INTRODUCTION
Changes in climatic conditions, such as warmer ambient temperature and increased frequency in heatwaves, are considered to be contributing factors to the emergence and re-emergence of infectious diseases [Reference Wu1]. Infectious gastrointestinal diseases, including those that are foodborne, are influenced by weather conditions, and this has been evident with increased cases and outbreaks of salmonellosis linked to elevated ambient temperature [Reference Grjibovski2–Reference Zhang, Bi and Hiller5]. While the relationship with ambient warmer temperature and increased numbers of cases of Salmonella infection has been established, Campylobacter infections have a less clear relationship with temperature and climate variability. Some studies report a positive association of increasing temperature and incidence of cases [Reference Allard6–Reference Yun13] and others an inverse or no relationship [Reference Bi14–Reference Lal16]. Of the studies, most have been reported from Europe, England and Wales [Reference Lake8–Reference Tam11, Reference Yun13, Reference Kovats15], the USA and Canada [Reference Allard6, Reference Fleury7, Reference White12]. Two studies have examined data from Australasia: one was a multi-city study [Reference Kovats15], and the other compared Adelaide (capital city of South Australia (SA)) and Brisbane (capital city of Queensland) [Reference Bi14].
Human infection with Campylobacter spp. is an important cause of food-borne illness, with major consequences for morbidity in individuals and populations [Reference Havelaar17]. In 2010, of the 550 million infectious gastroenteritis cases reported in the world, 96 million were attributed to Campylobacter spp. the most common bacterial infectious agent [Reference Havelaar17]. In Australia, of the estimated 16·6 million cases of acute gastroenteritis, 4·7% of cases was caused by Campylobacter spp. [Reference Gibney18]. In SA, around 2000 Campylobacter notifications occur each year. Heatwaves, with longer and higher temperatures, may increase the development pace of the pathogen, affect the reservoirs and also impact people's behaviour, including food storage and transportation. With temperature and the frequency of heatwaves predicted to rise, it is important to assess whether these climatic variables have an impact on Campylobacter notifications.
The aim of this study was to assess the relationship between reported cases of Campylobacter spp., and temperature and heatwaves. The findings will inform the design of public health messages and interventions aimed at improvements in food safety, prevention and control in an effort to reduce burden associated with warmer ambient temperature due to climate change.
METHODS
Adelaide experiences a Mediterranean climate with cool wet winters and hot dry summers with high temperatures in the warm season and heatwaves becoming more frequent, intense and of longer duration [Reference Deo19]. The average ambient temperature in SA has increased by 0·96 °C in the last 95 years. This increase in maximum temperature is occurring more rapidly than that observed nationally [Reference Suppiah20].
Data collection
Notifiable cases
Laboratory-confirmed Campylobacter cases notified to the Communicable Disease Control Branch (CDCB), SA Department for Health and Ageing between 1 January 1990 and 31 December 2012 were obtained from their notifiable disease surveillance system. Cases were included if they were a resident of metropolitan Adelaide. Information on demographic characteristics and illness was extracted for each case.
Temperature data
Recordings of daily maximum temperature (T max) in degrees Celsius (°C) from 1990 to 2012 were obtained from an Australian Bureau of Meteorology (BOM) weather monitoring station close to the Adelaide city centre. We used T max as our exposure variable because it is considered to be a better index of exposure than average or minimum temperature [Reference Milazzo3, Reference Xiang21]. This was also in keeping with previous studies in Adelaide, in which T max was used as a predictor of health outcomes related to heat exposure [Reference Hansen22, Reference Nitschke, Tucker and Bi23].
Heatwave definition
In this study, a heatwave was defined to have occurred when the daily T max reached or exceeded 35 °C for 3 or more consecutive days in a given period. This definition has been applied in previous heat-health studies conducted in Adelaide [Reference Milazzo4, Reference Hansen22–Reference Xiang25]. Because of the uncertainty about the characteristics that make heatwaves hazardous to health [Reference Tryhorn and Risbey26], we assessed the role of intensity (T max of ⩾35 °C during a heatwave event), duration (the length of a heatwave, in number of days) and timing (occurrence of a heatwave according to timing within the season, e.g. first, second and so on) on daily Campylobacter notifications.
Analysis
Temperature effects
A time-series Poisson regression model was fit to estimate the effect of T max on daily Campylobacter cases. The analysis dataset was restricted to the warm season from 1 October to 31 March, to control for the potential confounding effects of seasonal fluctuations [Reference Bhaskaran27].
We used Spearman's correlation coefficient to examine the relationship between T max and the daily number of Campylobacter notifications in the warm season. To identify any delayed effects of T max on daily Campylobacter notifications, we performed cross-correlation analyses and examined different lags in time. Sensitivity analyses of different lag times up to 28 days were conducted based on the cross-correlation results and were taken into account in the regression models. We controlled for autocorrelation (AC) of daily Campylobacter notifications based on the autocorrelation function (ACF) and partial ACF so as to identify the most appropriate autoregressive (AR) order.
To take into account potential confounders of temperature effects, we included day of the week (as a categorical variable with Sunday as the reference day), public holidays (as an indicator variable), and linear and quadratic terms for year to adjust for long-term trends. Lag values were also included to estimate the delayed effects of temperature on daily Campylobacter notifications. In the case of overdispersion, a negative binomial model was fitted. Goodness-of-fit tests were used to assess model fit.
Different temperature thresholds were examined to ascertain if a differential relationship across the temperature spectrum existed for a number of Campylobacter cases. A lowess smoother at a bandwidth of 0·8 was used to assess the shape of the exposure–response relationship between T max and counts of Campylobacter. Piecewise linear regression models were fitted with a single breakpoint at the identified temperature thresholds using the ‘hockey-stick’ nl command in Stata [Reference Muggeo28].
Heatwave effects
Poison regression models were used to examine the effect of heatwaves on daily Campylobacter cases. Generalized estimating equations were used to account for the clustering of observations within a heatwave. We accepted an exchangeable correlation structure within each cluster of heatwave days and used the quasi-likelihood under the independence model criterion to select the best working correlation structure. Similar to the models we used to assess the effect of T max on daily Campylobacter notifications, we included day of the week, public holidays and year and year2 in our statistical models.
Heatwave characteristics
As well as examining each of the three heatwave characteristics (intensity, duration and timing, as described earlier), we also estimated the overall effects of heatwaves on daily notifications by including a binary variable (heatwave and non-heatwave days). Separate models were fitted to examine the effects of each heatwave characteristic on number of daily Campylobacter cases. Heatwave day (e.g. days 3, 4 or 5) was used to examine the day which produced a greater risk of Campylobacter infection. Intensity was defined as daily T max recorded within heatwaves with four temperature ranges (35–36·9, 37–38·9, 39–40·9 and ⩾41 °C) included in the model. We examined duration by length of 3, 4 and 5 or more days. We also considered whether the duration was short (3 days) compared with long (4 or more days). We considered two different characteristics of timing. We defined timing by the first, second and third heatwave events within each warm season denoting the order of occurrence. We then examined whether timing differed by the occurrence of a heatwave event in the early part of the warm season (October to December) or later (January to March).
We report incidence rate ratios (IRRs) with 95% confidence intervals (CIs) with results interpreted as per cent (%) change in the number of daily Campylobacter counts per °C increase in T max and during heatwave periods compared with non-heatwave periods. A significance level of 0·05 was used for all statistical tests. Analyses were conducted using StataSE 13 (StataCorp LP, College Station, Texas, USA).
Ethics approval
Ethics approval was given by the Human Research Ethics Committees of the University of Adelaide (H-202–2011) and the SA Department for Health and Ageing (463/07/2014).
RESULTS
Descriptive statistics
In Adelaide, from 1990 to 2012, 35 601 Campylobacter cases were notified, with 18 570 (52%) reporting onset of illness in the warm season. During this period, there were no outbreaks detected, and hence no records were excluded from the analyses. Figure 1 shows the temporal distribution of daily Campylobacter spp. notifications over the entire study period with no obvious peaks occurring in the warmer months. Temperature summary statistics for the entire study period, by season, and by heatwaves are displayed in Table 1. The mean daily T max during the warm seasons was 26·5 °C (standard deviation (s.d.) = 6·1) and 38·4 °C (s.d. = 2·2) during heatwaves. Over the study period, 213 heatwave days across 50 distinct episodes were recorded. The length of heatwaves ranged from 3 to 15 days with a mean of 3·17 (s.d. = 2·40) days. During the study period, there were no recorded heatwaves in 1990, 1996 and 2005, and none occurred outside of the warm season. The proportion of days during the warm season with a recorded T max over 40 °C was 1%.
Fig. 1. Annual distribution of monthly notifications of Campylobacter infection, 1990–2012, Adelaide, South Australia.
Table 1. Daily maximum temperature (Tmax) by season, 1990 to 2012, Adelaide, South Australia
a Campylobacter cases (n=35,601) notified in the study period.
b Campylobacter cases (n=17,031) notified in the cool season (April to September).
c Campylobacter cases (n=18,570) notified in the warm season (October to March).
d Campylobacter cases (n=908) notified during heatwaves (within the warm season).
The final time-series Poisson regression model included an AR structure of order three and daily T max, based on the maximum correlation coefficients. We examined the data for overdispersion, and as there was none, negative binominal models were not fitted.
Effects of maximum temperature on Campylobacter infections
Correlation of T max with Campylobacter notifications was negligible. There was no lagged effect of T max on the number of daily Campylobacter cases. A third-order AC of the number of campylobacteriosis notifications was detected (IRR 1·037, 95% CI 1·032–1·042, P = <0·01). There was no substantiative effect of T max per 1 °C rise (IRR 0·995, 95% CI 0·993–0·997, P = <0·01) on Campylobacter cases.
Temperature thresholds
Figure 2 demonstrates the exposure–response relationship between T max and daily Campylobacter cases during the warm season. The relationship between temperature and Campylobacter notifications changed across the observed temperature range – as T max increased, the number of cases decreased. However, a clear temperature threshold was not detected.
Fig. 2. Exposure–response relationship between maximum temperature and daily Campylobacter notifications, reported in the warm season (October to March), 1990–2012, Adelaide, South Australia.
Effect of heatwaves and heatwave characteristics on Campylobacter notifications
As illustrated in Table 2, no association between heatwaves and an overall increase in daily Campylobacter cases was identified (IRR 0·906, 95% CI 0·800–1·026, P = 0·126). When examining heatwave characteristics, a 3-day heatwave compared with a heatwave with a duration of 4 and 5 days decreased the risk of infection on daily Campylobacter counts by 21% (IRR 0·795, 95% CI 0·689–0·918, P = 0·002). A 19% decrease in cases (IRR 0·818, 95% CI 0·679–0·987, P = 0·036) was estimated with the first heatwave in the season, thus reducing the risk compared with subsequent heatwaves in the season. The number of cases was lower in the early months of the warm seasons compared with the later months. Heatwave intensity within a temperature range of 39–40·9 °C on daily cases decreased the risk of infection (IRR 0·811, 95% CI 0·692–0·952, P = 0·010) by 19%.
Table 2. Effect estimates of heatwave characteristics on daily Campylobacter cases
IRR, incidence rate ratio; CI, confidence interval; P-value (0·05 significance level).
Adjusted for long-term trends, day of the week (reference day is Sunday) and public holidays. The reference group are non-heatwave days.
DISCUSSION
Our study found that there is no substantiative effect of T max on daily Campylobacter cases in the warm seasons. When examining the effect of heatwaves on daily Campylobacter notifications, there was little evidence of an effect of an increased risk of infection. These findings indicate that Campylobacter incidence in Adelaide may not be affected by temperature in the warm seasons or during heatwaves.
Few studies have examined the relationship between temperature and Campylobacter spp. in the warm seasons and none so far have considered the effects of heatwaves on cases. Among those studies that have been conducted, a positive association of increasing temperature and incidence of cases have been reported in studies from Europe [Reference Patrick10, Reference Yun13], the UK [Reference Lake8, Reference Louis9, Reference Tam11], the USA [Reference White12] and Canada [Reference Allard6, Reference Fleury7].
Our findings that temperature and heatwaves did not increase the risk of infection concur with a study in Australia comparing temperature effects and Campylobacter cases in Adelaide that has a temperate climate, with Brisbane a sub-tropical climate [Reference Bi14]. The study in Adelaide found an inverse relationship with temperature and Campylobacter cases, yet in Brisbane the effect was positive [Reference Bi14]. It is postulated that this difference could be associated with weather conditions specific to that area, which could have an impact on animal reservoirs or proccesses along the food chain [Reference Bi14]. Likewise a multi-jurisdictional study that compared the effects of temperature on Campylobacter infection across multiple continents of Europe, Canada, Australia and New Zealand did not find a strong effect of temperature on Campylobacter cases [Reference Kovats15]. This suggests that the impact of temperature on cases varies within and between geographical regions, thereby affecting disease transmission and environmental routes.
We found no lagged effect of temperature in the warm seasons on cases. Contrary to this, previous studies found lags ranged from 1 to 6 weeks. A positive association was found between Campylobacter and temperature in the current and previous weeks in a UK study [Reference Lake8]. Studies that identified long lags of 8 weeks or more were those that had a null effect of temperature on Campylobacter cases [Reference Bi14–Reference Lal16]. Bi et al. reported no discernible lag effect of temperature on Campylobacter cases in Adelaide but a lag of up to 6 weeks on cases in Brisbane [Reference Bi14]. The lack of a lag effect in Bi et al. and our study suggests that the main route of transmission may not be foodborne [Reference Bi14]. Generally, lag effects indicate when and where food contamination could have occurred, with short time lags pointing to food contamination closer to the time of consumption, and long lags indicating effects at the production processing stages [Reference Lake8]. In our study, we found that the number of campylobacteriosis notifications was related to the number occurring in the previous 3 days. As there was little evidence in our study of a lagged effect of temperature on daily counts in our study, we did not further explore this association with heatwaves.
We were unable to discern a temperature threshold, although by visually inspecting the plot, we observed a decrease in cases with a rise in T max above approximately 36 °C. These results need to be interpreted with caution and warrant further investigation. It may be that ambient temperature in the warm seasons and heatwaves are not linked to an increase in the risk of infection as indicated by daily Campylobacter notifications as the bacteria is sensitive to high temperature. The relationship between pathogen growth and temperature is non-linear [Reference Basu29], and there could be a temperature above which proliferation and survival of Campylobacter spp. in the environment will begin to decline. Campylobacter spp. does not multiply at temperatures below 30 °C; hence, the bacteria do not increase in foods kept at usual room temperatures in temperate regions [Reference Bronowski, James and Winstanley30].
Limitations in this study are similar to those reported in our related work concerned with effects of temperature and heatwaves on Salmonella cases [Reference Milazzo3, Reference Milazzo4]. Passive disease surveillance systems are likely to result in an under-reporting of Campylobacter notifications [Reference Hall31], but this is not likely to affect the estimates of the association between temperature and heatwaves in the warm seasons and the risk of Campylobacter infection. We did not exclude cases that travelled prior to becoming unwell because of the incomplete data recorded in the disease notification surveillance system. Only a small proportion of cases are expected to travel, and inclusion of cases that travelled is unlikely to affect the results [Reference Kovats32].
It remains unclear as to why Campylobacter infections have a less obvious relationship than Salmonella with temperature and climate variability. Transmission of campylobacteriosis is complex, as there are many routes and exposure pathways including environmental paths [Reference Domingues33, Reference Whiley34]. The seasonality with campylobacteriosis peaking in spring is not fully understood, but the environment is credited to play a role in this [Reference Bronowski, James and Winstanley30].
Campylobacter spp. is ubiquitous in the environment, and hosts include wild domestic animals and birds. It can be found in the gastrointestinal tract of cattle, sheep, goats, dogs, rabbits, cats, chickens, turkey, duck and pigs [Reference Craig and Batholomaeus35]. It is hypothesised that flies can act as a vector for the transmission of Campylobacter spp. to humans [Reference Ekdahl, Normann and Andersson36]. Food-borne transmission is another route as the primary source for Campylobacter infection in humans is the consumption of poultry meat [Reference Newell37]. A case–control study in Australia identified chicken consumption as the main risk factor for Campylobacter infection, with a population-attributable risk estimate of >50 000 cases each year [Reference Stafford38]. Although Campylobacter spp. is sensitive to high temperatures and dry environments, the bacteria survive well in poultry processing production stages [Reference Craig and Batholomaeus35] further supporting poultry meat as a high-risk food for Campylobacter transmission to humans.
Environmental transmission of Campylobacter infection via birds, farm and other wild animals to humans is multi-factorial. These many routes of transmission, some which may not be temperature-dependent, are multi-faceted. From our study, as well as those conducted previously, we know that varying environmental and climatic conditions prevail between continents, countries and regions, and may have an impact on disease transmission and case incidence [Reference Bi14]. This calls for further studies from different countries with different climatic conditions to ascertain a truer picture of the role of temperature in the incidence of Campylobacter infection.
Our study suggests that temperature has a limited role in increased incidence of Campylobacter notifications in the warm season, and we also found little evidence for an effect of heatwaves on cases. Limited understanding about the reservoirs and transmission routes for Campylobacter infection make it difficult to determine the role of ambient temperature in the warm season on disease incidence. Notwithstanding this, previous studies have established that there is a relationship with temperature and consequently the role of the environment, especially warmer temperature should not be ignored with the emerging evidence of climate change and predicted increase in the frequency of warmer days.
ACKNOWLEDGEMENTS
The authors thank staff from the Disease Surveillance and Investigation Section, Communicable Disease Control Branch, Department for Health and Ageing (SA Health) for their support in conducting this study.
DECLARATION OF INTEREST
We confirm that there is no conflict of interest.
