A Moran’s I Autocorrelation and Hot Spot Analysis for Identifying and Predicting Diarrheal Disease Cases around Sixty-Seven Community Wells in West Pokot County, Kenya

At the conclusion of the Millennium Development Goals in 2015, 663 million people still lacked an improved drinking water source and 2.4 billion people lacked access to an improved sanitation facility worldwide. Disparities were revealed as 8 of 10 people who lack an improved drinking water source and 7 of 10 people who lack an improve d sanitation facility live in rural areas [1]. The health effects of this lack of water, sanitation, and hygiene (WASH) infrastructure are staggering. Inadequate WASH causes a variety of diseases and disabilities including diarrheal disease from the ingestion of the minimum quantity of pathogenic organisms in drinking water. Inadequate WASH was estimated to be the cause of 842,000 diarrheal deaths globally representing 58% of diarrheal diseases in 2012 [2], which was approximately 2,300 deaths per day. Nine of 10 diarrheal deaths occur in children with the significant majority of deaths located in developing countries [3]. Persistent diarrhea among children is associated with malnutrition, cognitive impairment [4,5], and an increased risk of developing obesity later in life [6].


Introduction
At the conclusion of the Millennium Development Goals in 2015, 663 million people still lacked an improved drinking water source and 2.4 billion people lacked access to an improved sanitation facility worldwide. Disparities were revealed as 8 of 10 people who lack an improved drinking water source and 7 of 10 people who lack an improve d sanitation facility live in rural areas [1]. The health effects of this lack of water, sanitation, and hygiene (WASH) infrastructure are staggering. Inadequate WASH causes a variety of diseases and disabilities including diarrheal disease from the ingestion of the minimum quantity of pathogenic organisms in drinking water. Inadequate WASH was estimated to be the cause of 842,000 diarrheal deaths globally representing 58% of diarrheal diseases in 2012 [2], which was approximately 2,300 deaths per day. Nine of 10 diarrheal deaths occur in children with the significant majority of deaths located in developing countries [3]. Persistent diarrhea among children is associated with malnutrition, cognitive impairment [4,5], and an increased risk of developing obesity later in life [6].
In many areas of rural Kenya, improved drinking water sources and improved sanitation facilities are inadequate [1]. Therefore, rural village communities depend on untreated water from communal wells, rivers, and other surface water sources to meet their needs presenting the opportunity to consume pathogens and induce disease. Geographic information system (GIS) has been used to map the spread of diseases, including diarrheal diseases, as part of surveillance and control strategies [7][8][9][10]. Unfortunately, despite endemic diarrheal diseases throughout every Kenyan county, the application of GIS technologies to assist in the control and elimination of these diseases in Kenya is sparse in published literature and has not been reported for West Pokot County. The purpose of this research is to identify the reported cases of diarrheal diseases and to predict hot spots for other cases of diarrheal diseases in West Pokot County, Kenya using spatial autocorrelation and hot spot analysis.

Study area and population
West Pokot County is located in Northwest Kenya along the Ugandan border (  infrastructure in these rural agrovillages is generally inadequate. Community wells, boreholes, and other surface water sources are commonly used ( Figure 5). For sanitation, latrines are used however open defecation is still a common practice in many communities ( Figure 6). Livestock (e.g., cattle and goats) are an important agricultural and economic asset of Pokot culture and often graze near human habitations and water sources ( Figure 7) [13]. The close proximity of livestock to Pokot drinking water sources is a potential source of fecal contamination and enteric pathoge ns.

Data collection and classification of diarrheal diseases
Community data was provided by Harvester's International, a nonprofit organization based in South Carolina that works with indigenous Pokot leaders to install community wells and other development initiatives. Surveys were employed between June and September 2014 in communities where wells were installed recording well GPS coordinates and function as well as community health information including population and symptom-based disease cases. Data from 85 communities was provided of which 67 communities are in West Pokot County. Amoebiasis, cholera, dysentery, and typhoid were included in the reports and classified as diarrheal diseases for the purpose of this study.

Analytical tools and procedures
ArcMap GIS (Esri Inc., v.10.3, Redlands, CA) was used to generate maps. Spatial autocorrelation (Global Moran's I) and Hot Spot Analysis (Getis-Ord Gi*) were employed to assess level of significance of diarrheal disease cases around each community well installed by Harvester's International and to predict hot and cold spots across West Pokot County. Spatial autocorrelation was set to aggregate features within five kilometers. Hot Spot Analysis forms analysis based on case count and does not account for population or prevalence statistics.

Results
As illustrated in Figure 8, there are five hot spots of cases of diarrhea, one spot at 99% confidence and the other four spots at 95% confidence. There is only one cold spot, which was at 90% confidence. The prediction model has a high Z-value of 2.96 concentrated in the southwest region of the county and a low Z-value of -1.56 mostly located in the south-central and northwest regions of the county.    used in smaller geographical areas to illustrate disease case predictions. Our data set only had 67 spatial data points in West Pokot County, which is a limitation. However, this prediction model should be used as a priority guide to assist in the decision of which communities to target next. We recommend adding future community pidemiological assessment data to this prediction model to improve its reliability and validity for future intervention decision making.
The feasibility of adding to or replicating these models in the field is high because purchasing GIS band data was not required. The base maps and GIS tools were all included in the ArcGIS software package. Therefore, it is possible for a public health agency with limited resources to be able to make these models to identify and predict disease cases to inform their control and elimination interventions.
In summary, the purpose of this research was to identify the reported cases of diarrheal diseases and to predict hot and cold spots for other cases of diarrheal diseases in West Pokot County, Kenya using GIS spatial analysis. Figure 8 illustrates both the hot and cold spot trends based on the reported cases of diarrheal diseases from Harvester' s International and the prediction of the locations of other diarrheal disease cases across the entire county. However, the validity and reliability of this map may be confoundi ng due to the small number of data points to form the prediction. To improve prediction

Discussion and Conclusions
Using epidemiological and spatial data with GIS analysis, the results illustrated in Figure 8 reveal key areas of public health interest and intervention. The hot spot analysi s presents the locations where cases of diarrheal diseases are most prevalent. These locations should be given priority for control and elimination interventions to improve public health. Further assessment at each of these locations would be in order to determine the specific diarrheal disease and thus appropriate control and elimination strategies.
The prediction model illustrates the potential hot and cold spots of diarrheal disease across West Pokot County. This model appears to represent the data well as 9 of the top 10 locations of diarrheal disease cases are in red areas, which are predicted to be hot spots. However, the community of Samor, which had the sixth most diarrheal disease cases, is located in an area predicted as a cold spot. This is likely due to the relatively low number of spatial data points. In similar hot spot analyses [7,8,10], hundreds to thousands of spatial data points were   of hot spots especially in a large geographical area like West Pokot County, more data points is recommended for future studies.