OpenStreetMap (OSM) is a collaborative platform for mapping and data collection that allows contributors to express their perspectives on the world. However, this bottom-up approach results in varying levels of data quality and heterogeneity due to differences in mapping competence, local knowledge, and interests of the contributors. The lack of standardization and consistency in the tagging system of OSM further complicates the usability of the data, especially for end users such as field health workers. This research aims to develop a user-oriented assessment framework, focusing on OSM data related to health services, to improve the fitness-for-use of the data.
The research objectives include determining the parts of OSM datasets that need quality assessment, unifying the existing vocabulary using Semantic Web Technology, analyzing the correlation between mapping events/projects and data quality, and providing a web application for field health workers to assess and edit OSM data. The main research question addresses the effectiveness of a user-oriented approach in assisting health services in the Global South. The study focuses on two Kenyan counties, Nairobi and Wajir, and their specific needs for health-related data. It aims to evaluate the fitness of OSM data for the specific needs of health workers and analyze the vocabularies used in OSM. It also examines the quality of the data, particularly in relation to health facilities, water and sanitation points, and road networks. It identifies the role of humanitarian mapping projects in improving the quality and completeness of OSM data and proposes the use of ontologies and semantic web technologies to enhance the data analysis process. Additionally, the study explores the potential development of a web application for accessing OSM health services data.
The present research involves several stages in the way to reaching its objectives. The user’s needs analysis is the first part which builds the context for all the other parts and consists of a Questionnaire, several interviews, and a report analysis from the Kenyan Red Cross. Based on that the test datasets are were produced with the source being Kenya’s OSM dataset in RDF format. Using Qlever SPARQL engine and RDFlib for Python the aim is to construct the most complete test dataset possible containing all hidden information. These datasets were then compared to authoritative data to assess health services OSM data. The queries used for the assessment were also a guide to building an ontology for promoting existing vocabulary. Then by using OSMCha and querying for humanitarian hashtags, the goal was to find areas of the use cases where humanitarian mapping projects have been implemented and see if they are correlated with high-quality health services data. Finally, the setup of a web application was proposed using the outcomes of the previous stages.
The assessment of OSM data is guided by a users' needs analysis for health field workers, which prioritizes the completeness of health facilities data, their level according to the Kenya Essential Package for Health (KEPH), the services and equipment they provide, and their contact details. It finds out that the existing methods for assessing OSM data lack domain-level sensitivity, and therefore, a user-focused approach is adopted.
The analysis of the vocabularies used in OSM reveals that there is a lack of tags representing the classification of health facilities and the equipment they use in Kenya. While Nairobi has an extensive vocabulary for services and contact information, Wajir has a limited vocabulary but ongoing mapping projects from HOT in collaboration with the Kenyan Red Cross have contributed to improving the data quality. However, in areas where humanitarian mapping projects have already taken place, such as the informal settlements of Nairobi, the vocabulary is more extensive. The research suggests that mapping projects can improve the quality of OSM data by creating a taxonomy that future contributors can follow.
Quality assessments of OSM data are conducted to determine the completeness and attribute accuracy of health facilities, water and sanitation points, and road networks. The study compares OSM data with authoritative sources and finds that humanitarian mapping projects contribute significantly to the completeness of health facilities data, particularly in informal settlements. However, the attribute completeness of OSM data is relatively low, indicating room for improvement. In terms of water and sanitation points, OSM data in Nairobi's informal settlements are more complete than UN-Habitat data, while for Wajir, there is limited reference data available. The road networks in both Nairobi and Wajir counties are more extensive in OSM compared to authoritative data, suggesting that OSM can support navigation for health services.
The study highlights the importance of using all available data sources, including OSM and authoritative data, for a common humanitarian cause in the Global South. It proposes the use of humanitarian mapping projects as a proxy for assessing the quality of OSM data and suggests that the credibility of contributors can be an indicator of data quality. The research also explores the use of ontologies and SPARQL queries to build a classification system for health facilities and suggests the development of a web application for accessing OSM health services data.
In conclusion, this research emphasizes the importance of user-focused assessments, the role of humanitarian mapping projects in improving data quality, and the potential of ontologies and web applications to enhance the use of OSM data for supporting health services in the Global South. By leveraging these approaches, OSM can become a valuable tool for addressing healthcare challenges in vulnerable environments.