BACKGROUND: Organizational structure and process are thought to affect patient care outcomes beyond the effects of expert clinical care. OBJECTIVES: To test the hypothesis that a discretionary pattern of organizational structure and process factors is predictive of critical care unit performance, ie, desirable patient and organizational outcomes. METHODS: Quality-of-care patient and organizational outcomes were evaluated in 25 critical care units. A single value representing each critical care unit's euclidian distance from a theoretically ideal pattern of discretionary function was determined and correlated with unit-level measures of outcomes (standardized mortality ratio, severity-adjusted length of stay, patient satisfaction, quality of nursing care, and nursing retention). RESULTS: Distance from the ideal-type discretionary pattern predicted organizational but not clinical outcomes. Units closer to the ideal-type pattern had higher RN retention, and were viewed as better places to work, with higher-quality care by both nurses and physicians. Objectively measured quality of care, patient satisfaction, severity-adjusted mortality, and length of stay were not consistently related to better-structured units. CONCLUSIONS: With experienced critical care practitioners, unit-level structure and process factors were better predictors of organizational outcomes than of clinical outcomes.