Learning from Large-Scale Distributed Health Data : An Approximate Logistic Regression Approach