personnel, environment services) and mid- and upper-level management must promote a pervasive, patient-centered safety culture of adverse event prevention, not individual blame. In keeping with James Reason’s “swiss cheese model of error,” PHA and quality management programs must identify “latent” errors as well as the more apparent “active” errors. Systems redesign to prevent all such errors should then be based on a balanced utilization of evidence-based technology, training, ongoing education, and consensus standard operating procedures (SOPs) and “best practices,” keeping in mind each human’s inherent cognitive (e.g., memory recall) and physical (e.g., fatigue) limitations.
Lastly, health care must recognize that adverse event prevention is an ongoing process. Each new system intervention brings with it a whole new set of potential failure modes and contributing factors that should be similarly proactively analyzed and prioritized for intervention. This, combined with the ever-widening scope of system complexities due to an aging patient population, increased numbers of the immune compromised, and the need to “fast track” new and more effective technological advances in medicine, raises the need to handle health care’s current “patient safety paradox” with an organized, proactive collective consciousness.
How can data be employed to do prospective identification of risk points without waiting for a near miss? Pareto charts (histograms), run charts, control charts, and scatter grams are among the more widely used tools to exemplify performance data. Despite the inherent strong points and weaknesses of a respective tool, the reliability, defensibility, and reproducibility of the underlying performance data must be paramount. To maximize the accuracy and precision of such data and to facilitate standardized use throughout all health care, performance measures must be the result of a well-thought-out process to maximize efforts to exceed customer expectation and consistent error and failure definition.22
To facilitate and standardize measurement, Chang proffers an error taxonomy consisting of four subclassifications of error: impact, type, domain, and cause. The “impact” classification deals with the outcome or effect of the error; the “type” concerns the visible process that was in error; the “domain” is where the error occurred and who was involved; and the “cause” is the factors and agents that bring about error. Establishing subclasses for respective errors can not only help in defining and standardizing perfor-