation method (MSD), support post-season intensity assessment. We developed aedseo for early seasonal onset detection and within-season intensity assessment. Methodsaedseo fits a rolling quasi-Poisson generalised linear model to weekly counts; onset is declared when growth is significant and the five-week mean exceeds a disease-specific threshold (Tdisease) estimated from recent seasons. Intensity breakpoints (between intensity levels; very low to very high) are disease-specific; Tdisease defines very low; high is the 97.5th percentile of a log-normal fitted to three highest weekly counts per season; low and medium are equally log-spaced in-between. aedseo was developed on Danish respiratory surveillance data and validated with 63 data sources (influenza, RSV, ARI, ILI) from 21 European countries (seasons 2014/15-2023/24), benchmarked against mem, WHO-ACM and MSD. Resultsaedseo signalled onset in 60/63 sources and mem crossed its epidemic threshold in 55/63. When both signalled, aedseo was earlier in 45. Median lead times (weeks) were 6.5 for influenza, 4.5 for RSV, 22 for ARI and 5.5 for ILI; growth between signals was usually sustained, particularly for influenza, RSV and ILI. aedseo provided more consistent intensity categorisation across seasons than mem, WHO-ACM and MSD. Conclusionaedseo excels in early onset detection and within-season intensity assessment. It has demonstrated robust performance in the Danish national respiratory surveillance system and across European data sources, supporting timely planning and communication within surveillance frameworks.