Impact of Temperature and Rainfall on Typhoid/Paratyphoid Fever in Taizhou, China: Effect Estimation and Vulnerable Group Identification

ABSTRACT.

The impact of temperature and rainfall on the occurrence of typhoid/paratyphoid fever are not fully understood. This study aimed to characterize the effect of daily ambient temperature and total rainfall on the incidence of typhoid/paratyphoid in a sub-tropical climate city of China and to identify the vulnerable groups for disease prevention. Daily notified typhoid/paratyphoid fever cases and meteorological data for Taizhou from 2005 to 2013 were extracted from the National Notifiable Disease Surveillance System and the Meteorological Data Sharing Service System, respectively. Distributed lag nonlinear model was used to quantify the association between daily mean temperature, total rainfall, and typhoid/paratyphoid fever. Subgroup analyses by gender, age, and occupation were conducted to identify the vulnerable groups. A total of 625 typhoid fever cases and 1,353 paratyphoid fever cases were reported during the study period. An increased risk of typhoid fever was detected with the increase of temperature (Each 2°C rise resulted in 6%, 95% [confidence interval] CI: 2–10% increase in typhoid cases), while the increased risk was associated with the higher temperature for paratyphoid (the highest cumulative risk of temperature was 33.40 [95% CI: 12.23–91.19] at 33°C). After the onset of mild precipitation, the relative risk of typhoid fever increased in a short-lasting and with a 13–26 days delay, and the risk was no significant after the continuous increase of precipitation (the highest cumulative risk of rainfall was 24.96 [95% CI: 4.54–87.21] at 100 mm). Whereas the risk of paratyphoid fever was immediate and long lasting, and increase rapidly with the increase of rainfall (each 100 mm increase was associated with 26% increase in paratyphoid fever cases). Significant temperature-typhoid/paratyphoid fever and rainfall-typhoid/paratyphoid fever associations were found in both genders and those aged 0–4 years old, 15–60 years old, farmers, and children. Characterized with a lagged, nonlinear, and cumulative effect, high temperature and rainfall could increase the risk of typhoid/paratyphoid fever in regions with a subtropical climate. Public health interventions such as early warning and community health education should be taken to prevent the increased risk of typhoid/paratyphoid fever, especially for the vulnerable groups.

INTRODUCTION

Typhoid and paratyphoid fever, characterized by prolonged high fever, hepatosplenomegaly, and abdominal pain, are enteric infections caused by the bacterium Salmonella serotype Typhi or Paratyphi (A, B, C).^1,2 Based on the different surface antigens, the genus Salmonella can be divided into different serovars. Furthermore, the pathogenic strains of S. typhi and S. paratyphi C express the Vi antigen capsular polysaccharide.³ Although both diseases’ clinical features are similar, due to the different serotypes of pathogenic bacteria, the transmission mode and disease burden of the two diseases are different. A study conducted in Indonesian revealed that typhoid and paratyphoid fever may have two transmission modes: eating contaminated food and water through an immediate within-households “short cycle” or an environmental “long cycle.” Typhoid fever was mainly transmitted within the household, while paratyphoid fever was primarily transmitted outside the home.⁴ These different routes of transmission could lead to different disease burden. In 2017, for instance, there were an estimated 10.9 million typhoid fever cases, 116,800 related deaths and 3.4 million cases of paratyphoid fever,19,100 related deaths in the world.⁵ Although population vaccination and improvement of hygiene and sanitation have been implemented, typhoid/paratyphoid still remains a considerable public health concern worldwide, especially in low- and middle-income countries, where typhoid/paratyphoid were one of significant cause of morbidity and mortality.⁶ As one of the countries affected by typhoid and paratyphoid fever, 10,000–18,000 typhoid/paratyphoid cases being reported each year in China.⁷ Zhejiang Province was one of the typhoid/paratyphoid fever high-risk areas in China, with an incidence of 1.1 per 1,000,000 in 2013.⁸

Climate change has been considered as one of the biggest challenges threatening population health in the twenty-first century.⁹ World Health Organization (WHO) projected that, there will be a 10% increase in diarrheal diseases by 2030 due to climate change.¹⁰ According to the Intergovernmental Panel on Climate Change (IPCC) sixth assessment report, the global average surface temperature may increase 1.5°C by 2030 compared with 1986–2005.¹¹ In recent years, the health impact of temperature variations on infectious diseases has received increasing attention due to its increased burden on our society and economy.^12,13 There is growing evidence showing that climate change could lead to an increase in disease outbreaks.^14,15 It has been estimated that climate change-related diarrheal diseases resulted in more than 150,000 deaths worldwide in 2000.¹⁶

Generally, typhoid/paratyphoid fever epidemics occur in summer and autumn and the seasonal pattern varies across latitude,¹⁷ indicating that meteorological factors may play a role in their transmission.^4,18 Previous studies have found that the meteorological factors such as temperature and precipitation are associated with water/food-borne infectious diseases.^19–21 Enteric infections identified as being associated with temperature and precipitation include bacillary dysentery,²² cholera,²³ Salmonella infections,¹⁹ and other diarrheal diseases.²⁴ Many studies have revealed that increase of temperature and precipitation would promote the growth and reproduction of Salmonellae, lead to water source pollution and diffusion, influence host lifestyle, and immunity,^25–27 thereby increase the probability of water/food-borne diarrhea-associated diseases such as typhoid and paratyphoid fever. Several studies have reported that the climate-Salmonella relationships varied in different regions, indicating setting-specific infection dynamics.^17,28,29 However, much articles only research one type of Salmonella disease, and few study compare the time-lagged effects of meteorological factors on typhoid versus paratyphoid fever incidence.

With this background, it is therefore important to understand the impact of meteorological variables on the disease risk in high prevalence regions so relevant preventative actions could be taken. In this study, we undertook a case study using time-series data from 2005 to 2013 in Taizhou, China to explore the effect of temperature and rainfall on typhoid/paratyphoid fever and identify the vulnerable groups. Results of this study will enable a better understanding of how meteorological factors influences the incidence of typhoid/paratyphoid, which may guide local health authorities to take preventive and curative interventions and promote comprehensive prevention strategies in the context of climate change.

MATERIALS AND METHODS

Study area.

Taizhou (latitude 120°17′–121°56′N, longitude 28°01′–29°23′E) is a prefecture-level city, located in the middle of the coast of East China Sea in Zhejiang Province (Figure 1). It has an area of 10,500 km² and a population of 6 million by the end of 2019.³⁰ The city features a humid subtropical monsoon climate with four distinctive seasons. The average annual temperature is 17.40–18.20°C, and the average annual rainfall is 1,177–1,558 mm.³¹ In particular, Taizhou was selected as our study area for high typhoid and paratyphoid fever incidence in Zhejiang Province, with the annual incidence was 3.96 per 1,000,000 over the study period.⁸

Figure 1.

Location of Taizhou in Zhejiang Province, China. This figure appears in color at www.ajtmh.org.

Data collection.

Disease surveillance data.

Daily cases of typhoid/paratyphoid from 2005 to 2013 in Taizhou were obtained from the National Notifiable Disease Surveillance System (NNDSS).⁸ Since the launch of a national infectious diseases online reporting system in 2004,³² there is a routine annual inspection by local health departments to minimize the issue of under-reporting. An investigation from Taizhou revealed the under-reporting rate was 7.38% during 2013–2015.³³ Typhoid/paratyphoid fever are classified as a statutory notifiable group B infectious disease in China. According to the Law for the Prevention and Treatment of Infectious Diseases,³⁴ all typhoid/paratyphoid fever cases must be reported online to the NNDSS system within 24 hours after diagnosis. All typhoid/paratyphoid fever cases were diagnosed according to the unified national diagnostic criteria (WS280-2008),³⁵ which can be divided into clinically diagnosed and laboratory-confirmed cases. A laboratory-confirmed case was defined as a clinical suspected case combined with pathogenic examination (Widal agglutination test, positive blood cultures for S. Typhi or Paratyphi).³⁶ To better estimate the overall health impact of weather variables on typhoid and paratyphoid fever, both types of cases were included in this study. The variables included in the typhoid and paratyphoid fever dataset mainly comprised gender, age, occupation, case category, and date of onset.

Meteorological data.

Daily meteorological data for Taizhou, including mean temperature (°C) and total precipitation (mm) for the study period were downloaded from the China Meteorological Data Sharing Service System (http://data.cma.cn/). The meteorological data used in our study were from the national standard weather station (number, 58660) in Linhai District (28°51′N and 121°08′E), which was closest to the center of the sites and selected to represent the weather conditions of Taizhou city. The missing values were filled by a k-nearest neighbor (KNN) method.³⁷

Statistical analysis.

In the descriptive analysis, time series plots were drawn to describe the seasonal distribution of the daily typhoid and paratyphoid fever cases counts and meteorological factors during the study period in Taizhou. Spearman correlation analysis was used to examine the correlations between meteorological factors over the study period (Supplemental Table 1).³⁸ We tested the over-dispersion of disease data by Lagrange Multiplier Statistical test.

Given the meteorological factors often show nonlinear and delayed effects on disease,³⁹ distributed lag nonlinear model (DLNM) was conducted to quantify the lagged and cumulative effects of temperature/rainfall on typhoid/paratyphoid fever. The DLNM model can simultaneously represent nonlinear exposure–response dependencies and delayed effects,⁴⁰ which has been widely used to estimate the short effects of weather factors on infectious disease.^9,41 Quasi-Poisson distribution was used in DLNM to control over-dispersion in typhoid/paratyphoid fever cases (LM:7890.92/4544.79, P < 0.05). Cross-basis function was established for temperature and precipitation to captures the exposure-lag-response dependency simultaneously. The cross-basis is a bi-dimensional space of functions composed of two dimensions basis functions,⁴² which is composed of a natural cubic spline with 3 degree of freedom (df) for temperature, 5 df for rainfall, and a natural cubic spline with 3 df for lag response in this study. Meanwhile, to adjust the seasonality of typhoid/paratyphoid cases, a Fourier term up to the 6th harmonic, which were combined of a pair of sine and cosine term (harmonics) was integrated in the model.⁴³ Typically, the incubation period and infectious period of virus is 7–21 days (typhoid) and 3–10 days (paratyphoid fever), respectively.^36,44 Given the incubation period of typhoid/paratyphoid fever and the delayed transport of bacterium in the external environment, the lag period was set at 7–30 days (typhoid) and 3–20 days (paratyphoid), respectively to explore all possible lag association.⁴⁵ The model can be described as follow:

$$\log \left[\text{E}\left(Y_t\right)\right]=\text{β}+\text{cb}\left({\text{temp}}_{t,l}\right)\text{+cb}\left({\text{rain}}_{t,l}\right)+\text{as.factor}\left(\text{year}\right)\text{+holiday+time(Fourier},6\text{\hspace{0.17em}harmonics}/\text{year})+\text{lag(res},t)\log \left[\text{E}\left(Y_t\right)\right]$$

Here Yt is the daily number of Typhoid or paratyphoid fever cases on day t, β is the model intercept; cb(temp_t,l) and cb(rain_t,l) are the cross-basis function modeling the nonlinear lagged estimated effects of daily mean temperature and rainfall l days before the day of illness, respectively; year was a random effort to control long-term trend; holiday denotes the weekend that include public holiday; Fourier was the Fourier term; the term lag(res,t) means specific lagged model residuals to correct for partial autocorrelation. The df of the lag and meteorological variables for cross-basis matrices and the number of harmonics were selected according to Quasi-Akaike Information Criterion (Q-AIC).⁴⁶

Due to the availability of demographic variables in the dataset, we conducted age, gender, and occupation specific analyses to identify vulnerable subgroups. In subgroup analyses, typhoid/paratyphoid fever cases were divided into various strata by gender, age (0–4, 5–14, 15–60, and > 60 years) and occupation (peasants, children, students, workers, and others). DLNM models were performed to estimate the association between temperature and typhoid/paratyphoid fever in each subpopulation. In addition, the fractions of typhoid/paratyphoid fever cases attributable to high temperature/rainfall in each subgroup were calculated within the framework of DLNM.⁴⁷

Sensitivity analyses were conducted by 1) changing the df (3–5) for temperature in cross-basis matrices, 2) changing the df (3–5) for temperature lag in cross-basis matrices, 3) changing the df (3–5) for precipitation in cross-basis matrices, 4) changing the df (3–5) for precipitation lag in cross-basis matrices, 5) changing the number of harmonics (5–7), and 6) altering lag days. Autocorrelation function (ACF) and partial autocorrelation function (PACF) were also plotted to evaluate the autocorrelation of model residuals.

All statistical analyses were performed using “dlnm” package⁴⁰ in R 3.5.3. The statistical significance level was set as 0.05 for each test (two-sided).

RESULTS

Descriptive analysis.

A total of 625 typhoid fever cases and 1,353 paratyphoid fever cases were reported in Taizhou during the study period. The male-to-female ratio for typhoid fever and paratyphoid fever were 144:100 and 132:100, respectively. The majority of the cases were aged 15–60 years, accounting for more than two-thirds of the cases (Supplemental Table 2). Figure 2 display the time-series fluctuations of daily typhoid fever and paratyphoid fever cases and meteorological factors, characterized with distinct seasonal patterns. The number of typhoid fever and paratyphoid fever cases both shows a decreasing trend with more cases occurred in summer and autumn. The demographic characteristics and seasonal patterns were similar to other study conducted in Taizhou.⁴⁸ During the study period, the mean value of daily mean temperature, and daily precipitation were 18.24°C, 4.94 mm, respectively (Table 1).

Figure 2.

The distribution of daily typhoid/paratyphoid fever cases, daily mean temperature and daily total rainfall from 2005 to 2013 in Taizhou, China.

Table 1.

Summary statistic for daily meteorological variables and daily number of typhoid/paratyphoid fever cases from 2005 to 2013 in Taizhou, China

Variable	Mean	SD	Min	Percentiles			Max
				25	50	75
Case of typhoid	0.19	0.51	0.00	0.00	0.00	0.00	7.00
Case of paratyphoid	0.82	23.61	0.00	0.00	0.00	0.00	9.00
Mean temperature (°C)	18.24	8.50	−0.80	10.90	19.20	25.70	33.80
Precipitation (mm)	4.94	15.05	0.00	0.00	0.00	3.00	240.10
Wind speed (m/s)	1.26	0.74	0.00	0.80	1.10	1.50	6.40
Sunshine duration (h)	4.51	3.83	0.00	0.00	4.80	8.00	12.50
Atmospheric pressure (hPa)	101.54	0.86	98.51	100.82	101.54	102.21	103.77
Relative humidity (%)	72.67	11.87	20.0	66.00	74.00	81.00	100.00

Min = the minimum of the variables; Max = the maximum of the variables.

Temperature/rainfall-typhoid fever association.

A nonlinear relationship was observed between daily mean temperature and typhoid fever at different lags from Figure 3A–D. The relative risk (RR) of typhoid fever was positively associated with daily mean temperature, and the RR reached the maximum at lag 30 days. Results showed that when ambient temperature exceeded 18°C, each 2°C increase in average temperature led to 6% (95% confidence interval [CI]: 2%–10) increase in typhoid fever. Figure 3C shows the exposure–response relationship between daily average temperature and typhoid fever at lag 20 days. Results showed that with the mean temperature (18°C) as the reference, the highest risk was observed at 33°C (RR = 1.11, 95% CI: 1.01–1.23). Figure 3D shows the cumulative effects of mean temperature on typhoid fever up to 30 days (RR = 20.37, 95% CI: 3.83–58.32).

Figure 3.

Lagged and cumulative association between temperature/rainfall and typhoid fever. The three-dimensional effects diagram of the daily average temperature on typhoid fever at the lag of 7–30 days (A), Predictor-specific effects of temperature 20°C on typhoid fever (B), Lag-specific effects between temperature and typhoid fever at lag 20 days (C), the overall cumulative effects of daily average temperature on typhoid fever (D), The three-dimensional effects diagram of the daily total precipitation on typhoid fever at the lag of 7–30 days (E), Predictor-specific effects of precipitation 100 mm on typhoid fever (F), Lag-specific effects between precipitation and typhoid fever at lag 20 days (G), the overall cumulative effects of daily average temperature on typhoid fever (H), Taizhou, China, 2005–2013. This figure appears in color at www.ajtmh.org.

The exposure-lag-response relationship of daily rainfall on typhoid fever was shown in Figure 3E–H. The relationship was nonlinear with an approximate inverted “V” shape. The peak risk was 1.30 (95% CI: 1.16–1.47) at 100 mm compared with 0 mm as reference at lag 20 days (Figure 3F, Table 2). The rainfall with maximum cumulative risk was 100 mm (RR = 24.96, 95% CI: 4.54–87.21).

Table 2.

The effect of average temperature, rainfall on typhoid/paratyphoid fever cases along the lag days values relative to baseline temperature (18°C) and rainfall (0 mm)

	Temperature (20°C)	Temperature (25°C)	Temperature (30°C)	Rainfall (50 mm)	Rainfall (100 mm)	Rainfall (150 mm)	Rainfall (200 mm)
Typhoid fever
Lag7	0.98 (0.94, 1.03)	0.98 (0.87, 1.11)	1.02 (0.84, 1.25)	0.89 (0.74, 1.08)	0.78 (0.60, 1.01)	0.73 (0.49, 1.08)	0.71 (0.35, 1.48)
Lag13	1.01 (0.99, 1.03)	1.03 (0.98, 1.07)	1.03 (0.96, 1.11)	1.13 (1.05, 1.23)*	1.16 (1.03, 1.30)*	1.06 (0.85, 1.31)	0.90 (0.61, 1.34)
Lag19	1.03 (1.01, 1.05)*	1.08 (1.03, 1.14)*	1.10 (1.01, 1.19)*	1.20 (1.10, 1.30)*	1.30 (1.16, 1.47)*	1.22 (0.99, 1.51)	1.04 (0.70, 1.55)
lag25	1.05 (1.03, 1.06)*	1.14 (1.10, 1.20)*	1.21 (1.12, 1.30)*	1.12 (1.03, 1.20)*	1.22 (1.08, 1.37)*	1.21 (0.99, 1.49)	1.14 (0.79, 1.64)
lag30	1.06 (1.02, 1.10)*	1.20 (1.09, 1.32)*	1.33 (1.14, 1.55)*	1.00 (0.86, 1.17)	1.07 (0.86, 1.34)	1.14 (0.80, 1.63)	1.20 (0.64, 2.27)
Paratyphoid fever
lag 3	1.04 (1.00, 1.08)	1.14 (1.01, 1.29)*	1.20 (1.01, 1.44)*	1.10 (0.95, 1.26)	1.26 (1.07, 1.48)*	1.49 (1.22, 1.81)*	1.80 (1.27, 2.56)*
lag7	1.05 (1.03, 1.06)*	1.17 (1.12, 1.22)*	1.21 (1.14, 1.30)*	1.06 (0.99, 1.14)	1.20 (1.10, 1.32)*	1.43 (1.24, 1.65)*	1.75 (1.39, 2.22)*
lag 11	1.05 (1.03, 1.07)*	1.16 (1.10, 1.23)*	1.21 (1.12, 1.31)*	1.04 (0.97, 1.11)	1.17 (1.06, 1.29)*	1.40 (1.20, 1.62)*	1.73 (1.35, 2.23)*
lag 15	1.04 (1.02, 1.05)*	1.13 (1.09, 1.18)*	1.19 (1.12, 1.26)*	1.02 (0.96, 1.08)	1.14 (1.05, 1.25)*	1.37 (1.20, 1.58)*	1.73 (1.37, 2.18)*
lag 20	1.02 (0.99, 1.05)	1.08 (0.98, 1.19)	1.15 (1.10, 1.34)*	1.00 (0.89, 1.13)	1.12 (0.97, 1.29)	1.36 (1.13, 1.64)*	1.74 (1.25, 2.42)*

^*P < 0.05.

As shown in Table 3, we list the RRs of temperature and rainfall on daily typhoid fever cases among various subpopulations. Males shows a similar temperature-related and rainfall-related risk of typhoid fever as females. For temperature, an increased risk of typhoid fever was found in 0–4 years old group (RR = 1.61, 95% CI: 1.22–2.03) and 15–60 years old group (RR = 1.29, 95% CI: 1.08–1.53) at lag 30 days. In occupational categories, farmers (lag30, RR = 1.44, 95% CI: 1.14–1.81) and children (RR = 1.83, 95% CI: 1.16–2.90) were more vulnerable than other occupations. The rainfall-related typhoid risk vulnerable group were similar to the temperature-related.

Table 3.

The effect of daily average temperature and total rainfall on typhoid fever cases along the lag days in different subgroups

Category	Temperature			Rainfall
	Lag7	Lag19	lag30	Lag7	Lag19	lag30
Males	1.11 (0.85, 1.43)	1.09 (0.98, 1.21)	1.35 (1.12, 1.64)*	0.63 (0.44, 0.90)	1.27 (1.09, 1.48)*	0.99 (0.74, 1.32)
Females	1.08 (0.81, 1.45)	1.09 (0.96, 1.23)	1.41 (1.12, 1.76)*	1.05 (0.72, 1.53)	1.34 (1.12, 1.60)*	1.27 (0.91, 1.75)
0–4	1.08 (0.44, 2.69)	1.46 (1.07, 2.00)*	1.61 (1.22, 2.03)*	0.46 (0.14, 1.54)	1.36 (1.02, 1.83)*	1.56 (0.68, 3.58)
5–14	0.74 (0.36, 1.11)	0.88 (0.60, 1.29)	1.23 (0.56, 2.71)	0.24 (0.03, 2.23)	1.30 (0.95, 1.79)	1.09 (0.45, 2.67)
15–60	1.11 (0.89, 1.40)	1.08 (0.99, 1.19)	1.29 (1.08, 1.53)*	0.79 (0.59, 1.06)	1.28 (1.11, 1.47)*	0.92 (0.71, 1.19)
> 60	1.10 (0.62, 1.94)	1.06 (0.83, 1.35)	1.32 (0.83, 2.08)	1.12 (0.56, 2.21)	1.42 (1.04, 1.94)*	1.78 (1.06, 2.99)*
Farmers	1.22 (0.91, 1.63)	0.96 (0.85, 1.08)	1.44 (1.14, 1.81)*	0.91 (0.63, 1.32)	1.33 (1.12, 1.58)*	1.14 (0.84, 1.55)
Children	1.34 (0.76, 2.39)	1.22 (1.06, 1.37)*	1.83 (1.16, 2.90)*	0.86 (0.38, 1.91)	1.28 (1.11, 1.47)*	1.28 (1.05, 1.56)*
Students	0.78 (0.37, 1.68)	1.05 (0.76, 1.44)	1.24 (0.66, 2.32)	0.55 (0.20, 1.57)	1.36 (0.85, 2.16)	1.92 (0.96, 3.84)
Workers and others	1.06 (0.77,1.46)	1.21 (0.96, 1.47)	1.19 (0.94, 1.52)	0.76 (0.49, 1.16)	0.71 (0.32, 1.57)	0.93 (0.63, 1.38)

^*P < 0.05.

Temperature/rainfall-paratyphoid fever association.

The three-dimensional plot (Figure 4A) shows the relationship between mean temperature and paratyphoid fever cases at different lagged periods. Daily mean temperature was positively associated with paratyphoid fever, with risk ratio increased continuously when temperature above 18°C. Different from typhoid fever, when the temperature at a certain level, the RR of paratyphoid fever decline slightly with the extension of lag days (Figure 4B, Table 2). When temperature was 20°C, RR reached the highest value at lag 8 days (RR = 1.05, 95% CI: 1.03–1.07). Figure 4C shows the exposure-response relationship between daily average temperature and paratyphoid fever at lag 10 days. Similar with typhoid fever, results showed that the highest risk was observed at 33°C (RR = 1.22, 95% CI: 1.09–1.36). Figure 4D presents the cumulative effects of mean temperature on paratyphoid fever up to 20 days. With the mean temperature (18°C) as the reference, the cumulative risk of high temperature (33°C) was 33.40 (95% CI: 12.23–91.19).

Figure 4.

Lagged and cumulative association between temperature/rainfall and paratyphoid fever. The three-dimensional effects diagram of the daily average temperature on paratyphoid fever at the lag of 3–20 days (A), Predictor-specific effects of temperature 20°C on paratyphoid fever (B), Lag-specific effects between temperature and paratyphoid fever at lag 10 days (C), the overall cumulative effects of daily average temperature on paratyphoid fever (D), The three-dimensional effects diagram of the daily total precipitation on paratyphoid fever at the lag of 3–20 days (E), Predictor-specific effects of precipitation 100 mm on paratyphoid fever (F), Lag-specific effects between precipitation and typhoid fever at lag 10 days (G), the overall cumulative effects of daily average temperature on paratyphoid fever (H), Taizhou, China, 2005–2013. This figure appears in color at www.ajtmh.org.

Figure 4E–H shows the exposure-lag-response relationship of daily rainfall on typhoid fever. Results showed that when rainfall exceeded 0 mm, each 100 mm increase in rainfall was associated with appropriately 26% increase in the number of paratyphoid fever cases (Figure 4F, Table 2). Compare with typhoid fever, the RR of paratyphoid fever was positively associated with rainfall with an approximate “J” shape (Figure 4G), and the RR reached the maximum at 240 mm (RR = 2.08, 95% CI: 1.46–2.97). Meanwhile, the cumulative risk of rainfall on paratyphoid fever was much higher than that on typhoid fever.

Table 4 shows the RR of temperature/rainfall on paratyphoid fever by gender, age, and occupation during lag 3–20 days.

Table 4.

The effect of daily average temperature and total rainfall on paratyphoid fever cases along the lag days in different subgroups

Category	Temperature			Rainfall
	Lag3	Lag11	Lag20	Lag3	Lag11	Lag20
Males	1.14 (0.92, 1.41)	1.19 (1.08, 1.31)*	1.22 (1.03, 1.46)*	1.37 (1.11, 1.69)	1.17 (1.06, 1.29)*	1.12 (0.97, 1.29)
Females	1.06 (0.87, 1.30)	1.22 (1.10, 1.36)*	1.31 (1.03, 1.67)*	1.25 (0.98, 1.59)	1.15 (1.01, 1.31)*	1.18 (0.97, 1.42)
0–4	0.76 (0.40, 1.44)	1.55 (1.24, 1.94)*	1.05 (0.61, 1.78)	0.36 (0.05, 2.59)	1.18 (1.03, 1.36)*	1.04 (0.84, 1.30)
5–14	1.27 (0.78, 2.05)	1.29 (0.96, 1.74)	0.76 (0.50, 1.16)	1.08 (0.55, 2.09)	1.44 (0.80, 2.57)	0.45 (0.18, 1.17)
15–60	1.20 (0.99, 1.45)	1.19 (1.09, 1.29)*	1.06 (0.90, 1.24)	1.32 (1.10, 1.57)*	0.93 (0.63, 1.36)	1.48 (0.90, 2.45)
> 60	1.78 (0.99, 3.17)	1.05 (0.81, 1.36)	1.04 (0.65, 1.66)	0.84 (0.43, 1.62)	0.97 (0.77, 1.22)	1.14 (0.97, 1.34)
Farmers	1.17 (0.82, 1.67)	1.21 (1.03, 1.42)*	1.27 (0.95, 1.71)	1.47 (1.02, 2.10)*	1.07 (0.74, 1.55)	0.64 (0.33, 1.25)
Children	1.17 (0.84, 1.61)	1.78 (1.53, 2.08)*	0.70 (0.53, 0.94)	1.32 (0.85, 2.04)	1.16 (1.04, 1.30)*	1.06 (0.76, 1.48)
Students	0.8 (0.44, 1.47)	1.36 (1.04, 1.80)*	1.06 (0.64, 1.74)	0.39 (0.08, 1.8)	1.04 (0.78, 1.38)	1.27 (0.86, 1.88)
Workers and others	1.09 (0.99,1.20)	1.29 (1.05, 1.59)*	1.27 (1.07,1.51)*	1.16 (0.93, 1.44)	1.28 (0.79, 2.08)	0.50 (0.22,1.13)

^*P < 0.05.

The association between high temperature and paratyphoid fever were statistically significant among males (RR = 1.19, 95% CI: 1.08–1.31), females (RR = 1.22, 95% CI: 1.10–1.36), people aged 0–4 years old (RR = 1.55, 95% CI: 1.24–1.94), people aged 15–60 years old (RR = 1.19, 95% CI: 1.09–1.29), farmers (RR = 1.21, 95% CI: 1.03–1.42), children (RR = 1.78, 95% CI: 1.53–2.08), students (RR = 1.36, 95% CI: 1.04–180), and workers (RR = 1.29, 95% CI: 1.05–1.59). No significant association between temperature and paratyphoid fever was detected in people aged 5–14 and > 60 years old. However, the rainfall-related paratyphoid risk vulnerable groups were not found significant in students and workers.

Figure 5 presents the attributable fraction of high temperature and rainfall in different groups. The results showed that in each subgroup, the attributable risk of high rainfall was greater than that of high temperature. And the children and people aged 0–4 years old groups had highest attributable risk of high temperature (typhoid fever: 7.9% and 6.9%; paratyphoid fever: 10.2% and 11.6%) and high rainfall (typhoid fever: 8.8% and 8.2%; paratyphoid fever: 13.6% and 14.5%) among all subgroup.

Figure 5.

The fraction of typhoid/paratyphoid fever morbidity attribute to temperature/rainfall among different subgroups. This figure appears in color at www.ajtmh.org.

Residual and sensitivity analysis.

Our sensitivity analyses showed that the results were stable after changing the df in the temperature space, rainfall space, and lag space, respectively (Supplemental Figures 1–4). Similar results were also observed when changing the number of harmonics (Supplemental Figures 5 and 6) and altering lag days (Supplemental Figures 7 and 8). Supplemental Figures 9 and 10 show the ACF and PACF plots of residuals, which demonstrated no apparent autocorrelation.

DISCUSSION

In this study, we assessed the distinct and different lagged estimated effects of daily mean temperature and precipitation on the incidence of typhoid/paratyphoid fever, identified the vulnerable groups in Taizhou based on time-series surveillance data. An increased risk of typhoid fever with a 15–30 days delay was detected with the increase of temperature above a certain threshold from Taizhou, China. While in paratyphoid fever, the increased risk was associated with the higher temperature with a 3–18 days delay. After the onset of mild precipitation, the RR of typhoid fever increased in a short-lasting and with a 13–26 days delay, and the risk was no significant after the continuous increase of precipitation. Whereas the risk of paratyphoid fever was immediate and long lasting, and increase rapidly with the increase of precipitation. Such findings will provide scientific evidence for the prevention and control of typhoid/paratyphoid fever in the context of climate change with increases in temperature and precipitation.

Similar to our results, temperature and precipitation in a certain range could increase the risk of typhoid/paratyphoid fever has been reported in previous studies. A study in Australia discovered that each 1°C increase in daily mean temperature was associated with a 3.7% increase in salmonellosis cases⁴⁹ while the increase was 15% in New Zealand.⁵⁰ In China, some researchers also had assessed the associations between high temperature and different types of diarrhea, such as bacillary dysentery,⁵¹ noncholera diarrhea,⁵² and typhoid/paratyphoid fever.⁴⁵ A study in Guizhou, southwestern China indicated that high temperature was positively associated with a higher risk of typhoid/paratyphoid fever.⁴⁵ However, they combined the typhoid/paratyphoid fever as one disease, which may lead to some bias due to the different characteristics of the pathogens. Meanwhile, we did not find the statistically significant effect of extreme rainfall on typhoid fever. Deus et al. conducted a study in Blantyre revealed the similar result that extreme rainfall was associated with the reducing of the environmental reservoir of Salmonella.¹⁹ Nevertheless, this result is contradiction with several articles, which reported that extreme rainfall and flood would increase the number of typhoid fever cases,^18,53 stressing the possible importance of sample size and site-specific factors.

The typhoid/paratyphoid fever, often caused by ingestion of faecal-contaminated water and food, usually occur in the regions without clean water or adequate sanitation standard.⁵⁴ Although the incidence of typhoid/paratyphoid fever in Zhejiang province has declined since the implementation of renovation of tubing water and toilet, with the emergence of drug resistance and the Vi vaccine only effective against typhoid fever, it is still prevalent in some coastal regions.⁵⁵ As a coastal city, the residents of Taizhou have the habit of eating seafood, which may increase the risk of typhoid/paratyphoid infection as seafood is the natural carrier of Salmonella. However, the exact pathological and physiological mechanism of the association between temperature and precipitation and typhoid/paratyphoid fever have not been fully understood. High temperature may increase the susceptibility of people to typhoid/paratyphoid directly and indirectly in the following ways: 1) promoting the propagation rate of the pathogenic bacteria in the surrounding environment, such as contaminated food and water,⁵⁶ which was supported by an experimental study;⁵⁷ 2) affecting individual behaviors such as eating habits. During hot days, people are more likely to eat raw or under-cooked foods that may increase the chance of typhoid/paratyphoid infection;⁵⁸ and 3) high temperature may lead people to drink more water, resulting in diluted gastric acid and compromised bactericidal function.⁵⁹ This may trigger the prevalence of typhoid/paratyphoid fever among people.^60,61 Thus, public health actions should be actively taken before the start of high-temperature season to reduce the risks of typhoid/paratyphoid fever. Meanwhile, high rainfall can contaminate surface water in multiple ways, thereby increase the susceptibility of people to typhoid/paratyphoid. First, from a biological perspective, rainfall may provide available growth and reproduction sites for pathogenic bacteria such as pool, puddle, and river, increasing the transmission of water-borne pathogens.⁶² Second, high rainfall may increase pathogens in water in rural areas without sanitary toilets, which is caused by human faeces in rural household toilets.⁶³ As the faecal–oral transmission disease, people suffer from disease in high rainfall surroundings mainly by ingestion water and food contaminated by faeces.⁶⁴ Public education and awareness campaigns about gastrointestinal infectious disease and improvement of water and food supply and optimal allocation of sanitation resources will be effective in diarrheal disease prevention.

Lagged estimated effect of temperature and rainfall on typhoid/paratyphoid fever in our study is consistent with previous meteorological factors-diarrhea studies.^65–67 The days for the lag could include the time from bacteria reproduction in a suitable environment, transmission through contaminated food and water to the incubation period for the onset in infected hosts. Meanwhile, temperature and rainfall variation could influence the speed of disease transmission.⁶⁸ Therefore, temperature and rainfall could be used as an early forecasting indicator for typhoid/paratyphoid fever, as did in other climate sensitive infectious diseases.⁶⁹ In Taizhou, for example, public health intervention strategies against typhoid/paratyphoid fever could be implemented when daily average temperature reaches approximately 30 degrees to prevent the possible more cases after 2 weeks. Such measures should include early warning from local health authority, intensive seafood market inspection, restaurant hygienic standard reinforcement, and more importantly community health education.^70,71 In addition, the popularization of a registered typhoid/paratyphoid fever vaccine may significantly reduce typhoid/paratyphoid fever morbidity in the near future.⁷²

Exploring and comparing the associations between both forms of Salmonella disease and temperature/rainfall can help to develop more specific preventive strategies. For example, our results showed a more delayed effect of temperature on typhoid, which may due to slower clinical course and longer incubation period of typhoid. At the same time, with the expand of lagged days, the relative-risk of typhoid showed an increased trend. This could be explained by two different short- and long-transmission modes. For typhoid fever, the dose of infectious bacteria required is low (about 100 bacteria),⁷³ and is common transmitted from person-to-person.⁴ A higher dose of infectious bacteria is required for paratyphoid fever (about 1,000),⁷⁴ and food (in which Salmonella bacteria can multiply) is considered as the major vehicle for transmission. Furthermore, for paratyphoid fever, in high rainfall surroundings, the water reservoirs in the broader environment combined with the shorter incubation period can further promote the survival and reproduction of pathogens.⁴ The higher bacterium activity shortens the time of occurrence of paratyphoid fever and accelerates the transmission of pathogens between people and people. The different temperature/rainfall disease risk pattern suggests that the optimal control measures for the two diseases may not be the same.

Subgroup analyses suggested that those aged 0–4 years old group, aged 15–60 years old group, farmers, and children are at high risk of temperature/rainfall-attributable typhoid/paratyphoid fever. This could be because that people aged 15–60 years old are the main labor force and have more outdoor activities. Their behaviors such as more likely to drink cool water and eat under-cooked foods during hot days may enable them at high risk of typhoid/paratyphoid fever infection.⁵⁵ And during rainfall days, they may undertake more rescue and outdoor work than other groups, which may lead to a higher possibility of pathogen exposure. Compare with urban residents, farmers who live in rural area may have poor sanitation and health awareness, and they are more likely to eat contaminated food and water in high temperature and rainfall. Children were highly vulnerable to typhoid/paratyphoid fever may related to their immature immune system.

Limitations should be acknowledged in this study. First, under-reporting is inevitable for data extracted from a passive disease surveillance system. To minimized the impact of under-reporting, we included not only clinical but also laboratory-confirmed cases. In our study, the reporting quality remained stable at a high level without a significant variation during the study period.⁷⁵ Therefore, we consider the underreporting has a limited impact on our data analysis. Second, some nonmeteorological factors such as the variation of pathogen, the implementation of vaccination program, socio-economic status, and population immunity level also play an important role in typhoid/paratyphoid fever incidence. However, we did not take these factors into account due to the unavailability of relevant data. Therefore, to accurately quantify the burden of typhoid/paratyphoid fever attributable to temperature/rainfall, we calculated the attributable risk based on the model. Third, the relatively small sample size, especially for the subgroup analyses, dictates cautious interpretation of the results. Therefore, we only divided the cases in limited subgroups based on basic characteristics.

CONCLUSIONS

The positive association between high temperature and high rainfall and typhoid/paratyphoid fever in the regions with a subtropical climate suggests that relevant public health actions should be implemented, considering the lag period, targeting on the most vulnerable populations in the context of climate change. And the disease control or prevention strategies may be different for these two diseases in Taizhou.

Note: Supplemental figures and tables appear at www.ajtmh.org.