Estimating the under-reporting of norovirus illness in Germany utilizing enhanced awareness of diarrhoea during a large outbreak of Shiga toxin-producing E. coli O104:H4 in 2011 – a time series analysis

Abstract

BackgroundLaboratory-confirmed norovirus illness is reportable in Germany since 2001. Reported case numbers are known to be undercounts, and a valid estimate of the actual incidence in Germany does not exist. An increase of reported norovirus illness was observed simultaneously to a large outbreak of Shiga toxin-producing E. coli O104:H4 in Germany in 2011 – likely due to enhanced (but not complete) awareness of diarrhoea at that time. We aimed at estimating age- and sex-specific factors of that excess, which should be interpretable as (minimal) under-reporting factors of norovirus illness in Germany.MethodsWe used national reporting data on laboratory-confirmed norovirus illness in Germany from calendar week 31 in 2003 through calendar week 30 in 2012. A negative binomial time series regression model was used to describe the weekly counts in 8∙2 age-sex strata while adjusting for secular trend and seasonality. Overall as well as age- and sex-specific factors for the excess were estimated by including additional terms (either an O104:H4 outbreak period indicator or a triple interaction term between outbreak period, age and sex) in the model.ResultsWe estimated the overall under-reporting factor to be 1.76 (95% CI 1.28-2.41) for the first three weeks of the outbreak before the outbreak vehicle was publicly communicated. Highest under-reporting factors were here estimated for 20–29 year-old males (2.88, 95% CI 2.01-4.11) and females (2.67, 95% CI 1.87-3.79). Under-reporting was substantially lower in persons aged <10 years and 70 years or older.ConclusionsThese are the first estimates of (minimal) under-reporting factors for norovirus illness in Germany. They provide a starting point for a more detailed investigation of the relationship between actual incidence and reporting incidence of norovirus illness in Germany.

Under-reported data analysis with INAR-hidden Markov chains

“…Therefore, we have an efficient method to compute the likelihood function (13) using the recursive relation given in (12). Once we have implemented the likelihood function, it is maximized with respect to the parameters of interest through an iterative approach using the function nlm in R. The R script we used for this purpose is available in the Supporting Information.…”

“…which is a ratio of likelihoods. These likelihoods are computed using the forward algorithm described in (12), evaluating the likelihood function at the maximum likelihood estimators of the parameters. Then, the normal pseudo-residual segment [z − n , z + n ] can be obtained, where…”

“…In the context of biomedical research, the under-reporting of time series data has been studied for instance for norovirus illness [12]. The authors used a negative binomial model without taking into account the correlation structure of the time series, perhaps a reasonable assumption for their data, although they claimed that 'the independence assumption of the residuals was slightly violated'.…”

Norovirus Vaccines and Potential Antinorovirus Drugs: Recent Advances and Future Perspectives

“…A systematic review by the CDC analyzing 175 publications covering 48 countries and 25 years concluded that HuNoV was responsible for 18% cases of all gastroenteritis worldwide and 14–19% in developing countries and 20% in developed nations [13]. Though infections in the young and old are believed to be under-reported [18], these estimates are higher than previously thought, further highlighting the need for HuNoV vaccines and antivirals.…”

Monitoring Count Time Series inR: Aberration Detection in Public Health Surveillance

“…Furthermore, an increase in case numbers during the 2011 STEC O104:H4 outbreak is observed, which is explained by increased awareness and testing of many gastroenteritits pathogens during that period. The hypothesis is thus that there is no actual increased disease activity (Bernard, Werber, and Höhle 2014). Unfortunately, the German reporting system only records positive test results without keeping track of the number of actual tests performed -otherwise this would have been a natural adjustment variable.…”