Institutional gastroenteritis outbreaks and time to notify public health services
View Article PDF
Gastroenteritis is a non-specific term indicating pathological states of the gastrointestinal tract which manifest in diarrhoea, nausea, anorexia, fever, abdominal pain and/or vomiting.1 Children under five, adults over 65, pregnant women and immunocompromised people are at increased risk of developing gastroenteritis.2,3
Gastroenteritis in infants and children is a common cause of infant mortality in developing countries.1 Gastroenteritis incidence is lower in adults compared to children. However, it is well known that old age is a risk factor for gastroenteritis associated with a risk for death.4,5 The elderly are vulnerable to gastroenteritis because of pre-existing conditions such as chronic disease, weakened immune function, malnutrition, malabsorption and communal living in a long-term care facility.6,7 In developed countries including the US, residents of the long-term care facility are four times more likely to die from gastroenteritis than those in the community.8
A gastroenteritis outbreak is defined as an increase in cases of gastroenteritis which is beyond that normally expected.9 In 2014, gastroenteritis accounted for the majority of all outbreak notifications in New Zealand (95.0%, 820/863) and 37.3% (322/820) of these outbreaks were confirmed as due to the pathogen norovirus.10 Institutional outbreaks are those confined to the population of a specific residential or other institutional setting including aged care, early childhood education (ECE) centres, hospitals and defence facilities.10 Outbreaks in facilities have constituted about half the gastroenteritis outbreaks in New Zealand every year since 2006. Since then, the outbreak numbers have continued to increase.10 In 2014, 34.9% (301/863) of gastroenteritis outbreaks in New Zealand were notified from aged care institutions.10 Individuals living in aged care institutions are more vulnerable than the general population and communal living facilitates the spread of infection.
Norovirus is the most common cause of epidemic non-bacterial gastroenteritis worldwide.11 In New Zealand, norovirus has been the most common pathogen implicated in institutional gastroenteritis since 2007.10 Either foodborne or person-to-person contamination is the most common transmission route of norovirus outbreaks. The overall attack rate in New Zealand outbreaks is approximately 40–60% of the total population exposed, but can be higher in institutional outbreaks.12
Gastroenteritis outbreaks also cause a considerable burden to the economy. This is related to staff absenteeism due to illness and additional resourcing to implement appropriate controls in outbreaks, including staffing, cleaning, investigation, treatment and laboratory costs.
Thus, it is important to manage institutional gastroenteritis outbreaks. This includes the timely identification of outbreaks, implementation of controls and accessing expert advice through notification to public health services (PHSs). Early recognition of an outbreak and rapid implementation of appropriate control measures can reduce the impact of disease. This is supported by early notification to PHSs and identification of the likely causal organism. Confirmation of the casual organism provides reassurance that the best control measures are in place and improves knowledge around best practice to prevent and manage future outbreaks. The timely identification, notification and institution of control measures have been identified as important to limiting the size and duration of gastroenteritis outbreaks. To date, the present study collaborating with a North Island PHS is the first New Zealand study that aims to explore the seasonality and trend of institutional gastroenteritis outbreaks and to quantify the association between the length of time it takes for the facility to notify the PHS and the duration and size (incidence risk) of the outbreaks.
Methods
Ethical considerations for this project were evaluated by peer review and judged to be low risk. This has been recorded on the Massey University Low Risk Database.
Anonymised data provided by the PHS from the case logs of gastroenteritis outbreaks at institutions (1 January 2009–31 December 2014) were validated and standardised. The facility types included aged care, ECE, hospitals and defence facilities which fell under PHS’ remit.13 Time to notify PHS was the length of time it takes for the facility to notify PHS, ie, days between the date of the onset of symptoms of the second case and the date when an outbreak was first notified. Duration of outbreak was approximated by the number of days between the date of the onset of symptoms of the second case and the date of the onset of symptoms of the last case of an outbreak. Population at risk was the number of residents/attendees and staff members of the facility in which each outbreak happened. Calculation of the population at risk for each outbreak occurred during investigation of the outbreak. Incidence risk (IR) was calculated by dividing the number of gastroenteritis cases with the population at risk. All data analyses were performed using R version 3.1.0.14
Investigation of seasonality and trend
The date of onset of an outbreak was taken as being the date of onset of the second case in a given outbreak. Dates of onset were aggregated to the week, month and year and plotted as a time-series. Loess smoothing was applied to emphasise the trend and seasonality and to reduce distraction from random variation.15,16 A non-parametric Spearman-test was used to test if an increasing or decreasing trend existed in the time-series plot.17 Monthly box plots and periodograms of the raw data were produced to investigate seasonality and cyclicity.18
Modelling the duration of outbreaks
The ‘nlme’ package version 3.1-11719 in R version 3.1.014 was used to build a linear mixed-effects model of the association between time to notify PHS (main covariate) and log duration of outbreak (outcome). Other covariates available for inclusion in the model were pathogen, type of facility, number of gastroenteritis cases, size of the population at risk, year and sequence of outbreaks, ie, some facilities experienced multiple outbreaks over the course of the study. To adjust for repeated outbreaks within the same facility, a random effect term for facility was fitted. Bivariate analysis, analysis of collinearity and backward selection (multivariable models) were performed to select covariates. Those associated (p≤0.20) with the outcome in bivariate analysis were included in a preliminary multivariable model. The main covariate was maintained and other covariates were removed, one by one, while those with a p≤0.05 were retained and/or if removal altered the regression coefficient (β) estimate (>20%) or SE (>20%) of the main covariate. To determine if there was any interaction with the main covariate, interaction terms were tested for significance. The Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and log likelihood were used for model selection. The Goodness-of-fit tests in the ‘nlme’ package,19 focusing on the distribution of the random effects, were used to test the model fit.
Modelling the size of outbreaks
The ‘glmmADMB’ package version 0.8.020 in R version 3.1.014 was used to build a zero-truncated, negative-binomial, mixed-effect (ZTNBME) model of the association between time to notify PHS (main covariate) and size of the outbreak (IR). The ‘epi-bohning’ test of the ‘epiR’ package version 0.9-5821 was run to investigate over-dispersion of Poisson data. The procedure of inclusion of covariates in the ZTNBME model was the same as the procedure in the LME model. The diagnostic tests in the ‘lme4’ package,22 which was integrated to ‘glmmADMB’, were used to test the model fit.
Results
In 58 facilities, 175 outbreaks (with 141 of 175 having population at risk data available) were notified: 64 outbreaks (notified within 1 day), 46 outbreaks (2–3 days), 29 outbreaks (4–6 days) and 36 outbreaks (≥7 days). In total, 4,562 cases comprising 3,077 residents, 1,316 staff and 154 visitors were involved. In the multivariable analysis, 154 visitors were excluded. The facilities included 31 aged care, 23 ECE, three hospitals and a defence facility. Summary statistics for variables are displayed in Table 1.
Table 1: Descriptive statistics of main variables.
Norovirus was the most commonly identified pathogen (108 outbreaks), followed by rotavirus (14 outbreaks) and sapovirus (9 outbreaks). Other pathogens, ie, Clostridium difficile, Campylobacter spp. and Cryptosporidium spp. were identified in three outbreaks. The pathogen(s) were unidentified in 41 outbreaks.
Gastroenteritis outbreaks were mostly notified from aged care (n=98), followed by ECE (n=50). The rest of the outbreaks were reported from hospital (n=26) and defence facility (n=1). Multiple outbreaks were notified from 36 facilities. The largest number of outbreaks per facility was 21 and these were notified from a hospital.
Investigation of seasonality and trend
No evidence of cyclicity was observed in periodograms of gastroenteritis and norovirus outbreaks (not shown). In monthly boxplots there is a suggestion of higher counts of cases and increased variability in case numbers in the spring and autumn months (Figure 1). The non-parametric bootstrapped Spearman-test gave the value (ρ) of 0.23 (p=0.013), confirming the overall increasing significant trend in number of gastroenteritis outbreaks over the study period (1 January 2009–31 December 2014), whereas the value (ρ) of norovirus outbreak was 0.14 (p=0.099) over the same study period.
Figure 1: Raw monthly boxplot of gastroenteritis outbreak data (a) and norovirus outbreak data (b).
Modelling the duration of outbreaks
In the bivariate analysis (Table 2), covariates associated with the duration of outbreaks (p≤0.20) were time to notify, pathogen, type of facility and total cases. Covariates significant at p≤0.05 in the multivariable model (Table 3) were time to notify and pathogen. Compared with the baseline time to notify (0–1 day), the duration of outbreaks was longer by 1.2 days, 1.5 days and 3.4 days when the time to notify was respectively 2–3 days, 4–6 days and ≥7 days. These estimates were adjusted for pathogen and facility.
Table 2: Bivariate analysis: putative risk factors and statistical parameters for duration of gastroenteritis outbreaks (n=175) in facilities in PHS 2009 to 2014.
coef=coefficient; ref=reference categories.
Table 3: Multivariable analysis: risk factors and statistical parameters for duration of gastroenteritis outbreaks (n=175) in facilities in PHS 2009 to 2014.
coef=coefficient; ref=reference categories; AIC=400.49; BIC=431.67; log likelihood= -190.24.
Modelling the size of outbreaks
In the bivariate analysis (Table 4), covariates associated with the log size (IR) of outbreaks (p≤0.20) were time to notify, pathogen, type of facility, duration of outbreak, year and sequence of outbreak. Covariates significant at p≤0.05 in the multivariable model (Table 5) were time to notify and pathogen. Compared with the baseline (0–1 day), the IR was larger by 1.1 times, 1.1 times and 1.6 times when the time to notify was respectively 2–3 days, 4–6 days and ≥7 days. These estimates were adjusted for pathogen and facility.
Table 4: Bivariate analysis: putative risk factors and statistical parameters for size of gastroenteritis outbreaks (n=141) in facilities in PHS 2009 to 2014.
IRR= incidence risk-ratio; ref=reference categories.
Table 5: Multivariable analysis: risk factors and statistical parameters for size of gastroenteritis outbreaks (n=141) in facilities in PHS 2009 to 2014.
IRR=incidence risk-ratio; ref=reference categories; AIC=1062.60; log likelihood= -521.29; J (negative binomial dispersion parameter)=4.27; SE=0.61.
Model diagnostics
In both models (duration and size models), the random effects were normally distributed. The absolute value of the random effects for each facility was relatively low but some facilities had larger random effects than others.
Discussion
In our analysis of institutional gastroenteritis outbreaks norovirus was the most common pathogen and aged care was the most common institutional setting. A significant increasing trend in the number of institutional gastroenteritis outbreaks was observed over the study period. The model built to quantify the association between the main covariate, time to notify the PHS, and the main outcomes, duration and size of outbreaks identified that a shorter notification time to the PHS was significantly associated with shorter duration and smaller size of outbreaks.
Norovirus was identified as the most common pathogen in 61.7% (108/175) of the institutional outbreaks in MCPHS, which is similar to the 68% (39/60) of norovirus-attributable gastroenteritis found in an Australian study.23 Norovirus was also the most common pathogen in New Zealand annual outbreak surveillance reporting (eg, 37.3% (322/863) in 2014).10 Facilities such as aged care mostly consist of frail individuals who are more vulnerable to norovirus infection, so this may explain the difference in percentages between the New Zealand annual percentage and the rate in institutions.
Aged care and ECE were the most common settings of gastroenteritis outbreaks, comprising respectively 56.6% (99/175) and 28.6% (50/175) of outbreaks. Aged care was also the most common setting in New Zealand (2006–2014), and constituted about half of gastroenteritis outbreaks annually.10 Aged care is occupied by frail individuals who are more vulnerable to infections, thus it is reasonable to expect that a higher percentage of notified gastroenteritis outbreaks is reported from aged care facilities. In a 2013/14 UK and Ireland study of norovirus outbreaks in care settings, hospital (71.3%, 383/537) and aged care (21.4%, 115/537) were the most common settings.24
The results from the time-series plots of gastroenteritis outbreaks confirmed a significant increasing trend in the number of notified outbreaks. The increasing trend was also seen nationally in New Zealand (2005–2014)10 and in a study of gastroenteritis outbreaks in hospitals in the US (1996–2007).25 The increased ageing population (accompanied by a presumed rise in the number of people in residential aged care), the funded 20 hours ECE, the introduction of national guidelines of norovirus management in hospital and aged care on 2 January 2009 and the emergence of the virulent Sydney GII.4 strain of norovirus27 could have contributed to this increasing trend.
Although the time series analysis of our study showed no statistical evidence of a seasonal pattern, a trend was seen for more outbreaks in the Spring and Autumn. Previous research in the Northwest Territories of Canada28 and US29 reported seasonal patterns in gastroenteritis outbreaks. Gastroenteritis outbreaks peaked in spring and autumn in the Canadian study from community health facilities and this was suggested to be due to environmental and social factors such as higher temperatures, frequent travelling and surface water consumption.26 Other studies performed in hospitals in the US (1996–2007)25 and (2001–2009)27 reported that gastroenteritis outbreaks mostly happened in winter. Temporal patterns are likely associated with the health-seeking behaviour peculiar to each country and some infections which are community- and not hospital-acquired. It may also reflect how the surveillance data are managed and the notification of outbreaks.
The present study reported that 36.6% (64/175) outbreaks were notified to PHS within one day after the onset of symptoms of the second case, while 20.6% (36/175) outbreaks were notified later (≥7 days). In a similar study in residential care facilities (RCFs) in Queensland, Australia in 2008, 40% (24/60) were notified to the PHS within one day, and the latest notification was 18 days.23 The range of notification was 0 to 37 days in a similar study in nursing homes in Alsace, France.6. Approximately one of four persons at risk became cases in both this and Australian studies, and one of three persons at risk became cases in the French study.6,23 The duration of each outbreak as notified to the PHS in the current study was 1–55 days while in the Australian and French study the duration were 0–42 days and 2–26 days, respectively, of which Australian and French studies each had different definition of outbreak duration.6,23 In the current study, the start of the gastroenteritis outbreaks was calculated based on the onset of symptoms of the second case recorded by the PHS.
Shorter notification time was associated with shorter duration and smaller size (IR) of gastroenteritis outbreaks. For example, after adjusting for pathogen and facility, the duration of outbreaks was 3.4 days (p=0.001, 95% CI=3.1–3.7) longer than baseline (0–1 day), when time to notify was ≥7 days. Further, there is an association between the outcome variables (duration and size of outbreaks) as displayed in Figure 3. Modelling both duration and size of outbreaks via a single composite outcome variable could be a useful next step in investigating the association between notification to the PHS and the impact of the outbreak.
The finding of an association between shorter notification time to PHS and shorter duration of outbreaks is similar to the Australian study, which reported that shorter notification time was associated with shorter duration of outbreaks. However, different from this current study, the Australian study found that shorter notification was not associated with smaller size of outbreaks.23 The number of outbreaks notified in the Australian study was smaller (n=60)23 than in this study (n=175). A lack of statistical power might explain this lack of association.
The limitation of this study is the possibility that some cases were unreported, eg, in New Zealand, aged care has a better notification system than other facility types. Aged care institutions are likely to have their own health professional staff, have systems of recording illness, and to audit outbreak management procedures. This is unlike ECEs where a child can be absent without definite reason. It is more difficult in ECEs to ascertain smaller outbreaks and cases. In ECEs children and staff members do not reside in the facilities and attendance is only during school or working hours. These facilities usually do not have their own health professional staff, and their population is larger and mobile so outbreak management procedures are more difficult to implement. Therefore, more unreported cases were expected to come from ECE. This might be attributed with the length and size of the outbreaks. In the multivariate analysis, type of facility was not significantly associated with outbreak duration and size. The characteristics of each pathogen, eg, incubation period, might also become a confounder in the duration and size of outbreak. Furthermore, not enough information about the characteristics of each facility could be obtained from the dataset, eg, whether a staff member of a facility was more skilled and experienced than other staff that could contribute to their acts responding to the outbreak. Staff experienced with managing previous outbreaks are more likely to put in effective control measures and notify the PHS earlier than those inexperienced with outbreak management.
Prompt notification to the PHS appears to be one of the factors associated with reduced outbreak duration and size. The act of notification to PHS per se will help reduce the impact of the outbreaks more effectively if the facility’s procedures for controlling outbreaks have oversight by a regulatory authority.
If a facility notifies early then the PHS is able to provide earlier access to advice and action that support the facility’s procedures for controlling outbreaks, including support with implementing the Ministry of Health (MoH) Norovirus guidelines. This PHS actions include assigning a health protection officer to the outbreak; daily oversight of the case logs and epidemic curves to monitor outbreak progress; support and advice regarding control measures and identification of likely source; procedural reviews of controls and site visits if required; identification of the pathogens (through samples collection and laboratory submission); advice to reduce further transmission of the current outbreak; and prevention of outbreaks occurring again in the future through outbreak management training workshops based on MoH guidelines.
Conclusion
The models built to quantify the association between the main explanatory factor, time to notify the PHS, and the main outcomes of interest, duration and size of outbreaks identified that a shorter notification to the PHS was significantly associated with shorter duration and smaller size of outbreaks. Future studies should consider more complex modelling of the association between time to notify the PHS, the duration and the size of the outbreak. This should be combined with an investigation of the sensitivity of the definition of the start of the outbreak. For this analysis we chose the date of onset of the second case. Identification and modelling of a composite outcome variable that captures the shape of the epidemic curve (both size and duration of outbreak) is beyond the scope of this study but an important next step in better understanding the effect of time to notify the PHS.
Better data capture, both laboratory and epidemiological (eg, clear staff role identifications and days off work due to illness) is important: the former to provide pathogen specific interventions and the latter to more clearly estimate the cost of the outbreak. Improved identification of associated cases beyond the staff and residents/attendees (for example family members of staff and residents/attendees, visitors to the institutions) will help more clearly define the extent of the burden associated with institutional outbreaks.
Summary
Abstract
Aim
We report a quantification and visualisation of the association between the time to notify public health service (PHS) and the duration and size of institutional gastroenteritis outbreaks, and explore the seasonality and trend of the outbreaks.
Method
Descriptive analysis was performed on institutional gastroenteritis outbreak data from a North Island PHS (1 January 2009-31 December 2014). Time-series analysis was used to explore the seasonality and trend of outbreaks. Multivariate analyses were performed to quantify the association between the time to notify PHS and the duration and size of outbreaks.
Results
One hundred and seventy-five gastroenteritis outbreaks (from 58 facilities) were included in descriptive analyses. A significant increasing trend (p=0.01) without seasonal pattern was confirmed by time-series analysis. Shorter notification time was associated with shorter duration and smaller size of outbreaks, eg, duration of outbreaks when time to notify was 57 days, was 3.4 days (p=0.001, 95% CI=3.1-3.7) longer than baseline time to notify (0-1 day).
Conclusion
Prompt notification to the PHS appears to be a factor associated with reduced outbreak duration and size.
Author Information
Acknowledgements
Correspondence
Correspondence Email
Competing Interests
- South Island DHBs, Gastroenteritis - South Island, South Island Alliance, Place Published, 2015.
- European Commission, Opinion of the scientific committee on veterinary measures relating to public health on Listeria monocytogenes, Health and Consumer Protection Directorate-General, Place Published, 1999.
- Ministry of Primary Industries, Making food for vulnerable populations (YOPI), Manufacturers of food and beverages, Ministry of Primary Industries, Wellington, New Zealand, 2015, pp. 1.
- Calderon-Margalit R, Sheffer R, Halperin T, Orr N, Cohen D, Shohat T. A large-scale gastroenteritis outbreak associated with Norovirus in nursing homes, Epidemiology and Infection, 2005; 133:35–40.
- Kirk MD, Roberts L, Horvath J. Understanding gastroenteritis in elderly residents of aged-care facilities, Medical Journal of Australia, 2008; 189:476–477.
- Thouillot F, Delhostal C, Edel C, Bettinger A, Pothier P, Ambert-Balay K, Meffre C, Alsibai S. Gastroenteritis outbreaks in elderly homes in the east of France during winter 2009/10: Aetiology research for a series of 37 outbreaks, Euro surveillance: bulletin Europeen sur les maladies transmissibles = European communicable disease bulletin, 2012; 17.
- Greig JD, Lee MB. Enteric outbreaks in long-term care facilities and recommendations for prevention: a review, Epidemiology and Infection, 2009; 137:145–155.
- Jones TF. When diarrhea gets deadly: A look at gastroenteritis outbreaks in nursing homes, Clinical Infectious Diseases, 2010; 51:915–916.
- ESR, Guidelines for the investigation and control of disease outbreaks, The Institute of Environmental Science and Research Limited (ESR), Place Published, 2012.
- ESR, Annual summary of outbreaks in New Zealand 2014, The Institute of Environmental Science and Research Limited (ESR), Place Published, 2015.
- Tu ETV, Bull RA, Kim M-J, McIver CJ, Heron L, Rawlinson WD, White PA. Norovirus excretion in an aged-care setting, Journal of Clinical Microbiology, 2008; 46:2119–2121.
- ESR, Human norovirus, New Zealand Food Safety Authority, Place Published, 2010.
- MidCentral District Health Board, About MidCentral District Health Board, MidCentral District Health Board, Place Published, 2014.
- R Core Team, R: A language and environment for statistical computing, R Foundation for Statistical Computing, Vienna, Austria, 2014.
- Bowman AW, Azzalini A. Applied smoothing techniques for data analysis: The kernel approach with S-Plus illustrations, Oxford University Press, Place Published, 1997.
- Cleveland WS. Robust locally weighted regression and smoothing scatterplots, Journal of the American Statistical Association, 1979; 74:829–836.
- Siegel S, Castellan Jr J. Nonparametric statistics for the behavioral sciences, McGraw-Hill, Place Published, 1988.
- Chatfield C. The analysis of time series: An introduction, CRC press, Place Published, 2013.
- Pinheiro J, Bates D, DebRoy S, Sarkar D. {nlme}: Linear and nonlinear mixed effects models, 2014.
- Fournier DA, Skaug HJ, Ancheta J, Ianelli J, Magnusson A, Maunder M, Nielsen A, Sibert J, Bolker B. {glmmADMB}: Generalized linear mixed models using automatic differentiation model builder, AD Model Builder: Using automatic differentiation for statistical inference of highly parameterized complex nonlinear models, Optim. Methods Softw, 2014.
- Stevenson M, Nunes T, Heuer C, Marshall J, Sanchez J, Thornton R, Reiczigel J, Robison-Cox J, Sebastiani P, Solymos P, Yoshida K. {epiR}: An R package for the analysis of epidemiological data, Tools for the analysis of epidemiological data, 2014.
- Bates D, Maechler M, Bolker BM, Walker S. {lme4}: Linear mixed-effects models using Eigen and S4, Fitting Linear Mixed-Effects Models using lme4, ArXiv e-print, 2015.
- Davis C, Vally H, Bell R, Sheehan F, Beard F. Viral gastrointestinal outbreaks in residential care facilities: An examination of the value of public health unit involvement, Australian and New Zealand journal of public health, 2014; 38:177–183.
- Curran ET, Wilson J, Haig CE, McCowan C, Leanord A, Loveday H. The where is norovirus control lost (WINCL) study: An enhanced surveillance project to identify norovirus index cases in care settings in the UK and Ireland, Journal of Infection Prevention, 2016; 17:8–14.
- Lopman BA, Hall AJ, Curns AT, Parashar UD. Increasing rates of gastroenteritis hospital discharges in US adults and the contribution of norovirus, 1996–2007, Clinical Infectious Diseases, 2011; 52:466–474.
- Pardhan-Ali A, Berke O, Wilson J, Edge VL, Furgal C, Reid-Smith R, Santos M, McEwen SA. A spatial and temporal analysis of notifiable gastrointestinal illness in the Northwest Territories, Canada, 1991–2008, International Journal of Health Geographics, 2012; 11.
- Gastanaduy PA, Hall AJ, Curns AT, Parashar UD, Lopman BA. Burden of norovirus gastroenteritis in the ambulatory setting-United States, 2001–2009, Journal of Infectious Diseases, 2013; 207:1058–1065.
For the PDF of this article,
contact nzmj@nzma.org.nz
Institutional gastroenteritis outbreaks and time to notify public health services
View Article PDF
Gastroenteritis is a non-specific term indicating pathological states of the gastrointestinal tract which manifest in diarrhoea, nausea, anorexia, fever, abdominal pain and/or vomiting.1 Children under five, adults over 65, pregnant women and immunocompromised people are at increased risk of developing gastroenteritis.2,3
Gastroenteritis in infants and children is a common cause of infant mortality in developing countries.1 Gastroenteritis incidence is lower in adults compared to children. However, it is well known that old age is a risk factor for gastroenteritis associated with a risk for death.4,5 The elderly are vulnerable to gastroenteritis because of pre-existing conditions such as chronic disease, weakened immune function, malnutrition, malabsorption and communal living in a long-term care facility.6,7 In developed countries including the US, residents of the long-term care facility are four times more likely to die from gastroenteritis than those in the community.8
A gastroenteritis outbreak is defined as an increase in cases of gastroenteritis which is beyond that normally expected.9 In 2014, gastroenteritis accounted for the majority of all outbreak notifications in New Zealand (95.0%, 820/863) and 37.3% (322/820) of these outbreaks were confirmed as due to the pathogen norovirus.10 Institutional outbreaks are those confined to the population of a specific residential or other institutional setting including aged care, early childhood education (ECE) centres, hospitals and defence facilities.10 Outbreaks in facilities have constituted about half the gastroenteritis outbreaks in New Zealand every year since 2006. Since then, the outbreak numbers have continued to increase.10 In 2014, 34.9% (301/863) of gastroenteritis outbreaks in New Zealand were notified from aged care institutions.10 Individuals living in aged care institutions are more vulnerable than the general population and communal living facilitates the spread of infection.
Norovirus is the most common cause of epidemic non-bacterial gastroenteritis worldwide.11 In New Zealand, norovirus has been the most common pathogen implicated in institutional gastroenteritis since 2007.10 Either foodborne or person-to-person contamination is the most common transmission route of norovirus outbreaks. The overall attack rate in New Zealand outbreaks is approximately 40–60% of the total population exposed, but can be higher in institutional outbreaks.12
Gastroenteritis outbreaks also cause a considerable burden to the economy. This is related to staff absenteeism due to illness and additional resourcing to implement appropriate controls in outbreaks, including staffing, cleaning, investigation, treatment and laboratory costs.
Thus, it is important to manage institutional gastroenteritis outbreaks. This includes the timely identification of outbreaks, implementation of controls and accessing expert advice through notification to public health services (PHSs). Early recognition of an outbreak and rapid implementation of appropriate control measures can reduce the impact of disease. This is supported by early notification to PHSs and identification of the likely causal organism. Confirmation of the casual organism provides reassurance that the best control measures are in place and improves knowledge around best practice to prevent and manage future outbreaks. The timely identification, notification and institution of control measures have been identified as important to limiting the size and duration of gastroenteritis outbreaks. To date, the present study collaborating with a North Island PHS is the first New Zealand study that aims to explore the seasonality and trend of institutional gastroenteritis outbreaks and to quantify the association between the length of time it takes for the facility to notify the PHS and the duration and size (incidence risk) of the outbreaks.
Methods
Ethical considerations for this project were evaluated by peer review and judged to be low risk. This has been recorded on the Massey University Low Risk Database.
Anonymised data provided by the PHS from the case logs of gastroenteritis outbreaks at institutions (1 January 2009–31 December 2014) were validated and standardised. The facility types included aged care, ECE, hospitals and defence facilities which fell under PHS’ remit.13 Time to notify PHS was the length of time it takes for the facility to notify PHS, ie, days between the date of the onset of symptoms of the second case and the date when an outbreak was first notified. Duration of outbreak was approximated by the number of days between the date of the onset of symptoms of the second case and the date of the onset of symptoms of the last case of an outbreak. Population at risk was the number of residents/attendees and staff members of the facility in which each outbreak happened. Calculation of the population at risk for each outbreak occurred during investigation of the outbreak. Incidence risk (IR) was calculated by dividing the number of gastroenteritis cases with the population at risk. All data analyses were performed using R version 3.1.0.14
Investigation of seasonality and trend
The date of onset of an outbreak was taken as being the date of onset of the second case in a given outbreak. Dates of onset were aggregated to the week, month and year and plotted as a time-series. Loess smoothing was applied to emphasise the trend and seasonality and to reduce distraction from random variation.15,16 A non-parametric Spearman-test was used to test if an increasing or decreasing trend existed in the time-series plot.17 Monthly box plots and periodograms of the raw data were produced to investigate seasonality and cyclicity.18
Modelling the duration of outbreaks
The ‘nlme’ package version 3.1-11719 in R version 3.1.014 was used to build a linear mixed-effects model of the association between time to notify PHS (main covariate) and log duration of outbreak (outcome). Other covariates available for inclusion in the model were pathogen, type of facility, number of gastroenteritis cases, size of the population at risk, year and sequence of outbreaks, ie, some facilities experienced multiple outbreaks over the course of the study. To adjust for repeated outbreaks within the same facility, a random effect term for facility was fitted. Bivariate analysis, analysis of collinearity and backward selection (multivariable models) were performed to select covariates. Those associated (p≤0.20) with the outcome in bivariate analysis were included in a preliminary multivariable model. The main covariate was maintained and other covariates were removed, one by one, while those with a p≤0.05 were retained and/or if removal altered the regression coefficient (β) estimate (>20%) or SE (>20%) of the main covariate. To determine if there was any interaction with the main covariate, interaction terms were tested for significance. The Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and log likelihood were used for model selection. The Goodness-of-fit tests in the ‘nlme’ package,19 focusing on the distribution of the random effects, were used to test the model fit.
Modelling the size of outbreaks
The ‘glmmADMB’ package version 0.8.020 in R version 3.1.014 was used to build a zero-truncated, negative-binomial, mixed-effect (ZTNBME) model of the association between time to notify PHS (main covariate) and size of the outbreak (IR). The ‘epi-bohning’ test of the ‘epiR’ package version 0.9-5821 was run to investigate over-dispersion of Poisson data. The procedure of inclusion of covariates in the ZTNBME model was the same as the procedure in the LME model. The diagnostic tests in the ‘lme4’ package,22 which was integrated to ‘glmmADMB’, were used to test the model fit.
Results
In 58 facilities, 175 outbreaks (with 141 of 175 having population at risk data available) were notified: 64 outbreaks (notified within 1 day), 46 outbreaks (2–3 days), 29 outbreaks (4–6 days) and 36 outbreaks (≥7 days). In total, 4,562 cases comprising 3,077 residents, 1,316 staff and 154 visitors were involved. In the multivariable analysis, 154 visitors were excluded. The facilities included 31 aged care, 23 ECE, three hospitals and a defence facility. Summary statistics for variables are displayed in Table 1.
Table 1: Descriptive statistics of main variables.
Norovirus was the most commonly identified pathogen (108 outbreaks), followed by rotavirus (14 outbreaks) and sapovirus (9 outbreaks). Other pathogens, ie, Clostridium difficile, Campylobacter spp. and Cryptosporidium spp. were identified in three outbreaks. The pathogen(s) were unidentified in 41 outbreaks.
Gastroenteritis outbreaks were mostly notified from aged care (n=98), followed by ECE (n=50). The rest of the outbreaks were reported from hospital (n=26) and defence facility (n=1). Multiple outbreaks were notified from 36 facilities. The largest number of outbreaks per facility was 21 and these were notified from a hospital.
Investigation of seasonality and trend
No evidence of cyclicity was observed in periodograms of gastroenteritis and norovirus outbreaks (not shown). In monthly boxplots there is a suggestion of higher counts of cases and increased variability in case numbers in the spring and autumn months (Figure 1). The non-parametric bootstrapped Spearman-test gave the value (ρ) of 0.23 (p=0.013), confirming the overall increasing significant trend in number of gastroenteritis outbreaks over the study period (1 January 2009–31 December 2014), whereas the value (ρ) of norovirus outbreak was 0.14 (p=0.099) over the same study period.
Figure 1: Raw monthly boxplot of gastroenteritis outbreak data (a) and norovirus outbreak data (b).
Modelling the duration of outbreaks
In the bivariate analysis (Table 2), covariates associated with the duration of outbreaks (p≤0.20) were time to notify, pathogen, type of facility and total cases. Covariates significant at p≤0.05 in the multivariable model (Table 3) were time to notify and pathogen. Compared with the baseline time to notify (0–1 day), the duration of outbreaks was longer by 1.2 days, 1.5 days and 3.4 days when the time to notify was respectively 2–3 days, 4–6 days and ≥7 days. These estimates were adjusted for pathogen and facility.
Table 2: Bivariate analysis: putative risk factors and statistical parameters for duration of gastroenteritis outbreaks (n=175) in facilities in PHS 2009 to 2014.
coef=coefficient; ref=reference categories.
Table 3: Multivariable analysis: risk factors and statistical parameters for duration of gastroenteritis outbreaks (n=175) in facilities in PHS 2009 to 2014.
coef=coefficient; ref=reference categories; AIC=400.49; BIC=431.67; log likelihood= -190.24.
Modelling the size of outbreaks
In the bivariate analysis (Table 4), covariates associated with the log size (IR) of outbreaks (p≤0.20) were time to notify, pathogen, type of facility, duration of outbreak, year and sequence of outbreak. Covariates significant at p≤0.05 in the multivariable model (Table 5) were time to notify and pathogen. Compared with the baseline (0–1 day), the IR was larger by 1.1 times, 1.1 times and 1.6 times when the time to notify was respectively 2–3 days, 4–6 days and ≥7 days. These estimates were adjusted for pathogen and facility.
Table 4: Bivariate analysis: putative risk factors and statistical parameters for size of gastroenteritis outbreaks (n=141) in facilities in PHS 2009 to 2014.
IRR= incidence risk-ratio; ref=reference categories.
Table 5: Multivariable analysis: risk factors and statistical parameters for size of gastroenteritis outbreaks (n=141) in facilities in PHS 2009 to 2014.
IRR=incidence risk-ratio; ref=reference categories; AIC=1062.60; log likelihood= -521.29; J (negative binomial dispersion parameter)=4.27; SE=0.61.
Model diagnostics
In both models (duration and size models), the random effects were normally distributed. The absolute value of the random effects for each facility was relatively low but some facilities had larger random effects than others.
Discussion
In our analysis of institutional gastroenteritis outbreaks norovirus was the most common pathogen and aged care was the most common institutional setting. A significant increasing trend in the number of institutional gastroenteritis outbreaks was observed over the study period. The model built to quantify the association between the main covariate, time to notify the PHS, and the main outcomes, duration and size of outbreaks identified that a shorter notification time to the PHS was significantly associated with shorter duration and smaller size of outbreaks.
Norovirus was identified as the most common pathogen in 61.7% (108/175) of the institutional outbreaks in MCPHS, which is similar to the 68% (39/60) of norovirus-attributable gastroenteritis found in an Australian study.23 Norovirus was also the most common pathogen in New Zealand annual outbreak surveillance reporting (eg, 37.3% (322/863) in 2014).10 Facilities such as aged care mostly consist of frail individuals who are more vulnerable to norovirus infection, so this may explain the difference in percentages between the New Zealand annual percentage and the rate in institutions.
Aged care and ECE were the most common settings of gastroenteritis outbreaks, comprising respectively 56.6% (99/175) and 28.6% (50/175) of outbreaks. Aged care was also the most common setting in New Zealand (2006–2014), and constituted about half of gastroenteritis outbreaks annually.10 Aged care is occupied by frail individuals who are more vulnerable to infections, thus it is reasonable to expect that a higher percentage of notified gastroenteritis outbreaks is reported from aged care facilities. In a 2013/14 UK and Ireland study of norovirus outbreaks in care settings, hospital (71.3%, 383/537) and aged care (21.4%, 115/537) were the most common settings.24
The results from the time-series plots of gastroenteritis outbreaks confirmed a significant increasing trend in the number of notified outbreaks. The increasing trend was also seen nationally in New Zealand (2005–2014)10 and in a study of gastroenteritis outbreaks in hospitals in the US (1996–2007).25 The increased ageing population (accompanied by a presumed rise in the number of people in residential aged care), the funded 20 hours ECE, the introduction of national guidelines of norovirus management in hospital and aged care on 2 January 2009 and the emergence of the virulent Sydney GII.4 strain of norovirus27 could have contributed to this increasing trend.
Although the time series analysis of our study showed no statistical evidence of a seasonal pattern, a trend was seen for more outbreaks in the Spring and Autumn. Previous research in the Northwest Territories of Canada28 and US29 reported seasonal patterns in gastroenteritis outbreaks. Gastroenteritis outbreaks peaked in spring and autumn in the Canadian study from community health facilities and this was suggested to be due to environmental and social factors such as higher temperatures, frequent travelling and surface water consumption.26 Other studies performed in hospitals in the US (1996–2007)25 and (2001–2009)27 reported that gastroenteritis outbreaks mostly happened in winter. Temporal patterns are likely associated with the health-seeking behaviour peculiar to each country and some infections which are community- and not hospital-acquired. It may also reflect how the surveillance data are managed and the notification of outbreaks.
The present study reported that 36.6% (64/175) outbreaks were notified to PHS within one day after the onset of symptoms of the second case, while 20.6% (36/175) outbreaks were notified later (≥7 days). In a similar study in residential care facilities (RCFs) in Queensland, Australia in 2008, 40% (24/60) were notified to the PHS within one day, and the latest notification was 18 days.23 The range of notification was 0 to 37 days in a similar study in nursing homes in Alsace, France.6. Approximately one of four persons at risk became cases in both this and Australian studies, and one of three persons at risk became cases in the French study.6,23 The duration of each outbreak as notified to the PHS in the current study was 1–55 days while in the Australian and French study the duration were 0–42 days and 2–26 days, respectively, of which Australian and French studies each had different definition of outbreak duration.6,23 In the current study, the start of the gastroenteritis outbreaks was calculated based on the onset of symptoms of the second case recorded by the PHS.
Shorter notification time was associated with shorter duration and smaller size (IR) of gastroenteritis outbreaks. For example, after adjusting for pathogen and facility, the duration of outbreaks was 3.4 days (p=0.001, 95% CI=3.1–3.7) longer than baseline (0–1 day), when time to notify was ≥7 days. Further, there is an association between the outcome variables (duration and size of outbreaks) as displayed in Figure 3. Modelling both duration and size of outbreaks via a single composite outcome variable could be a useful next step in investigating the association between notification to the PHS and the impact of the outbreak.
The finding of an association between shorter notification time to PHS and shorter duration of outbreaks is similar to the Australian study, which reported that shorter notification time was associated with shorter duration of outbreaks. However, different from this current study, the Australian study found that shorter notification was not associated with smaller size of outbreaks.23 The number of outbreaks notified in the Australian study was smaller (n=60)23 than in this study (n=175). A lack of statistical power might explain this lack of association.
The limitation of this study is the possibility that some cases were unreported, eg, in New Zealand, aged care has a better notification system than other facility types. Aged care institutions are likely to have their own health professional staff, have systems of recording illness, and to audit outbreak management procedures. This is unlike ECEs where a child can be absent without definite reason. It is more difficult in ECEs to ascertain smaller outbreaks and cases. In ECEs children and staff members do not reside in the facilities and attendance is only during school or working hours. These facilities usually do not have their own health professional staff, and their population is larger and mobile so outbreak management procedures are more difficult to implement. Therefore, more unreported cases were expected to come from ECE. This might be attributed with the length and size of the outbreaks. In the multivariate analysis, type of facility was not significantly associated with outbreak duration and size. The characteristics of each pathogen, eg, incubation period, might also become a confounder in the duration and size of outbreak. Furthermore, not enough information about the characteristics of each facility could be obtained from the dataset, eg, whether a staff member of a facility was more skilled and experienced than other staff that could contribute to their acts responding to the outbreak. Staff experienced with managing previous outbreaks are more likely to put in effective control measures and notify the PHS earlier than those inexperienced with outbreak management.
Prompt notification to the PHS appears to be one of the factors associated with reduced outbreak duration and size. The act of notification to PHS per se will help reduce the impact of the outbreaks more effectively if the facility’s procedures for controlling outbreaks have oversight by a regulatory authority.
If a facility notifies early then the PHS is able to provide earlier access to advice and action that support the facility’s procedures for controlling outbreaks, including support with implementing the Ministry of Health (MoH) Norovirus guidelines. This PHS actions include assigning a health protection officer to the outbreak; daily oversight of the case logs and epidemic curves to monitor outbreak progress; support and advice regarding control measures and identification of likely source; procedural reviews of controls and site visits if required; identification of the pathogens (through samples collection and laboratory submission); advice to reduce further transmission of the current outbreak; and prevention of outbreaks occurring again in the future through outbreak management training workshops based on MoH guidelines.
Conclusion
The models built to quantify the association between the main explanatory factor, time to notify the PHS, and the main outcomes of interest, duration and size of outbreaks identified that a shorter notification to the PHS was significantly associated with shorter duration and smaller size of outbreaks. Future studies should consider more complex modelling of the association between time to notify the PHS, the duration and the size of the outbreak. This should be combined with an investigation of the sensitivity of the definition of the start of the outbreak. For this analysis we chose the date of onset of the second case. Identification and modelling of a composite outcome variable that captures the shape of the epidemic curve (both size and duration of outbreak) is beyond the scope of this study but an important next step in better understanding the effect of time to notify the PHS.
Better data capture, both laboratory and epidemiological (eg, clear staff role identifications and days off work due to illness) is important: the former to provide pathogen specific interventions and the latter to more clearly estimate the cost of the outbreak. Improved identification of associated cases beyond the staff and residents/attendees (for example family members of staff and residents/attendees, visitors to the institutions) will help more clearly define the extent of the burden associated with institutional outbreaks.
Summary
Abstract
Aim
We report a quantification and visualisation of the association between the time to notify public health service (PHS) and the duration and size of institutional gastroenteritis outbreaks, and explore the seasonality and trend of the outbreaks.
Method
Descriptive analysis was performed on institutional gastroenteritis outbreak data from a North Island PHS (1 January 2009-31 December 2014). Time-series analysis was used to explore the seasonality and trend of outbreaks. Multivariate analyses were performed to quantify the association between the time to notify PHS and the duration and size of outbreaks.
Results
One hundred and seventy-five gastroenteritis outbreaks (from 58 facilities) were included in descriptive analyses. A significant increasing trend (p=0.01) without seasonal pattern was confirmed by time-series analysis. Shorter notification time was associated with shorter duration and smaller size of outbreaks, eg, duration of outbreaks when time to notify was 57 days, was 3.4 days (p=0.001, 95% CI=3.1-3.7) longer than baseline time to notify (0-1 day).
Conclusion
Prompt notification to the PHS appears to be a factor associated with reduced outbreak duration and size.
Author Information
Acknowledgements
Correspondence
Correspondence Email
Competing Interests
- South Island DHBs, Gastroenteritis - South Island, South Island Alliance, Place Published, 2015.
- European Commission, Opinion of the scientific committee on veterinary measures relating to public health on Listeria monocytogenes, Health and Consumer Protection Directorate-General, Place Published, 1999.
- Ministry of Primary Industries, Making food for vulnerable populations (YOPI), Manufacturers of food and beverages, Ministry of Primary Industries, Wellington, New Zealand, 2015, pp. 1.
- Calderon-Margalit R, Sheffer R, Halperin T, Orr N, Cohen D, Shohat T. A large-scale gastroenteritis outbreak associated with Norovirus in nursing homes, Epidemiology and Infection, 2005; 133:35–40.
- Kirk MD, Roberts L, Horvath J. Understanding gastroenteritis in elderly residents of aged-care facilities, Medical Journal of Australia, 2008; 189:476–477.
- Thouillot F, Delhostal C, Edel C, Bettinger A, Pothier P, Ambert-Balay K, Meffre C, Alsibai S. Gastroenteritis outbreaks in elderly homes in the east of France during winter 2009/10: Aetiology research for a series of 37 outbreaks, Euro surveillance: bulletin Europeen sur les maladies transmissibles = European communicable disease bulletin, 2012; 17.
- Greig JD, Lee MB. Enteric outbreaks in long-term care facilities and recommendations for prevention: a review, Epidemiology and Infection, 2009; 137:145–155.
- Jones TF. When diarrhea gets deadly: A look at gastroenteritis outbreaks in nursing homes, Clinical Infectious Diseases, 2010; 51:915–916.
- ESR, Guidelines for the investigation and control of disease outbreaks, The Institute of Environmental Science and Research Limited (ESR), Place Published, 2012.
- ESR, Annual summary of outbreaks in New Zealand 2014, The Institute of Environmental Science and Research Limited (ESR), Place Published, 2015.
- Tu ETV, Bull RA, Kim M-J, McIver CJ, Heron L, Rawlinson WD, White PA. Norovirus excretion in an aged-care setting, Journal of Clinical Microbiology, 2008; 46:2119–2121.
- ESR, Human norovirus, New Zealand Food Safety Authority, Place Published, 2010.
- MidCentral District Health Board, About MidCentral District Health Board, MidCentral District Health Board, Place Published, 2014.
- R Core Team, R: A language and environment for statistical computing, R Foundation for Statistical Computing, Vienna, Austria, 2014.
- Bowman AW, Azzalini A. Applied smoothing techniques for data analysis: The kernel approach with S-Plus illustrations, Oxford University Press, Place Published, 1997.
- Cleveland WS. Robust locally weighted regression and smoothing scatterplots, Journal of the American Statistical Association, 1979; 74:829–836.
- Siegel S, Castellan Jr J. Nonparametric statistics for the behavioral sciences, McGraw-Hill, Place Published, 1988.
- Chatfield C. The analysis of time series: An introduction, CRC press, Place Published, 2013.
- Pinheiro J, Bates D, DebRoy S, Sarkar D. {nlme}: Linear and nonlinear mixed effects models, 2014.
- Fournier DA, Skaug HJ, Ancheta J, Ianelli J, Magnusson A, Maunder M, Nielsen A, Sibert J, Bolker B. {glmmADMB}: Generalized linear mixed models using automatic differentiation model builder, AD Model Builder: Using automatic differentiation for statistical inference of highly parameterized complex nonlinear models, Optim. Methods Softw, 2014.
- Stevenson M, Nunes T, Heuer C, Marshall J, Sanchez J, Thornton R, Reiczigel J, Robison-Cox J, Sebastiani P, Solymos P, Yoshida K. {epiR}: An R package for the analysis of epidemiological data, Tools for the analysis of epidemiological data, 2014.
- Bates D, Maechler M, Bolker BM, Walker S. {lme4}: Linear mixed-effects models using Eigen and S4, Fitting Linear Mixed-Effects Models using lme4, ArXiv e-print, 2015.
- Davis C, Vally H, Bell R, Sheehan F, Beard F. Viral gastrointestinal outbreaks in residential care facilities: An examination of the value of public health unit involvement, Australian and New Zealand journal of public health, 2014; 38:177–183.
- Curran ET, Wilson J, Haig CE, McCowan C, Leanord A, Loveday H. The where is norovirus control lost (WINCL) study: An enhanced surveillance project to identify norovirus index cases in care settings in the UK and Ireland, Journal of Infection Prevention, 2016; 17:8–14.
- Lopman BA, Hall AJ, Curns AT, Parashar UD. Increasing rates of gastroenteritis hospital discharges in US adults and the contribution of norovirus, 1996–2007, Clinical Infectious Diseases, 2011; 52:466–474.
- Pardhan-Ali A, Berke O, Wilson J, Edge VL, Furgal C, Reid-Smith R, Santos M, McEwen SA. A spatial and temporal analysis of notifiable gastrointestinal illness in the Northwest Territories, Canada, 1991–2008, International Journal of Health Geographics, 2012; 11.
- Gastanaduy PA, Hall AJ, Curns AT, Parashar UD, Lopman BA. Burden of norovirus gastroenteritis in the ambulatory setting-United States, 2001–2009, Journal of Infectious Diseases, 2013; 207:1058–1065.
For the PDF of this article,
contact nzmj@nzma.org.nz
Institutional gastroenteritis outbreaks and time to notify public health services
View Article PDF
Gastroenteritis is a non-specific term indicating pathological states of the gastrointestinal tract which manifest in diarrhoea, nausea, anorexia, fever, abdominal pain and/or vomiting.1 Children under five, adults over 65, pregnant women and immunocompromised people are at increased risk of developing gastroenteritis.2,3
Gastroenteritis in infants and children is a common cause of infant mortality in developing countries.1 Gastroenteritis incidence is lower in adults compared to children. However, it is well known that old age is a risk factor for gastroenteritis associated with a risk for death.4,5 The elderly are vulnerable to gastroenteritis because of pre-existing conditions such as chronic disease, weakened immune function, malnutrition, malabsorption and communal living in a long-term care facility.6,7 In developed countries including the US, residents of the long-term care facility are four times more likely to die from gastroenteritis than those in the community.8
A gastroenteritis outbreak is defined as an increase in cases of gastroenteritis which is beyond that normally expected.9 In 2014, gastroenteritis accounted for the majority of all outbreak notifications in New Zealand (95.0%, 820/863) and 37.3% (322/820) of these outbreaks were confirmed as due to the pathogen norovirus.10 Institutional outbreaks are those confined to the population of a specific residential or other institutional setting including aged care, early childhood education (ECE) centres, hospitals and defence facilities.10 Outbreaks in facilities have constituted about half the gastroenteritis outbreaks in New Zealand every year since 2006. Since then, the outbreak numbers have continued to increase.10 In 2014, 34.9% (301/863) of gastroenteritis outbreaks in New Zealand were notified from aged care institutions.10 Individuals living in aged care institutions are more vulnerable than the general population and communal living facilitates the spread of infection.
Norovirus is the most common cause of epidemic non-bacterial gastroenteritis worldwide.11 In New Zealand, norovirus has been the most common pathogen implicated in institutional gastroenteritis since 2007.10 Either foodborne or person-to-person contamination is the most common transmission route of norovirus outbreaks. The overall attack rate in New Zealand outbreaks is approximately 40–60% of the total population exposed, but can be higher in institutional outbreaks.12
Gastroenteritis outbreaks also cause a considerable burden to the economy. This is related to staff absenteeism due to illness and additional resourcing to implement appropriate controls in outbreaks, including staffing, cleaning, investigation, treatment and laboratory costs.
Thus, it is important to manage institutional gastroenteritis outbreaks. This includes the timely identification of outbreaks, implementation of controls and accessing expert advice through notification to public health services (PHSs). Early recognition of an outbreak and rapid implementation of appropriate control measures can reduce the impact of disease. This is supported by early notification to PHSs and identification of the likely causal organism. Confirmation of the casual organism provides reassurance that the best control measures are in place and improves knowledge around best practice to prevent and manage future outbreaks. The timely identification, notification and institution of control measures have been identified as important to limiting the size and duration of gastroenteritis outbreaks. To date, the present study collaborating with a North Island PHS is the first New Zealand study that aims to explore the seasonality and trend of institutional gastroenteritis outbreaks and to quantify the association between the length of time it takes for the facility to notify the PHS and the duration and size (incidence risk) of the outbreaks.
Methods
Ethical considerations for this project were evaluated by peer review and judged to be low risk. This has been recorded on the Massey University Low Risk Database.
Anonymised data provided by the PHS from the case logs of gastroenteritis outbreaks at institutions (1 January 2009–31 December 2014) were validated and standardised. The facility types included aged care, ECE, hospitals and defence facilities which fell under PHS’ remit.13 Time to notify PHS was the length of time it takes for the facility to notify PHS, ie, days between the date of the onset of symptoms of the second case and the date when an outbreak was first notified. Duration of outbreak was approximated by the number of days between the date of the onset of symptoms of the second case and the date of the onset of symptoms of the last case of an outbreak. Population at risk was the number of residents/attendees and staff members of the facility in which each outbreak happened. Calculation of the population at risk for each outbreak occurred during investigation of the outbreak. Incidence risk (IR) was calculated by dividing the number of gastroenteritis cases with the population at risk. All data analyses were performed using R version 3.1.0.14
Investigation of seasonality and trend
The date of onset of an outbreak was taken as being the date of onset of the second case in a given outbreak. Dates of onset were aggregated to the week, month and year and plotted as a time-series. Loess smoothing was applied to emphasise the trend and seasonality and to reduce distraction from random variation.15,16 A non-parametric Spearman-test was used to test if an increasing or decreasing trend existed in the time-series plot.17 Monthly box plots and periodograms of the raw data were produced to investigate seasonality and cyclicity.18
Modelling the duration of outbreaks
The ‘nlme’ package version 3.1-11719 in R version 3.1.014 was used to build a linear mixed-effects model of the association between time to notify PHS (main covariate) and log duration of outbreak (outcome). Other covariates available for inclusion in the model were pathogen, type of facility, number of gastroenteritis cases, size of the population at risk, year and sequence of outbreaks, ie, some facilities experienced multiple outbreaks over the course of the study. To adjust for repeated outbreaks within the same facility, a random effect term for facility was fitted. Bivariate analysis, analysis of collinearity and backward selection (multivariable models) were performed to select covariates. Those associated (p≤0.20) with the outcome in bivariate analysis were included in a preliminary multivariable model. The main covariate was maintained and other covariates were removed, one by one, while those with a p≤0.05 were retained and/or if removal altered the regression coefficient (β) estimate (>20%) or SE (>20%) of the main covariate. To determine if there was any interaction with the main covariate, interaction terms were tested for significance. The Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and log likelihood were used for model selection. The Goodness-of-fit tests in the ‘nlme’ package,19 focusing on the distribution of the random effects, were used to test the model fit.
Modelling the size of outbreaks
The ‘glmmADMB’ package version 0.8.020 in R version 3.1.014 was used to build a zero-truncated, negative-binomial, mixed-effect (ZTNBME) model of the association between time to notify PHS (main covariate) and size of the outbreak (IR). The ‘epi-bohning’ test of the ‘epiR’ package version 0.9-5821 was run to investigate over-dispersion of Poisson data. The procedure of inclusion of covariates in the ZTNBME model was the same as the procedure in the LME model. The diagnostic tests in the ‘lme4’ package,22 which was integrated to ‘glmmADMB’, were used to test the model fit.
Results
In 58 facilities, 175 outbreaks (with 141 of 175 having population at risk data available) were notified: 64 outbreaks (notified within 1 day), 46 outbreaks (2–3 days), 29 outbreaks (4–6 days) and 36 outbreaks (≥7 days). In total, 4,562 cases comprising 3,077 residents, 1,316 staff and 154 visitors were involved. In the multivariable analysis, 154 visitors were excluded. The facilities included 31 aged care, 23 ECE, three hospitals and a defence facility. Summary statistics for variables are displayed in Table 1.
Table 1: Descriptive statistics of main variables.
Norovirus was the most commonly identified pathogen (108 outbreaks), followed by rotavirus (14 outbreaks) and sapovirus (9 outbreaks). Other pathogens, ie, Clostridium difficile, Campylobacter spp. and Cryptosporidium spp. were identified in three outbreaks. The pathogen(s) were unidentified in 41 outbreaks.
Gastroenteritis outbreaks were mostly notified from aged care (n=98), followed by ECE (n=50). The rest of the outbreaks were reported from hospital (n=26) and defence facility (n=1). Multiple outbreaks were notified from 36 facilities. The largest number of outbreaks per facility was 21 and these were notified from a hospital.
Investigation of seasonality and trend
No evidence of cyclicity was observed in periodograms of gastroenteritis and norovirus outbreaks (not shown). In monthly boxplots there is a suggestion of higher counts of cases and increased variability in case numbers in the spring and autumn months (Figure 1). The non-parametric bootstrapped Spearman-test gave the value (ρ) of 0.23 (p=0.013), confirming the overall increasing significant trend in number of gastroenteritis outbreaks over the study period (1 January 2009–31 December 2014), whereas the value (ρ) of norovirus outbreak was 0.14 (p=0.099) over the same study period.
Figure 1: Raw monthly boxplot of gastroenteritis outbreak data (a) and norovirus outbreak data (b).
Modelling the duration of outbreaks
In the bivariate analysis (Table 2), covariates associated with the duration of outbreaks (p≤0.20) were time to notify, pathogen, type of facility and total cases. Covariates significant at p≤0.05 in the multivariable model (Table 3) were time to notify and pathogen. Compared with the baseline time to notify (0–1 day), the duration of outbreaks was longer by 1.2 days, 1.5 days and 3.4 days when the time to notify was respectively 2–3 days, 4–6 days and ≥7 days. These estimates were adjusted for pathogen and facility.
Table 2: Bivariate analysis: putative risk factors and statistical parameters for duration of gastroenteritis outbreaks (n=175) in facilities in PHS 2009 to 2014.
coef=coefficient; ref=reference categories.
Table 3: Multivariable analysis: risk factors and statistical parameters for duration of gastroenteritis outbreaks (n=175) in facilities in PHS 2009 to 2014.
coef=coefficient; ref=reference categories; AIC=400.49; BIC=431.67; log likelihood= -190.24.
Modelling the size of outbreaks
In the bivariate analysis (Table 4), covariates associated with the log size (IR) of outbreaks (p≤0.20) were time to notify, pathogen, type of facility, duration of outbreak, year and sequence of outbreak. Covariates significant at p≤0.05 in the multivariable model (Table 5) were time to notify and pathogen. Compared with the baseline (0–1 day), the IR was larger by 1.1 times, 1.1 times and 1.6 times when the time to notify was respectively 2–3 days, 4–6 days and ≥7 days. These estimates were adjusted for pathogen and facility.
Table 4: Bivariate analysis: putative risk factors and statistical parameters for size of gastroenteritis outbreaks (n=141) in facilities in PHS 2009 to 2014.
IRR= incidence risk-ratio; ref=reference categories.
Table 5: Multivariable analysis: risk factors and statistical parameters for size of gastroenteritis outbreaks (n=141) in facilities in PHS 2009 to 2014.
IRR=incidence risk-ratio; ref=reference categories; AIC=1062.60; log likelihood= -521.29; J (negative binomial dispersion parameter)=4.27; SE=0.61.
Model diagnostics
In both models (duration and size models), the random effects were normally distributed. The absolute value of the random effects for each facility was relatively low but some facilities had larger random effects than others.
Discussion
In our analysis of institutional gastroenteritis outbreaks norovirus was the most common pathogen and aged care was the most common institutional setting. A significant increasing trend in the number of institutional gastroenteritis outbreaks was observed over the study period. The model built to quantify the association between the main covariate, time to notify the PHS, and the main outcomes, duration and size of outbreaks identified that a shorter notification time to the PHS was significantly associated with shorter duration and smaller size of outbreaks.
Norovirus was identified as the most common pathogen in 61.7% (108/175) of the institutional outbreaks in MCPHS, which is similar to the 68% (39/60) of norovirus-attributable gastroenteritis found in an Australian study.23 Norovirus was also the most common pathogen in New Zealand annual outbreak surveillance reporting (eg, 37.3% (322/863) in 2014).10 Facilities such as aged care mostly consist of frail individuals who are more vulnerable to norovirus infection, so this may explain the difference in percentages between the New Zealand annual percentage and the rate in institutions.
Aged care and ECE were the most common settings of gastroenteritis outbreaks, comprising respectively 56.6% (99/175) and 28.6% (50/175) of outbreaks. Aged care was also the most common setting in New Zealand (2006–2014), and constituted about half of gastroenteritis outbreaks annually.10 Aged care is occupied by frail individuals who are more vulnerable to infections, thus it is reasonable to expect that a higher percentage of notified gastroenteritis outbreaks is reported from aged care facilities. In a 2013/14 UK and Ireland study of norovirus outbreaks in care settings, hospital (71.3%, 383/537) and aged care (21.4%, 115/537) were the most common settings.24
The results from the time-series plots of gastroenteritis outbreaks confirmed a significant increasing trend in the number of notified outbreaks. The increasing trend was also seen nationally in New Zealand (2005–2014)10 and in a study of gastroenteritis outbreaks in hospitals in the US (1996–2007).25 The increased ageing population (accompanied by a presumed rise in the number of people in residential aged care), the funded 20 hours ECE, the introduction of national guidelines of norovirus management in hospital and aged care on 2 January 2009 and the emergence of the virulent Sydney GII.4 strain of norovirus27 could have contributed to this increasing trend.
Although the time series analysis of our study showed no statistical evidence of a seasonal pattern, a trend was seen for more outbreaks in the Spring and Autumn. Previous research in the Northwest Territories of Canada28 and US29 reported seasonal patterns in gastroenteritis outbreaks. Gastroenteritis outbreaks peaked in spring and autumn in the Canadian study from community health facilities and this was suggested to be due to environmental and social factors such as higher temperatures, frequent travelling and surface water consumption.26 Other studies performed in hospitals in the US (1996–2007)25 and (2001–2009)27 reported that gastroenteritis outbreaks mostly happened in winter. Temporal patterns are likely associated with the health-seeking behaviour peculiar to each country and some infections which are community- and not hospital-acquired. It may also reflect how the surveillance data are managed and the notification of outbreaks.
The present study reported that 36.6% (64/175) outbreaks were notified to PHS within one day after the onset of symptoms of the second case, while 20.6% (36/175) outbreaks were notified later (≥7 days). In a similar study in residential care facilities (RCFs) in Queensland, Australia in 2008, 40% (24/60) were notified to the PHS within one day, and the latest notification was 18 days.23 The range of notification was 0 to 37 days in a similar study in nursing homes in Alsace, France.6. Approximately one of four persons at risk became cases in both this and Australian studies, and one of three persons at risk became cases in the French study.6,23 The duration of each outbreak as notified to the PHS in the current study was 1–55 days while in the Australian and French study the duration were 0–42 days and 2–26 days, respectively, of which Australian and French studies each had different definition of outbreak duration.6,23 In the current study, the start of the gastroenteritis outbreaks was calculated based on the onset of symptoms of the second case recorded by the PHS.
Shorter notification time was associated with shorter duration and smaller size (IR) of gastroenteritis outbreaks. For example, after adjusting for pathogen and facility, the duration of outbreaks was 3.4 days (p=0.001, 95% CI=3.1–3.7) longer than baseline (0–1 day), when time to notify was ≥7 days. Further, there is an association between the outcome variables (duration and size of outbreaks) as displayed in Figure 3. Modelling both duration and size of outbreaks via a single composite outcome variable could be a useful next step in investigating the association between notification to the PHS and the impact of the outbreak.
The finding of an association between shorter notification time to PHS and shorter duration of outbreaks is similar to the Australian study, which reported that shorter notification time was associated with shorter duration of outbreaks. However, different from this current study, the Australian study found that shorter notification was not associated with smaller size of outbreaks.23 The number of outbreaks notified in the Australian study was smaller (n=60)23 than in this study (n=175). A lack of statistical power might explain this lack of association.
The limitation of this study is the possibility that some cases were unreported, eg, in New Zealand, aged care has a better notification system than other facility types. Aged care institutions are likely to have their own health professional staff, have systems of recording illness, and to audit outbreak management procedures. This is unlike ECEs where a child can be absent without definite reason. It is more difficult in ECEs to ascertain smaller outbreaks and cases. In ECEs children and staff members do not reside in the facilities and attendance is only during school or working hours. These facilities usually do not have their own health professional staff, and their population is larger and mobile so outbreak management procedures are more difficult to implement. Therefore, more unreported cases were expected to come from ECE. This might be attributed with the length and size of the outbreaks. In the multivariate analysis, type of facility was not significantly associated with outbreak duration and size. The characteristics of each pathogen, eg, incubation period, might also become a confounder in the duration and size of outbreak. Furthermore, not enough information about the characteristics of each facility could be obtained from the dataset, eg, whether a staff member of a facility was more skilled and experienced than other staff that could contribute to their acts responding to the outbreak. Staff experienced with managing previous outbreaks are more likely to put in effective control measures and notify the PHS earlier than those inexperienced with outbreak management.
Prompt notification to the PHS appears to be one of the factors associated with reduced outbreak duration and size. The act of notification to PHS per se will help reduce the impact of the outbreaks more effectively if the facility’s procedures for controlling outbreaks have oversight by a regulatory authority.
If a facility notifies early then the PHS is able to provide earlier access to advice and action that support the facility’s procedures for controlling outbreaks, including support with implementing the Ministry of Health (MoH) Norovirus guidelines. This PHS actions include assigning a health protection officer to the outbreak; daily oversight of the case logs and epidemic curves to monitor outbreak progress; support and advice regarding control measures and identification of likely source; procedural reviews of controls and site visits if required; identification of the pathogens (through samples collection and laboratory submission); advice to reduce further transmission of the current outbreak; and prevention of outbreaks occurring again in the future through outbreak management training workshops based on MoH guidelines.
Conclusion
The models built to quantify the association between the main explanatory factor, time to notify the PHS, and the main outcomes of interest, duration and size of outbreaks identified that a shorter notification to the PHS was significantly associated with shorter duration and smaller size of outbreaks. Future studies should consider more complex modelling of the association between time to notify the PHS, the duration and the size of the outbreak. This should be combined with an investigation of the sensitivity of the definition of the start of the outbreak. For this analysis we chose the date of onset of the second case. Identification and modelling of a composite outcome variable that captures the shape of the epidemic curve (both size and duration of outbreak) is beyond the scope of this study but an important next step in better understanding the effect of time to notify the PHS.
Better data capture, both laboratory and epidemiological (eg, clear staff role identifications and days off work due to illness) is important: the former to provide pathogen specific interventions and the latter to more clearly estimate the cost of the outbreak. Improved identification of associated cases beyond the staff and residents/attendees (for example family members of staff and residents/attendees, visitors to the institutions) will help more clearly define the extent of the burden associated with institutional outbreaks.
Summary
Abstract
Aim
We report a quantification and visualisation of the association between the time to notify public health service (PHS) and the duration and size of institutional gastroenteritis outbreaks, and explore the seasonality and trend of the outbreaks.
Method
Descriptive analysis was performed on institutional gastroenteritis outbreak data from a North Island PHS (1 January 2009-31 December 2014). Time-series analysis was used to explore the seasonality and trend of outbreaks. Multivariate analyses were performed to quantify the association between the time to notify PHS and the duration and size of outbreaks.
Results
One hundred and seventy-five gastroenteritis outbreaks (from 58 facilities) were included in descriptive analyses. A significant increasing trend (p=0.01) without seasonal pattern was confirmed by time-series analysis. Shorter notification time was associated with shorter duration and smaller size of outbreaks, eg, duration of outbreaks when time to notify was 57 days, was 3.4 days (p=0.001, 95% CI=3.1-3.7) longer than baseline time to notify (0-1 day).
Conclusion
Prompt notification to the PHS appears to be a factor associated with reduced outbreak duration and size.
Author Information
Acknowledgements
Correspondence
Correspondence Email
Competing Interests
- South Island DHBs, Gastroenteritis - South Island, South Island Alliance, Place Published, 2015.
- European Commission, Opinion of the scientific committee on veterinary measures relating to public health on Listeria monocytogenes, Health and Consumer Protection Directorate-General, Place Published, 1999.
- Ministry of Primary Industries, Making food for vulnerable populations (YOPI), Manufacturers of food and beverages, Ministry of Primary Industries, Wellington, New Zealand, 2015, pp. 1.
- Calderon-Margalit R, Sheffer R, Halperin T, Orr N, Cohen D, Shohat T. A large-scale gastroenteritis outbreak associated with Norovirus in nursing homes, Epidemiology and Infection, 2005; 133:35–40.
- Kirk MD, Roberts L, Horvath J. Understanding gastroenteritis in elderly residents of aged-care facilities, Medical Journal of Australia, 2008; 189:476–477.
- Thouillot F, Delhostal C, Edel C, Bettinger A, Pothier P, Ambert-Balay K, Meffre C, Alsibai S. Gastroenteritis outbreaks in elderly homes in the east of France during winter 2009/10: Aetiology research for a series of 37 outbreaks, Euro surveillance: bulletin Europeen sur les maladies transmissibles = European communicable disease bulletin, 2012; 17.
- Greig JD, Lee MB. Enteric outbreaks in long-term care facilities and recommendations for prevention: a review, Epidemiology and Infection, 2009; 137:145–155.
- Jones TF. When diarrhea gets deadly: A look at gastroenteritis outbreaks in nursing homes, Clinical Infectious Diseases, 2010; 51:915–916.
- ESR, Guidelines for the investigation and control of disease outbreaks, The Institute of Environmental Science and Research Limited (ESR), Place Published, 2012.
- ESR, Annual summary of outbreaks in New Zealand 2014, The Institute of Environmental Science and Research Limited (ESR), Place Published, 2015.
- Tu ETV, Bull RA, Kim M-J, McIver CJ, Heron L, Rawlinson WD, White PA. Norovirus excretion in an aged-care setting, Journal of Clinical Microbiology, 2008; 46:2119–2121.
- ESR, Human norovirus, New Zealand Food Safety Authority, Place Published, 2010.
- MidCentral District Health Board, About MidCentral District Health Board, MidCentral District Health Board, Place Published, 2014.
- R Core Team, R: A language and environment for statistical computing, R Foundation for Statistical Computing, Vienna, Austria, 2014.
- Bowman AW, Azzalini A. Applied smoothing techniques for data analysis: The kernel approach with S-Plus illustrations, Oxford University Press, Place Published, 1997.
- Cleveland WS. Robust locally weighted regression and smoothing scatterplots, Journal of the American Statistical Association, 1979; 74:829–836.
- Siegel S, Castellan Jr J. Nonparametric statistics for the behavioral sciences, McGraw-Hill, Place Published, 1988.
- Chatfield C. The analysis of time series: An introduction, CRC press, Place Published, 2013.
- Pinheiro J, Bates D, DebRoy S, Sarkar D. {nlme}: Linear and nonlinear mixed effects models, 2014.
- Fournier DA, Skaug HJ, Ancheta J, Ianelli J, Magnusson A, Maunder M, Nielsen A, Sibert J, Bolker B. {glmmADMB}: Generalized linear mixed models using automatic differentiation model builder, AD Model Builder: Using automatic differentiation for statistical inference of highly parameterized complex nonlinear models, Optim. Methods Softw, 2014.
- Stevenson M, Nunes T, Heuer C, Marshall J, Sanchez J, Thornton R, Reiczigel J, Robison-Cox J, Sebastiani P, Solymos P, Yoshida K. {epiR}: An R package for the analysis of epidemiological data, Tools for the analysis of epidemiological data, 2014.
- Bates D, Maechler M, Bolker BM, Walker S. {lme4}: Linear mixed-effects models using Eigen and S4, Fitting Linear Mixed-Effects Models using lme4, ArXiv e-print, 2015.
- Davis C, Vally H, Bell R, Sheehan F, Beard F. Viral gastrointestinal outbreaks in residential care facilities: An examination of the value of public health unit involvement, Australian and New Zealand journal of public health, 2014; 38:177–183.
- Curran ET, Wilson J, Haig CE, McCowan C, Leanord A, Loveday H. The where is norovirus control lost (WINCL) study: An enhanced surveillance project to identify norovirus index cases in care settings in the UK and Ireland, Journal of Infection Prevention, 2016; 17:8–14.
- Lopman BA, Hall AJ, Curns AT, Parashar UD. Increasing rates of gastroenteritis hospital discharges in US adults and the contribution of norovirus, 1996–2007, Clinical Infectious Diseases, 2011; 52:466–474.
- Pardhan-Ali A, Berke O, Wilson J, Edge VL, Furgal C, Reid-Smith R, Santos M, McEwen SA. A spatial and temporal analysis of notifiable gastrointestinal illness in the Northwest Territories, Canada, 1991–2008, International Journal of Health Geographics, 2012; 11.
- Gastanaduy PA, Hall AJ, Curns AT, Parashar UD, Lopman BA. Burden of norovirus gastroenteritis in the ambulatory setting-United States, 2001–2009, Journal of Infectious Diseases, 2013; 207:1058–1065.
Contact diana@nzma.org.nz
for the PDF of this article
Institutional gastroenteritis outbreaks and time to notify public health services
View Article PDF
Gastroenteritis is a non-specific term indicating pathological states of the gastrointestinal tract which manifest in diarrhoea, nausea, anorexia, fever, abdominal pain and/or vomiting.1 Children under five, adults over 65, pregnant women and immunocompromised people are at increased risk of developing gastroenteritis.2,3
Gastroenteritis in infants and children is a common cause of infant mortality in developing countries.1 Gastroenteritis incidence is lower in adults compared to children. However, it is well known that old age is a risk factor for gastroenteritis associated with a risk for death.4,5 The elderly are vulnerable to gastroenteritis because of pre-existing conditions such as chronic disease, weakened immune function, malnutrition, malabsorption and communal living in a long-term care facility.6,7 In developed countries including the US, residents of the long-term care facility are four times more likely to die from gastroenteritis than those in the community.8
A gastroenteritis outbreak is defined as an increase in cases of gastroenteritis which is beyond that normally expected.9 In 2014, gastroenteritis accounted for the majority of all outbreak notifications in New Zealand (95.0%, 820/863) and 37.3% (322/820) of these outbreaks were confirmed as due to the pathogen norovirus.10 Institutional outbreaks are those confined to the population of a specific residential or other institutional setting including aged care, early childhood education (ECE) centres, hospitals and defence facilities.10 Outbreaks in facilities have constituted about half the gastroenteritis outbreaks in New Zealand every year since 2006. Since then, the outbreak numbers have continued to increase.10 In 2014, 34.9% (301/863) of gastroenteritis outbreaks in New Zealand were notified from aged care institutions.10 Individuals living in aged care institutions are more vulnerable than the general population and communal living facilitates the spread of infection.
Norovirus is the most common cause of epidemic non-bacterial gastroenteritis worldwide.11 In New Zealand, norovirus has been the most common pathogen implicated in institutional gastroenteritis since 2007.10 Either foodborne or person-to-person contamination is the most common transmission route of norovirus outbreaks. The overall attack rate in New Zealand outbreaks is approximately 40–60% of the total population exposed, but can be higher in institutional outbreaks.12
Gastroenteritis outbreaks also cause a considerable burden to the economy. This is related to staff absenteeism due to illness and additional resourcing to implement appropriate controls in outbreaks, including staffing, cleaning, investigation, treatment and laboratory costs.
Thus, it is important to manage institutional gastroenteritis outbreaks. This includes the timely identification of outbreaks, implementation of controls and accessing expert advice through notification to public health services (PHSs). Early recognition of an outbreak and rapid implementation of appropriate control measures can reduce the impact of disease. This is supported by early notification to PHSs and identification of the likely causal organism. Confirmation of the casual organism provides reassurance that the best control measures are in place and improves knowledge around best practice to prevent and manage future outbreaks. The timely identification, notification and institution of control measures have been identified as important to limiting the size and duration of gastroenteritis outbreaks. To date, the present study collaborating with a North Island PHS is the first New Zealand study that aims to explore the seasonality and trend of institutional gastroenteritis outbreaks and to quantify the association between the length of time it takes for the facility to notify the PHS and the duration and size (incidence risk) of the outbreaks.
Methods
Ethical considerations for this project were evaluated by peer review and judged to be low risk. This has been recorded on the Massey University Low Risk Database.
Anonymised data provided by the PHS from the case logs of gastroenteritis outbreaks at institutions (1 January 2009–31 December 2014) were validated and standardised. The facility types included aged care, ECE, hospitals and defence facilities which fell under PHS’ remit.13 Time to notify PHS was the length of time it takes for the facility to notify PHS, ie, days between the date of the onset of symptoms of the second case and the date when an outbreak was first notified. Duration of outbreak was approximated by the number of days between the date of the onset of symptoms of the second case and the date of the onset of symptoms of the last case of an outbreak. Population at risk was the number of residents/attendees and staff members of the facility in which each outbreak happened. Calculation of the population at risk for each outbreak occurred during investigation of the outbreak. Incidence risk (IR) was calculated by dividing the number of gastroenteritis cases with the population at risk. All data analyses were performed using R version 3.1.0.14
Investigation of seasonality and trend
The date of onset of an outbreak was taken as being the date of onset of the second case in a given outbreak. Dates of onset were aggregated to the week, month and year and plotted as a time-series. Loess smoothing was applied to emphasise the trend and seasonality and to reduce distraction from random variation.15,16 A non-parametric Spearman-test was used to test if an increasing or decreasing trend existed in the time-series plot.17 Monthly box plots and periodograms of the raw data were produced to investigate seasonality and cyclicity.18
Modelling the duration of outbreaks
The ‘nlme’ package version 3.1-11719 in R version 3.1.014 was used to build a linear mixed-effects model of the association between time to notify PHS (main covariate) and log duration of outbreak (outcome). Other covariates available for inclusion in the model were pathogen, type of facility, number of gastroenteritis cases, size of the population at risk, year and sequence of outbreaks, ie, some facilities experienced multiple outbreaks over the course of the study. To adjust for repeated outbreaks within the same facility, a random effect term for facility was fitted. Bivariate analysis, analysis of collinearity and backward selection (multivariable models) were performed to select covariates. Those associated (p≤0.20) with the outcome in bivariate analysis were included in a preliminary multivariable model. The main covariate was maintained and other covariates were removed, one by one, while those with a p≤0.05 were retained and/or if removal altered the regression coefficient (β) estimate (>20%) or SE (>20%) of the main covariate. To determine if there was any interaction with the main covariate, interaction terms were tested for significance. The Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and log likelihood were used for model selection. The Goodness-of-fit tests in the ‘nlme’ package,19 focusing on the distribution of the random effects, were used to test the model fit.
Modelling the size of outbreaks
The ‘glmmADMB’ package version 0.8.020 in R version 3.1.014 was used to build a zero-truncated, negative-binomial, mixed-effect (ZTNBME) model of the association between time to notify PHS (main covariate) and size of the outbreak (IR). The ‘epi-bohning’ test of the ‘epiR’ package version 0.9-5821 was run to investigate over-dispersion of Poisson data. The procedure of inclusion of covariates in the ZTNBME model was the same as the procedure in the LME model. The diagnostic tests in the ‘lme4’ package,22 which was integrated to ‘glmmADMB’, were used to test the model fit.
Results
In 58 facilities, 175 outbreaks (with 141 of 175 having population at risk data available) were notified: 64 outbreaks (notified within 1 day), 46 outbreaks (2–3 days), 29 outbreaks (4–6 days) and 36 outbreaks (≥7 days). In total, 4,562 cases comprising 3,077 residents, 1,316 staff and 154 visitors were involved. In the multivariable analysis, 154 visitors were excluded. The facilities included 31 aged care, 23 ECE, three hospitals and a defence facility. Summary statistics for variables are displayed in Table 1.
Table 1: Descriptive statistics of main variables.
Norovirus was the most commonly identified pathogen (108 outbreaks), followed by rotavirus (14 outbreaks) and sapovirus (9 outbreaks). Other pathogens, ie, Clostridium difficile, Campylobacter spp. and Cryptosporidium spp. were identified in three outbreaks. The pathogen(s) were unidentified in 41 outbreaks.
Gastroenteritis outbreaks were mostly notified from aged care (n=98), followed by ECE (n=50). The rest of the outbreaks were reported from hospital (n=26) and defence facility (n=1). Multiple outbreaks were notified from 36 facilities. The largest number of outbreaks per facility was 21 and these were notified from a hospital.
Investigation of seasonality and trend
No evidence of cyclicity was observed in periodograms of gastroenteritis and norovirus outbreaks (not shown). In monthly boxplots there is a suggestion of higher counts of cases and increased variability in case numbers in the spring and autumn months (Figure 1). The non-parametric bootstrapped Spearman-test gave the value (ρ) of 0.23 (p=0.013), confirming the overall increasing significant trend in number of gastroenteritis outbreaks over the study period (1 January 2009–31 December 2014), whereas the value (ρ) of norovirus outbreak was 0.14 (p=0.099) over the same study period.
Figure 1: Raw monthly boxplot of gastroenteritis outbreak data (a) and norovirus outbreak data (b).
Modelling the duration of outbreaks
In the bivariate analysis (Table 2), covariates associated with the duration of outbreaks (p≤0.20) were time to notify, pathogen, type of facility and total cases. Covariates significant at p≤0.05 in the multivariable model (Table 3) were time to notify and pathogen. Compared with the baseline time to notify (0–1 day), the duration of outbreaks was longer by 1.2 days, 1.5 days and 3.4 days when the time to notify was respectively 2–3 days, 4–6 days and ≥7 days. These estimates were adjusted for pathogen and facility.
Table 2: Bivariate analysis: putative risk factors and statistical parameters for duration of gastroenteritis outbreaks (n=175) in facilities in PHS 2009 to 2014.
coef=coefficient; ref=reference categories.
Table 3: Multivariable analysis: risk factors and statistical parameters for duration of gastroenteritis outbreaks (n=175) in facilities in PHS 2009 to 2014.
coef=coefficient; ref=reference categories; AIC=400.49; BIC=431.67; log likelihood= -190.24.
Modelling the size of outbreaks
In the bivariate analysis (Table 4), covariates associated with the log size (IR) of outbreaks (p≤0.20) were time to notify, pathogen, type of facility, duration of outbreak, year and sequence of outbreak. Covariates significant at p≤0.05 in the multivariable model (Table 5) were time to notify and pathogen. Compared with the baseline (0–1 day), the IR was larger by 1.1 times, 1.1 times and 1.6 times when the time to notify was respectively 2–3 days, 4–6 days and ≥7 days. These estimates were adjusted for pathogen and facility.
Table 4: Bivariate analysis: putative risk factors and statistical parameters for size of gastroenteritis outbreaks (n=141) in facilities in PHS 2009 to 2014.
IRR= incidence risk-ratio; ref=reference categories.
Table 5: Multivariable analysis: risk factors and statistical parameters for size of gastroenteritis outbreaks (n=141) in facilities in PHS 2009 to 2014.
IRR=incidence risk-ratio; ref=reference categories; AIC=1062.60; log likelihood= -521.29; J (negative binomial dispersion parameter)=4.27; SE=0.61.
Model diagnostics
In both models (duration and size models), the random effects were normally distributed. The absolute value of the random effects for each facility was relatively low but some facilities had larger random effects than others.
Discussion
In our analysis of institutional gastroenteritis outbreaks norovirus was the most common pathogen and aged care was the most common institutional setting. A significant increasing trend in the number of institutional gastroenteritis outbreaks was observed over the study period. The model built to quantify the association between the main covariate, time to notify the PHS, and the main outcomes, duration and size of outbreaks identified that a shorter notification time to the PHS was significantly associated with shorter duration and smaller size of outbreaks.
Norovirus was identified as the most common pathogen in 61.7% (108/175) of the institutional outbreaks in MCPHS, which is similar to the 68% (39/60) of norovirus-attributable gastroenteritis found in an Australian study.23 Norovirus was also the most common pathogen in New Zealand annual outbreak surveillance reporting (eg, 37.3% (322/863) in 2014).10 Facilities such as aged care mostly consist of frail individuals who are more vulnerable to norovirus infection, so this may explain the difference in percentages between the New Zealand annual percentage and the rate in institutions.
Aged care and ECE were the most common settings of gastroenteritis outbreaks, comprising respectively 56.6% (99/175) and 28.6% (50/175) of outbreaks. Aged care was also the most common setting in New Zealand (2006–2014), and constituted about half of gastroenteritis outbreaks annually.10 Aged care is occupied by frail individuals who are more vulnerable to infections, thus it is reasonable to expect that a higher percentage of notified gastroenteritis outbreaks is reported from aged care facilities. In a 2013/14 UK and Ireland study of norovirus outbreaks in care settings, hospital (71.3%, 383/537) and aged care (21.4%, 115/537) were the most common settings.24
The results from the time-series plots of gastroenteritis outbreaks confirmed a significant increasing trend in the number of notified outbreaks. The increasing trend was also seen nationally in New Zealand (2005–2014)10 and in a study of gastroenteritis outbreaks in hospitals in the US (1996–2007).25 The increased ageing population (accompanied by a presumed rise in the number of people in residential aged care), the funded 20 hours ECE, the introduction of national guidelines of norovirus management in hospital and aged care on 2 January 2009 and the emergence of the virulent Sydney GII.4 strain of norovirus27 could have contributed to this increasing trend.
Although the time series analysis of our study showed no statistical evidence of a seasonal pattern, a trend was seen for more outbreaks in the Spring and Autumn. Previous research in the Northwest Territories of Canada28 and US29 reported seasonal patterns in gastroenteritis outbreaks. Gastroenteritis outbreaks peaked in spring and autumn in the Canadian study from community health facilities and this was suggested to be due to environmental and social factors such as higher temperatures, frequent travelling and surface water consumption.26 Other studies performed in hospitals in the US (1996–2007)25 and (2001–2009)27 reported that gastroenteritis outbreaks mostly happened in winter. Temporal patterns are likely associated with the health-seeking behaviour peculiar to each country and some infections which are community- and not hospital-acquired. It may also reflect how the surveillance data are managed and the notification of outbreaks.
The present study reported that 36.6% (64/175) outbreaks were notified to PHS within one day after the onset of symptoms of the second case, while 20.6% (36/175) outbreaks were notified later (≥7 days). In a similar study in residential care facilities (RCFs) in Queensland, Australia in 2008, 40% (24/60) were notified to the PHS within one day, and the latest notification was 18 days.23 The range of notification was 0 to 37 days in a similar study in nursing homes in Alsace, France.6. Approximately one of four persons at risk became cases in both this and Australian studies, and one of three persons at risk became cases in the French study.6,23 The duration of each outbreak as notified to the PHS in the current study was 1–55 days while in the Australian and French study the duration were 0–42 days and 2–26 days, respectively, of which Australian and French studies each had different definition of outbreak duration.6,23 In the current study, the start of the gastroenteritis outbreaks was calculated based on the onset of symptoms of the second case recorded by the PHS.
Shorter notification time was associated with shorter duration and smaller size (IR) of gastroenteritis outbreaks. For example, after adjusting for pathogen and facility, the duration of outbreaks was 3.4 days (p=0.001, 95% CI=3.1–3.7) longer than baseline (0–1 day), when time to notify was ≥7 days. Further, there is an association between the outcome variables (duration and size of outbreaks) as displayed in Figure 3. Modelling both duration and size of outbreaks via a single composite outcome variable could be a useful next step in investigating the association between notification to the PHS and the impact of the outbreak.
The finding of an association between shorter notification time to PHS and shorter duration of outbreaks is similar to the Australian study, which reported that shorter notification time was associated with shorter duration of outbreaks. However, different from this current study, the Australian study found that shorter notification was not associated with smaller size of outbreaks.23 The number of outbreaks notified in the Australian study was smaller (n=60)23 than in this study (n=175). A lack of statistical power might explain this lack of association.
The limitation of this study is the possibility that some cases were unreported, eg, in New Zealand, aged care has a better notification system than other facility types. Aged care institutions are likely to have their own health professional staff, have systems of recording illness, and to audit outbreak management procedures. This is unlike ECEs where a child can be absent without definite reason. It is more difficult in ECEs to ascertain smaller outbreaks and cases. In ECEs children and staff members do not reside in the facilities and attendance is only during school or working hours. These facilities usually do not have their own health professional staff, and their population is larger and mobile so outbreak management procedures are more difficult to implement. Therefore, more unreported cases were expected to come from ECE. This might be attributed with the length and size of the outbreaks. In the multivariate analysis, type of facility was not significantly associated with outbreak duration and size. The characteristics of each pathogen, eg, incubation period, might also become a confounder in the duration and size of outbreak. Furthermore, not enough information about the characteristics of each facility could be obtained from the dataset, eg, whether a staff member of a facility was more skilled and experienced than other staff that could contribute to their acts responding to the outbreak. Staff experienced with managing previous outbreaks are more likely to put in effective control measures and notify the PHS earlier than those inexperienced with outbreak management.
Prompt notification to the PHS appears to be one of the factors associated with reduced outbreak duration and size. The act of notification to PHS per se will help reduce the impact of the outbreaks more effectively if the facility’s procedures for controlling outbreaks have oversight by a regulatory authority.
If a facility notifies early then the PHS is able to provide earlier access to advice and action that support the facility’s procedures for controlling outbreaks, including support with implementing the Ministry of Health (MoH) Norovirus guidelines. This PHS actions include assigning a health protection officer to the outbreak; daily oversight of the case logs and epidemic curves to monitor outbreak progress; support and advice regarding control measures and identification of likely source; procedural reviews of controls and site visits if required; identification of the pathogens (through samples collection and laboratory submission); advice to reduce further transmission of the current outbreak; and prevention of outbreaks occurring again in the future through outbreak management training workshops based on MoH guidelines.
Conclusion
The models built to quantify the association between the main explanatory factor, time to notify the PHS, and the main outcomes of interest, duration and size of outbreaks identified that a shorter notification to the PHS was significantly associated with shorter duration and smaller size of outbreaks. Future studies should consider more complex modelling of the association between time to notify the PHS, the duration and the size of the outbreak. This should be combined with an investigation of the sensitivity of the definition of the start of the outbreak. For this analysis we chose the date of onset of the second case. Identification and modelling of a composite outcome variable that captures the shape of the epidemic curve (both size and duration of outbreak) is beyond the scope of this study but an important next step in better understanding the effect of time to notify the PHS.
Better data capture, both laboratory and epidemiological (eg, clear staff role identifications and days off work due to illness) is important: the former to provide pathogen specific interventions and the latter to more clearly estimate the cost of the outbreak. Improved identification of associated cases beyond the staff and residents/attendees (for example family members of staff and residents/attendees, visitors to the institutions) will help more clearly define the extent of the burden associated with institutional outbreaks.
Summary
Abstract
Aim
We report a quantification and visualisation of the association between the time to notify public health service (PHS) and the duration and size of institutional gastroenteritis outbreaks, and explore the seasonality and trend of the outbreaks.
Method
Descriptive analysis was performed on institutional gastroenteritis outbreak data from a North Island PHS (1 January 2009-31 December 2014). Time-series analysis was used to explore the seasonality and trend of outbreaks. Multivariate analyses were performed to quantify the association between the time to notify PHS and the duration and size of outbreaks.
Results
One hundred and seventy-five gastroenteritis outbreaks (from 58 facilities) were included in descriptive analyses. A significant increasing trend (p=0.01) without seasonal pattern was confirmed by time-series analysis. Shorter notification time was associated with shorter duration and smaller size of outbreaks, eg, duration of outbreaks when time to notify was 57 days, was 3.4 days (p=0.001, 95% CI=3.1-3.7) longer than baseline time to notify (0-1 day).
Conclusion
Prompt notification to the PHS appears to be a factor associated with reduced outbreak duration and size.
Author Information
Acknowledgements
Correspondence
Correspondence Email
Competing Interests
- South Island DHBs, Gastroenteritis - South Island, South Island Alliance, Place Published, 2015.
- European Commission, Opinion of the scientific committee on veterinary measures relating to public health on Listeria monocytogenes, Health and Consumer Protection Directorate-General, Place Published, 1999.
- Ministry of Primary Industries, Making food for vulnerable populations (YOPI), Manufacturers of food and beverages, Ministry of Primary Industries, Wellington, New Zealand, 2015, pp. 1.
- Calderon-Margalit R, Sheffer R, Halperin T, Orr N, Cohen D, Shohat T. A large-scale gastroenteritis outbreak associated with Norovirus in nursing homes, Epidemiology and Infection, 2005; 133:35–40.
- Kirk MD, Roberts L, Horvath J. Understanding gastroenteritis in elderly residents of aged-care facilities, Medical Journal of Australia, 2008; 189:476–477.
- Thouillot F, Delhostal C, Edel C, Bettinger A, Pothier P, Ambert-Balay K, Meffre C, Alsibai S. Gastroenteritis outbreaks in elderly homes in the east of France during winter 2009/10: Aetiology research for a series of 37 outbreaks, Euro surveillance: bulletin Europeen sur les maladies transmissibles = European communicable disease bulletin, 2012; 17.
- Greig JD, Lee MB. Enteric outbreaks in long-term care facilities and recommendations for prevention: a review, Epidemiology and Infection, 2009; 137:145–155.
- Jones TF. When diarrhea gets deadly: A look at gastroenteritis outbreaks in nursing homes, Clinical Infectious Diseases, 2010; 51:915–916.
- ESR, Guidelines for the investigation and control of disease outbreaks, The Institute of Environmental Science and Research Limited (ESR), Place Published, 2012.
- ESR, Annual summary of outbreaks in New Zealand 2014, The Institute of Environmental Science and Research Limited (ESR), Place Published, 2015.
- Tu ETV, Bull RA, Kim M-J, McIver CJ, Heron L, Rawlinson WD, White PA. Norovirus excretion in an aged-care setting, Journal of Clinical Microbiology, 2008; 46:2119–2121.
- ESR, Human norovirus, New Zealand Food Safety Authority, Place Published, 2010.
- MidCentral District Health Board, About MidCentral District Health Board, MidCentral District Health Board, Place Published, 2014.
- R Core Team, R: A language and environment for statistical computing, R Foundation for Statistical Computing, Vienna, Austria, 2014.
- Bowman AW, Azzalini A. Applied smoothing techniques for data analysis: The kernel approach with S-Plus illustrations, Oxford University Press, Place Published, 1997.
- Cleveland WS. Robust locally weighted regression and smoothing scatterplots, Journal of the American Statistical Association, 1979; 74:829–836.
- Siegel S, Castellan Jr J. Nonparametric statistics for the behavioral sciences, McGraw-Hill, Place Published, 1988.
- Chatfield C. The analysis of time series: An introduction, CRC press, Place Published, 2013.
- Pinheiro J, Bates D, DebRoy S, Sarkar D. {nlme}: Linear and nonlinear mixed effects models, 2014.
- Fournier DA, Skaug HJ, Ancheta J, Ianelli J, Magnusson A, Maunder M, Nielsen A, Sibert J, Bolker B. {glmmADMB}: Generalized linear mixed models using automatic differentiation model builder, AD Model Builder: Using automatic differentiation for statistical inference of highly parameterized complex nonlinear models, Optim. Methods Softw, 2014.
- Stevenson M, Nunes T, Heuer C, Marshall J, Sanchez J, Thornton R, Reiczigel J, Robison-Cox J, Sebastiani P, Solymos P, Yoshida K. {epiR}: An R package for the analysis of epidemiological data, Tools for the analysis of epidemiological data, 2014.
- Bates D, Maechler M, Bolker BM, Walker S. {lme4}: Linear mixed-effects models using Eigen and S4, Fitting Linear Mixed-Effects Models using lme4, ArXiv e-print, 2015.
- Davis C, Vally H, Bell R, Sheehan F, Beard F. Viral gastrointestinal outbreaks in residential care facilities: An examination of the value of public health unit involvement, Australian and New Zealand journal of public health, 2014; 38:177–183.
- Curran ET, Wilson J, Haig CE, McCowan C, Leanord A, Loveday H. The where is norovirus control lost (WINCL) study: An enhanced surveillance project to identify norovirus index cases in care settings in the UK and Ireland, Journal of Infection Prevention, 2016; 17:8–14.
- Lopman BA, Hall AJ, Curns AT, Parashar UD. Increasing rates of gastroenteritis hospital discharges in US adults and the contribution of norovirus, 1996–2007, Clinical Infectious Diseases, 2011; 52:466–474.
- Pardhan-Ali A, Berke O, Wilson J, Edge VL, Furgal C, Reid-Smith R, Santos M, McEwen SA. A spatial and temporal analysis of notifiable gastrointestinal illness in the Northwest Territories, Canada, 1991–2008, International Journal of Health Geographics, 2012; 11.
- Gastanaduy PA, Hall AJ, Curns AT, Parashar UD, Lopman BA. Burden of norovirus gastroenteritis in the ambulatory setting-United States, 2001–2009, Journal of Infectious Diseases, 2013; 207:1058–1065.
Contact diana@nzma.org.nz
for the PDF of this article
Institutional gastroenteritis outbreaks and time to notify public health services
Subscriber Content
The full contents of this pages only available to subscribers.
Login, subscribe or email nzmj@nzma.org.nz to purchase this article.
