We performed a systems analysis of 10 complex reconstructive operations at BWH, evaluating system performance as a function of clock time, process time, scheduling cycles, information/material resource utilization, case stage and intensity, and concurrent activities. Using standard clinical outcome process variables, we identified 11 adverse events. Reconstructing the system state at intervals, we found that most adverse events were the result of: [1] previously unrecognized and poorly-synchronized interactions-interdependencies between individual system components; [2] interruptions in information flow or degradations in the quality of data; [3] byproducts of 'standard' protocols that increased, rather than decreased complexity. As in other complex systems (e.g., neural dynamics of auditory cortex, Boltzmann machines, etc.) global performance depended on a large number of locally-interacting elements, back-propagation and boundary effects. We used the data to develop new models of risk and to redesign scheduling cycles, protocols, technology use, and information flow to control complexity, optimize system performance and reduce the incidence or severity of adverse events in these settings (i.e., identify control laws that stabilize the system, keeping state vectors bounded).