A range of machine-learning approaches can be used, including classical models such as logistic regression, support vector machines and random forests, as well as more advanced deep-learning models. The choice depends largely on the type and complexity of the available data. Importantly, disease-trait modelling should go beyond clinical characteristics and, where possible, integrate genetic, imaging, environmental and temporal data to provide a more holistic understanding of disease patterns. Model performance should then be assessed using appropriate validation and accuracy measures suited to the particular problem and dataset. Environmental and geographical differences are also important because disease patterns may vary across populations and regions. The availability of the specific malaria dataset in OpenMRS would need to be confirmed separately.

AI can strengthen telemedicine by extending specialist expertise to health facilities and patients in remote areas. In our cervical cancer work at Makerere AI Health Lab, for example, we have developed both web-based and offline applications. The web platform enables specialists, including pathologists, to remotely review and validate AI-generated diagnostic results, while the offline smartphone-based tool supports on-site diagnosis where internet connectivity is limited. This is particularly useful in facilities without specialists, as health workers can obtain expert interpretation without requiring the specialist to be physically present. Beyond diagnostics, AI-enabled telehealth platforms can support teleconsultations, patient monitoring, pathology reporting, virtual assistants and treatment support, ultimately bringing quality healthcare services closer to patients regardless of their location.