Skip to main content
Annals of Saudi Medicine logoLink to Annals of Saudi Medicine
. 2023 Oct 5;43(5):263-276. doi: 10.5144/0256-4947.2023.263

Climate change and cutaneous leishmaniasis in the province of Ghardaïa in Algeria: A model-based approach to predict disease outbreaks

Yasmine Saadene a,, Amina Salhi a, Feriel Mliki a, Zihad Bouslama b
PMCID: PMC10560365  PMID: 37805813

Abstract

BACKGROUND:

Cutaneous leishmaniasis (CL) is a vector-borne disease prevalent in Algeria since 2000. The disease has significant impacts on affected communities, including morbidity and social stigma.

OBJECTIVE:

Investigate the association between environmental factors and the incidence of CL in the province of Ghardaïa and assess the predictive capacity of these factors for disease occurrence.

DESIGN:

Retrospective

SETTING:

The study area included both urban and rural communities.

METHODS:

We analyzed a dataset on CL in the province of Ghardaïa, Algeria, spanning from 2000 to 2020. The dataset included climatic variables such as temperature, average humidity, wind speed, rainfall, and the normalized difference vegetation index (NDVI). Using generalized additive models, we examined the relationships and interactions between these variables to predict the emergence of CL in the study area.

MAIN OUTCOME MEASURES:

The identification of the most significant environmental factors associated with the incidence and the predicted incidence rates of CL in the province of Ghardaïa, Algeria.

SAMPLE SIZE AND CHARACTERISTICS:

252 monthly observations of both climatic and epidemiological variables.

RESULTS:

Relative humidity and wind speed were the primary climatic factors influencing the occurrence of CL epidemics in Ghardaïa, Algeria. Additionally, NDVI was a significant environmental factor associated with CL incidence. Surprisingly, temperature did not show a strong effect on CL occurrence, while rainfall was not statistically significant. The final fitted model predictions were highly correlated with real cases.

CONCLUSION:

This study provides a better understanding of the long-term trend in how environmental and climatic factors contribute to the emergence of CL. Our results can inform the development of effective early warning systems for preventing the transmission and emergence of vector-borne diseases.

LIMITATIONS:

Incorporating additional reservoir statistics such as rodent density and a human development index in the region could improve our understanding of disease transmission.

INTRODUCTION

Leishmaniasis is a vector-borne disease caused by protozoan parasites belonging to the Leishmania genus.1,2 With its continuous expansion, it has become a significant public health concern, affecting nearly 100 countries worldwide. The most common form of leishmaniasis is cutaneous leishmaniasis, accounting for 0.7 to 1.2 million cases per year.3 The first identification of the parasite in Algeria occurred in 1914 in the province of Biskra, but the disease only became prevalent in the 2000s.4 Since then CL has been designated as a notifiable disease, making Algeria the second most affected country after Afghanistan.3 In 2005, the province of Ghardaïa was identified as one of the most high risk areas in Algeria for CL due to the high incidence rates, with almost 2000 cases recorded.5 This disease remains a major cause of morbidity and disfigurement worldwide.6 According to Max Sorre, the association between Leishmania species, their vertebrate hosts, and the sand-fly vector constitutes a pathogenic complex.7 The epidemiological potential of leishmaniasis is influenced by the ecology and behavior of this “Pathogen Complex,” including the natural hosts and vectors of the parasite, which are themselves influenced by environmental factors.8 The life cycle of sandflies, driven by biogeographic circumstances such as climate and vegetation coverage, further supports the sensitivity of leishmaniasis to environmental conditions.913 Understanding the impact of the environment on disease distribution is crucial, especially with the increasing concern about climate change and its potential effects on vector-borne diseases like leishmaniasis, West Nile fever, chikungunya fever, malaria, and Rift Valley fever.1418 Predicting the distribution of these diseases requires a thorough understanding of the transmission environment.

To this end, numerous studies have used modeling techniques to assess and predict the impact of environmental and climatic factors on the transmission of vector-borne diseases. Statistical models have emerged as essential tools in the field of epidemiological prevention.1923 However, in Algeria little is known about climate change and climate impact on disease transmission. To the best of our knowledge, this is the first study in Algeria to examine a large dataset from 2000 to 2020 on CL in the province of Ghardaïa that includes all the climatic variables and the vegetation index. Our study aimed to analyze the prevalence of CL in the Ghardaïa province of Algeria over a period of two decades. Furthermore, we sought to identify the key climatic determinants that influence the distribution and incidence of CL in this region. By providing valuable insights into the potential trajectory of the disease, our findings will contribute to the development of targeted prevention strategies and enhance our understanding of optimal intervention timing within the epidemiological chain of CL.

METHODS

The province of Ghardaïa, located in the center of northern Algeria, is surrounded by desert with an area of 86 650 km2 and a population of 488 965 spread over 13 communes (Figure 1).24 Its climate is known for being arid with poor vegetation cover where rainfall is low and irregular, while temperatures are highest in July (36°C) and lowest in January (12°C).26 In this region, CL is caused by Leishmania major and Leishmania killicki transmitted by the sand fly Phlebotomus papatasi and Phlebotomus sergenti and the main reservoirs are rodents: Meriones shawi, Psammomys and Massoutiera mzabi.27

Figure 1.

Figure 1.

Geographical map of the province of Ghardaïa, Algeria (Quantum Geographic Information System).

The number of monthly CL cases recorded in Ghardaïa and throughout all of Algeria from 1 January 2000 to 1 December 2020 was transmitted in Excel format from the Algerian Ministry of Health as part of the collection of epidemiological data on the disease. Monthly meteorological data for the province of Ghardaïa (minimum temperature, maximum temperature, average temperature, humidity, rainfall and wind speed) from 1 January 2000 to 1 December 2020 were collected from the meteorological station 605660 (DAUG) on the website TuTiempo.net.28

The normalized difference vegetation index (NDVI) is a commonly used vegetation index that can provide insights into changes in vegetation cover over time. For this study, NDVI data was obtained from Copernicus Global Land Services.29 The data covered multiple municipalities in the Ghardaïa Province (Ghardaïa, Metlili, Sebseb, Bounoura, Dhayet Ben Houa, El Atteuf, Bounoura, Zelfana) for which the mean NDVI of each month was calculated from January 2000 to June 2020. The calculations were based on 10-daily raster data, which means that the data provides an average NDVI value for every 10-day period. These values were derived from satellite images captured by the Sentinel-2 sensor, with a resolution of 1 km. The satellite images capture the reflectance of near-infrared and visible light wavelengths.

As an initial step toward modeling the relationship between CL and environmental variables in the province, we conducted a descriptive analysis of the data spanning 20 years (Table 1). This involved calculating the minimum, maximum, mean and standard deviation values of the environmental factors, as well as determining correlation coefficients (Table 2) to identify any potential relationships. These results served as a foundation for our subsequent modeling analysis.

Table 1.

Summary of cutaneous leishmaniasis cases and associated environmental factors in the province of Ghardaïa over 20 years.

Variable Cutaneous leishmaniasis Average temperature (°C) Maximum temperature (°C) Minimum temperature (°C) Humidity (%) Precipitation (mm) Wind speed (km/h) NDVI
Count (number of observations) 252 252 252 252 252 252 252 252
Median (IQR) 17 (36) 22.5 (15.0) 28.5 (15.4) 16.2 (13.8) 33.5 (17.9) 2.0 (7.2) 13.2 (4.4) 0.14 (0.01)
Minimum 0 8.6 13.4 3.3 15.3 0 4.5 0.1127
Maximum 383 38.0 43.9 31.4 64.7 74.18 21.1 2.197

NVDI: normalized difference vegetation index.

Table 2.

Computed correlations and statistical significance by the Spearman method with listwise deletion.

CL AT MXT MNT HM PRC WS NDVI
Average temperature (°C) ‒0.442 (<.001)
Maximum temperature (°C) ‒0.454 (<.001) 0.998 (<.001)
Minimum temperature (°C) ‒0.418 (<.001) 0.996 (<.001) 0.992 (<.001)
Relative humidity (%) 0.512 (<.001) ‒0.853 (<.001) ‒0.859 (<.001) ‒0.822 (<.001)
Precipitation (mm) 0.002 (‒975) ‒0.067 (.292) ‒0.078 (.218) ‒0.040 (.525) 0.268 (<.001)
Wind speed (km/h) ‒0.365 (<.001) ‒0.034 (.586) ‒0.039 (.536) ‒0.054 (.396) ‒0.183 (.004) ‒0.015 (‒816)
Normalized difference vegetation index 0.505 (<.001) ‒0.697 (<.001) ‒0.698 (<.001) ‒0.682 (<.001) 0.643 (<.001) 0.004 (.954) ‒0.221 (<.001)

We employed a generalized additive model (GAM) to meet the requirements of both challenging predictive tasks and explanatory ones. Explanatory analysis aimed to identify the relevant risk factors that contribute to the emergence of CL and to understand the dynamics of the disease. The predictive model was employed to assist policymakers in making informed decisions to prevent the spread of the epidemic.

GAMs were developed by Hastie and Tibshirani.30 They are a widely used statistical tool in ecology for modeling the relationships between predictor variables and the response variable.3134 They provide a flexible approach for estimating smooth functional relationships, which allows for a more accurate representation of the complex associations that often exist in ecological systems.

The GAMs were represented by Wood by the following formula:38

graphic file with name 0256-4947.2023.263-eq1.jpg

Where g is a link function, yi denotes a response variable, Xi* is a row of the model matrix for any strictly parametric model components, the vector 0 contains fixed parameters and the fi are smooth splines of the explanatory covariates (environmental factors) Xi. The specification of smooth functions can be achieved by using either piecewise polynomial functions or basis functions. The junctions where the polynomial segments are joined together are known as knots, which are commonly denoted by k. In the case of time series studies where climatic factors are the predictors, cubic spline functions are the most widely adopted.38

In these splines, if bk is the ki cubic basis function, then f is represented as:

graphic file with name 0256-4947.2023.263-eq2.jpg

Where k is the total number of nodes and with each node a corresponding set of unknown parameters represented by Bk. In the case of using cubic splines, the resultant curve will be composed of a sequence of cubic polynomials that exhibit continuity in terms of their value from the first and to the second derivative.39

The data were divided into training and testing sets to develop and evaluate a model for predicting CL incidence. The training set contained 80% of the data and was used to develop the models, while the remaining 20% was used to evaluate the model’s performance. Several environmental covariates, including rainfall, vegetation index, humidity, temperature, and wind speed, were selected as predictors of CL incidence. To identify the best lag of these covariates, a cross-correlation analysis was conducted with a maximum lag of 12 months and the only lagged variable selected was the rainfall factor (PRC_Lag6) with a lag of six months.

Two GAMs, GAM_01 and GAM_02, were fitted to the data. GAM_01 served as the baseline model containing only environmental predictors, while GAM_02 included an additional predictor; a time variable (year). The inclusion of the time variable aimed to capture any potential temporal trends or yearly variations in the CL incidence.

GAMs were employed, incorporating smooth terms for all environmental covariates and the lagged rainfall variable. The smoothness of these terms was determined using cubic regression splines with knots placed at mid-months of the year. The models were fitted using the restricted maximum likelihood method and the negative binomial distribution to account for the count nature of the outcome variable. The final model performance was evaluated using various metrics such as the Akaike information criterion (AIC), the Bayesian information criterion (BIC), R-squared value (R2), restricted maximum likelihood score (REML) and the root-mean-square error (RMSE). The analysis was conducted using the mgcv (mixed GAM computation vehicle with automatic smoothness estimation) package in R Studio 4.2.3 with R package ggplot2 3.4.2.

RESULTS

Cutaneous leishmaniasis has been reported in Algeria since at least 1860, with historical records documenting its presence in the country. However, from 2000 onwards, there has been an apparent increase in the number of reported cases, indicating a resurgence of the disease in Algeria.2

During the study period from 2000 to 2020, the Algerian Ministry Of Health recorded a total of 217 741 cases of human CL in Algeria (Figure 2), while the province of Ghardaïa reported a total of 9328 identified cases, indicating that approximately 4.3% of the overall cases have been concentrated in this area (Figure 3). The maximum annual incidence of CL cases in Algeria occurred in 2005, 2010, 2011 and 2017 with 23 072, 21043, 16585 and 11771 cases, respectively during a period of increasing population size from 3 023 510 in 2000 to 44 300 000 in 2020. The same pattern of incidence of CL in Ghardaïa is shown in the graph where the peaks occurred in 2005 and 2011 as well as a little increase in 2020 in a population size from 344 429 to 488 965. Figure 4 reveals a clear seasonal trend in Ghardaïa, with a higher incidence of cases occurring in the winter months from October to March, and a lower incidence in the summer months from April to September. Specifically, the highest number of cases was reported in November, December, and January, with the maximum number of cases ever recorded in this region with 383 in November 2005, while the lowest number of cases was recorded between the periods of April to September. These findings suggest that seasonal factors may play an important role in the incidence of CL.

Figure 2.

Figure 2.

Incidence of cutaneous leishmaniasis in Algeria by year (2000–2020).

Figure 3.

Figure 3.

Incidence of cutaneous leishmaniasis in Ghardaïa by year (2000–2020).

Figure 4.

Figure 4.

Cutaneous leishmaniasis incidence in Ghardaïa by months 2000–2020.

In total, 252 monthly observations for both epidemiological and environmental variables were registered, from 1 January 2000 to 1 December 2020, providing sufficient statistical power for analysis. Focusing on CL cases and multiple climatic factors in the Ghardaïa region of Algeria as temperature (°C) (maximum, minimum and average) (MXT, MNT, AT), relative humidity (%) (HM), precipitation (mm) (PRC) and wind speed (km/h) (WS). In addition, the normalized difference vegetation index (NDVI) was used to assess the extent of green vegetation coverage in the study area.

The median (IQR) values for average, maximum and minimum temperatures, humidity, precipation, wind speed, and NDVI are reported in Table 1. These statistics provide valuable insights into the environmental conditions in the study area and serve as a basis for further analysis of the relationship between CL cases and environmental factors.

The correlation coefficients showed that the number of CL cases and temperatures (AT, MXT, MNT) were negatively correlated (P<.001) (Table 2). Conversely, the number of CL cases was positively correlated with HM and NDVI (P<.001). WS had weak negative correlations with the number of CL cases (P<.001), but the PRC correlation was nonsignificant. Our results suggest that environmental factors such as temperature, humidity, wind speed and vegetation cover are important predictors of the incidence of CL in the study area of Ghardaïa.

These significant correlations identified between environmental factors and CL incidence were further explored through GAMs to develop a more comprehensive understanding of the occurrence of CL. We fitted two GAMs, denoted as GAM_01 and GAM_02. GAM_01 included only environmental predictors, while GAM_02 incorporated a temporal variable to assess the influence of time trends on model performance in predicting disease incidence. The results show that certain environmental factors had differential effects between the two models. In GAM_01, humidity, wind speed, and the vegetation index were statistically significant predictors of the response variable (P<.005). In contrast, average temperature and rainfall lagged by six months and were not significant predictors of the response variable (P>.005). In GAM_02, which included a temporal variable “year” in addition to the environmental predictors, humidity, vegetation index and the temporal predictor were statistically significant predictors (P<.005), while the effect of wind speed was weakly significant in this model (P>.005). The GAM_01 model explained 35% of the deviance, while the GAM_02 model explained 67% of the deviance.

There were some notable differences between GAM_01 and GAM_02 models, as indicated by the estimated degrees of freedom with their respective chi-squared statistics and P values for GAM_01 and GAM_02 (Table 3). First, the addition of a time variable is reflected in the estimates and standard error of the intercept, with GAM_02 having a slightly lower estimate than GAM_01 (3.1291 vs 3.36440) and a smaller standard error. Second, in smooth terms, both models include the same environmental variables (HM, AT, WS, NDVI, and PRC_Lag6). However, the estimated degrees of freedom (EDF) and corresponding chi-squared test statistics differ between the two models. For instance, the smooth term for HM has an EDF of 3.591 in GAM_01 and 4.282 in GAM_02. With the exception of the HM predictor, the smooth terms for all other variables in GAM_01 exhibit higher EDFs compared to those in GAM_02. In general, it appears that the smooth terms in GAM_02 have smaller EDFs and larger chi-squared test statistics compared to those in GAM_01. Moreover, the P values for the smooth terms in GAM_02 are generally smaller than those in GAM_01, indicating stronger evidence against the null hypothesis of no smooth effect. For instance, the smooth term for HM in GAM_02 has a P value of <2e-16, while it is 0.000111 in GAM_01.

Table 3.

Summary of environmental and temporal effects on response variables using generalized additive models.

Model Intercept Smooth term Effective degrees of freedom Chi-squared P value
GAM_01 Environmental variables Estimate (SE) 3.36440 (0.06937) P<2e-16 HM 3.591 18.30 .0001
AT 0.242 0.28 .2855
WS 2.140 10.29 .0021
NDVI 3.436 22.72 3.35e-05
PRC_Lag6 0.005 0.01 .347312
GAM_02 Environmental variables + Time variable Estimate (SE) 3.1291 (0.0514) P<2e-16 HM 4.282 38.71 <2e-16
AT 3.500e-04 0.00 .8085
WS 1.709 5.09 .0289
NDVI 3.294 15.22 .001
PRC_Lag6 4.583e-04 0.00 .9335
year 13.50 161.50 <2e-16

Effective degrees of freedom of the smooth function terms>1 = nonlinear relationships

Overall, it seems that the GAM_02 model, which includes the time variable, has smaller P values for the smooth terms and may provide a better fit to the data than GAM_01.

Figures 5 and 6 depict the fitted functions for the models GAM_01 and GAM_02, respectively, which demonstrate the association between environmental factors and the incidence of CL. The results show that during wet seasons, when relative humidity is between 25% and 55%, the incidence of the disease increases. However, when the relative humidity exceeds 55%, the incidence of the disease decreases. Wind speed is another important factor, with a wind speed less than 10 km/h favoring the dispersion of the disease. Conversely, a wind speed of more than 10 km/h was not conducive to disease transmission. The relationship between CL and vegetation index is complex, but a small presence of vegetation index above 0.1 can be a risk factor for the emergence of CL. The incidence of the disease is slightly higher when the average temperature is between 20°C and 30°C. However, the fitted function for rainfall does not provide any insight into its relationship with the disease.

Figure 5.

Figure 5.

Relationship between environmental variables and cutaneous leishmaniasis incidence in the province of Ghardaïa, Algeria 2000–2020 based on GAM_01.

Figure 6.

Figure 6.

Relationship between environmental variables and cutaneous leishmaniasis incidence in Ghardaïa, Algeria 2000–2020 based on GAM_02.

Table 4 presents the model fit statistics for the two fitted GAMs. The Akaike information criteria and Bayesian information criteria values for GAM_02 are lower compared to GAM_01, suggesting an improved model fit. The R-squared adjusted value is also higher for GAM_02 compared to GAM_01, indicating that GAM_02 explains more of the variability in the data. Additionally, the REML value is higher for GAM_01 than for GAM_02, indicating that GAM_02 is a better model fit.

Table 4.

Model fit statistics for the fitted generalized additive models (GAM_01 and GAM_02).

Model Akaike information criteria Bayesian information criteria Adjusted R squared Restricted maximum likelihood score
GAM_01 1744.724 1790.518 0.198 879.55
GAM_02 1631.247 1723.84 0.613 839.24

Both GAM_01 and GAM_02 showed no significant correlations in their ACF and PACF plots, with no lines crossing the significance level (Figure 7). This suggests that the data is stationary and there is no autocorrelation between the observations at any lag, indicating that the models are a good fit for the data. The residual plots of GAM_01 and GAM_02 (Figure 8) were evaluated to assess the goodness-of-fit of the models. The residual plots showed no apparent patterns or trends, indicating that the models fit the data well. In GAM_01, the residual plot appeared to have a random scatter of points around the zero line, with no visible outliers. The residual plot for GAM_02 showed a random scatter of points around the zero line, with a few visible outliers. However, these outliers were deemed to be normal since count data tend to have some level of over dispersion, and removing them would result in a loss of important information. Overall, the residual plots of both models suggested that they were good fits for the data.

Figure 7A.

Figure 7A.

Autocorrelation and partial autocorrelation function plots for GAM_01 autocorrelation function.

Figure 8A.

Figure 8A.

Residual plots of the fitted model GAM_01.

Figure 7B.

Figure 7B.

Autocorrelation and partial autocorrelation function plots for GAM_02 autocorrelation function.

Figure 8B.

Figure 8B.

Residual plots of the fitted model GAM_02.

Predictions

To assess the adequacy and predictive performance of our models, we conducted a correlation analysis between the predicted values generated by the fitted models and the actual CL cases with a separate test dataset that represents the remaining 20% of the data used in our analysis. This allowed us to determine how closely the models predictions aligned with the actual observed data. We also calculated the RMSE to measure the average magnitude of the errors in the model predictions. The combination of these measures gives us a comprehensive understanding of how well the models fit the data and how accurately they can predict future outcomes.

Based on the correlation coefficient and RMSE values, both models are reasonably accurate in predicting the number of CL cases (Table 5). However, the correlation coefficient of GAM_02 is higher than GAM_01, indicating that GAM_02 has a stronger linear relationship with the test data. Additionally, the RMSE of GAM_02 is lower than GAM_01, indicating that GAM_02 predictions have less error than GAM_01.

Table 5.

Cutaneous leishmaniasis case prediction model metrics.

Model Correlation coefficient Root mean square error
GAM_01 0.621 52.511
GAM_02 0.906 29.726

The prediction plots (Figure 9) indicate that both models follow a similar pattern, with the predicted values tracking closely to the actual values. However, the predictions of GAM_02 appear to fit the actual data more closely than GAM_01 as the predicted line in the GAM_02 plot is closer to the actual values. This further supports the fact that GAM_02 is a more accurate model for predicting the number of CL cases. Additionally, we can see from the confidence intervals that the predictions of GAM_02 are more precise than GAM_01.

Figure 9A.

Figure 9A.

Cutaneous leishmaniasis cases prediction plots for GAM_01 with 95% predictive interval using new data (20% of the original data).

Figure 9B.

Figure 9B.

Cutaneous leishmaniasis cases prediction plots for GAM_02 with 95% predictive interval using new data (20% of the original data).

DISCUSSION

Cutaneous leishmaniasis has been prevalent in Algeria since 2000, especially in the province of Ghardaïa. While the occurrence of CL is not stable, it reappears cyclically every 5–7 years across most regions of the country which is in line with the findings of Toumi et al.42 The distribution of CL cases exhibited a distinct seasonal pattern providing insight about the months when the disease is most prevalent in humans. However, it is important to consider that the incubation period of the pathogen responsible for symptomatic CL can range from weeks to months. Therefore, while the seasonal pattern offers useful information, it is crucial to take a long-term view when studying the epidemiology of CL.43,44

The analysis using GAMs has revealed that the most significant climatic factor affecting the occurrence of CL epidemics is relative humidity. We found a highly significant correlation coefficient and both GAMs confirmed the relationship with very significant P values and EDFs. These findings suggest that CL cases are highly responsive to humidity levels in the environment. In fact, previous research by has demonstrated that relative humidity is a better predictor of CL incidence than rainfall.42 Indeed, the survival and reproduction of sandflies, the vectors of various diseases including leishmaniasis, are greatly influenced by humidity levels and they are more likely to thrive and multiply in higher humidity environments, increasing the risk of disease transmission.45

According to a study conducted in Morocco, higher relative humidity levels were positively correlated with increased sandflies abundance which is in line with our findings that’s suppose that a range of humidity from 25% to 55% increases enormously the number of CL cases in the Ghardaïa region.46 Another study conducted in Biskra province in Algeria found out the same correlation coefficient between humidity and CL cases.47 However, a study in Golestan province, Iran has confirmed that precipitation is the most important factor in the density of CL vectors.48 Sandflies need a certain amount of moisture for their development and survival, and heavy rainfall can also kill adults and immature stages of sandflies, which is why rainfall is generally negatively associated with CL occurrence.4951 Due to the complex and indirect effects of precipitation on vegetation, rodents, and vectors, the resulting consequences are difficult to capture, and they may vary from one region to another.52 Our GAMs analysis did not reveal any significant association between rainfall and the occurrence of CL cases. It is likely that the amount and duration of rainfall in the desert region of Ghardaïa over our study period of 20 years were not sufficient to have a significant impact on the incidence of CL. Moreover, recent years have seen even less rainfall in the region, further diminishing the potential impact of rainfall on CL transmission. These findings suggest that other factors, such as relative humidity and temperature may be more important in influencing the incidence of CL in Ghardaïa, a new drought zone.

The analysis of NDVI data revealed that the region of Ghardaïa has a very low vegetation cover, with a maximum NDVI of 0.21. Spearman correlation showed that vegetation cover is positively associated with the number of CL cases. The limited impact of vegetation cover on the incidence of CL can be observed from the GAM outputs, where the effect is only prominent within the range of 0.01 to 0.2, potentially indicating the true dynamic of the disease as some other studies proved that the limited vegetation cover creates favorable conditions for sand flies by increasing temperature and evaporation.53 In fact a study on the presence of P papatasi, the major vector of CL, was conducted using a genetic algorithm for rule-set production and maximum entropy techniques. Both models revealed that the central and northern regions of the province, with lowland areas, were more susceptible to P papatasi propagation compared to the southern parts, characterized by mountainous and forested areas.54 A study conducted in Iran in 2014 revealed that the southern parts of Golestan province, where vegetation indices were high, had the lowest incidences of CL. Conversely, the northern parts of the Golestan province, which have arid and semi-arid climates and low vegetation indices, had townships with the highest CL incidence rates. This suggests that the lack or low coverage of vegetation in these areas provides the favorable ecological niche for the occurrence of CL.55 Actually, sand flies are attracted to decomposed organic material, which serves as the primary source of nourishment for the larvae of these insects.56 They breed in dark, humid environments that have a supply of organic matter that serves as food for the larvae. However, they have a strong preference for different sources of decomposing organic matter, often from fecal sources.57

The results of our study indicated that wind speed had an impact on the incidence of CL, as evidenced by the negative correlation coefficient. Our findings are consistent with those of a previous study conducted in southern Iran, which also reported a negative correlation between wind speed and the spatial distribution of leishmaniasis.58 However, we also found that wind speed has a complex effect on CL incidence, with wind speeds between 5 to 10 km/h favoring an increase in cases, while speeds above 10 km/h led to a decrease in cases. It is important to consider the strength of wind speed in disease control and prevention strategies, as it can influence the transmission of CL in various ways. Wu et al in 2016 proposed that wind speed has a dual effect on CL cases, decreasing the likelihood of insect bites while increasing their flight range.59 Similarly, another study conducted in Iran in 2018 confirmed the role of wind speed as one of the multiple climatic factors affecting the emergence of CL cases in the study area, using a decision tree model.53

The correlation analysis between temperatures (average, maximum, and minimum) and CL cases revealed a significant negative relationship. These findings suggest that while ambient temperature is a good condition for disease transmission, an increase in temperature is not conducive to the development of CL. However, the GAM results showed that the temperature impact on the emergence of CL is limited, with effects concentrated in a narrow range of 20°C to 30°C. This range of temperature appears to be particularly favorable for CL incidence due to environmental conditions that promote the proliferation of phlebotomine vectors. These vectors prefer temperatures in the same range so temperature plays a crucial role in the transmission of CL.60 As demonstrated by studies from different parts of the world, such as Panama, Afghanistan, Algeria and Tunisia.19,6163 However, temperatures outside the range of 20°C to 30°C can also have a significant impact on the emergence of CL. For example, a small increase of 1°C to 3°C in temperature can affect the growth of vector populations and decrease their density for the next year.64 This decrease could result in a reduction in the number of CL cases in the Ghardaïa region. However, this also highlights the fact that global warming in North African cities and Europe poses a significant risk for the transmission of CL particularly and vector-borne diseases in general in the near future.

The GAMs results showed that adding the time trend improved the model fit statistics and the accuracy of the model predictions. Specifically, the REML value improved, which indicates a better fit of the model to the data, meaning that it explains more of the variability in the data. Additionally, the AIC value decreased suggesting a better trade-off between model complexity and goodness of fit. A lower AIC value indicates that the model is simpler while still explaining a large proportion of the variability in the data.65 The adjusted R squared also increased. An R squared of 0.19 indicates that only 19% of the variability in the response variable can be explained by the predictors included in the model, while an R squared of 0.6 explains 60% of the variability in the response variable. This suggests that the additional predictors included in the model have significantly improved the ability of the model to explain the response variable.

We compared the predictive performance of both models. The results showed that GAM_02 had a stronger linear relationship between its predictions and the test data, while GAM_01 had a lower correlation coefficient. This indicates that GAM_02 had a better predictive performance than GAM_01. Further evidence of this can be observed in the prediction plots, where the red line represents the predicted values of the models. For GAM_02, the red line follows the true trend of the disease, indicating that GAM_02 accurately predicted the disease dynamics.

However, for GAM_01, the red line almost explains the variation of the disease but does not accurately capture the true dynamics between 2005 to 2010. This suggests that while GAM_01 performed well in predicting disease outcomes from ecological variables in the model, it failed to detect any temporal aspects of the CL incidence. This underscores the significance of considering the time trend of any disease when making predictions and assessing incidences. Incorporating the time trend as a predictor adds a crucial element to the model, enhancing its predictive capabilities. These findings are consistent with previous research that has demonstrated the importance of model selection in accurately predicting outcomes.66

By examining the interactions between multiple predictors, we identified specific values and thresholds of each factor that may trigger the transmission of the disease. The most significant predictors of CL in Ghardaïa were one environmental factor (vegetation index) and two climatic factors (relative humidity and wind speed). Our findings also highlight the importance of considering temporal variability when analyzing disease incidence. However, this study could be improved by incorporating additional reservoir statistics such as rodent density and economic/human development index in the region of Ghardaïa. This would allow for the development of a more comprehensive predictive model and an effective early warning system for decision-making and preventing the transmission and emergence of the disease in new cities.

Funding Statement

None

Footnotes

CONFLICT OF INTEREST: None.

REFERENCES

  • 1.Desjeux P. Leishmaniasis: Current situation and new perspectives. Comp Immunol Microbiol Infect Dis. 2004. Sep;27(5):305–318. doi: 10.1016/j.cimid.2004.03.004. [DOI] [PubMed] [Google Scholar]
  • 1a.Mokni M. Leishmanioses cutanées. Ann Dermatol Venereol. 2019. Mar;146(3):232–246. doi: 10.1016/j.annder.2019.02.002. [DOI] [PubMed] [Google Scholar]
  • 2.Alvar J, Vélez ID, Bern C, Herrero M, Desjeux P, Cano J, Jannin J, den Boer M; WHO Leishmaniasis Control Team. Leishmaniasis worldwide and global estimates of its incidence. PLoS One. 2012;7(5):e35671. doi: 10.1371/journal.pone.0035671. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Sergent ED, Sergent ÉT, Parrot L, Donatien A, Beguet M.. Transmission du clou de Biskra par le phlébotome (Phlebotomus papatasi Scop.). CR Acad Sci. 1921;173:1030. [Google Scholar]
  • 4.www.insp.dz/images/PDF/Epidemio/REM%20annuel2017f.pdf (5)
  • 5.World Health Organization. Control of the leishmaniases. Report of a meeting of the WHO Expert Committee. Last visit : 29 April 2023 [Google Scholar]
  • 6.Sorre M. Complexes pathogènes et géographie médicale. Ann Geogr. 1933;42:1–18. [Google Scholar]
  • 7.Dedet JP. Leishmaniasis: A growing problem due to environmental changes and human behaviors. Bull Acad Natl Med. 2007;191(8):1579–1588. doi: 10.1016/s0001-4079(19)32909-7 [DOI] [PubMed] [Google Scholar]
  • 8.Pigott DM, Bhatt S, Golding N, Duda KA, Battle KE, Brady OJ, et al.. Global distribution maps of the leishmaniases. Elife. 2014;3:e02851. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Guernier V, Hochberg ME, Gue Guernier V, Hochberg ME, Guégan J-F.. Ecology drives the worldwide distribution of human diseases. PLoS Biol. 2004. Jun 15;2(6):e141. doi: 10.1371/journal.pbio.0020141 PMID: 15208708. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Chaves LF, Pascual M.. Climate cycles and forecasts of Cutaneous Leishmaniasis, a nonstationary vector-borne disease. PLoS Med. 2006;3(8):e295. doi: 10.1371/journal.pmed.0030295 PMID: 16903778; PMCID: PMC1539092. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Li Y, Zheng C.. Associations between meteorological factors and visceral leishmaniasis outbreaks in Jiashi County, Xinjiang Uygur Autonomous Region, China, 2005–2015. Int J Environ Res Public Health. 2019;16:1775. doi: 10.3390/ijerph16101775 PMID: 31137482; PMCID: PMC6571646. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Nikonahad A, Khorshidi A, Ghaffari HR, Aval HE, Miri M, Amarloei A, et al.. A time series analysis of environmental and meteorological factors’ impact on cutaneous leishmaniasis incidence in an endemic area of Dehloran, Iran. Environ Sci Pollut Res. 2017;24:14117–23. doi: 10.1007/s11356-017-8962-0 PMID: 28417326. [DOI] [PubMed] [Google Scholar]
  • 13.Semenza JC, Menne B:. Climate change and infectious diseases in Europe. [DOI] [PubMed] [Google Scholar]
  • 14.DH: Health effects of climate change in the UK 2008: an update of the department of health report 2001/2002. [Internet]. Available from: http://webarchive.nationalarchives.gov.uk/20130107105354/ http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/PublicationsPolicyAndGuidance/DH_080702.
  • 15.Hunter PR. Climate change and water-borne and vector-borne disease. J Appl Microbiol. 2003;94(Suppl):37S–46S. [DOI] [PubMed] [Google Scholar]
  • 16.McMichael AJ, Woodruff RE, Hales S.. Climate change and human health: present and future risks. Lancet. 2006;367:859–869. [DOI] [PubMed] [Google Scholar]
  • 17.WHO: Protecting health from climate change: global research priorities. [Internet]. Available from: http://www.who.int/global-change/publications/9789241598187/en/.
  • 18.Talmoudi K, Bellali H, Ben-Alaya N, Saez M, Malouche D, Chahed MK.. Modeling zoonotic cutaneous leishmaniasis incidence in central Tunisia from 2009-2015: Forecasting models using climate variables as predictors. PLoS Negl Trop Dis. 2017;11(8):e0005844. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Cheong YL, Burkart K, Leitão PJ, Lakes T.. Assessing weather effects on dengue disease in Malaysia. Int J Environ Res Public Health. 2013;10:6319–6334. PMID: 24287855. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Colon-Gonzalez FJ, Fezzi C, Lake IR, Hunter PR.. The effects of weather and climate change on dengue. PLoS Negl Trop Dis. 2013;7(11):e2503. PMID: 24244765. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Bouzid M, Colón-González FJ, Lung T, Lake IR, Hunter PR.. Climate change and the emergence of vector-borne diseases in Europe: case study of dengue fever. BMC Public Health. 2014;14:781. PMID: 25149418. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Ma W, Sun X, Song Y, Tao F, Feng W, et al.. Applied Mixed Generalized Additive Model to Assess the Effect of Temperature on the Incidence of Bacillary Dysentery and Its Forecast. PLoS One. 2013;8(4):e62122. PMID: 2363797827. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Abid, P. L.(n.d.). La couverture sanitaire de la wilaya de Ghardaïa
  • 24.Benyoucef B. Le M’zab:. Les pratiques de l’espace. Alger: Entreprise nationale du livre; 2004. [Google Scholar]
  • 25.Harrat, Z., Chawki, S., Pratlong, F., Benikhlef, R., Selt, B., Pierre, J., Ravel, C., & Belkaid, M. (2009). Description of a dermatotropic Leishmania close to L. killicki (Rioux, Lanotte & Pratlong 1986) in Algeria. Trans R Soc Trop Med Hyg, 103(7), 716–720. 10.1016/j.trstmh.2009.04.013 [DOI] [PubMed] [Google Scholar]
  • 26.Garni R, Tran A, Guis H, Baldet T, Benallal K, Boubidi S, Harrat Z.. Remote sensing, land cover changes, and vector-borne diseases: Use of high spatial resolution satellite imagery to map the risk of occurrence of cutaneous leishmaniasis in Ghardaïa, Algeria. Infect Genet Evol. 2014;28:725–734. [DOI] [PubMed] [Google Scholar]
  • 27.Climate Data for Ghardaïa. Tutiempo. Available from: https://en.tutiempo.net/climate/ws-605660.html. Accessed February 28, 2023. [Google Scholar]
  • 28.Normalized vegetation index for Ghardaïa. Copernicus global land services. Available from: https://land.copernicus.vgt.vito.be/PDF/portal/Application.html#Browse;Root=513186;Collection=1000321;Time=NORMAL,NORMAL,-1,,,-1. Accessed March 15, 2023.
  • 29.Hastie T, Tibshirani R.. Generalized Additive Models. London: Chapman & Hall; 1990. [DOI] [PubMed] [Google Scholar]
  • 30.Aidoo EN, Adebanji AO, Awashie GE, Appiah SK.. The effects of weather on the spread of COVID-19: evidence from Ghana. Bull Natl Res Cent. 2021;45(1). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Pedersen EJ, Miller DL, Simpson GL, Ross N.. Hierarchical generalized additive models in ecology: An introduction with mgcv. Peer J. 2019;5. doi: 10.7717/peerj.6876. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Ma W, Sun X, Song Y, Tao F, Feng W, He Y, Zhao N.. Applied Mixed Generalized Additive Model to Assess the Effect of Temperature on the Incidence of Bacillary Dysentery and Its Forecast. PLoS ONE. 2013;8(4):1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Khouloud T, Hedia B, Nissaf BA, Marc S, Dhafer M, Kouni CM.. Comparative Performance Analysis for Generalized Additive and Generalized Linear Modeling in Epidemiology Methods of Evaluation for Modeling Disease Incidence. IJACSA Int J Adv Comput Sci Appl. 2017;8(12). [Google Scholar]
  • 34.Wood SN.. Generalized Additive Models: An Introduction with R. New York: Chapman & Hall/CRC; 2006. [Google Scholar]
  • 35.Wood SN. Generalized Additive Models: An Introduction with R. 2nd edition. Chapman & Hall/CRC Press; Taylor & Francis; 2017. [Google Scholar]
  • 36.Ma W, Sun X, Song Y, Tao F, Feng W, He Y, Zhao N.. Applied Mixed Generalized Additive Model to Assess the Effect of Temperature on the Incidence of Bacillary Dysentery and Its Forecast. PLoS ONE. 2013;8(4):1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Wood SN, Pya N, Säfken B.. Smoothing Parameter and Model Selection for General Smooth Models. J Am Stat Assoc. 2016;111(516):1548–1575. [Google Scholar]
  • 38.R Development Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2015. [Google Scholar]
  • 39.Kuhn M, Johnson K.. Applied Predictive Modeling. Springer; 2013. [Google Scholar]
  • 40.Ready PD. Biology of phlebotomine sand flies as vectors of disease agents. Annual Review of Entomology. 2013;58:227–250. [DOI] [PubMed] [Google Scholar]
  • 41.Toumi A, Chlif S, Bettaieb J, Alaya NB, Boukthir A, Ahmadi ZE, Salah AB.. Temporal dynamics and impact of climate factors on the incidence of Zoonotic Cutaneous Leishmaniasis in central Tunisia. PLoS Neglected Tropical Diseases. 2012;6(5). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Jones TC, Johnson WD Jr, Barretto AC, Lago E, Badaro R, Cerf B, Reed SG, Netto EM, Tada MS, Franca TF.. Epidemiology of American cutaneous leishmaniasis due to Leishmania braziliensis. J Infect Dis. 1987. Jul;156(1):73–83. [DOI] [PubMed] [Google Scholar]
  • 43.Melby PC. Experimental leishmaniasis in humans: review. Rev Infect Dis. 1991;13(5):1009. [DOI] [PubMed] [Google Scholar]
  • 44.Talbi FZ, El Ouali Lalami A, Fadil M, Najy M, Ech-Chafay H, Lachhab M, Lotfi S, Nouayti N, Lahouiti K, Faraj C, Janati Idrissi A.. Entomological Investigations, Seasonal Fluctuations and Impact of Bioclimate Factors of Phlebotomines Sand Flies (Diptera: Psychodidae) of an Emerging Focus of Cutaneous Leishmaniasis in Aichoun, Central Morocco. J Parasitol Res. 2020;2020:6495108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Selmane S. Dynamic relationship between climate factors and the incidence of cutaneous leishmaniasis in Biskra Province in Algeria. Annals of Saudi Medicine. 2015;35(6):445–449. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Hosseini SH, Allah-Kalteh E, Sofizadeh A.. The Effect of Geographical and Climatic Factors on the Distribution of Phlebotomus papatasi (Diptera: Psychodidae) in Golestan Province, an Endemic Focus of Zoonotic Cutaneous Leishmaniasis in Iran, 2014. J Arthropod Borne Dis. 2021. Jun 30;15(2):225–235. PMID: 35111860; PMCID: PMC8782746. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Kasap OE, Alten B.. Comparative demography of the sand fly Phlebotomus papatasi (Diptera: Psychodidae) at constant temperatures. J Vector Ecol. 2006. Dec;31(2):378–385. [DOI] [PubMed] [Google Scholar]
  • 48.Simsek FM, Alten B, Caglar SS, Ozbel Y, Aytekin AM, Kaynas S, Belen A, Kasap OE, Yaman M, Rastgeldi S.. Distribution and altitudinal structuring of phlebotomine sandflies (Diptera: Psychodidae) in southern Anatolia, Turkey: their relation to human cutaneous leishmaniasis. J Vector Ecol. 2007. Dec;32(2):285–291. [DOI] [PubMed] [Google Scholar]
  • 49.Kasap OE, Alten B.. Laboratory estimation of degree-day developmental requirements of Phlebotomus papatasi (Diptera: Psychodidae). J Vector Ecol. 2005. Dec;30(2):328–333. [PubMed] [Google Scholar]
  • 50.Fichet-Calvet E, Jomaa I, Ben Ismail R, Ashford RW.. Leishmania major infection in the fat sand rat Psammomys obesus in Tunisia: interaction of host and parasite populations. Ann Trop Med Parasitol. 2003. Oct;97(7):593–603. [DOI] [PubMed] [Google Scholar]
  • 51.Ramezankhani R, Sajjadi N, Esmaeilzadeh RN, Jozi SA, Shirzadi MR.. Application of decision tree for prediction of cutaneous leishmaniasis incidence based on environmental and topographic factors in Isfahan province, Iran. Geospatial Health. 2018;13(1):172–178. [DOI] [PubMed] [Google Scholar]
  • 52.Mollalo A, Alimohammadi A, Shahrisvand M, Reza Shirzadi M, Reza Malek M.. Spatial and statistical analyses of the relations between vegetation cover and incidence of cutaneous leishmaniasis in an endemic province, northeast of Iran. Asian Pac J Trop Dis. 2014;4(3):176–180. [Google Scholar]
  • 53.Romo Bechara N, Wasserberg G, Raymann K.. Microbial ecology of sand fly breeding sites: aging and larval conditioning alter the bacterial community composition of rearing substrates. Parasit Vectors. 2022;15(1). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Šlapeta J.. Sand flies of animals. In: Merck Veterinary Manual (Professional Version). Last review/revision Aug 2022 | Modified Oct 2022. Retrieved May 12, 2023.
  • 55.Ali-Akbarpour M, Mohammadbeigi A, Tabatabaee SHR, Hatam G.. Spatial analysis of eco-environmental risk factors of cutaneous leishmaniasis in southern Iran. J Cutan Aesthet Surg. 2012;5:30–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Wu X, Lu Y, Zhou S, Chen L, Xu B.. Impact of climate change on human infectious diseases: Empirical evidence and human adaptation. Environ Int. 2016;86:14–23. [DOI] [PubMed] [Google Scholar]
  • 57.Depaquit J, Grandadam M, Fouque F, Andry PE, Peyrefitte C, Kamhawi S.. Arthropod-borne viruses transmitted by phlebotomine sandflies in Europe: a review. Eurosurveillance. 2009;14(40):19330. [PubMed] [Google Scholar]
  • 58.Chaves LF, Calzada JE, Valderrama A, Saldaña A.. Cutaneous Leishmaniasis and Sand Fly Fluctuations Are Associated with El Niño in Panamá. PLoS Negl Trop Dis. 2014;8(10). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Adegboye OA, Adegboye M.. Spatially correlated time series and ecological niche analysis of cutaneous leishmaniasis in Afghanistan. Int J Environ Res Public Health. 2017;14(3). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Khezzani B, Bouchemal S.. Demographic and spatio-temporal distribution of cutaneous leishmaniasis in the Souf oasis (Eastern South of Algeria): Results of 13 years. Acta Trop. 2017;166:74–80. [DOI] [PubMed] [Google Scholar]
  • 61.Bounoua L, Kahime K, Houti L, Blakey T, Ebi KL, Zhang P, Imhoff ML, Thome KJ, Dudek C, Sahabi SA, Messouli M, Makhlouf B, el Laamrani A, Boumezzough A.. Linking climate to incidence of zoonotic cutaneous leishmaniasis (L. major) in pre-Saharan North Africa. Int J Environ Res Public Health. 2013. Aug;10(8):3172–3191. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.Bhattacharya PK, Burman P.. Linear Models. In: Bhattacharya PK, Burman P, editors. Theory and Methods of Statistics. Academic Press; 2016. p. 309–382. ISBN 9780128024409. [Google Scholar]
  • 63.Chowdhury MZI, Turin TC.. Variable selection strategies and its importance in clinical prediction modelling. Fam Med Community Health. 2020. Feb 16;8(1):e000262. doi: 10.1136/fmch-2019-000262. PMID: 32148735; PMCID: PMC7032893. [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from Annals of Saudi Medicine are provided here courtesy of King Faisal Specialist Hospital and Research Centre

RESOURCES