validation, grounds the predictions of the model with observations in the real world and is conducted across expected use domains. An operational validation using the SimCare model to predict randomized clinical trial outcomes was conducted over three main aspects of type 2 diabetic patients: A1c, SBP and lipids panel. Populations of individuals were generated and treatments simulated based on RCT regimens. Results were aggregated per cohort and correlations between predicted and actual outcomes compared.
, elucidates the structure and workings of the model. The SimCare model is a relatively simple model (i.e., not a biophysiological model) and uses functional mathematical representations to compute simulations. The second aspect of validation,
SimCare is a computational model of individual patients with type 2 diabetes. SimCare represents individuals (i.e., not populations) receiving and responding to medical treatments over time. In conjunction with a team of physicians, 12 key health indicators of diabetic populations were identified and included as patient state attributes. Patient states are generated on a daily basis and are affected by previous attribute values, psychosocial conditions of the patient (e.g., adherence) and the time between clinical visits. Values are generated by functional descriptions of dose and time response relationships between attributes and treatments. SimCare represents individual patients by instantiating the model with initial values and administering treatments for the patient on a visit-by-visit basis. Populations of patients may be simulated by instantiating N patients and treating each of them. Two complementary aspects of model validation enable its confident use. The first aspect,