As the use of bandwidth hungry applications such as video conferencing and video on demand increases, ensuring Quality of Service (QoS) becomes increasingly important. In wireless access networks such as WiFi and WiMax, admission control is necessary to control QoS. Many admission control methods use the mean bit rate of a flow to determine if it can be allowed, however there is a lack of methods that accurately assess variable bit rate flows. We propose a simple alternative method that extends our previous work on admission control. It uses the bit rate statistics of each flow to predict the percentage of packets that are excessively delayed. A new flow is permitted only if this predicted percentage is sufficiently low. Using a simulation study we show that the proposed method can accurately predict the proportion of delayed packets and hence can control the load of the network to achieve the desired delay bounds.