As part of ongoing research to investigate the impact of patient characteristics, organization characteristics and patient unit characteristics on safety and quality outcomes, we used a computational modeling program, OrgAhead, to model patient care units' achievement of patient safety (medication errors and falls) and quality outcomes. We tuned OrgAhead using data we collected from 32 units in 12 hospitals in Arizona. Validation studies demonstrated acceptable levels of correspondence between actual and virtual patient units. In this paper, we report how we used OrgAhead to develop testable hypotheses about the kinds of innovations that nurse managers might realistically implement on their patient care units to improve quality and safety outcomes. Our focus was on unit-level innovations that are likely to be easier for managers to implement. For all but the highest performing unit (for which we encountered a ceiling effect), we were able to generate practical strategies that improved performance of the virtual units that could be implemented by actual units to improve safety and quality outcomes. Nurse managers have responded enthusiastically to the additional decision support for quality improvement.