Conference Abstract | Volume 8, Abstract NACNDC/19JASH014 (Oral) | Published: 18 Nov 2025
Drake Amutuheire1,&, Kelvin Bwambale1, Simon Kigozi1, Nazarius Mbona Tumwesigye1, Sarah Zalwango2, Robert Kakaire3,4, Noah Kiwanuka1, Christopher Curtis Whalen3,4
1School of Public Health, Makerere University, Kampala, Uganda; 2Kampala Capital City Authority, Department of Public Health Service and Environment, Kampala, Uganda; 3Department of Epidemiology and Biostatistics, College of Public Health, University of Georgia, Athens, GA, USA; 4Global Health Institute, College of Public Health, University of Georgia, Athens, GA, USA
&Corresponding author: Drake Amutuheire, School of Public Health, Makerere University, Kampala, Uganda, Email: drakeamutuheire@gmail.com ORCID: https://orcid.org/0009-0003-4914-4179
Received: 20 Sept 2025, Accepted: 20 Oct 2025, Published: 18 Nov 2025
Domain: Infectious Disease Epidemiology
Keywords: Tuberculosis; mobility patterns; Artificial Intelligence; network analysis, Kampala
©Drake Amutuheire et al. Journal of Interventional Epidemiology and Public Health (ISSN: 2664-2824). This is an Open Access article distributed under the terms of the Creative Commons Attribution International 4.0 License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Cite this article: Drake Amutuheire et al., Characterisation of mobility patterns of tuberculosis cases in Lubaga and Kawempe Divisions of Kampala using artificial intelligence and network analysis. Journal of Interventional Epidemiology and Public Health. 2025;8(ConfProc6):00014. https://doi.org/10.37432/JIEPH-CONFPRO6-000014
Tuberculosis is a major public health problem in Kampala. Lubaga and Kawempe report high case numbers that strain control efforts. The disease persists because people move through crowded markets, transport points, slum zones, and health facilities, creating many contact opportunities each day. Routine surveillance does not capture these movements. This gap limits targeted interventions and slows progress. This study used Natural Language Processing and network analysis to extract mobility patterns from cellphone records of confirmed TB patients.
The study used four years of call detail record metadata from 307 bacteriologically confirmed TB patients in the Mapping Tuberculosis Transmission Study. Data preprocessing included geocoding, timestamp cleaning, and removal of routine movements. Feature extraction used TF-IDF weights and Doc2Vec embeddings. The DBSCAN algorithm grouped mobility events into clusters. A directed weighted mobility network measured movement flows. Centrality metrics identified locations with repeated convergence. Model evaluation used Jaccard similarity and silhouette scores.
Patients visited an average of 54 unique locations. Movement patterns showed strong heterogeneity. DBSCAN identified one dominant cluster that captured most mobility events, with few noise points. The mobility network revealed a small set of highly connected nodes that linked large segments of patient trajectories. Markets, transport stages, slum zones, and health facilities in Kisekka, Bwaise, and Mulago showed the highest centrality. The mean Jaccard score was 0.941. The silhouette score was 0.387. These results show that the model reconstructed trajectories with high accuracy and captured clear structure in movement behavior.
NLP and network analysis provide a practical way to understand mobility patterns of TB patients in crowded urban areas. The approach identifies locations that attract repeated visits and may amplify transmission. These findings support targeted active case finding in the most connected urban zones in Lubaga and Kawempe. The method can strengthen surveillance in similar African cities that lack detailed movement data.
Menu