Temporal Trends and Climatic Factors Associated with Bacterial Enteric Diseases in Vietnam, 1991–2001

Objective In Vietnam, shigellosis/dysentery, typhoid fever, and cholera are important enteric diseases. To better understand their epidemiology, we determined temporal trends, seasonal patterns, and climatic factors associated with high risk periods in eight regions across Vietnam. Methods We quantified monthly cases and incidence rates (IR) for each region from national surveillance data (1991–2001). High- and low-disease periods were defined from the highest and lowest IRs (1 SD above and below the mean) and from outbreaks from positive outliers (4 SDs higher in 1 month or 2 SDs higher in ≥ 2 consecutive months). We used general linear models to compare precipitation, temperature, and humidity between high- and low-risk periods. Results Shigellosis/dysentery was widespread and increased 2.5 times during the study period, with the highest average IRs found between June and August (2.1/100,000–26.2/100,000). Typhoid fever was endemic in the Mekong River Delta and emerged in the Northwest in the mid-1990s, with peaks between April and August (0.38–8.6). Cholera was mostly epidemic along the central coast between May and November (0.07–2.7), and then decreased dramatically nationwide from 1997 onward. Significant climate differences were found only between high- and low-disease periods. We were able to define 4 shigellosis/dysentery, 14 typhoid fever, and 8 cholera outbreaks, with minimal geotemporal overlap and no significant climatic associations. Conclusions In Vietnam, bacterial enteric diseases have distinct temporal trends and seasonal patterns. Climate plays a role in defining high- and low-disease periods, but it does not appear to be an important factor influencing outbreaks.

In Vietnam, shigellosis (bacillary dysentery), typhoid fever, and cholera are enteric diseases of significant public health concern (DeRoeck et al. 2005). They are primarily caused by the bacterial pathogens Shigella spp., Salmonella typhi, and Vibrio cholerae, respectively, and transmission occurs through fecal contamination of food or water or by person-to-person contact (Bhan et al. 2005;Crump et al. 2004;Kindhauser 2003;Kotloff et al. 1999;Lanata et al. 2002). Infection rates and outbreaks are highest where the standards of living, water supply, and human behaviors related to personal hygiene and food preparation are poor. The distribution and ecologic determinants of shigellosis/dysentery, typhoid fever, and cholera have recently been described from surveillance data in Vietnam (Kelly-Hope et al. 2007). The data show that each disease varies in magnitude and has a distinct spatial pattern, which appears to be driven by a combination of human and environmental factors, including poverty, water sources, and climate.
Many infectious diseases, including shigellosis/dysentery, typhoid fever, and cholera, are influenced by climate. Specifically, climate plays an important role in the transmission process and can influence spatial and seasonal distributions, as well as interannual variability and long-term trends [Burke et al. 2001; Kovats et al. 2003;World Health Organization (WHO) 2004]. Although climate is one aspect of the complex epidemiology of these enteric diseases, it can help to define high-risk periods. Few studies conducted in Asia have described the temporal patterns and outbreaks of shigellosis/dysentery and typhoid fever, and no study has specifically examined the impact of climate on these diseases. In general, cholera has been studied more widely, and formal and informal listings of outbreaks and putative risk factors are available from various sources (Griffith et al. 2006;Kelly-Hope et al. 2007;WHO 2003WHO , 2005WHO , 2006. Studies have shown associations of V. cholerae with climate, including rainfall, flooding, water temperature and depth, sea surface temperatures, and the El Niño Southern Oscillation (ENSO) (Huq et al. 2005;Koelle et al. 2005b;Lipp et al. 2002;Lobitz et al. 2000;Pascual et al. 2000;Rodo et al. 2002).
In Vietnam, monthly shigellosis/dysentery, typhoid fever, and cholera surveillance data have been collated for 1991-2001. We used these national data to determine the long-term temporal trends and seasonal patterns of shigellosis/dysentery, typhoid fever, and cholera in eight geographic regions of Vietnam, and to examine climatic factors associated with high-risk periods. and confirmed by serology and stool culture. Provincial data were pooled to provide estimates for each of the eight study regions.
Temporal trends and seasonal patterns. To determine long-term temporal trends and seasonal patterns of shigellosis/dysentery, typhoid fever, and cholera, we quantified the monthly number of cases and average incidence rates (IRs) per 100,000 population for each region. Population data for 1995-2001 were obtained from the General Statistics Office of Vietnam (2005), and population estimates for 1991-1994 were extrapolated from the fitted cubic spline of the known years (Eubank 1999) in order to obtain regional population estimates and crude IRs for each study year.
To identify distinct seasonal variations, we detrended (with a fourth-degree polynomial) and log-transformed monthly IRs in each region for each disease, and defined "high" and "low" disease periods based on the months with the highest and lowest rates (months with values at least 1 SD above and below the mean, respectively). Outbreak periods were detected similarly, but we defined them empirically as the positive outliers that were 4 SDs higher in 1 month or 2 SDs higher in ≥ 2 consecutive months from the modeled Fourier function of the time series (Bloomfield 2000;Pollock 1999), which was performed on each time series, accounting for disease seasonality.
Climate data and analysis. Monthly climatic data were obtained from worldwide climate maps generated by the interpolation of data from ground-based meteorologic stations with a monthly temporal resolution and 0.5°( latitude) by 0.5°(longitude) spatial resolution (Mitchell and Jones 2005). The climatic variables used were precipitation; average daily minimum, maximum, and mean temperatures; vapor pressure; and number of wet days. Monthly climate data during 1991-2001 were extracted from the pixels containing the centroid of each province and clustered according to the eight regional divisions of Vietnam. To calculate climatic averages for the eight regions, we used the climatic values for each province weighted by its respective population (to account for the proportional relevance of the diseases of each province within the regions, so the climatology of places where few people live would, in fact, account proportionally less in the regional analyses than places with a large demographic concentration).
To explore climatic factors associated with high-risk times, we examined differences between high-and low-disease periods and outbreak and non-outbreak periods. First, we used a general linear model to test significant differences between high-and low-disease periods with time lags from 0 to 2 months. Because multiple tests were conducted (four climatic variables tested at three time lags of 0, 1, and 2 months, thus yielding 12 tests for each disease at each region), significance levels were adjusted with the Bonferroni correction (Sokol and Rohlf 1995); we considered p-values < 0.05/12 significant.
Second, we compared climate data corresponding to the outbreak period in each region with climate data for the same months in previous years when outbreaks did not occur (i.e., the non-outbreak period), with time lags from 0 to 2 months. We used general linear models with the climate variables as dependent variables, outbreak presence as a fixed factor, and region as a random factor.

Temporal trends and seasonal patterns.
The monthly numbers of shigellosis/dysentery, typhoid fever, and cholera cases reported in Vietnam during 1991-2001 are shown in Figure 2. Shigellosis/dysentery was the most prevalent disease and increased approximately 2.5 times during the study period, with 16,976 cases (annual IR of 25.3 per 100,000) reported in 1991 compared with 46,292 cases (IR, 58.8) in 2001. The annual number of typhoid fever cases was similar at the beginning (7,592 cases; IR, 11.3) and end (9,614 cases; IR, 12.2) of the study period; however, there was a 3-fold increase during 1994 to 1997, with an average of 24,553 cases (IR, 33.8) reported annually. Overall, there were fewer cholera cases, which appeared episodically during 1991-1996, with four main peaks in May 1992 (1,851 cases; IR, 2.7), August-September 1993 (943-1,054 cases; IR, 1.4-1.5), May 1994 (1,127 cases; IR, 1.6), and June-July 1995 (1,097-1,492 cases; IR, 1.5-2.1). From January 1997 onward, the number of cholera cases reported nationwide decreased significantly, with only two minor peaks reported in January-February 1999 (188 cases; IR, 0.25) and September-October 2000 (166 cases; IR, 0.21). Figure 3A shows the monthly IRs of shigellosis/dysentery, typhoid fever, and cholera for each region during 1991-2001. This figure highlights the widespread incidence of shigellosis/dysentery and its increase in the Central Highlands and the South Central Coast, the endemicity of typhoid fever in the Mekong River Delta and its emergence in the Northwest region, and the significant decline of cholera nationwide.
Overall, we found distinct seasonal variations in each region, as shown by the average monthly IRs in Figure 4. Shigellosis/dysentery rates peaked in the northern regions of the country (Northeast, Northwest, Red River Delta, North Central Coast) between June and August (IR range, 2.1-7.8), and in the southern regions (South Central Coast, Central Highlands, Southeast, Mekong River Delta)   In total, 26 enteric outbreaks were identified-4 shigellosis/dysentery, 14 typhoid fever, and 8 cholera-during 1991-2001 ( Figure 3B). Apart from typhoid and cholera in the Mekong River Delta in June 1995, no disease outbreak coincided temporally with any other disease outbreak in any region. However, typhoid outbreaks in the Northeast, Red River Delta, North Central Coast, South Central Coast, and Southeast regions in 1996 overlapped temporally, with outbreak months ranging from March to July. Overall, outbreaks occurred most commonly in the months of May, June, and July, followed by April, August, and September. No outbreaks occurred in December, and only one to three outbreaks occurred in October-March.
In our climate analyses we found no significant differences in the climatic conditions between the months during or preceding each outbreak period compared with non-outbreak periods in previous years. The data in Table 2 highlight the range of climate conditions under which enteric outbreaks occurred. Overall, precipitation ranged from 37 to 311 mm; for the majority (> 80%) of the outbreaks, > 100 mm was recorded. All mean temperatures were > 21.9°C (majority > 25°C); the number of wet days ranged from 4.9 to 20.3 (majority > 11); and most outbreaks occurred in months with an average vapor pressure > 26 hPa.

Discussion
This is the first time that temporal patterns of endemic and epidemic shigellosis/dysentery, typhoid fever, and cholera have been defined concurrently on such a large scale. In the present study we used surveillance data to highlight the different magnitudes and epidemiologic patterns of each disease in Vietnam during 1991-2001, and we offer some insight into the role of climate. Notwithstanding the inherent limitations associated with surveillance data, this large data set is probably the most comprehensive available in any developing country, and provides the basis for more specific and well-defined hypotheses in relation to climate and disease.
Overall, we found that the incidence of shigellosis/dysentery was widespread and increased significantly during the study period, especially in the Central Highlands and South Central Coast. The reported dysentery could have been caused by other pathogens such as Campylobacter or Escherichia coli (Isenbarger et al. 2001(Isenbarger et al. , 2002Ngan et al. 1992); however, Shigella spp. are the most common cause of dysentery, with four distinct species able to exist in a range of ecologic niches (Kotloff et al. 1999). Also, new variants have potentially emerged in Vietnam (Isenbarger et al. 2001). In addition, the increase in shigellosis/dysentery may be related to widespread antibiotic resistance (Anh et al. 2001;Isenbarger et al. 2001Isenbarger et al. , 2002Nguyen et al. 2005;Vinh et al. 2000) and the fact that no vaccines or alternative treatments are available. Thus, shigellosis is potentially one of the most important enteric pathogens in Vietnam.
We found typhoid fever concentrated in three regions of the country, each with differing temporal patterns. In the Mekong River Delta the disease was endemic and rates were among the highest in the country, which supports previous studies (Lin et al. 2000;Luxemburger et al. 2001;Nguyen et al. 1993). In the central region of Vietnam, especially the North Central Coast, South Central Coast, and Southeast, a substantial increase occurred between 1995 and 1998, which may account for the high number of cases reported nationwide and the series of outbreaks we identified during this period. In the Northwestern region, typhoid fever first appeared in 1996-1997 and remained endemic thereafter (Tran et al. 2005). The reason for its emergence and persistence in this remote rural region is unclear. It is possible that ENSO, which resulted in extremely hot conditions across the country in 1997-1998, somehow enhanced the transmission of Salmonella typhi in this region or endemicity is related to new border openings.
In contrast, cholera decreased dramatically from 1997 onward, and many regions reported no further cases after years of epidemic and endemic activity (Dalsgaard et al. 1999). This sudden widespread reduction in cholera may be attributable to several factors, including interannual variability, immunity, economic development, and improvements and interventions in hygiene and sanitation. The initial decline probably reflects the episodic nature of cholera. Other studies have also shown that interannual variability is common and is affected by climate and events such as the ENSO, as well as by levels of immunity within populations (Koelle et al. 2005a(Koelle et al. , 2005bLipp et al. 2002;Lobitz et al. 2000;Pascual et al. 2000;Rodo et al. 2002).
However, the fact that cholera numbers remained low from 1997 to 2001 may be related to the introduction of a new locally produced vaccine in 1997 (Trach et al. 1997;Vu et al. 2003) instead of ENSO influence, given that in 1996 there were already virtually no reported cases of cholera (with the exception of the outbreak in the Northeast region). Public health campaigns and > 5 million doses of the cholera vaccine targeting both V. cholerae 01 and 0139 pathogens have since been distributed primarily to epidemic-prone regions via the national vaccine program, thus influencing the epidemiology of cholera (Thiem et al. 2006). It is impossible to know which factor is most responsible for this decline and almost disappearance of cholera in Vietnam, but this success is undoubtedly due to a combination of public health interventions, including water and sanitation improvements, vaccine delivery to high-risk populations, and changes in public awareness, as well as cyclical population immunity. Identifying peak periods of disease helps to focus local interventions. We were able to better define the seasonality of each disease and found that, on average, the highest IRs of shigellosis/dysentery occurred between May and August; of typhoid fever between April and September; and of cholera between May and November. For all diseases, the highest monthly IRs occurred earlier (April/May to July) in the southern regions than in the northern regions (May/June to November) of the country, which may be indicative of the different climatic patterns of the north and south. In particular, the tropical conditions of the south may help local health authorities implement timely interventions because peak periods of disease coincided with the onset of the wet season.
Distinct climatic differences were evident between the high-and low-disease periods, with hotter, wetter, and more humid conditions associated with an increased incidence of disease. Climatic associations, however, were not strong, and we found significant differences mainly when we compared the highand low-disease periods (0-month lags) and not the months leading up to (2-month lag) each specific period. This may be because high and low periods occurred during more extreme climate conditions (i.e., wet and dry seasons) and because climate conditions outside these parameters are more variable and not specific enough to dramatically increase or decrease disease transmission.
The overall weak association with climate could also be related to the quality of surveillance data, which are inevitably flawed because of underreporting, misdiagnosis, and misclassification. In Vietnam, adequate diagnostic facilities are not universally available, and detection can be difficult and may be biased to those individuals with severe symptoms or better access to health centers (Dalsgaard et al. 1999;Hong et al. 2003). Further, other factors such as poor socioeconomic conditions play a role (Fewtrell et al. 2005: Kelly-Hope et al. 2007 and are also likely to be as important, if not more important, than climate. This theory is supported by our analysis of outbreaks, which found no significant climatic differences in the same months between years with outbreaks and years without outbreaks. Using a robust method, we were able to define statistically 4 outbreaks of shigellosis/ dysentery, 14 of typhoid fever, and 8 of cholera. We found little or no overlap between outbreaks of the three diseases within each region, which suggests that a combination of different factors triggered each event in each region, and that competition may have occurred between these enteric microbes for available hosts (Rabbani and Greenough 1999). Comparisons of climatic factors between outbreak and nonoutbreak periods indicated that no specific or unusual climate conditions preceded any outbreak. However, most outbreaks occurred within certain periods and climatic parameters, with May, June, and July being the most common outbreak months, followed by April, August, and September.
We acknowledge that climate is only one aspect of a multitude of complex interactions that cause disease. Although the role of climate is limited, we believe that climate factors help define high-and low-risk periods and potentially provide some clues into the ecology and epidemiology of these enteric diseases. It is reasonable to expect that the different pathogens, as well as humans, respond to seasonal changes in the environment and that some conditions are more favorable than others for disease transmission.

CORRECTION
In the original manuscript published online, "vapor pressure" and "number of wet days" were incorrectly labeled in the "Results" and in Tables 1 and 2 because of an error in the climate data set. These have been corrected here.

Climate measures Region
Year