The models tend to be close to equivalent in their inputs: either total or LDL cholesterol (small-particle LDL is the one that drives atherosclerosis, but if the model includes HDL cholesterol, then it doesn't make a big difference in accuracy whether the other variable is total or LDL cholesterol), HDL cholesterol (seems to cause a protective effect, although randomized trials of HDL-boosting drugs haven't shown improvements, so may not be causal), blood pressure, and smoking. People also sometimes include newer biomarkers like triglycerides or hs-CRP, but they don't consistently show accuracy gains (LDL, triglycerides, and hs-CRP are highly correlated, so each one individually is a good predictor but having all three doesn't necessarily give you a more accurate model).
There's still a lot of active research to find out which of these variables are causal and which are merely associated with heart disease. For example, LDL-lowering drugs like statins do reduce the risk of heart attacks in randomized trials.