We present a hybrid quantum–classical Deep Q-Learning framework for short-horizon blood-pressure control and evaluate it in a reproducible environment derived from openly released, de-identified ICU trajectories. Our approach inserts a variational quantum circuit as a nonlinear feature map inside an otherwise standard DQN pipeline, keeping replay, target networks, and \(\epsilon \) -greedy exploration unchanged to enable fair comparisons with classical baselines. Taken together, the results provide an open, lightweight testbed and a competitive baseline for studying quantum-enhanced function approximation in safety-critical sequential decision making, while clarifying where hybrid models can add value without claiming absolute quantum advantage.