In this paper, adaptive modulation schemes are designed that yieldthe minimum outage probability for wireless systems with strict delayconstraints, under the assumption of perfect causal channel stateinformation (CSI) at the transmitter and receiver. Using the channel characterization and delay constraint parameters imposed by the application, a recursive set of equations is derived for the outage probability. This set of equations is used to derive an optimal adaptive uncoded quadrature amplitude modulation scheme, and its performance is compared against that of traditional variable rate techniques designed to maximize average rate subject to a bit error probability constraint. A suboptimal scheme is also presented. Numerical results indicate that the proposed schemes significantly outperform adaptive schemes designed to maximize average rate.