The human body is incredibly complex so most expert-derived models are less accurate than purely statistical methods. There are ways to marry the two, like Baye

The human body is incredibly complex so most expert-derived models are less accurate than purely statistical methods. There are ways to marry the two, like Bayesian networks, but they take much longer to develop and aren't typically more accurate. The thing is, a model doesn't have to be accurate to be useful. Expert-derived systems tend to operate closer to how doctors understand the problem, making them more interpretable.