Hostname: page-component-7c8c6479df-7qhmt Total loading time: 0 Render date: 2024-03-27T01:36:09.326Z Has data issue: false hasContentIssue false

Climate variations and salmonellosis in northwest Russia: a time-series analysis

Published online by Cambridge University Press:  04 April 2012

A. M. GRJIBOVSKI*
Affiliation:
Norwegian Institute of Public Health, Oslo, Norway International School of Public Health, Northern State Medical University, Arkhangelsk, Russia
V. BUSHUEVA
Affiliation:
Federal Service for Surveillance on Consumer Rights Protection and Human Wellbeing in the Arkhangelsk Region, Arkhangelsk, Russia
V. P. BOLTENKOV
Affiliation:
Federal Service for Surveillance on Consumer Rights Protection and Human Wellbeing in the Arkhangelsk Region, Arkhangelsk, Russia
R. V. BUZINOV
Affiliation:
Federal Service for Surveillance on Consumer Rights Protection and Human Wellbeing in the Arkhangelsk Region, Arkhangelsk, Russia
G. N. DEGTEVA
Affiliation:
Institute of Polar Medicine, Northern State Medical University, Arkhangelsk, Russia
E. D. YURASOVA
Affiliation:
WHO Office in the Russian Federation, Moscow, Russia
J. NURSE
Affiliation:
WHO European Centre for Environment and Health, Rome Office, Rome, Italy
*
*Author for correspondence: Professor A. M. Grjibovski, Norwegian Institute of Public Health, Postbox 4404 Nydalen, 0403 Oslo, Norway. (Email: andrei.grjibovski@fhi.no)
Rights & Permissions [Opens in a new window]

Summary

Associations between monthly counts of all laboratory-confirmed cases of salmonellosis in Arkhangelsk, northern Russia, from 1992 to 2008 and climatic variables with lags 0–2 were studied by three different models. We observed a linear association between the number of cases of salmonellosis and mean monthly temperature with a lag of 1 month across the whole range of temperatures. An increase of 1 °C was associated with a 2·04% [95% confidence interval (CI) 0·25–3·84], 1·84% (95% CI 0·06–3·63) and 2·32% (95% CI 0·38–4·27) increase in different models. Only one of the three models suggested an increase in the number of cases, by 0·24% (95% CI 0·02–0·46) with an increase in precipitation by 1 mm in the same month. Higher temperatures were associated with higher monthly counts of salmonellosis while the association with precipitation was less certain. The results may have implications for the future patterns of enteric infections in northern areas related to climate change.

Type
Original Papers
Copyright
Copyright © Cambridge University Press 2012

INTRODUCTION

Salmonellosis is a bacterial disease commonly manifested by an acute enterocolitis with sudden onset of headache, fever, abdominal cramps, diarrhoea, nausea and vomiting caused by Salmonella bacteria. The first symptoms usually occur 12–72 h after infection. While most individuals recover without treatment, morbidity and associated costs of salmonellosis are high [Reference Adak, Long and O'Brien1]. Although deaths from foodborne infections are uncommon, Salmonella infection causes more deaths than any other foodborne pathogen in England and Wales [Reference Chin2]. According to estimates, more than 5 million cases occur every year in the USA alone, although less than 1% of cases are usually reported even in industrialized countries [Reference Adak, Long and O'Brien1]. In Russia, there were more than 118 000 registered cases of salmonellosis in 1992, but since then the overall incidence has decreased from 79·9/100 000 in 1992 to 36·2/100 000 in 2008 [3].

Several studies have reported associations between climatic variables and the number of cases of salmonellosis [Reference D'Souza, Becker, Hall and Moodie4Reference Lake9] or food poisoning [Reference Bentham and Langford10, Reference Bentham and Langford11]. Most of the studies assessed the effects of temperature alone [Reference D'Souza, Becker, Hall and Moodie4, Reference Kovats7Reference Lake9] while a few also included data on precipitation in the models [Reference D'Souza, Becker, Hall and Moodie4, Reference Zhang, Bi and Hiller5]. D'Souza et al. observed that a 1 °C increase in temperature in the previous month was associated with an increase in the number of cases of salmonellosis ranging between 4·1% in Perth and 11·0% in Brisbane [Reference D'Souza, Becker, Hall and Moodie4]. Similar associations, but with a lag of 2 weeks have been reported by Zhang et al. [Reference Zhang, Bi and Hiller5]. The effect of temperature varied between the settings in Australia: while in subtropical Brisbane the temperature lag was 2 weeks, in tropical Townsville the association with temperature was observed in the same month [Reference Zhang, Bi and Hiller6]. In European countries, however, no obvious pattern related to either geography or mean summer temperatures was found [Reference Kovats7]. Moreover, while in Australia, no threshold was apparent and the relationship was approximately linear across the whole temperature range [Reference D'Souza, Becker, Hall and Moodie4Reference Zhang, Bi and Hiller6], in a few European countries, certain thresholds below which there was no effect of temperature were detected [Reference Kovats7]. In Canada, the results varied between the provinces – while positive, although less pronounced than in Australia, an association was observed between temperature with lags 0–6 weeks and weekly salmonellosis counts in Alberta, while no association was found in Newfoundland-Labrador [Reference Fleury8]. The threshold value of temperature in Alberta was −10 °C while in European countries it varied between −2 °C in the Czech Republic and 13 °C in Estonia [Reference Kovats7, Reference Fleury8]. The above-mentioned studies raised concern that climate change in general and rising average temperatures in particular may increase the risk of foodborne infections in the future. This concern was addressed in the study by Lake et al., who observed that in England and Wales the effect of temperature had decreased over time suggesting that adaptation strategies aimed at reducing pathogen concentration in food and improvement of food hygiene could counterbalance the effects of global warming [Reference Lake9].

Although most studies have reported associations between temperature and salmonellosis, the existing differences in data aggregation, modelling approaches, treatment of outbreaks and imported cases as well as the fact that almost all the studies have been performed in Australia, North America or EU countries warrants replication of the results in other settings.

This study aimed to investigate associations between the number of reported cases of salmonellosis and ambient air temperature and precipitation in one of the northernmost Russian cities.

METHODS

Arkhangelsk is a regional capital city in northwest Russia (64° 32′ N, 40° 32′ E) with a population of ∼348 000 in 2010 (Fig. 1). It lies on both banks of the northern Dvina River near to its exit into the White Sea and according to a Köppen–Geiger classification has a sub-arctic climate characterized by long, usually cold winters, and short, cool to mild summers [Reference Peel, Finlayson and McMahon12].

Fig. 1. Map of the Arctic region. Arkhangelsk is indicated by an arrow.

Monthly counts of all laboratory-confirmed cases of Salmonella infection in the city for 1992–2008 were obtained from the Regional Infectious Diseases Surveillance Centre (Rospotrebnadzor). Data on a weekly or daily basis were not collected for the whole study period and therefore were not used in the study. Cases linked to outbreaks were identified in the records and excluded from the analyses since the effects of climatic factors on outbreaks may differ from their effects on sporadic cases [Reference D'Souza, Becker, Hall and Moodie4Reference Zhang, Bi and Hiller5, Reference Kovats7]. The population of the city for each year was obtained from the regional Medical Information Analytical Centre (MIAC). Data on mean monthly ambient air temperature and monthly precipitation were retrieved from the regional branch of the Russian Hydrometeorological Service (Roshydromet).

Associations between mean monthly temperature, precipitation and salmonellosis notifications were studied by negative binomial regression to allow for overdispersion in the data [Reference Long and Freese13]. Monthly counts of laboratory-confirmed cases were used as a dependent variable. A logarithm of the population size was included in the model as an offset. Given that the effects of high temperature on case counts may be delayed up to 9 weeks [Reference Kovats7], we used mean monthly temperature with lags of 0–2 months. Similarly, the monthly amount of precipitation was included with lags of 0–2 months. Year-to-year variations during the study period were modelled by fitting a polynomial of time. A cubic polynomial was sufficient to model a long-term trend and higher-order polynomials did not further improve the model fit. Seasonal variations were modelled using sine and cosine functions for a period of 12 months. Robust standard errors were calculated for all estimates to adjust for heterogeneity in the model. To control for autocorrelation in the outcome variable, first- and second-order autoregressive terms were included in the model (model 1). The analyses were repeated using indicator dummy variables for each month and year (model 2) as an alternative method to control for long-term and seasonal effects as in other studies [Reference Kovats7].

In addition, as in previous studies [Reference Zhang, Bi and Hiller5, Reference Kovats7] a ‘hockey-stick’ model was fitted to the data. This model assumes no effect of temperature on salmonellosis counts below the threshold and a linear relationship above the threshold. We used the ‘nl’ program in Stata to estimate a threshold temperature [Reference Bi, Zhang and Parton14]. Moreover, a curvilinear relationship between temperature, precipitation and number of notified cases of salmonellosis was modelled by fitting cubic splines with three knots placed at quartiles of the distribution of each continuous variable using the ‘uvrs’ estimation program [Reference Royston and Sauerbrei15].

Finally, we repeated our analyses using detrended and deseasonalized logarithms of case counts as dependent variables and deseasonalized data on temperature and precipitation as independent variables in least squares regression (model 3) as described in detail previously [Reference Lake9]. Interaction terms between sequential time variables and climatic variables with lags were added to test whether the effect of temperature had changed over time [Reference Lake9].

All analyses were performed using Stata 10.0 software (StataCorp, USA).

RESULTS

The overall number of laboratory-confirmed cases of salmonellosis during the 17-year period of observation was 4627. After excluding cases linked to outbreaks this number decreased to 4585. The average monthly temperature ranged between −11·6 °C in February and 16·3 °C in July. The lowest mean monthly temperature was −21·7 °C in February 2008 while the highest was 19·7 °C, registered in July 2003. Monthly precipitation varied from 1·4 mm in April 2002 to 151·7 mm in July 1995; mean annual precipitation was 616 mm. No long-term trends were observed for either temperature or precipitation during the 17-year period.

Figure 2 illustrates a clear periodic pattern of salmonellosis notifications in Arkhangelsk. Moreover, the number of cases decreased in both the early and mid-1990s and mid-2000s with stagnation over a few years in between. The highest average number of cases occurred in August, 1 month after the peak temperature in July. The second less pronounced peak occurred in April. The lowest number of cases was observed for December–February. Figure 3 summarizes the seasonal pattern averaged across 17 years of observation.

Fig. 2. Number of cases of salmonellosis and estimated long-term trend in Arkhangelsk, 1992–2008.

Fig. 3. Seasonal pattern of mean monthly number cases of salmonellosis and mean monthly temperature in Arkhangelsk, averaged for each month for the period 1992–2008.

Multivariable modelling of the salmonellosis counts using polynomial long-term trend and seasonality modelled by trigonometric functions (model 1) showed that temperature in the previous month was significantly related to outcome. An increase of 1 °C was associated with an increase in the monthly number of cases by 2·04% [95% confidence interval (CI) 0·25–3·84]. Model 2, in which year-to-year and seasonal variations were modelled using indicator dummy variables, yielded similar results: an increase of 1 °C was associated with an increase in the monthly number of cases by 1·84% (95% CI 0·06–3·63). In addition, in this model a 1-mm increase in precipitation was associated with a 0·24% (95% CI 0·02–0·46) increase in salmonellosis counts in the same month.

No thresholds for the effects of either temperature or precipitation were detected by the hockey-stick model. The results of modelling curvilinear relationship between the studied climatic variables and the monthly number of salmonellosis cases are presented in Figure 4. No significant deviations from linearity were detected for any of the climatic variables.

Fig. 4. Relationship between (a) salmonellosis and temperature with lag 1 and (b) relationship between salmonellosis and precipitation with lag 0. Both models are adjusted for covariates as in Table 1. Grey areas represent 95% confidence intervals.

Analysis of associations between detrended deseasonalized logarithms of salmonellosis counts and deseasonalized values of temperature and precipitation yielded results close to those in model 1: an increase in mean monthly temperature of 1 °C was associated with a 2·3% (95% CI 0·38–4·27) increase in the number of cases of salmonellosis in the next month. No interactions between climatic variables and time were observed. The results of all three models are presented in detail in Table 1.

Table 1. Percent change in monthly salmonellosis counts per 1 °C increase in mean temperature and 1 mm increase in precipitation per month

CI, Confidence interval.

* Negative binomial regression adjusted for variables in the table, first- and second-order autocorrelations, seasonal variations and long-term trend modelled using trigonometric functions and polynomials.

Negative binomial regression adjusted for variables in the table, first- and second-order autocorrelations, seasonal variation and long-term trend modelled as dummy variables.

Linear regression on logarithmically transformed detrended deseasonalized counts and deseasonalized climatic variables adjusted for first- and second-order autocorrelations.

DISCUSSION

The results of this northernmost study, and the first Russian study, suggest a linear relationship between monthly counts of salmonellosis and mean temperature in the previous month across the whole temperature spectrum with no threshold values. Moreover, one of the models suggested an association between monthly counts of salmonellosis and precipitation, but this association was not replicated in other models.

Our results are generally in line with findings from other parts of the world. D'Souza et al. [Reference D'Souza, Becker, Hall and Moodie4] also observed an association between monthly cases of salmonellosis and temperature in the previous month in five Australian cities, although the effect was much more pronounced and varied between 5% and 10% increase in the number of cases for each 1 °C increase in temperature. Other Australian studies have also reported positive associations between temperature and salmonellosis, although both the effect and lags varied between the settings: in subtropical Brisbane, the effect of temperature was delayed 2 weeks while in tropical Townsville the effect of temperature was observed in the same month [Reference Zhang, Bi and Hiller6]. Moreover, much greater effects than those found in our study were observed using several different statistical models [Reference Zhang, Bi and Hiller5]. Interestingly, similar to studies from Australia [Reference D'Souza, Becker, Hall and Moodie4Reference Zhang, Bi and Hiller6], we could not identify the threshold under which the effect of temperature was not present. However, no thresholds were found in many European countries, e.g. Spain, Switzerland, Slovak Republic, Poland, Scotland and Denmark [Reference Kovats7]. Moreover, no associations between temperature and salmonellosis were found in Denmark, Slovak Republic and the Canadian province of Newfoundland-Labrador [Reference Kovats7, Reference Fleury8]. The latter was the northernmost study until now, but it included only 986 cases of Salmonella infection and did not have sufficient power to detect small effects, which were observed in Alberta (1·2% increase for each 1 °C increase). Lake et al. have recently provided evidence not only on the effect of ambient air temperature on salmonellosis and other enteric infections in England and Wales, but also calculated that this effect has reduced over the last decades as a result of preventive measures aimed at reducing pathogen levels in major food groups and improving food hygiene at the domestic and institutional levels, which can be considered as effective adaptation strategies to threats posed by climate change [Reference Lake9]. Our results suggest that the effect of temperature on the number of salmonellosis cases in Arkhangelsk was constant over time.

The associations between precipitation and enteric infections are much less certain even in the same country: while in Adelaide, an inverse association between salmonellosis and rainfall in the same month was observed, the same authors reported a positive association in both Brisbane and Townsville [Reference Zhang, Bi and Hiller6]. Contamination of drinking water as a consequence of heavy rainfall is likely to be a plausible explanation of that positive association. However, the association observed in the present study may also be spurious since it was found only in one of the three models. Moreover, almost no cases of salmonellosis were linked to water in Arkhangelsk during the study period. Further, the growth of Salmonella is greatly reduced in temperatures <15 °C [Reference Doyle and Mazzotta16]. Given that the temperatures in Arkhangelsk are below this range for the most part of the year, transmission through drinking water seems unlikely.

The main strength of our study is the use of a complete database for all laboratory-confirmed cases in the city from the same reliable source for a 17-year period. Moreover, given that the effects of climatic factors on outbreak-related cases may be different from those on sporadic cases, we identified all registered outbreaks and all outbreak-related cases were excluded from the analysis.

Another advantage of our study is the use of several statistical models to ensure comparability of the results with most of the published studies. Using trigonometric functions has been considered appropriate for controlling for seasonal fluctuations [Reference Zhang, Bi and Hiller5] as well as using polynomials for modelling long-term trends [Reference D'Souza, Becker, Hall and Moodie4]. However, using binary variables for each year and month provided better goodness-of-fit in our study, although at the expense of statistical power. Comparison of predictive abilities of the models is beyond the scope of this paper and is presented in detail elsewhere [Reference Zhang, Bi and Hiller5]. While there is no consensus about which model is best, several analytical strategies including different approaches to modelling long-term and seasonal variations have been applied in different studies [Reference D'Souza, Becker, Hall and Moodie4Reference Lake9]. Zhang et al. reported that the SARIMA model was superior to Poisson or linear models [Reference Zhang, Bi and Hiller5], but limited sample size in their study did not allow us to use this method. The fact that the association between temperature and salmonellosis in Arkhangelsk was found in all models including model 3, which produces the most conservative estimates [Reference Lake9], suggests robustness of the association.

Nevertheless, we recommend interpreting the results with caution taking into account potential limitations of the study. Underreporting of the number of cases of salmonellosis is a common limitation of all similar studies which use data obtained from passive surveillance systems. Although we used laboratory-confirmed data from a reliable source, it is likely that the numbers of cases used in this study represent only the tip of the iceberg. The degree of underreporting varies both between and within countries. For example, only one out of 15 cases of salmonellosis is reported in Australia [Reference Hall, D'Souza and Kirk17] and in many industrialized countries only 1% of all cases are registered [Reference Adak, Long and O'Brien1]. It can be speculated that this proportion of reported cases is unlikely to be higher in Russia, although no studies have been published to support this hypothesis. Moreover, the registered cases may not be representative of all cases of salmonellosis in the city [Reference Tam, Rodrigues and O'Brien18]. Given the tremendous social changes in Russia after the break-up of the Soviet Union, the completeness of reporting may vary over time. However, the use of corrections for long-term trends could at least partly eliminate this effect on the estimates. It is also unlikely that underreporting correlated with temperature or precipitation [Reference Lake9, Reference Bentham and Langford10]. Registration routines changed several times during the study period, but they were mostly associated with the inclusion of increasing amounts of detailed information and changes in laboratory techniques and reporting forms rather than in changes in reporting the number of cases per se. Lack of data on outbreaks or incomplete identification of them was mentioned as a limitation in several studies [Reference D'Souza, Becker, Hall and Moodie4, Reference Zhang, Bi and Hiller6Reference Kovats7]. In our study, we manually searched for all outbreaks using all available medical documentation and excluded all cases related to outbreaks from the analysis, although given changes in the routines of reporting outbreaks during the study period, some of the outbreaks might remain unregistered and therefore unidentified. The data available for analysis did not allow identification of travel-related cases, but these were almost non-existent in Arkhangelsk during the study period. Travel has a very clear seasonal pattern in Arkhangelsk with peaks during summer months, thus adjustment for seasonality in the analysis would remove the effects of travelling. Moreover, in other studies, where this information was available, exclusion of travel-related cases did not influence the results [Reference Kovats7]. Contrary to the most recent studies, in which weekly counts of salmonellosis were used [Reference Zhang, Bi and Hiller5Reference Bentham and Langford11], we were only able to use monthly counts because all data for routine reporting from the regional surveillance centre is aggregated by month. However, given the relatively lower number of weekly cases, monthly aggregation seems appropriate and was used in other studies [Reference D'Souza, Becker, Hall and Moodie4, Reference Zhang, Bi and Hiller6Reference Kovats7]. In addition, there are more than 2000 serotypes of Salmonella with different sensitivity to climatic variables [Reference Hall, D'Souza and Kirk17]. Although the present study does not distinguish between serotypes, more than 70% of all reported cases of salmonellosis in Arkhangelsk are attributable to S. Enteritidis, making contribution of other salmonellae to the observed pattern rather limited.

Average temperatures in the Russian Arctic are increasing faster than in the rest of the country and our results may have public health implications in terms of potential increase of the number of cases of salmonellosis and other enteric infections in the future if the temperature continues to rise. Temperature may affect contamination at any point along the food chain. Association with temperature with lags of 0–1 week are considered to reflect bacterial growth near the place of consumption while longer lags, e.g. 1 month as in our study may suggest deficiencies in food hygiene during production, processing and distribution rather than food preparation before consumption [Reference Lake9, Reference Bentham and Langford11]. However, lagged association in our study may also be partly attributed to reporting delays, because the surveillance system uses the date of laboratory confirmation, not the date when the first symptoms occurred before aggregating the data; however, in most cases the difference between the dates did not exceed 1 week. Warmer outdoor temperatures, particularly during the warm season, may also increase the likelihood of getting salmonellosis or other enteric infections through increased consumption of barbecued food, fresh salads or through outdoor recreational activities (e.g. hiking, swimming) or other activities that increase the likelihood of contact with salmonellae in the environment. The Russian tradition of preparation of large amounts of foods in advance, combined with a potential for inadequate storing, particularly in rural areas, may also contribute to an increase in the number of cases of salmonellosis in the future. However, the incidence of salmonellosis changes more rapidly than the climate and the baseline incidence of infection is the most important determinant of the effect of climate on enteric infections [Reference Tirado19]. For example, in most countries, including Russia, a considerable decrease in the incidence of salmonellosis was observed during the last decade due to the introduction of measures to control S. Enteritidis in poultry [Reference Kovats, Menne and Ebi20].

Replications of our results are required before generalizing the findings to other urban areas in high latitudes. Using the exact dates of symptom onset instead of monthly aggregated data to reduce the influence of delays is recommended for future studies. Moreover, given that more information is now collected in Russia than was available for the whole observation period in this study, studies from larger settings including information on patients' socio-demographic factors and Salmonella serotypes are warranted.

CONCLUSIONS

Higher temperatures may be associated with higher monthly counts of salmonellosis even in high latitudes while the association with precipitation is less certain. The results may have implications for the future patterns of enteric infections in the northern areas related to climate change.

ACKNOWLEDGEMENTS

This work is part of a seven-country initiative of WHO/Europe and has been funded by the German Federal Ministry for the Environment, Nature Conservation and Nuclear Safety (BMU). The project aims to protect health from climate change through addressing adaptation, strengthening of health systems and building institutional capacity. WHO/Europe coordinates the projects, contributing to the implementation of the WHO regional work plan on climate change and health. It also has provided technical assistance, guidance, training and expertise. In each country, a multisectoral steering committee is established, and a project coordinator oversees implementation at the national level. Country coordinators are supported by WHO/Europe. All activities are being implemented in collaboration with the BMU and the national Governments of the seven countries.

DECLARATION OF INTEREST

None.

References

REFERENCES

1.Adak, G, Long, SM, O'Brien, S. Trends in indigenous foodborne disease and deaths, England and Wales, 1992–2000. Gut 2002; 51: 832841.Google Scholar
2.Chin, J. Control of Communicable Diseases Manual, 17th edn.American Public Health Association, Washington DC, 2000.Google Scholar
3.WHO. Centralized information system for infectious diseases (CISID) (http://data.euro.who.int/CISID/). Accessed 7 July 2011.Google Scholar
4.D'Souza, RM, Becker, NG, Hall, G, Moodie, KB. Does ambient temperature affect foodborne disease? Epidemiology 2004; 15: 8692.Google Scholar
5.Zhang, Y, Bi, P, Hiller, J. Climate variations and salmonellosis transmission in Adelaide, South Australia: a comparison between regression models. International Journal of Biometeorology 2008; 52: 179187.Google Scholar
6.Zhang, Y, Bi, P, Hiller, JE. Climate variations and Salmonella infection in Australian subtropical and tropical regions. Science of the Total Environment 2010; 408: 524530.Google Scholar
7.Kovats, RS, et al. The effect of temperature on food poisoning: a time-series analysis of salmonellosis in ten European countries. Epidemiology and Infection 2004; 132: 443453.Google Scholar
8.Fleury, M, et al. A time series analysis of the relationship of ambient temperature and common bacterial enteric infections in two Canadian provinces. International Journal of Biometeorology 2006; 50: 385391.CrossRefGoogle ScholarPubMed
9.Lake, IR, et al. A re-evaluation of the impact of temperature and climate change on foodborne illness. Epidemiology and Infection 2009; 137: 1538–1147.Google Scholar
10.Bentham, G, Langford, IH. Climate change and the incidence of food poisoning in England and Wales. International Journal of Biometeorology 1995; 39: 8186.Google Scholar
11.Bentham, G, Langford, IH. Environmental temperatures and the incidence of food poisoning in England and Wales. International Journal of Biometeorology 2001; 45: 2226.CrossRefGoogle ScholarPubMed
12.Peel, MC, Finlayson, BL, McMahon, TA. Updated world map of the Köppen-Geiger climate classification. Hydrology and Earth System Sciences 2007; 11: 16331644.Google Scholar
13.Long, S, Freese, J. Regression Models for Categorical Dependent Variables Using Stata. StataCorp., 2006, 527 pp.Google Scholar
14.Bi, P, Zhang, I, Parton, K. Weather variables and Japanese encephalitis in the metropolitan area on Jinan city, China. Journal of Infection 2007; 55: 551556.Google Scholar
15.Royston, P, Sauerbrei, W. Multivariable modelling with cubic regression splines: a principled approach. Stata Journal 2007; 7: 4570.CrossRefGoogle Scholar
16.Doyle, ME, Mazzotta, AS. Review of studies on the thermal resistance of salmonellae. Journal of Food Protection 2000; 63: 779795.Google Scholar
17.Hall, GV, D'Souza, RM, Kirk, MD. Foodborne disease in the new millennium: out of the frying pan and into the fire? Medical Journal of Australia 2005; 177: 614618.CrossRefGoogle Scholar
18.Tam, CC, Rodrigues, LC, O'Brien, S. The study of infectious intestinal disease in England: what risk factors for presentation to general practice tell us about potential for selection bias in case control studies of reported cases of diarrhoea. International Journal of Epidemiology 2003; 32: 99105.Google Scholar
19.Tirado, C.WHO surveillance programme for control of foodborne diseases in Europe. Trends on foodborne diseases during the last 25 years. WHO, Copenhagen.Google Scholar
20.Kovats, RS, et al. Climate, weather and enteric disease. In: Menne, B, Ebi, K, eds. Climate Change and Adaptation Strategies for Human Health. Darmstadt: Steinkopff Verlag, 2006, pp. 269295.Google Scholar
Figure 0

Fig. 1. Map of the Arctic region. Arkhangelsk is indicated by an arrow.

Figure 1

Fig. 2. Number of cases of salmonellosis and estimated long-term trend in Arkhangelsk, 1992–2008.

Figure 2

Fig. 3. Seasonal pattern of mean monthly number cases of salmonellosis and mean monthly temperature in Arkhangelsk, averaged for each month for the period 1992–2008.

Figure 3

Fig. 4. Relationship between (a) salmonellosis and temperature with lag 1 and (b) relationship between salmonellosis and precipitation with lag 0. Both models are adjusted for covariates as in Table 1. Grey areas represent 95% confidence intervals.

Figure 4

Table 1. Percent change in monthly salmonellosis counts per 1 °C increase in mean temperature and 1 mm increase in precipitation per month