Sentinel events, that is, events whose single occurrence is of sufficient concern to trigger systematic response, recently were advocated as an important component in managing quality under Continuous Quality Improvement (CQI). Ideally, sentinel events are exceedingly rare and invariably indicate preventable deficiencies (eg, operation on the wrong patient, which should never occur, denotes a lack of management control requiring corrective action). However, although single cases of nosocomial group A streptococcal surgical wound infection or of tuberculosis are harbingers of worse to come unless prompt intervention is initiated, sentinel events generally have not served infection surveillance programs adequately. This is because nosocomial infection is the result of multifactorial chains of events that produce a probability, not a certainty, of infection, and our knowledge is incomplete. Without detailed knowledge of these probabilistic chains, the absolute minimum rates of infection that may be attained, and the optimal statistical approaches to define predictive outbreak “warning” or “action” levels, we will not be able to define meaningful sentinel events. How, then, can we best assure health services quality?

Nearly 30 years ago, Drucker suggested that three elements comprise effective management decisions. They involve determining 1) whether a specific situation is generic or an exception, 2) clear specification of what the decision has to accomplish, and 3) what is right rather than simply what is acceptable.