Radiation Hazards and Challenges in Reporting Adverse Events: A Human Factors Perspective

• 3/5/2010
• Author: Steven Berman
• Category: The Journal Blog
• 14635 Views
• 0 Comments
• 

In his January 29, 2010 guest blog, Kurt Herzer commented on the two recent investigative articles in the New York Times on the incidence and impact of rare accidents in radiation therapy. Shortly after, the American Society for Radiation Oncology issued a six-point plan for improving safety and quality and reducing the chances of medical errors—including working with other stakeholders to create a database for the reporting of linear accelerator– and computed tomography–based medical errors. This week, Emily S. Patterson, Ph.D., Ohio State University Medical Center, School of Allied Medical Professions, and a member of the Journal’s Editorial Advisory Board, discusses challenges faced in the reporting of adverse events.We welcome your comments.

The New York Times articles have served as a wake-up call for the human suffering that results from negative, unintended consequences of using increasingly complex technology in a largely unregulated, competitive marketplace to perform potentially life-saving radiation treatments. A particularly troublesome aspect of these stories is the potential for overradiated and underradiated patients to suffer harm that goes undetected for years, and likely in many cases, forever. The articles, as well as Kurt Herzer, proposed increase reporting and monitoring of adverse events as one solution. Although certainly an increased understanding of the hazards and patterns in incidents would be extremely helpful in improving patient safety, I agree with him that achieving this “will require vision, creativity, and collaboration.”

The patients’ stories in the New York Times articles hint at some of the challenges that I believe will be faced in effectively harnessing the potential of reporting systems to improve patient safety. First, there are substantial issues in obtaining high-quality data.   Many have noted that there is no such thing as a “mandatory” reporting system—all reporting systems are, at some level of analysis, voluntary. In addition, particularly with such horrific patient outcomes, there will predictably be enormous pressure to concurrently use the data for “accountability” purposes. Experience suggests that a single breach of confidentiality directly and immediately results in a widespread lack of trust that reporting is safe, thereby eroding the volume and quality of data. This lack of trust creates  pressure to blame individuals and recommend sanctions, training, or revised policies and procedures—at the expense of attention to systemic flaws. Hospitals will be concerned about lawsuits and loss of reputation, device companies will be concerned about loss of market share and lawsuits, health care providers will be concerned about losing their ability to provide health care and maintain their employment, and the administration personnel of the reporting systems themselves will be concerned about their own ability to keep their employment in the face of criticism for failing to appropriately communicate issues and intervene, even when their sphere of influence over device design and regulation is limited. Finally, it is likely that access to the richly detailed stories, even if they are obtainable, will be highly limited because of concerns about breaching confidentiality of the reporters for “outsiders” interested in analyzing the data.  A second challenge area is data analysis. Because “tags” of some kind will be needed to organize the collected data, an easy pathway will be to create summary statistics using those tags. The tags will need to be easily understood by those entering the data, by definition, so it is unlikely that they will be in a language conducive to recognizing patterns that reveal how to improve systems. For example, the geographic region in which incidents occurred are not as informative as learning that fail-safe mechanisms are lacking in automated filter placement software, but it is unlikely that such a tag would be used to classify cases from the outset. Similarly, outcome-based tags, such as whether a patient was overradiated to the point of mortality within six months, are probably not the best grouping for finding patterns, since processes and outcomes are not necessarily correlated. 

The most important recommendation that I can give for a reporting system is to have at a minimum one full-time dedicated Ph.D.-level human factors engineer and one full-time M.S.-level usability specialist on staff conducting analyses on cases that are found to be related on the basis of searching text within the report itself, not just the tagged fields. To truly understand the reported cases, these specialists should have continuous access to all devices in the field and be able to replicate reported issues with phantom patients. Even independent of reported incidents, these personnel could conduct heuristic evaluations, usability testing, multidevice integration testing, and systematic identification of potential hazards (as Kurt Herzer described in his blog, and, earlier, in his article, “A practical framework for patient care teams to prospectively identify and mitigate clinical hazards.”  In just this way, previously unknown hazards have been discovered in the past that were believed impossible by the device manufacturers, such as an IV pump that allowed free flow of medications when in demo mode and a handheld device for wireless medication administration that corrupted electronic medical record databases. 

Finally, the largest challenge for adverse event—or error—reporting is the fact that it is a fundamentally slow and painful approach to improving systems. Can it be done? Probably. But let me suggest an alternative on the basis of my highly positive experiences with the Veterans Health Administration’s (VHA’s) Office of Information Patient Safety Program Office (PSPO) reporting system for information technology issues. Although this system does ask whether a patient was harmed based on the reported safety concern, it is not a required data field. Instead, reports mainly emphasize “unsafe conditions”: that are attributed to software design concerns, such as information technology that allowed primary care providers who had been terminated to be assigned to new patients. This framing, in addition to reducing barriers to reporting, has also lent itself to easier tracking of whether reported issues were related and had been addressed and has reduced incentives for manufacturers to deny that changes are needed. Another reason why improving systems is so difficult is that, in some legal cases, changes made to devices have been used as evidence of wrongdoing—because if a change was needed, perhaps it wasn’t safe before, leading to the response, “The device worked as intended. It was human error.”