Table of Contents 

Introduction 1

Findings 2

Geolocation Error Identification 2

Methodology 2

Conclusion 5

Table of Figures  

Figure 1, A google Earth Image showing the location of Dandora I Health Centre according to the existing SSA Data & its correct location 3

Figure 2, A street view google map showing Dandora I Health Centre 3

Figure 3, An ArcMap snip showing the correction of the location of Dandora I Health Centre 4

Introduction

In this report, I present the methodology and findings of a comprehensive analysis and correction of geolocation errors in a dataset of health facilities in Nairobi County that was done for hospitals across Sub-Saharan Africa. The dataset initially comprised 242 health facilities, which required verification for both geolocation accuracy and establishment dates (2013 and the years before). My task involved identifying and rectifying inaccuracies to enhance the dataset's reliability for further analysis.

Findings

Upon examining the dataset, I identified that out of the total 242 health facilities, 11 were established after 2013. These facilities included Mananja Health Centre, Brother Andre Clinic, Dandora PCEA Clinic, Family Health Medical Clinic, Kahawa Garrison Health Centre, Kariobangi EDARP Dispensary, Likoni Road SDA Health Services Clinic, Ngundo PCEA Clinic, Njiru Dispensary, Redemeed Health Centre, and Woodley Clinic.

Additionally, there were 3 facilities that I could not locate on the maps. These included Dandora Provide International Clinic (assumed to be Kayole Provide International Clinic due to a naming error), Lina Medical Services (located on an empty field), Silangi Community Clinic (assumed to be a repetition of Silangi MSF Belgium Dispensary) and the Vision People Inter Health Centre (located on a road).

Geolocation Error Identification

Out of the 242 facilities, I identified 71 with inaccuracies in their geolocation data. These errors varied in significance, with most being less critical compared to other datasets. The next step was to identify and adopt an approach to correct the errors in these 71 facilities.

Methodology

1. Initial Verification Using Google Earth:

The first step involved visually confirming the locations of the health facilities using Google Earth. I checked the names and exact locations and their existence in the year 2013 and prior. Look at the example below. This is a google earth photo showing where Dandora I Health Centre was located according to our dataset (1) and its actual location (2).

Figure 1, A google Earth Image showing the location of Dandora I Health Centre according to the existing SSA Data & its correct location

I utilized street view where available for additional verification. For instance, Dandora I Health Centre was found listed as Dandora Health Centre on Google Earth, but the street view confirmed its name as Dandora I Health Centre.

Figure 2, A street view google map showing Dandora I Health Centre

2. Data Handling with ArcMap:

Due to limitations in moving KML files on Google Earth without losing attributes, I used ArcMap for the geolocation corrections. I began by loading the SSA shapefiles containing hospital locations into ArcMap. Then, I loaded the base map to visually align the locations with those identified on Google Earth. Using ArcMap, I moved the facilities to their correct locations as verified on Google Earth, ensuring that all attributes remained intact. For example, I corrected the location of Dandora I Health Centre in ArcMap after confirming its true position on Google Earth. Look at the example below, (1) is where the hospital was indicated to be located and (2) is where hospital was moved to be located which is the correct position.

Figure 3, An ArcMap snip showing the correction of the location of Dandora I Health Centre

This process was repeated for a total of 74 hospitals.

3. Handling Unlocatable Facilities:

For facilities that could not be located, I assumed potential naming errors or duplications. For example, Silangi Community Clinic was treated as a duplicate of Silangi MSF Belgium Dispensary. This assumption helped in rationalizing the dataset and focusing on the correct facilities.

Conclusion

In conclusion, the process of identifying and correcting geolocation errors in the Sub-Saharan Africa health facilities dataset significantly improved its accuracy and reliability. A total of 74 hospitals were corrected, and the methodology adopted can serve as a standard for future geolocation verifications and error corrections. This report documents each step meticulously, ensuring transparency and replicability in future projects.

Recommendations:

• Future geolocation verifications should adopt a similar methodology to ensure systematic and verifiable corrections.
• Further investigation is recommended for unlocatable facilities to resolve any remaining ambiguities.

By following this detailed methodology, the dataset's reliability is greatly enhanced, making it a more robust resource for health facility analysis in Sub-Saharan Africa.

Summary Of Health Facilities That Have Errors

Table 1, Summary of the health facilities that have errors