The occurrence of emergency situations in high-rise buildings, daily hosting hundreds of people, may force the massive evacuation of their occupants with the ultimate goal of preventing the loss of lives. In this paper, we propose an adaptive algorithm for dynamically computing safe evacuation routes, while the load of people is balanced between the accesses of each floor of the building, thus avoiding the accumulation of people in hotstpots. The combination of this algorithm with sensors to detect the events of interest and a navigation system to guide the evacuees turns this solution into a specially suited approach for evacuation in high-rise buildings. Simulation results demonstrate that the yield evacuation routes drive all people outdoor with a similar average path length for the different risk scenarios addressed.