Improving Caregiver Training and Healthcare Workforce in Kenya Through Machine Learning and Artificial Intelligence

Caregiving for both children and older people is now a major concern in Kenya with the growing population, despite the shortage of qualified caregivers. This study leverages machine learning to enhance caregiver workforce planning by determining the key drivers of caregiver transition to caregiving roles. Data used in this research were obtained from a vocational training institution for caregivers in Kenya (2023–2025), supplemented by national population data for demographic growth trends (20202025). Subsequently, after data pre–processing, data analysis was done to uncover the patterns from which we found that many students prefer eldercare compared to childcare with a disparity in regional distribution of caregivers. Statistical analysis presented a significant difference in final scores between transitioned and non-transitioned caregivers (t =11.67, p < 0.001) with majority of students being youths below 35 years and female at 87.4% in care course. The modeling ready dataset was then used to train classification models from which Catboost emerged the best on hyperparameter tuning using randomized search with 5-folds. Analysis of feature importance with SHapley Additive exPlanations (SHAP) revealed that caregiver performance in classwork and practicals plays a big part in caregiver transition with less importance from geographical features. The optimized model was then deployed and made available in a web application in the institution's servers to assist in decision making. This approach gives significant insights to the educational institutions and healthcare policy makers to enhance training of caregivers, make skillful manpower distribution more effective and, in the end, guarantee quality healthcare delivery, thus satisfying the increasing need for skilled caregivers in Kenya.