Spatial and seasonal analysis of human leptospirosis in the District of Gampaha, Sri Lanka

Introduction: Leptospirosis is a zoonotic infectious disease, caused by a pathogenic species of the Genus Leptospira. In recent years, a markedly increased number of leptospirosis cases have been reported in the District of Gampaha, in the Western Province of Sri Lanka. Typically, the risk of the disease in the district is seasonal with a small spike which occurs in March to May and a large spike occurring from October to December. Objectives: To analyze spatial and seasonal patterns of human leptospirosis and to predict the leptospirosis epidemic trend in the District of Gampaha, Sri Lanka. Methods: All Divisional Secretariats (DS) of the District of Gampaha were selected for the study. Epidemiological data were obtained from the Regional Epidemiological Unit, Gampaha. The leptospirosis cases were georeferenced according to the DS where these cases were reported. The cumulative incidence and the fatality were calculated for each DS. Results: Of the georeferenced data, highest mean (±S.E.) of the number of leptospirosis cases (72.60 ±15.54) were observed from DS of Mirigama. The highest mean cumulative incidence (4.97±1.10) and case fatality rate (3.88±2.42) were observed from DS of Divulapitiya and Katana respectively. According to the past 10 years data on leptospirosis, highest mean numbers of leptospirosis cases were reported in March (51.00±12.99) and November (56.80±8.27). A predictive model for clinically confirmed human leptospirosis was designed for the district by using TSA package of the statistical software R. Conclusions: This study provides an evidence base for reducing disease burden by improving understanding of the dynamic patterns of the disease in the District of Gampaha, Sri Lanka.


Introduction
Leptospirosis is a globally important zoonotic disease which affects humans and animals in both developing and developed countries with humid, tropical and subtropical climates. 1,2 It is caused by infection with pathogenic spirochetes of the Genus Leptospira. 1 Leptospirosis is an important emerging infectious disease in Sri Lanka. 3 This disease has a significant effect on livelihoods of people living in the endemic areas of the country. Approximately 4300 cases per year and several outbreaks of leptospirosis have been reported in the country. 4 Scientifically sound data and scientific publications are lacking in Sri Lanka due to lack of laboratory diagnostic facilities and baseline data on transmission of leptospirosis. 5 Laboratory confirmation of patients with leptospirosis is important for preventing severe complications and death. 1 Most feral and domestic mammals may serve as major reservoir hosts for Leptospira. 6,7 Further, the knowledge on reservoir animals and potential risk factors among general population are also lacking in Sri Lanka. The environment, economic activities, hygienic conditions, recent ecological changes and climate in Sri Lanka contribute to provide an ideal environment for leptospirosis transmission 5 and led to an upsurge of leptospirosis cases in the country. Laboratory diagnosis is limited to a few laboratories in the country and awareness programs on leptospirosis are generally confined to the period of an outbreak in the country. 8 Environmental data, Geographical Information System (GIS), spatial statistical analysis, predictive risk maps and models have been used for the investigation and management of a range of infectious diseases including leptospirosis. 5,9 Spatial and seasonal analysis of human leptospirosis in the District of Gampaha is needed as the district has become one of the endemic/epidemic areas in the island. 4,10 Risk maps identify geographic areas with high disease prevalence and/or risk of outbreaks and are useful for guiding allocation of scarce public health resources and interventions. Such maps are particularly useful where disease surveillance data are poor or lacking. To date, no predictive risk maps and models have been produced for leptospirosis in the District of Gampaha and there is a need for a leptospirosis early warning system to predict when and where leptospirosis epidemics may occur within the district. Further, unusual meteorological conditions, such as high rainfall and flooding are often cited retrospectively as the precipitating factors for epidemics and might be useful in predicting the number of cases of leptospirosis. Consequently, if risk periods can be predicted, early steps can be taken to prevent infections, thus reducing the monthly number of cases in the district.
The objectives of this study were to analyze spatial and seasonal distribution of leptospirosis, identify possible ecological components affecting transmission of leptospirosis and to provide an evidence base for reducing disease burden by improving the understanding of the environmental drivers of the disease in the District of Gampaha, Sri Lanka.

Materials and Methods
The District of Gampaha is located in the Western Province and Wet Zone. All 13 Divisional Secretariats (DS) of the district were selected for the study. Monthly number of clinically confirmed cases and deaths of leptospirosis reported in each DS of the District of Gampaha from 2004 to 2013 was obtained from the Regional Director of Health Office, Gampaha after obtaining permission from the Director. Meteorological data such as minimum, maximum and average temperature in degrees Celsius, average rainfall in millimeters and average humidity of the District of Gampaha from 2004 to 2013 was obtained from the Meteorological Department of Colombo.
The leptospirosis cases were georeferenced according to the DS where these cases were reported during the past five years in the District of Gampaha. The cumulative incidence (number of cases/population 11 living in the DS×10,000) and the case fatality rate (number of deaths/number of cases×100) were calculated manually for each DS 12 from the data of the past five years of each DS. Pearson Correlation Coefficient (r) between climatic data such as mean, maximum, minimum, temperature, mean rainfall, mean relative humidity and leptospirosis cases were calculated using Minitab 17.0 software package. P value of 0.05 or lower was considered to be significant.
A predictive time series model with exogenous variables, an Auto-Regressive Integrated Moving-Average with eXogenous variable (ARIMAX) model was prepared to predict the leptospirosis epidemic trend in the District of Gampaha using the TSA package of the statistical software R. 13,14 It was based on clinically confirmed leptospirosis epidemic data and meteorological data (mean, maximum and minimum temperature in degrees Celsius, mean rainfall in millimeters and mean relative humidity) of the District of Gampaha. According to the regression analysis, highest cross-correlation value with significant lags of each meteorological time series data was selected to fit a time series regression (ARIMAX) model for leptospirosis cases. Significant lags of meteorological parameters were then selected based on p-values. Auto Correlation Function (ACF) and Partial Autocorrelation Function (PACF) were analyzed to identify the seasonal and cyclic trends in time series of cases of leptospirosis. After the identification of time series pattern, p the number of Auto Regressive (AR), d the number of Differencing (D) and q the number of Moving Average (MA) parameters were determined based on residual analysis of ACF and PACF. All possible ARIMAX models were analyzed. The best-fitting model was selected based on the goodness of fit criteria, Akaike's Information Criterion (AIC) and diagnostic plot values. The predictive analysis was then performed on the cases of leptospirosis in the following year using the most appropriate model and the average values of monthly meteorological parameters of the previous 10 years.

Results
Disease frequency and geographical distribution: Of the georeferenced data, highest mean (±S.E.) number of leptospirosis cases (72.6 ±15.54) were observed from Mirigama DS (Figure 1). The highest mean incidence (4.97±1.10) and case fatality rate (3.88±2.42) were observed from DS of Divulapitiya and Katana respectively (Figures 2 and 3). According to the past 10 years data on leptospirosis, the highest mean numbers of leptospirosis cases were reported in March     (Table 1). Significant positive cross-correlation was detected between the number of cases of leptospirosis and monthly minimum temperature measured 4 months previously (r=0.194, P<0.05) but significant strongest negative cross-correlation was detected in lag 9 months. Significant negative cross-correlation was detected between the number of cases of leptospirosis and monthly mean rainfall measured 2, 8, 9 months previously but significant strongest negative cross-correlation was detected in lag 8 months (r=-0.251, P<0.05). Significant negative cross-correlation was detected between the number of cases of leptospirosis and monthly mean relative humidity measured 7, 8 and 9 months previously but significant strongest negative cross-correlation was detected in lag 8 months (r=-0.392, P<0.05) ( Table 1).
Predictive model for human leptospirosis: Monthly number of clinically confirmed leptospirosis cases of the past 10 years was graphed. As can be seen in the ACF plot, the time series appears to be stationary by Dickey-fuller test (P< 0.01) ( Figure 5) without differencing (d=0) the time series. Further, there is no seasonality of the time series data because there is no cyclic pattern of the peak of ACF plot). ACF and PACF plots were used identify the order of MA and AR terms ( Figure 5 and 6). According to that ACF and PACF plots analysis p was 2, q was 1 and d was zero. Then all possible models were considered. Cross-correlation between the pre-whitened climatic series and leptospirosis series were determined to incorporate the climatic factors as input series in the regression model. Of the highest significant correlations, only lagged 2 maximum temperature, lagged 8 relative humidity, and lagged 8 rainfall were significant in the model. Several orders of non-seasonal ARIMAX models were tested with significant climatic factors. Most appropriate model was selected based on smallest AIC and diagnostic plot. Finally, ARIMAX (0, 0, 1) model with lagged 2 maximum temperature, lagged 8 relative humidity, and lagged 8 rainfall was selected as the best model to predict the leptospirosis cases (Figure 7). Actual and predicted number of leptospirosis cases based on the predicted model for 2014 are shown in the table 2.

Figure 5
Autocorrelation function for number of leptospirosis cases (with 5% significant limits of the autocorrelation)

Figure 6
Partial autocorrelation function for number of leptospirosis cases (with 5% significant limits of the partial autocorrelation)

Discussion
According to the analyzed data, leptospirosis cases have been reported in all 13 DS of the District of Gampaha every year but geographic location revealed that the disease risk was higher in rural areas where paddy cultivation is high compared to the urban areas such as DS of Mirigama, Attanagalla and Divulapitiya. Less than 10 cases have been reported from urban/slum DSs such as Wattala, Ja-Ela, Negombo and Kelaniya.   Correlation analysis results show the link between numbers of cases of leptospirosis with the considered meteorological parameters. This seasonal pattern of cases of leptospirosis in the district follows the seasonal variations in rainfalls characterized by monsoon seasons (Yala and Maha). In Yala season, rainfall was more than 5000 mm and this heavy rainfall results in an increased transmission of leptospirosis. During the rainy season the soil remains moist and can lead to formation of pools of water which help leptospires survive for longer periods of time, and ultimately can lead to an increase in human and animal exposure to the bacteria. 9,15 During dry periods, high concentrations of Leptospira in the soil are limited to a few meters around waste accumulation sources. 12 In the District of Gampaha, the rainy season period is about 6-8 months every year. 16 Climatic factors of lag months may also have an effect on the rodent population. In a rainy and humid environment, rodent reproduction will increase because of availability of excess food resources. Few months later, an increase in the rodent population may lead to competition for food resources which may result in wide dispersal of rodents in the environment. 17,18. Therefore, people in the district may be exposed to Leptospira during agricultural activities.
During floods, Leptospira can reach distant areas through water which increases the possibility of contact with a wider spectrum of the population resulting in the possibility of an outbreak. Hence, Public Health authorities need to increase awareness programmes. However, it is difficult to take preventive measures during the period of monsoon rains, especially in endemic areas in the district.
Generally, temperature fluctuation in the district varies between 23 ºC and 31 ºC throughout the year. 16 Mean monthly temperature of the district ranges between 26 ºC and 29 ºC. A temperature of between 28 ºC and 30 ºC is optimum for growth of Leptospira. Leptospires are able survive longer periods in a warm and humid environment. In the present study, correlation between distinct climatic factors with incidence of leptospirosis is considerable even though combination of meteorological parameters rather than any distinct meteorological parameter might greatly affect the number of human leptospirosis cases.
Climatic factors (rainfall, number of rainy days, mean temperature and mean humidity) showed statistically non-significant correlation with leptospirosis cases in a previous study. 19 According to their study, two districts (Ratnapura and Anuradhapura) showed a significant correlation of cases of leptospirosis only with rainfall. High humidity and heavy rain fall played an important role in the 2011 leptospirosis outbreak in Sri Lanka, possibly because the extensive rainfall that year resulted in flooding in many areas of the country. 19 Other studies have also shown high positive correlation between leptospirosis cases and climatic factors. 12,20 In general, considerable correlation between the number of leptospirosis cases and rainfall, relative humidity, temperature were observed in the present study. These results would be valuable for district level health managers in targeting effective disease control activities.
The prepared model in this study can show monthly tendency of the leptospirosis incidence variations but cannot predict extreme incidences of leptospirosis due to environmental and climatic parameters which may greatly influence the number of cases. In the current study, the actual data matched the predicted data of the model perfectly, within the predicted 95% confidence interval. Although the prediction accuracy of the ARIMAX model was high the model can be modified including other environmental, human behavioral and climatic factors that influence the survival of Leptospira such as dissolved oxygen in water and pH of water. However, ARIMAX model can be used for short-term forecasts with useful guidance for timely prevention and control measures to reduce the disease burden of the district.
Analysis of georeferenced maps, modeling and forecasting of expected leptospirosis cases could be useful for development and implementation of public health education and control of disease incidence. In addition, several environmental management practices and educational campaigns can be launched to limit activities that contribute to the transmission of leptospira in the endemic areas of the districts. Further, this type of georeferenced maps and model can be used for other endemic areas of the country with similar environmental and climatic conditions.
The current study has some limitations because the epidemiological data that was obtained from the Epidemiological Unit were clinically confirmed leptospirosis cases and not laboratory confirmed as leptospirosis. Further, the study was an ecological study conducted using available secondary climatic data. There are several limitations of getting climatic data because of biased distribution of data collecting centers in the district. Ecology research studies covering the whole country is needed to test the findings of the current study and to take control and preventive measure to reduce disease burden in the country.