Quality of Service (QoS) guarantees in mobile wireless networks are typically provided through queuing analysis, which uses an idealized model for the physical layer, leading to optimistic guarantees. A queued-code combines ideas from queuing theory and information/coding theory, to allow rate adaptation, while enabling accurate modeling of the physical layer. Previous work showed the benefit of using a queued-code, in achieving lower error probabilities while satisfying a fixed delay bound. This paper analyzes the queued-code when some function of delay statistics must be optimized while guaranteeing a fixed error probability. A practical implementation of queued-codes based on low density parity check (LDPC) convolutional codes is also presented. Numerical and simulation results demonstrate the benefit of using the queued-code in obtaining a better QoS.