Introduction

Leptospirosis is one of the most widespread zoonotic disease globally [1]. The Leptospira bacteria present a complex transmission cycle that can include several species of mammalian reservoirs. These reservoirs shed pathogenic serovars through their urine, contaminating water and soil where the bacteria can survive for long periods [2]. The disease is considered to be primarily an occupational infection, associated with groups who handle animals or animal tissues and subsistence farming in tropical countries. In these regions, leptospirosis exhibits an endemic pattern of occurrence mainly in rural settings [3]. It is also considered to be an emerging infectious disease with outbreaks frequently associated with extreme rainfall events, such as hurricanes and flooding [2,4,5], and often cited as an urban slum health problem in resource-poor countries [6,7].

Annually, leptospirosis causes an estimated one million cases and around 60,000 deaths worldwide [8], though the actual burden may be higher due to underreporting. The reference method for serological diagnosis is the microscopic agglutination test (MAT), which can be time-consuming, and availability is frequently limited, especially in low and middle-income countries [1]. Tropical islands are considered a particular high-risk setting for human infection [8,9] and a recent study finding one-third of all globally reported outbreaks occurred in Latin America and the Caribbean (LAC) [10]. While recent outbreaks of the disease have been primarily associated with floods and hurricanes, this pattern coexists with the endemic form in this region. Drivers and risk factors for each epidemiological context are specific and remain poorly explored in most countries [11].

The adoption of spatial modelling methods for studying infectious diseases has contributed to a better understanding of the distribution of seroprevalence and risk factors and drivers of many diseases [12–14], including leptospirosis [15]. Cluster analysis methods are spatial statistical tests that assess geographical variation in disease risk and/or occurrence, and identify areas where the number of events (e.g., participants testing positive for leptospirosis seromarkers) exceeds expectations based on the size of the population analysed [16]. Increasing evidence suggests a spatial clustering of leptospirosis infection [17–21], indicating that some areas may provide more favourable conditions for sustaining transmission cycles than others.

The complex interactions between ecology, epidemiology, multiple animal hosts, and human conditions in urban and rural settings, identifying characteristics associated with populations living in high-risk areas can provide insights to guide targeted public health action. However, the importance of risk factors and drivers can vary across space [22] due to local differences in the complex transmission cycle [23], such as the presence of different reservoir mammals and their intrinsic relationship with serovars [5], vaccination rate of livestock herds [24], suitable conditions for pathogen survival in the environment [2] and opportunities for human exposure [1,5]. The variation of suitable conditions for transmission can occur within and between regions, with high-risk areas experiencing more localised outbreaks [17,20,21]. By pinpointing high-risk areas, spatial cluster analysis enables healthcare managers to generate hypotheses for understanding higher prevalence in these areas. This study aims to identify clusters of leptospirosis seroprevalence in the Dominican Republic (DR) and determine the factors associated with these high-risk areas.

Methods

Data acquisition and processing

Survey.

A three-stage cross-sectional national serosurvey was conducted in the DR, between 30 June and 12 October 2021. The survey sampling design has already been described previously [29]. In summary, a total of 10 communities were sampled from Espaillat Province and 13 from SPM. In both Espaillat and SPM, 60 households were selected from each community. All household members above 5 years of age were invited to participate. Participants included in the study were interviewed by a trained local fieldwork team. The questionnaire included individual and household-level questions. Venous blood was collected from all participants, and household Global Positioning System (GPS) coordinates were recorded.

Venous blood was processed as sera, and frozen at -80°C. Samples of participants from the two provinces were tested by MAT to detect anti-Leptospira antibodies. Serological analyses were performed at the Centers for Disease Control and Prevention’s Zoonoses and Select Agent Laboratory, Bacterial Special Pathogens Branch, Atlanta, USA. A panel of 20 pathogenic serovars were selected for the MAT panel, and titres ≥1:100 were considered seropositive and indicative of prior infection. Serovars identified by MAT were used to determine the putative serogroups associated with infections, as previously described [30]. In summary, analyses were conducted on the serogroup level. If a sample reacted to multiple serogroups, the serogroup with the highest titre was considered the main one. When the highest titre was observed across two or more serogroups the sample was recorded as ‘mixed’.

Spatial data.

Environmental, sociodemographic and census data was obtained from publicly available sources. All spatial data considered for the analyses are described in Table 1. These data were extracted for the household location of each participant and incorporated into the survey data.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Environmental and sociodemographic data considered for the characterisation of high-risk areas of leptospirosis seroprevalence in the Dominican Republic, Jun-Oct, 2021.

https://doi.org/10.1371/journal.pntd.0013103.t001

Results

Across the two provinces included in this analysis, there were 2091 participants enrolled in the survey. Thirteen records were excluded because there were incomplete data from the questionnaire, resulting in a final sample of 2,078 participants (802 in Espaillat and 1,276 in SPM). A total of 237 seropositive participants were identified (seroprevalence of 11.4%); 127 (15.8%) in Espaillat and 110 (8.6%) in SPM.

In Espaillat, the four main serogroups were Icterohaemorragiae (58.5%), Australis (21.2%), Canicola (14.1%) and Djasiman (9.7%). In SPM, Icterohaemorrhagiae was the most prevalent serogroup (33.3%) followed by Australis (17.2%) and Pyrogenes, Djasiman and Canicola (all three with 11 cases, 10.0%). Across the two provinces, all but one Pyrogenes cases occurred in SPM (S2 Table).

Local clustering

FleXScan identified one statistically significant cluster in Espaillat and one marginally statistically significant in SPM, for all serogroups combined (Figs 1 and 2). In Espaillat, the identified cluster was located in the southwest of the province and included eight households, with a maximum distance between households of 3.1km (p-value 0.044). (Fig 1). The households inside the cluster represent 1.5% of the households surveyed and were home to 7.1% of all seropositive participants in this province. For participants living in a household within the identified cluster, RR was 6.7 (95%CI 5.7-7.9) compared to participants living in a household outside the identified cluster. For participants living in a community where the cluster was identified, the RR was 1.6 (95% CI 1.1-3.2) compared to participants living in a non-cluster community. An additional 11 secondary (non-significant) clusters were identified (S5 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Location of clusters identified by flexible spatial scan in Espaillat, Dominican Republic, Jun-Oct 2021.

A) Household’s link based on nearest neighbour determined from the Thiessen Polygon, households identified as a hotspot and their connection are highlighted in red. B) Sampled communities in Espaillat with the cluster communities highlighted in red. Base layer from: https://data.humdata.org/dataset/cod-ab-dom.

https://doi.org/10.1371/journal.pntd.0013103.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Location of clusters identified by flexible spatial scan in San Pedro de Macoris, Dominican Republic, Jun-Oct 2021.

A) Household’s link based on nearest neighbour determined from the Thiessen Polygon, households identified as a hotspot and their connection are highlighted in red. B) Sampled communities in Espaillat with the cluster communities highlighted in red. Base layer from: https://data.humdata.org/dataset/cod-ab-dom.

https://doi.org/10.1371/journal.pntd.0013103.g002

In SPM, the identified cluster was located in the northeast of the province and included six households, with a maximum distance between households of 10.5km (p-value 0.08). (Fig 2) The households inside the cluster represent 0.8% of the households surveyed and were home to 6.4% of all cases in this province. For participants living in a household within the identified cluster, RR was 9.6 (95%CI 5.6-14.2) compared to participants living in households outside the identified cluster. For participants living in a community where the cluster was identified the RR was 2.0 (1.3-2.9) compared to participants living in a non-cluster community. Seven secondary (non-significant) clusters were identified (S6 Fig). Local cluster analysis for main serogroups in both provinces did not identify significant clusters.

Logistic regression

Results from the non-parametric regression identified that when compared to participants outside the cluster, participants inside the cluster in Espaillat were significantly older (OR 2.22; p-value 0.026), worked in outdoor work environments (OR 4.05, p-value 0.019), were more likely to live in households with no access to piped water (OR 4.72, p-value 0.041), at greater distance to water bodies (OR 1.44, p-value 0.023) and health facilities (OR 1.86, p-value<0.001 by motorized travel time and OR 1.36, p-value 0.001 by walking travel time), and surrounded by greater areas of range land (OR 1.8, p-value 0.007) (Table 2). Participants inside the cluster were also less likely to live in households at greater distance to major roads (OR 0.42, p-value<0.001), at higher altitudes (OR 0.58, p-value<0.001), located in areas of higher population density (OR 0.76, p-value<0.001) and higher gross domestic product (GDP) at the community level (OR 0.70, p-value<0.001), compared to participants outside the cluster.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Odds ratio of each variable associated with household from a hotspot for all serogroups combined by province, Dominican Republic, Jun-Oct, 2021.

https://doi.org/10.1371/journal.pntd.0013103.t002

Participants inside the cluster in SPM, compared with participants outside the cluster, were significantly more likely to be male (OR 3.89, p-value 0.030) and live in households at greater distance to health facilities (OR 4.41, p-value 0.044 by motorized travel time and OR 4.67, p-value 0.021 by walking travel time) and educational facilities (OR 4.36, p-value 0.017), located at higher altitudes (OR 1.15, p-value 0.036) and greater soil moisture (OR 1.97, p-value<0.001). Participants inside the cluster were also less likely to live in households located in areas of higher population density (OR 0.29, p-value<0.001), higher GDP at the community level (OR 0.42, p-value<0.001), and surrounded by a greater percentage of built-up areas (OR 0.32, p-value 0.019).

Discussion

Our study identified two clusters of leptospirosis seroprevalence in each of the two provinces included in the analysis, with significant associations between cluster and sociodemographic and environmental characteristics. Across both provinces, participants living in households located farther from health facilities and in areas with lower population density and GDP were more likely to belong to the identified clusters. These shared characteristics highlight the role of poverty-associated factors in shaping the epidemiological profile of leptospirosis transmission. However, specific risk factors varied between provinces, reflecting the specific sociodemographic, economic, and ecological contexts. By examining these provincial differences, this study underscores the complexity of leptospirosis transmission dynamics and emphasizes the need for tailored public health interventions to address local drivers of disease.

Characteristics associated with participants living inside the clusters identified by FleXScan can be divided into two groups: common to both provinces and specific to each province. In both provinces, participants living in households located at greater distances from education and health facilities were more likely to be inside the clusters, and participants living in households located in areas of higher population density and higher GDP were less likely to be inside the clusters. These associations suggest that the households inside the identified clusters presented characteristics frequently associated with impoverished communities. The association between leptospirosis and poverty have been extensively studied [2,47–50]. Populations living in resource-poor communities are more prone to overcrowded housing conditions, and have inadequate access to safe water, sanitation, waste management and other infrastructure [7]. These conditions can facilitate the transmission of the Leptospira bacteria and offer limited or non-existent barriers to prevent human exposure to contaminated environments [1,7].

In this study, specific risk factors associated with clusters in each province reflected local differences and the complexity of the transmission cycle. Specific characteristics of the population living inside the Espaillat cluster, such as outdoor work environment, households with limited access to piped water and located at a greater distance to major roads might reinforce the characterisation of impoverished communities [7], but can also suggest a rural setting. Similarly, the association of clusters with a higher percentage of range land surrounding the households indicates a rural setting, implying the presence of livestock as an important reservoirs [5]. In this province, those findings can also mirror a higher percentage of population residing in rural setting [27,51,52]. Although the specific factors associated with the SPM cluster provide fewer factors to characterise the high-risk area, lower percentage of built-up areas surrounding the households and greater distance to the province capital were observed. In SPM, as the population residing in rural settings represents a lower percentage of the total population [27,51,52] and other economic activities have greater importance, these findings might suggest that subsistence farmers, further away from larger cities are more vulnerable to leptospirosis transmission. Finally, clusters in each province presented opposite directions of association with elevation. While Espaillat is located inland, SPM is a coastal province. The SPM cluster is located in the northeast of the province, with greater distance to the coast compared to the other communities included in the study, which may impact the higher median elevation of the cluster. Additionally, the difference in the median elevation between households inside and outside the cluster in Espaillat is small, with a non-negligible overlap of variation range between them. In the two provinces included in this study, although the association was significant, it was not linear and should be interpreted based on the geographical context of each province.

This study has several limitations. The relationship between serogroups and reservoir mammals can be highly specific, and in turn, routes of human exposure are related to exposure to the reservoir mammals. Thus, characteristics of clusters of high leptospirosis seroprevalence are likely to vary based on the predominant serogroup. In our analysis, the small sample size limited local cluster analysis based on the serogroup predominant in each province, which could have provided more comprehensive insights into specific risk factors and drivers of transmission. The main serogroups identified in Espaillat (results shown in the S2 Table), Australis, associated with wild and domestic animals, rats and mice, and Icterohaemorrhagie, mainly associated with rodents [3,53]. In this province, clusters were not associated with self-reported rat exposure, yet several characteristics associated with clusters were indicative of remote impoverished, agricultural communities, which could be associated with the presence of reservoirs of Australis and Icterohaemorrhagie. In SPM, the main serovar identified was Australis, followed by Canicola, Djasiman, Icterohaemorrhagie and Pyrogenes equally distributed, which suggests a broader range of reservoirs driving transmission in the province. Our data were obtained from a cross-sectional survey, although this study design has the advantage of providing a more extensive population-level seroprevalence characterisation, it was limited to one point in time. Water-borne diseases such as leptospirosis are strongly affected by climatic conditions, and transmission might vary according to seasonal variations in rainfall and temperature [1,9]. However, our study did not identify association between clusters and precipitation. The survey design may have impacted the semivariogram results. The sampling design aimed for a national representative and widespread spatial distribution of sampled communities as described previously [29]. However, by sampling few communities in each province, few households in each community, and including all members of each household sampled, the spatial distribution of the final sample was, to some extent, clustered. To identify positive spatial autocorrelation, the seroprevalence clustering needed to overcome the clustered spatial distribution of our sample, which might have been limited due the reduced sample size in each province. A clustered spatial pattern of leptospirosis cases has been reported on several occasions, by both global and local clustering methods [17–21,54].

This study focused on leptospirosis seroprevalence in Espaillat and SPM, yet the concepts and methods could be applied to other locations in the country and the other regions, as well as to other infectious diseases strongly driven by environmental and sociodemographic factors in these provinces. Our results reinforce the findings of previous studies that identified clustered spatial pattern of leptospirosis and provide empirical evidence that drivers of transmission can be highly specific to the context of each province. In both provinces, the identified clusters of leptospirosis seroprevalence were associated with characteristics which suggests higher transmission in impoverished communities and most likely in rural settings. However, the translation of this insight into modifiable factors varied between provinces. Aiming to achieve cost-effective prevention and control measures will require public health and environmental interventions specific to each context. Further studies exploring risk factors and drivers associated with predominant serogroups could help enhance future targeted interventions.

Supporting information

Acknowledgments

We would like to thank the many study participants who volunteered to participate in this study. We would also like to thank the study staff who collected the field data, the Dominican Republic Ministry of Health and Social Assistance, and the Pedro Henriquez Ureña National University, for their commitment and support for the study. Finally, we would like to thank Dr Gregorio Antonio Rosario Michel and the valuable team working in the Servicio Geologico Nacional for providing the flooding-risk map.