Healthcare process usually involves multiple aspects (disease development, treatment application, economics, logistics and many others) and actors (patients, physicians, hospital staff, visitors, etc.). While processing of patient flow, the process evolves in many scales (from city to single patient), undergoes various external processes and environmental conditions (weather, ecology, traffic). The analysis and predictive modeling of such complex process are important for decision making both within the treatment of a single patient and in a context of process control and policymaking. Moreover, lately, popular concepts of P4-medicine (personalized, predictive, preventive, participatory) and value-based healthcare introduces higher requirements in the detailed analysis of this process. Still, the important aspect is that the decision-making process itself is a part of the healthcare process. This includes a) data acquisition to describe the disease development, treatment history; b) predictive modelling of the process with model calibration, selection, data assimilation; c) decision support with knowledge processing, machine learning and other techniques involved; d) decision making as human-centered procedure followed by certain action taken by different actors of the healthcare process (including patients). Within this work, we are focused on two important aspects of the healthcare process. First, processing of the information set which may be extended during iterative decision making (e.g., while a doctor requests certain clinical tests to be performed for more accurate diagnosis). Second, both patient and decision makers are humans. This leads to personalized behavioral patterns (on both sides), mental health, etc. are to be considered as an important ingredient of the healthcare process. Under such conditions predictive modelling could be considered not only as a tool for extending available information, but as a tool for intelligent decision support. We consider three conceptual layers (system, data, and model) introduces previously for analysis of healthcare process. In a context of decision making and process control, control loops are defined through these levels having predictive the process modeling as a source for a) clinical decision making (e.g., selection of treatment plan), b) data acquisition and analysis (e.g., requesting additional data), and c) complex model modification. This multi-layer loop integrated into process of healthcare enable enhanced automation of clinical decision making. To validate our approach, we consider decision support in two classes of care process which has a significant difference in structure. Both cases are considered for patients of Federal Almazov North-West Medical Research Center (Saint Petersburg, Russia), one of the leading cardiological research centers in Russia, and described with a set of electronic health records. First process class is providing care for in-patients in the acute state (considered with acute coronary syndrome as an example). The second class is providing care for out-patients with chronic diseases (considered with arterial hypertension as an example).