Here's the problem: the human body is an immensely complex system, with millions of factors influencing its status and well-being. Untangling these factors correctly and producing an accurate and sophisticated statistical theory of the body would require a comparably large number of parameters - on the order of millions or more.
Unfortunately, modern medical science relies on low-N observational or clinical trials, with N on the order of hundreds or thousands. In this radically low-data regime it is impossible to justify the use of complex models. If you try to use a complex model, you will just get overfitting. You can use a simple model to avoid overfitting, but there's no reason to believe that a simple model will produce a good approximation of the underlying dynamics.