Model estimates of hospitalization discharge rates for norovirus gastroenteritis in Europe, 2004–2015

Norovirus is an important cause of acute gastroenteritis (AGE) globally across all age groups [1‐3]. Populations at increased risk of norovirus gastroenteritis (NGE) are the very young, the immunocompromised, and those living in closed communities such as long-term care facilities [4, 5]. A recent systematic review estimated that norovirus caused 677 million AGE cases worldwide in 2010 [2]. Despite being a major contributor of diarrheal disease worldwide, norovirus is rarely laboratory confirmed or recorded as a cause of hospitalization. As an example, from July 2007 through June 2013 90% of all gastroenteritis hospital discharges in England were of undetermined etiology [6].

To overcome these issues, in recent years there has been an increasing interest and use of medical databases and mathematical models to indirectly estimate the incidence of norovirus-associated disease [7‐11]. In-patient datasets are usually recorded as mandatory national health registers in most European countries. These registers contain information on hospital discharges by diagnostic codes commonly using the International Classification of Diseases, Ninth/tenth Revision (ICD-9/10). Upon request, register data holders may release statistics for specific scientific research purposes.

In order to better understand the incidence of NGE hospitalizations in Europe, we conducted a retrospective, observational study of national in-patient discharge registers from 2004 through 2015 of countries in Europe with the aim of estimating the incidence of community-onset AGE and NGE requiring hospitalization in Europe. This information can serve to inform public health decisions related to the potential introduction of norovirus immunization in Europe.

National hospital discharge registers in all 28 European Union (EU28) countries and all 4 European Free Trade Association (EFTA) countries were contacted and invited to participate in the study. Procedures for data application were fulfilled as required. Hospital statistics data were included if available upon request without apparent restrictions, with or without charge, and when episode records could be classified by AGE-related ICD-9 or ICD-10 codes (see Supplemental Table 1 for a comparison of the codes in the two systems). As the first-listed ICD9/10 code on discharge records presents the principal diagnosis and hence condition leading to a person being admitted to hospital, analyses were limited to discharge records with AGE as the first listed diagnosis in order to capture the burden of hospitalizations for community-onset AGE. Only finalized discharges of inpatients were included. A hospital discharge was defined as the release of a patient who was hospitalized for a minimum of one night: same-day discharges were excluded. When national registers did or could not release individual data, data aggregated by the following diagnostic categories were requested: norovirus, rotavirus, ‘other viral’, Clostridium difficile, ‘other bacterial’, not otherwise specified, and parasitic. The study included all subjects of all ages, and both genders registered in hospital discharge statistics between 2004 through 2015.

Population estimates were obtained from Eurostat for EU and EFTA countries, and from the UK’s Office for National Statistics for England. Denominators were obtained by multiplying the total population estimate by the coverage percentage of the dataset. The representativeness of national hospital statistics was obtained from data providers.

Using national in-patient registers from 15 European countries, we estimated the incidence of all-cause gastroenteritis and norovirus-associated gastroenteritis leading to hospitalization in Europe from 2004 to 2015. The lack of etiologic data for the cause of gastroenteritis in more than three-quarters of in-patients was addressed by using an indirect method as previously applied [6‐11, 17]. We estimated that nearly 200,000 hospital discharges are attributable to norovirus every year across the 32 EU and EFTA countries. This corresponds to an annual NGE hospital discharge rate of 3.9 per 10,000 population.

Few studies have been carried out to assess the incidence of hospitalizations for NGE in Europe. Kowalzik et al. [19] calculated the NGE hospitalization rate in Germany based on aggregate data for hospital discharges with NGE as first-listed diagnosis from the German Federal Statistics Office for 2007–12. Our observed NGE coded hospital discharge rates for the same period match perfectly, as was to be expected when utilizing the same data source. However, when applying the modelling approach, we estimated that the incidence of NGE hospitalizations in Germany in this period was 2.4-fold higher (56,000 predicted compared with 22,000 observed). This suggests a significant underreporting of NGE hospitalizations, even in Germany where testing for norovirus and diagnosis of norovirus as a cause of AGE is relatively high (7.1% of all AGE hospitalizations) compared with other European countries.

The comparability with previous studies that used modelling approaches to estimate the incidence of NGE hospitalizations is limited due to the differing methods used. In studies conducted in England [6] and the United States (US) [8] all discharge records coded with a gastroenteritis diagnosis were considered, irrespective of whether this was the principal or secondary diagnosis. As our aim was to estimate community-onset norovirus hospitalizations, we only included AGE diagnostic codes in the first-listed (principal) position. In an earlier study we performed on the incidence of NGE in England, using a similar modeling method and using data from the Clinical Practice Research Datalink (CPRD) linked to the Hospital Episode Statistics (HES) database, we found a rate of 7.1 NGE hospital discharges per 10,000 person-years [6], compared with 1.8 per 10,000 using only the HES database in the current study. The main differences between the previous CPRD study and current study are that in the first we included all discharges with an AGE code irrespective of whether this was the principal or a secondary diagnoses, thus counting both community-onset and hospital-acquired norovirus infections and including day-cases, while in the current study we only included discharges with AGE as principal diagnosis and defined hospital discharges as the release of a patient who stayed hospitalized for a minimum of one night, which represents 47.3% of all hospitalizations for all causes in the UK [20].

Our study estimated that norovirus was responsible for 17% of all AGE hospital discharges in Europe from 2004 through 2015. These results are in line with the findings of a large systematic literature review that estimated that norovirus was associated with 18% [95% CI, 17–20] of all diarrheal disease worldwide in any setting, and 17% [95% CI, 15–19] specifically in in-patient settings [21].

We found a wide variation in NGE discharge rates between European countries included in this analysis, with a 9-fold difference between the lowest rate in Italy and the highest rate in Lithuania. These differences can be explained to a certain extent by differences in overall hospitalization rates between countries: namely, when comparing estimated NGE hospitalization rates with overall hospitalization rates as reported to Eurostat [20], we found that Lithuania and Romania had the highest estimated NGE discharge rates in children aged < 5 years (87.6 and 58.4 per 10,000 person-years, respectively) and also had the highest overall hospital discharge rates in that age group (4454 and 3817 discharges per 10,000 children aged < 5 years, respectively). Similarly, Austria and Germany had the highest estimated NGE discharge rates in elderly adults aged 80 ≥ years (23.5 and 22.4 per 10,000, respectively) and also had the highest overall hospital discharge rates in this age group (8731 and 6685 discharges per 10,000 elderly). Another source of variability between the estimated NGE rates may be related to the completeness and accuracy of the hospital discharge registers upon which we relied, as well as the robustness of the norovirus coding used in the models.

We estimated a higher number of norovirus discharges during the 2014/15 season followed by seasons 2007/08, 2009/10 and 2013/14. The 2007/08 and 2009/10 peaks coincide with the emergence of new pandemic GII.4 strains (Den Haag 2006B [22, 23] and New Orleans 2009 [24, 25], while the later peaks in 2013–2015 are likely related to the Sydney 2012 strain. The Sydney 2012 strain was originally identified in Australia and was first noted in Europe in the winter of 2012 at relatively low levels [26, 27]. Our data suggest that this strain may have had a delayed impact in Europe.

Our models further suggest that, as expected, hospital discharges for AGE attributable to norovirus peaked during the winter seasons [4, 28, 29]; however, for some of the more recent years a weaker seasonality was observed, with a few distinct peaks of NGE discharges during summer months. The magnitude of these summer peaks varied. Specifically, our model estimated a sharp increase in NGE discharge rates in the summers of 2010 and 2013. As previously suggested by others, this rather atypical increase in norovirus activity during the summer may be associated with the emergence of new pandemic strains, [26, 28, 30‐32]. Escaping population immunity may result in atypical seasonality, but this would depend on when in the norovirus season the emerging strain first appeared [28].

Some limitations need to be taken into consideration when interpreting our results. Firstly, the statistical methods applied assumed that any residual seasonality not otherwise captured in the model was due to norovirus, which may lead to an overestimation of norovirus-associated discharge rates. Secondly, to estimate the rate of community-onset norovirus infections requiring hospitalization, we limited our analyses to discharges with AGE codes in the first position. Although the first-listed diagnostic code is considered the principal diagnosis, or medical condition for which a person is admitted to hospital, we cannot exclude that discharges with AGE as secondary diagnosis may have included cases where AGE was in fact the principal reason for hospitalization, and hence we may have underestimated the true incidence of hospitalizations for community-onset norovirus infections. Lastly, we assumed that the countries for which we obtained data are representative of the 32 countries in Europe. We believe this is a fair assumption as the countries for which we estimated NGE represent 68% of the total population and are geographically widely spread.

By using routinely collected hospital discharge records from a range of European countries, we estimated an overall hospital discharge rate for NGE of 3.9 per 10,000 person-years, or 1 in 400 hospital discharges, in Europe. Norovirus affects all ages, but estimated NGE discharge rates were the highest for children aged < 5 and older adults aged ≥80 years. Estimated NGE discharge rates varied between included European countries, which may be partly explained by countries having different data collection, coding and hospitalization practices.

Our study demonstrates that the burden of NGE on hospitals may be underestimated. Norovirus is a common cause of severe gastroenteritis requiring hospitalization. Increased diagnostic testing for norovirus in hospital settings is needed to better assess the incidence of the disease and to monitor the changing epidemiology of the virus. This information can serve to inform public health decisions related to the health and economic benefit of the potential introduction of norovirus immunization in Europe and to measure the impact of its introduction.