The University of Maryland group has just published two new papers on the safety of contact precautions. One is an analysis of the data from the BUGG study (published in Clinical Infectious Diseases), which was a multicenter study of intensive care unit patients, and the other is a matched prospective cohort analysis at an academic medical center (published in Infection Control & Hospital Epidemiology).
Here's a quick summary of the 2 papers:
The authors' conclusions for both papers are essentially the same: concerns about the safety of contact precautions should not limit the implementation of contact precautions. One paper showed no difference in adverse events and the other paper showed that contact precautions was actually safer than not being in contact precautions, which I think is a counterintuitive finding for many of us.
The problem that I have with both of these papers is the use of the Institute for Healthcare Improvement (IHI) Global Trigger Tool (GTT) to determine the rates of adverse events. I have 3 concerns:
1. Are all of the events identified by the GTT causally linked to the exposure? For example, in the ICHE study, the strongest association was found for surgical events, and the association was highly counterintuitive: postoperative hemorrhage was found in 1 in148 patients in contact precautions vs 12 in 148 patients not in contact precautions. Given what we know about contact precautions, particularly that it reduces healthcare worker visits, what could be the causal mechanism for a higher rate of post-operative hemorrhage in patients not in contact precautions? Is there a confounder at play here? There may very well be a causal mechanism but I'm hard pressed to think of one.
2. The IHI GTT only looks for errors of commission, not errors of omission, and it seems likely that errors of omission would occur more commonly when a barrier to care is implemented (contact precautions).
3. The IHI GTT sets a 20-minute time limit for the detection of errors when reviewing a patient chart whether the patient was hospitalized for 1 day or 100 days. In the ICHE study, the patients were matched on length of stay; however, I don't think this resolves the problem. The GTT's ability to detect error is time dependent. So when comparing two groups, if one group truly has a significantly higher rate of adverse events, a lower proportion of the events will be detected in the group that has more events, reducing the difference between the two groups. In effect we have surveillance bias.
Even the IHI openly admits that this is an issue. Here's a direct quote from the GTT manual:
The IHI Global Trigger Tool was never intended to identify all adverse events. Experienced reviewers familiar with the IHI Global Trigger Tool will identify almost all events greater than category E in a patient record that can be reviewed completely within the 20-minute time limit. Events in the E category of harm involve more judgment and at times are not as obvious, so these are less easily identified and may be missed. When the 20-minute time limit is enforced not all adverse events are expected to be identified.
In both of the papers, the 3 issues cited above are likely biasing towards the null, making contact precautions to appear to be safer than it is. So I continue to be concerned that contact precautions may result in less safe care. I think that most people in the hospital epidemiology community regard the University of Maryland researchers as some of the best methodologists in the field, and I certainly hold them in very high regard. Nonetheless, I'm curious about their love for the Global Trigger Tool, and hope that they share their views with us on this blog.
Michael B. Edmond, MD, FACP, is a hospital epidemiologist in Iowa City, IA, with a focus on improving the quality and safety of health care, and sees patients in the inpatient and outpatient settings. This post originally appeared at the blog Controversies in Hospital Infection Prevention.