A key concept here is how predictive healthcare models are built. Predictive models in healthcare rely on supervised learning, where the algorithm learns to map patient features to an outcome that has been labeled in the data (like readmission, mortality, or disease progression). Neural networks, decision trees, and random forests are all common supervised models used for predicting such outcomes because they can capture complex relationships between inputs and the target variable. K-means clustering is different. It’s an unsupervised method that groups patients into clusters based on similarity, without using any labeled outcome to guide the grouping. It’s great for discovering patient phenotypes or segmenting a population for targeted interventions, but by itself it doesn’t produce a direct prediction for a specific outcome on a new patient. So it isn’t listed as a common predictive model in healthcare—neural networks, decision trees, and random forests are.