Risk managers are assigned the task of preventing sentinel events from reoccurring. Despite their efforts, there are limited data on reduction of adverse events. Medication errors, wrong-site surgery, unintended retention of foreign objects, falls, wrong-patient procedures, and many other adverse events continue to occur at seemingly the same rates as before. Some of these adverse events (eg, medication errors) are not even rare. The presence of adverse events is the norm and not the exception in our system. We have lived with these events as if these accidents are inevitable. In a way, we have come to accept them as the price we have to pay for access to our health care system. An independent observer looking at today's health care system would be aghast at how much harm is being done. Other industries, such as airlines, have radically reduced their accidents. Health care has not.
Why has success been so elusive? The answer is not simple. There could be many reasons. One could hypothesize that it is because of too many actors, too much technology, care processes that are too complicated, too many points at which the system could break down, too little data on adverse events, and so on. One possible reason is the way we have approached the problem. Adverse and sentinel events have traditionally been analyzed case by case, as if it is a legal court proceeding. Of course, implicit in the case by case approach is the false assumption that patterns of adverse events do not matter. A failure model is described, itemizing many events that could have led to the adverse event. A root cause analysis (RCA) is done to articulate the possible factors that could have led to the adverse event. Yet, no one examines patterns of failure over time and across sites and settings of care. No one verifies whether the hypothesized root causes fit more general experience. The analysis of adverse and sentinel events is mostly an exercise in logical reasoning, with little or no effort to check these intuitions against data. The claims that certain events have caused the adverse event are not checked against either small or big data. Such analyses reinforce commonsense intuitions and seldom lead to new insights.
It is time we reexamine our approach to sentinel events. In this issue of the journal, our special section on Sentinel Event Investigation and Risk Management approaches brings together a number of related articles that try to do so. In this issue, Khaleghzadegan and colleagues describe the history and steps in Failure Mode and Effect Analysis (FMEA) and RCA. They describe some of the challenges of these two methods and how a leading health care institution is using these tools. The article by Pirouzi and colleagues shows the application of FMEA to adverse events in operating rooms. The researchers combine FMEA with quantitative verification of hazards associated with various types of failures. The article by Abd El-fattah Mohamed Aly Aly uses new types of errors in blood administration to enhance FMEA. The article by Slade and colleagues combines RCA with improvement strategies to optimize venous thromboembolism prophylaxis. These articles improve FMEA or RCA by combining them with other quantitative methods. As Allen suggests, one could combine RCA with implementation science to make both techniques more effective.
The article by Landsittel and colleagues takes a fundamentally different approach. It is not focused on improving FMEA or RCA but replacing these methods entirely. It reviews the literature on causal analysis and shows that alternative approaches to FMEA and RCA are available. These alternative approaches are data driven. They provide novel insights on why adverse and sentinel events occur. In the last article in the section, I offer a peer-reviewed tutorial on how network models can be used to analyze causes of emergency department (ED) excessive boarding (ie, waiting >6 hours in the ED). The intent of this article is to make risk managers more familiar with methods of understanding root causes that rely heavily on data.
Readers may be concerned with availability of data on adverse and sentinel events. Some adverse events, such as medication errors, are not rare and data are available on these events. The Hospital Compare Web site provides rates of errors for many adverse outcomes. Other adverse outcomes are rare. Yet, even these rare events can be analyzed using the new causal methods. The claim that sentinel event investigations cannot rely on data may be a red herring, an excuse from days when data were not routinely collected in electronic health records.
Another concern could be that clinicians and administrators may have difficulty understanding changes in very small probabilities. The probability of a sentinel event that occurs once in 3 years is less than .001; most people will not understand changes in these very small probabilities. One solution is to report days to events. The construction industry reports days between accidents. It might be reasonable to do the same in health care; in fact, there are intensive care units that track and share “Days since last central line–associated bloodstream infection.” Clear communication of even rare events can increase transparency and motivate everyone to remain vigilant. The point is that contrary to claims, the needed data are often available, new methods of analysis can help make sense out of the available data, and analysis of rare events can be effectively communicated.
In the coming months, the Journal is looking forward to hearing from risk managers about their use of data in reducing adverse and sentinel events. We would like to see reports that these sentinel events are getting closer to zero. We look forward to reviewing reports of application of modern causal methods in analysis of adverse outcomes. We will continue to publish FMEA and RCA applications, but we will apply a new standard. These reports must report data showing that the effort has led to reduction in adverse outcomes. And, we hope that as Quality Management in Health Care goes, so goes the field.
—Farrokh Alemi, PhD
Special Section Editor and Associate Editor
Quality Management in Health Care
Professor of Health Informatics
Department of Health Administration and Policy
College of Health & Human Services
George Mason University