Users in OFDMA-based WiMAX are subject to Adaptive Modulation and Coding (AMC) which implies more robust and hence less frequency efficient modulation and coding for users experiencing bad radio conditions. When users are mobile, this implies lower throughput as they move away from the base station for instance. This may be acceptable for data users, but not for streaming ones which require a constant bit rate throughout the cell. We propose, in this work, a new QoS-oriented resource allocation strategy wherein streaming flows experiencing bad radio conditions are allocated more sub-carriers so as to keep their bit rate constant as they move around in the cell. The risk is however an increase in the dropping rate. To minimize the latter without increasing blocking too much, they are second subject to a state-dependent admission control scheme where the degree of acceptance depends on the density of the users in a given location. We develop an analytical model which allows us to derive some performance measures such as blocking and dropping probabilities. Our results quantify how these metrics vary with the load as well as the admission strategies.