Automating the prediction of the next activity in a clinical pathway has the potential to improve patient outcomes and healthcare operations by supporting timely interventions and better management. Accurate forecasts help clinicians stay one step ahead by planning treatments and managing resources more effectively. However, such predictions are not easy due to the complexity of hospital environments and wide variety in patient conditions and care plans. Predictive Process Monitoring (PPM) addresses this challenge by combining clinical and administrative data with process-related information from past patient journeys. Process Mining (PM) of clinical events data can add valuable context to improve prediction accuracy. In this study, we used the open access MIMIC-IV dataset to test how well PPM can predict the next step in a patient’s care. We focused on two key points where prediction has the greatest utility: the first recorded admission event in the patient’s pathway and the moment when a patient is transferred to a ward. We used demographic, clinical and process features to build an extended event log and then employed it for next activity prediction. We introduced a method for events aggregation and tested stochastic transition, Machine Learning (ML), Deep Learning (DL) and Deep Sequence (DS) models within a PPM framework. Our results demonstrate how predictions can become more accurate as patient progress along a clinical pathway, demonstrating the value of including such information in predictive models. This work contributes to the expanding application of PPM in healthcare, offering a data-driven approach to enhance clinical decision-making and resource planning.