1-dimensional splines as building blocks for improving accuracy of risk outcomes models