Asten, L. van, Siebenga, J., Wijngaard, C. van den, Verheij, R., Vliet, H. van, Kretzschmar, M., Boshuizen, H.,
Pelt, W. van, Koopmans, M. Unspecified gastroenteritis illness and deaths in the elderly associated with
norovirus epidemics. Epidemiology: 2011, 22(3), 336-343
This is a NIVEL certified Post Print, more info at http://www.nivel.eu
Postprint Version 1.0
Journal website http://journals.lww.com/epidem/pages/articleviewer.aspx?year=2011&issue=05
000&article=00013&type=abstract
Pubmed link http://www.ncbi.nlm.nih.gov/pubmed/21358409
DOI 10.1097/EDE.0b013e31821179af
This is a NIVEL certified Post Print, more info at http://www.nivel.eu
Unspecified Gastroenteritis Illness and Deaths in the
Elderly Associated With Norovirus Epidemics
L
ISELOTTE VAN ASTEN,
A
JOUKJE SIEBENGA,
A,D
CEES VAN DEN WIJNGAARD,
A
ROBERT VERHEIJ,
C
HANS VAN
VLIET,
A
MIRJAM KRETZSCHMAR,
A
HENDRIEK BOSHUIZEN,
B
WILFRID VAN PELT,
A
AND MARION
KOOPMANS
A,D
From the aCentre for Infectious Disease Control Netherlands, National Institute for Public Health and the
Environment, Bilthoven, The Netherlands; bExpertise Centre for Methodology and Information Services,
National Institute for Public Health and the Environment, Bilthoven, The Netherlands; cNetherlands
Institute of Health Services Research (NIVEL), Utrecht, The Netherlands; and dDepartment of Virology,
Erasmus Medical Center, Rotterdam, The Netherlands.
Background: New variant strains of norovirus have emerged worldwide in recent years,
evolving by mutation much like influenza viruses. These strains have been associated with a
notable increase in the number of annual norovirus outbreaks. However, the impact of such
increased norovirus activity on morbidity and mortality is not clear because norovirus infection
is rarely specifically registered.
Methods: We studied trends of gastroenteritis with unspecified cause in medical registrations
(ie, general practitioner [GP] visits, hospitalizations, and deaths) and their association with
known temporal trends in norovirus outbreaks in the Netherlands. Using weekly counts in the
elderly (aged 65+ years) from 1999 through 2006, we applied Poisson regression analyses
adjusted for additional pathogens and seasonal trends (linear, sine, and cosine terms).
Results: In the elderly, each notified norovirus outbreak was associated with an estimated 26
unspecified gastroenteritis GP visits (95% confidence interval = 17–34), 2.2 unspecified
gastroenteritis hospitalizations (1.6–2.7), and 0.14 unspecified gastroenteritis deaths (0.08–
0.21). For the heaviest norovirus season (2004 –2005), these models attributed up to 3804
unspecified gastroenteritis GP visits, 318 unspecified gastroenteritis hospitalizations, and 21
unspecified gastroenteritis deaths to norovirus outbreaks among a total elderly population of
2.3 million.
Discussion: The recent increase in norovirus outbreak activity is associated with increases of
unspecified gastroenteritis morbidity and even deaths in the elderly. Norovirus should not be
regarded as an infection with trivial health risks.
Diarrheal disease has been linked to increased mortality among the elderly, especially in winter.1 This
seasonal rise suggests that, in regions of temperate climate, a putative cause may be infections with
norovirus, which typically peak in the winter.2 Norovirus infection, or “winter vomiting disease,” is a
common cause of community-acquired sporadic cases and outbreaks of acute gastroenteritis, with vomiting
and diarrhea being the characteristic symptoms.3–5 Noroviruses are highly infectious and are notorious for
causing outbreaks in semi-closed communities such as nursing homes.6 Although symptomatic infection is
generally regarded as mild, data indicate that severe disease—including dehydrating diarrhea requiring
Asten, L. van, Siebenga, J., Wijngaard, C. van den, Verheij, R., Vliet, H. van, Kretzschmar, M., Boshuizen, H.,
Pelt, W. van, Koopmans, M. Unspecified gastroenteritis illness and deaths in the elderly associated with
norovirus epidemics. Epidemiology: 2011, 22(3), 336-343
This is a NIVEL certified Post Print, more info at http://www.nivel.eu
hospitalization, and perhaps even death—occurs in vulnerable populations such as infants in developing
countries, the elderly, and the immunocompromised.7–10 The extent of morbidity and mortality posed by
norovirus infections among the elderly is unclear and probably underestimated.11 Norovirus epidemiology
has changed remarkably in recent years. In 2002, a genetic shift in circulating strains was observed with the
global emergence of the GII.4-2002 norovirus variant. This coincided with an unusually high number of
outbreaks worldwide in winter 2002–2003, preceded by off-seasonal norovirus activity in spring 200212,13
with reported outbreaks more than doubling across Europe.14 In 2004, another genetic variant (GII.4-2004
strain) emerged, followed by high norovirus activity, and in 2006, 2 novel variants emerged (GII.4- 2006a
and -2006b), which again coincided with an exceptionally high number of outbreaks starting early spring
2006.6,15,16 Information from our ongoing norovirus-outbreak surveillance captured 3 recent winters with
heavy norovirus activity and included information on genotypes. This permitted unique exploration of
norovirus outbreak activity and its association with nationwide gastroenteritis morbidity and mortality
based on medical registrations of general practitioner (GP) consultations, hospitalizations, and deaths for an
8-year period (1999–2006).
Because norovirus laboratory testing is uncommon and the health burden of norovirus infections is
unknown, a study of medical diagnoses and causes of death of unspecified gastroenteritis (ie, of unknown
causative pathogen) allowed us to estimate the proportion of unspecified gastroenteritis that is attributable
to norovirus activity, in particular for the heavy outbreak seasons of 2002–2003, 2004–2005, and 2005–
2006.
METHODS
To study the association of norovirus independent of other agents with morbidity and mortality at the
population level, we analyzed time series from 5 data sources (Table 1).
These series reflect weekly counts of selected gastrointestinal complaints or diagnoses from registries in
the Netherlands.
Because norovirus testing17 is not common in general practices or in hospitals (coding systems do not
refer specifically to norovirus infection), cases are rarely registered, even if recognized. Instead, they are
coded into a category for infectious or viral gastroenteritis (probably most labeled to be of unknown
etiology). We have grouped such categories under “unspecified gastroenteritis,” as seen in 3 of the 5 data
sources that follow.
1) Norovirus Outbreak Surveillance
Information for 13 years of norovirus outbreaks (1994– 2006) was acquired from the Dutch national
norovirus outbreak surveillance system, initiated in 1994. A cluster of two or more epidemiologically
linked cases was considered an outbreak. In this surveillance, a minimal set of data is collected for each
reported outbreak and combined with results from molecular-biologic detection and typing
techniques.6,18,19
2) Unspecified Gastroenteritis in General Practitioner Consultations
A diagnosis of a “suspected gastro-intestinal infection” (International Classification of Primary Care
_ICPC_ code D73) was defined as unspecified gastroenteritis. Vomiting and diarrhea (both common in
norovirus infection, ICPCcodes D10 and D11) were considered separately. GP consultation data came from
a representative sentinel network of GPs, the Netherlands Information Network of General Practice,
covering 2% of the Dutch population between 2001 and 2006.20,21 Virtually all Dutch citizens, including
those in homes for the elderly, are registered with a GP (ie, family physician), their first point of contact
with the Dutch health care system. Persons in nursing homes are alternatively grouped with a nursing home
GP.
3) Unspecified Gastroenteritis in Hospitalizations
A hospital discharge diagnosis (primary or secondary) of gastroenteritis designated to be of viral or of
infectious etiology but not clearly specified was defined as unspecified gastroenteritis (International
Classification of Diseases, ninth revision _ICD9_, codes 0086, 0088, 0090–0091, 0059, and 5589). We also
included discharge diagnoses registered as “Other and unspecified noninfectious gastroenteritis” (ICD9:
5589), as this code is known to consist of a large viral gastroenteritis infectious component, not registered
Asten, L. van, Siebenga, J., Wijngaard, C. van den, Verheij, R., Vliet, H. van, Kretzschmar, M., Boshuizen, H.,
Pelt, W. van, Koopmans, M. Unspecified gastroenteritis illness and deaths in the elderly associated with
norovirus epidemics. Epidemiology: 2011, 22(3), 336-343
This is a NIVEL certified Post Print, more info at http://www.nivel.eu
or recognized as such.22 Data were obtained from the Dutch National Medical Register, which covered
99% of the Dutch population (16.3 million), between 1999 and 2006.
4) Unspecified Gastroenteritis in Mortality Data
Causes of death attributed to viral gastroenteritis (A081–A089) and infectious gastroenteritis (A091–
A099, International Classification of Diseases and Related Health Problems, or ICD10) were defined as
unspecified gastroenteritis deaths, with data from 1999 to 2006. Both primary and secondary causes of
death were considered. The code A081 designates norovirus as a cause of death but was applied to only one
person during the whole study period (eAppendix 1, http://links.lww.com/EDE/A459). Our mortality
statistics originate from Statistics Netherlands and cover the total Dutch population.
5) Classic Laboratory Surveillance
From laboratory surveillance, we used positive results of diagnostic tests for infectious pathogens that can
cause gastrointestinal complaints. We used time series (1999–2006) for influenza (estimated proportion of
all diagnostics captured by national surveillance 73%), rotavirus (38%), Salmonella (64%), Campylobacter
(50%), and Shigella (100%).23 The time series are assumed representative for the Dutch population of all
ages.
Study Population
Graphic representation and analyses of trends were restricted to persons 65 years and older, except for
laboratory surveillance data and norovirus outbreak data (which do not provide data on age). However,
most norovirus outbreaks occur in nursing homes, and thus in the elderly. The gastroenteritis morbidity
pattern in the elderly differs from that in children, whose gastroenteritis hospitalization rates are dominated
by rotavirus activity.22,24
Statistical Analyses
We plotted monthly time series of unspecified gastroenteritis trends in the elderly. On weekly data, we
used Poisson regression models (which included linear and periodic components) to characterize various
trends: model (1) unspecified gastroenteritis GP consultations explained by a combination of a linear trend,
a seasonal trend, norovirus outbreak activity, and other pathogens from laboratory surveillance; model (2)
unspecified gastroenteritis hospitalizations explained by the same variables as given for model 1; and model
(3) unspecified gastroenteritis deaths explained by the same variables as given for model 1.
We used the following equation:
Unspecified gastroenteritis
t
~ Poisson(λ
t
)
λ
t =
ß
0 +
ß
1t +
ß
2
sin(2π
t
/52) + ß
3
cos(2π/52) + ß
4
P
1, (t-lagP1)
+ ß
5
P
2, (t-lagP2)
+ … + ß
m
P
k, (t-lagPk)
(See eAppendix 2 _http://links.lww.com/EDE/A459 for the definitions of all terms.)
[
TABLE 1]
[
FIGURE 1]
This statistical model is similar to time series models for other illnesses, such as hospital admissions
attributable to rotavirus activity,22,25 and deaths attributable to influenza and respiratory syncytial virus.26
We incorporated seasonal trends because many health variables show systematic variation over the course
of a year, even if these variables are not causally related.27 We used a generalized linear model with an
overdispersed Poisson-distributed error, and an identity link (with the addition of a scale parameter to take
overdispersion into account). We assumed that each weekly pathogen count was associated with a constant
number of unspecified gastroenteritis cases throughout the study period (ie, regression coefficients: ß4, ß5,
etc not varying with time).
ß0 is the estimated constant level of weekly registered gastroenteritis not explained by the variation in
pathogen trends or the seasonal trends included in the model. We did not include the population at risk in
the model, as the population increase in this period is moderate. Any increasing trend will be taken into
account by the linear term in our model (ß1).
For each parameter, we computed Wald 95% confidence intervals (CIs).
Asten, L. van, Siebenga, J., Wijngaard, C. van den, Verheij, R., Vliet, H. van, Kretzschmar, M., Boshuizen, H.,
Pelt, W. van, Koopmans, M. Unspecified gastroenteritis illness and deaths in the elderly associated with
norovirus epidemics. Epidemiology: 2011, 22(3), 336-343
This is a NIVEL certified Post Print, more info at http://www.nivel.eu
For each model, we checked for a statistically significant increasing or decreasing linear trend with time
and a significant seasonal trend (sine and cosine terms). Using a forward stepwise selection, we checked
which additional explanatory variables (norovirus outbreaks and other pathogen counts) contributed
significantly to the trends in the outcome variables (defined as unspecified gastroenteritis GP consultations,
unspecified gastroenteritis hospitalizations, or unspecified gastroenteritis mortality). We also evaluated the
association with the lagged values of norovirus outbreaks and the other pathogens (up to 4 weeks
backwards in time), building each increment in the model by adding all possible lags of all pathogens and
selecting the lag with the best fit (assessed with the deviance), until no more pathogens contributed
significantly to the model. In the final model, each pathogen, appropriately lagged, was included in the
model only once. To avoid over-modeling of the data, negative associations were not included with the
underlying consideration that pathogens can cause disease but generally do not decrease disease burden.
RESULTS
Norovirus Outbreaks on the Rise
Data on the frequency and starting dates of reported norovirus outbreaks, available since 1994, are
detailed elsewhere.
6 Outbreaks occur mostly in the winter season (October– March). The 3 recent winter seasons of 2002–
2003, 2004–2005, and 2006–2007 stand out in these time series, with well over 100 confirmed norovirus
outbreaks in each season. Although only the beginning of the 2006–2007 winter was included in our study
period, the weekly number of reported outbreaks was also relatively high (Fig. 1). The norovirus time series
did not coincide with known trends in any other monitored enteric pathogen, or with influenza (which
occasionally causes gastroenteritis symptoms) (Fig.
1). For most norovirus outbreaks (95%), the month of occurrence was known; for 72%, the week was also
available. As the missing weeks were distributed equally over the observed years, the trend is similar
whether based on monthly or weekly data (data not shown).
Trends in Unspecified Gastroenteritis Illness
(Seen at GPs and Hospitalizations)
Marked elevations of unspecified gastroenteritis among the elderly were seen in both GP consultations and
in hospitalizations, coinciding with all 3 heavy epidemic norovirus peaks (winter 2002–2003, 2004–2005,
and 2005–2006, eAppendices 3 and 4 _http://links.lww.com/EDE/A459_). The average monthly incidence
of unspecified gastroenteritis registered by sentinel physicians increased 70%–240% above average at the
height of the 3 unusual norovirus peaks (overall average: 91/100,000 elderly individuals, increasing to 150,
171, and 217, respectively) (eAppendix 3). A similar trend, although less extreme, was seen for
hospitalizations, with incidence of unspecified gastroenteritis increasing more than 30% above average
during the 3 epidemic norovirus outbreaks (the overall average monthly incidence of unspecified
gastroenteritis was 19/100,000, peaking at 24, 27, and 26 per 100,000 elderly in the epidemic seasons,
eAppendix 4).
Such peaks in unspecified morbidity were seen at no other time in unspecified gastroenteritis
hospitalizations, and at only one other time in unspecified gastroenteritis GP consultations (April 2004,
coinciding with rotavirus season).
In multivariate regression models, norovirus outbreaks were a significant predictor of both the number of
unspecified gastroenteritis GP consultations and the number of unspecified gastroenteritis hospitalizations
(see Table 2 for model parameters). In the model of unspecified gastroenteritis GP consultations, the beta
for norovirus was 0.51 (95% CI = 0.33– 0.67). With the GP-sentinel data representing 2% of the Dutch
population, we estimate this beta to be approximately 50 times higher for the total population (25.7, ie,
estimating approximately 25 unspecified gastroenteritis GP visits for every norovirus outbreak, Table 2),
thus attributing over 11,000 unspecified gastroenteritis GP visits to norovirus activity over a 5-year period
(the majority occurring in the 3 epidemic years: eg, 3804 attributed to the epidemic year of 2004–2005,
Table 3). The beta for unspecified gastroenteritis hospitalizations was 2.15 (1.57–2.74), thus attributing
215– 318 unspecified gastroenteritis hospitalizations to norovirus activity in the 3 epidemic years (Table 3).
Asten, L. van, Siebenga, J., Wijngaard, C. van den, Verheij, R., Vliet, H. van, Kretzschmar, M., Boshuizen, H.,
Pelt, W. van, Koopmans, M. Unspecified gastroenteritis illness and deaths in the elderly associated with
norovirus epidemics. Epidemiology: 2011, 22(3), 336-343
This is a NIVEL certified Post Print, more info at http://www.nivel.eu
While the unspecified gastroenteritis trend in both GP and hospital data clearly coincided with norovirus
activity, the GP consultations for vomiting or diarrhea (both symptoms that can occur with numerous
illnesses) showed no clear trends.
[
TABLE 2]
Trends in Unspecified Gastroenteritis Deaths
A total of 551 deaths with unspecified gastroenteritis as primary or secondary cause were registered in
1999–2006 in the elderly (an average monthly incidence of 2.6 deaths per million inhabitants 65 years and
older). As with unspecified gastroenteritis morbidity, unspecified gastroenteritis deaths also rose during the
3 norovirus epidemics (eAppendix 5, http://links.lww.com/EDE/A459), with the monthly rate rising to
above 6 per million during the norovirus seasons (even peaking at 8.2 per million in the 2002–2003 winter).
Unspecified gastroenteritis was coded as the primary cause of death (n = 437) more often than as secondary
cause (n = 114, eAppendix 2, http://links.lww.com/EDE/A459), but both showed the same temporal trend.
The number of deaths in the 2 subclassifications of unspecified gastroenteritis deaths (A08, “viral intestinal
infection” and A09, “gastroenteritis of presumed infectious origin”) were roughly equal, with 26% more
deaths coded in the latter (280 vs. 354 for all age groups combined) during the 8-year study period.
The peaks in mortality during the 2 norovirus epidemics did not seem to coincide with known seasons of
influenza or infections with rotavirus, Salmonella, Shigella, or Campylobacter (Fig. 1). Indeed, when
modeling-unspecified gastroenteritis deaths by laboratory counts of enteric pathogens (and correcting for a
linear and seasonal trend with time) norovirus activity (1 weeks previously) remained as the single
pathogen with significant predictive value (Fig. 2). The beta for norovirus was 0.14 (95% CI, 0.08–0.21;
Table 3), ie, one death registered as unspecified gastroenteritis for every 7 outbreaks. Of all 551 unspecified
gastroenteritis deaths, 79 in the elderly were thus statistically attributed to norovirus over the 8-year period
(Table 3 and Fig. 2). The remaining gastropathogens did not significantly improve the model, nor did
influenza. Whether these deaths were among persons with extreme comorbidity is not known, as we had no
data on comorbidity.
[
TABLE 3]
[
FIGURE 2]
DISCUSSION
Norovirus gastroenteritis is generally viewed as a trivial illness of short duration. This study, using
population-based databases, suggests that the overall public health consequences of norovirus have been
underestimated: extreme outbreak activity (in the 2002–2003, 2004–2005, and 2005–2006 winters) due to
the introduction of new strains12 coincided with sharp increases of unspecified gastroenteritis morbidity,
hospitalizations, and even deaths in the Dutch elderly population.
We know of no other published data that estimate the impact of norovirus outbreaks at various public
health levels: from milder illness (presenting to general practitioners), to severe illness (presenting at or
during hospitalizations), and deaths. That norovirus is associated not only with morbidity but with deaths in
the elderly may come as a surprise, considering that norovirus is regarded as an illness with a low case-
fatality rate.28 A few recent—mostly anecdotal—reports have suggested that norovirus infections may be
more severe in the elderly. Nursing homes in Japan,29 the United States, Israel, and the Netherlands
reported unexpected numbers of deaths during norovirus outbreaks in the absence of any influenza
activity.30–33 A study in England and Wales estimated only one death for every 50 outbreaks,9 but these
data were acquired before the worldwide emergence of the recent epidemic norovirus variants12 and
considered only persons within the outbreaks, not the total exposed and infected population.
In the Netherlands, gastroenteritis deaths suddenly increased after the striking epidemic norovirus season
of 2002– 2003.34 In total, our models estimated up to 21 yearly unspecified gastroenteritis deaths, 318
hospitalizations, and 3804 GP visits attributable to norovirus activity in the heavy 2004– 2005 norovirus
season. This number of deaths is comparable to a recent estimate from the United Kingdom.35 Although
our study suggests morbidity and mortality attributable to norovirus activity, our estimates may still
represent an underestimation of the actual situation. Illness and deaths due to infectious gastroenteritis are
poorly recognized and underreported.
36–38 Frenzen36 suggests a roughly 12-fold underestimation.