Effects of daily mean temperature and other meteorological variables on bacillary dysentery in Beijing-Tianjin-Hebei region, China

Abstract

Background: Although previous studies have shown that meteorological factors such as temperature are related to the incidence of bacillary dysentery (BD), researches about the non-linear and interaction effect among meteorological variables remain limited. The objective of this study was to analyze the effects of temperature and other meteorological variables on BD in Beijing-Tianjin-Hebei region, which is a high-risk area for BD distribution.

Methods: Our study was based on the daily-scale data of BD cases and meteorological variables from 2014 to 2019, using generalized additive model (GAM) to explore the relationship between meteorological variables and BD cases and distributed lag non-linear model (DLNM) to analyze the lag and cumulative effects. The interaction effects and stratified analysis were developed by the GAM.

Results: A total of 147,001 cases were reported from 2014 to 2019. The relationship between temperature and BD was approximately liner above 0 °C, but the turning point of total temperature effect was 10 °C. Results of DLNM indicated that the effect of high temperature was significant on lag 5d and lag 6d, and the lag effect showed that each 5 °C rise caused a 3% [Relative risk (RR) = 1.03, 95% Confidence interval (CI): 1.02-1.05] increase in BD cases. The cumulative BD cases delayed by 7 days increased by 31% for each 5 °C rise in temperature above 10 °C (RR = 1.31, 95% CI: 1.30-1.33). The interaction effects and stratified analysis manifested that the incidence of BD was highest in hot and humid climates.

Conclusions: This study suggests that temperature can significantly affect the incidence of BD, and its effect can be enhanced by humidity and precipitation, which means that the hot and humid environment positively increases the incidence of BD.

Keywords: Bacillary dysentery; Distributed lag non-linear model; Generalized additive model; Nonlinear and interaction effects.

Fig. 1

The time-series distributions of BD and mean temperature in Beijing-Tianjin-Hebei region. From the peak of the cases of BD and the mean temperature, there is a certain lag effect between them.

Fig. 2

The relationship curves between meteorological variables and BD. The daily mean temperature had a nonlinear effect on BD and showed an approximately linear trend above 0 °C. Abbreviations: Temp, daily mean temperature; Wind, daily mean wind speed; Pre, precipitation; RH, relative humidity; SSD, sunshine duration.

Fig. 3

The total effect of daily mean temperature. The risk of BD had a hardly increase when the temperature was lower than 10 °C, while it caused a substantial increase when higher than 10 °C.

Fig. 4

The visualization of the lag effect. The 3D graph (a) and contour plot (b) of lag effect show that the lag effect is strongest on the 5th and 6th days. The concrete lag effect on lag 5d and cumulative effect in 15 °C (c) indicate that BD cases on lag 5d increased by 3% and the cumulative BD cases delayed by 7 days increased by 31% for each 5 °C rise.

Fig. 5

The modification and stratified analysis by other meteorological variables on the effect of mean temperature. The modification effect of relative humidity on temperature is positive (a), and that of wind speed is reverse (c). There is a positive relationship between the incidence of BD and precipitation during 0∼20 mm and reverse relationship with precipitation exceeding 20 mm (b). In our research, the precipitation in the Beijing-Tianjin-Hebei region was defined as the total precipitation of these three places, and its distribution was significantly left skewed, concentrate in 0∼20 mm. In the total of 2191 days from 2014 to 2019, the number of days with precipitation exceeding 20 mm was 138 days, accounting for only 6%. Thus, the number of BD cases with precipitation exceeding 20 mm was not representative, and to a certain extent could not represent the incidence of BD in heavy precipitation. However, the risk of BD with precipitation is still higher than that without precipitation.

Fig. 6

Relative risk (RR) stratified by wind speed, precipitation, relative humidity and sunshine duration. Squares represent the RR and error bars indicate the 95% CI.

Fig. 7

The 3D graph (a∼c), contour plot (d∼f) of lag effect in Beijing, Tianjin and Hebei, respectively.