Prediction bounds are relatively important when an interval considering the error of the regression model and not only sampling error comparing to the confidence interval.
prediction intervals are only concerned with past and future observations, rather than unobservable population parameters, they are advocated as a better method than confidence intervals by some statisticians,