Estimating excess length of stay due to healthcare-associated infections by applying and comparing three time-varying approaches: multistate model, survival regression and matched case control methods.