Accuracy and Computational Cost in Cardiovascular Modelling

Cardiovascular models describe processes ranging from whole-circulation pressure regulation to local flow separation and valve deformation. Their usefulness depends on selecting a mathematical representation that resolves the intended output without introducing unnecessary computational or identification demands. This critical narrative review examines accuracy–cost trade-offs across lumped-parameter models, one-dimensional arterial networks, computational fluid dynamics and fluid–structure interaction. Selected studies by Hassani and collaborators provide application examples connecting electrical analogues, pathological vascular geometries and clinically informed valve simulations. These studies are considered alongside methodological work on numerical benchmarking, boundary conditions, calibration, uncertainty propagation and recent surrogate modelling. The central distinction is between structural detail and demonstrated predictive adequacy: increasing dimensionality expands the phenomena a model can represent, but does not establish its accuracy for a particular use. Conversely, a reduced model may reproduce global haemodynamics while remaining unsuitable for local mechanical quantities. A practical assessment should therefore consider the quantity of interest, numerical error, parameter identifiability, independent validation and the full cost of repeated simulation. Recent research on calibration and physics-informed emulation expands the feasibility of inverse analysis, although approximation errors and restricted training domains remain relevant. The review proposes an output-specific framework for model selection and reporting. Cardiovascular simulation is most defensible when complexity is justified by a measurable improvement in the intended prediction and when the remaining uncertainty is reported alongside that prediction.