Evaluating of machine learning models in predicting COVID-19 vaccine uptake in Kenya
Journal of Kenya Assocation of Physicians · 08.04.2026 · 8 (1) · 38-47
DOI: 10.4314/jkap.v8i1.5 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Kenya Assocation of Physicians, online erschienen am 08.04.2026. Verfasst von David Andai und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Background
Machine learning algorithms are employed to create prognostic and diagnostic models, aiding in public health decision-making and intervention strategies.
Objective
This study aimed to explore predictive models for the COVID-19 vaccination uptake using machine learning algorithms.
Methods
Supervised machine learning algorithms were utilised to predict accuracy. We employed an evaluation framework to assess the performance of Random Forest (RF), Support Vector Machine (SVM) and Extreme Gradient Boost (XGB) models in predicting COVID-19 vaccine hesitancy. The data used was based on a cross-sectional study conducted between November 2021 and January 2022 among the general adult public seeking care at six different healthcare facilities in Kenya. The survey, in English, consisted of questions based on demographics, knowledge, and attitudes, including knowledge towards the COVID-19 vaccine. Model effectiveness was evaluated by comparing accuracy, F1 score, sensitivity, specificity and AUROC.
Results
Of the 3996 surveys collected, approximately 68.8% reported being vaccinated with at least one dose. Three models were established using 56 predicting features. Among the three models, RF and XGB yielded better overall performance, with an equal AUROC of 0.88, compared to SVM, which achieved an AUROC of 0.82. Additionally, protection against COVID-19 variables had the highest importance to vaccine uptake, and the following features were healthcare facility and recommendation to the family respectively.
Conclusion
The study findings underscore the potential of leveraging machine learning algorithms in public health strategies to address vaccine hesitancy. Our results indicated that random forest could be a useful predictive tool to identify vaccine uptake which may facilitate effective strategies and further optimize resources. These algorithms were selected for their widespread use, transferability, and proven performance in healthcare prediction tasks, aiming to enhance the predictive accuracy and offer actionable insights into combating vaccine hesitancy.
Zitieren
APA
Andai, D., Shah, S., Muchungi, K., Ali, S. K., & Shah, J. (2026). Evaluating of machine learning models in predicting COVID-19 vaccine uptake in Kenya. Journal of Kenya Assocation of Physicians, 8(1), 38-47. https://doi.org/10.4314/jkap.v8i1.5
Vancouver
Andai D, Shah S, Muchungi K, Ali SK, Shah J. Evaluating of machine learning models in predicting COVID-19 vaccine uptake in Kenya. Journal of Kenya Assocation of Physicians. 2026;8(1):38-47. doi:10.4314/jkap.v8i1.5
BibTeX
@article{Andai2026,
author = {Andai, David and Shah, Sheena and Muchungi, Kendi and Ali, Sayed Karar and Shah, Jasmit},
title = {Evaluating of machine learning models in predicting COVID-19 vaccine uptake in Kenya},
journal = {Journal of Kenya Assocation of Physicians},
year = {2026},
volume = {8},
number = {1},
pages = {38--47},
doi = {10.4314/jkap.v8i1.5}
}
EUVIMED