Afib, Severe Sepsis, and the Risk for Stroke and Death

Andrew F. Shorr, MD, MPH


December 22, 2011

This feature requires the newest version of Flash. You can download it here.

This is Andy Shorr from Washington, DC, with a pulmonary and critical care literature update. I would like to bring your attention to an article in the November 23rd issue of JAMA that dealt with critically ill patients in atrial fibrillation.[1] The article by Walkey and colleagues tried to look at the relationships between severe sepsis and septic shock, atrial fibrillation, new stroke in terms of an embolic phenomenon, and in-hospital mortality. The authors looked at this from an administrative claims perspective, using data from California. They initially looked at more than 3 million hospitalizations and then focused on about 50,000 patients with severe sepsis.

The authors used International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) coding to identify severe sepsis and septic shock. They also relied on ICD-9-CM coding to identify new-onset atrial fibrillation and new embolic phenomena. After conducting very careful statistical and sensitivity analyses, the authors concluded that the incidence of atrial fibrillation or new-onset atrial fibrillation in [patients with] severe sepsis is not trivial. In fact, a good number of patients in whom new-onset atrial fibrillation developed in the hospital did have severe sepsis, and patients with severe sepsis seemed to be at a substantially increased risk for new-onset atrial fibrillation developing, even after adjusting for comorbidities and demographics.

Beyond that, the authors saw that not only were patients with severe sepsis at an increased risk for new-onset atrial fibrillation, but they were also at increased risk for a new embolic phenomenon if they did have a new-onset atrial fibrillation event. Not only were you more likely to get atrial fibrillation if you had severe sepsis, but if you did have severe sepsis and it was new-onset atrial fibrillation that was complicating it, you were more likely to have embolic phenomena.

The authors found that, overall, selected variables were independently associated. I want to be cautious here. I don't want to use the term "risk factor" because this was an observational dataset. However, the [authors] saw independent factors that were associated with the development of new-onset atrial fibrillation in severe sepsis. Demographics (men are at higher risk than women), certain comorbidities (ie, underlying heart disease), and measures of acute severity of illness (ie, organ failures) were also independently associated with a greater probability of atrial fibrillation developing in patients who have severe sepsis. Again, we are concerned about new-onset atrial fibrillation.

When the authors adjusted for as many covariants as they could and looked at hospital mortality, new-onset atrial fibrillation in severe sepsis seemed to independently increase the risk for death by about 7%. That is not trivial.

One of the limitations of using claims data is that it is not known how good the claims data are for what you are looking at. Fortunately, for severe sepsis, it seems that the claims data tend to be rather reliable in terms of coding. However, when it comes to atrial fibrillation, we have to be more cautious. Atrial fibrillation may or may not be coded by the intensivist because he or she may not think that it is an important event if it was not very long in duration or if it did not require an intervention and resolved on its own. Whenever you look at observational datasets that rely on claims data, you have to be cautious when the diagnosis or the events that you are interested in are only based on ICD-9-CM codes.

To adjust for all this, the authors did a series of very cautious sensitivity analyses, being much more restrictive in terms of which ICD-9-CM codes they looked at. They really focused in on atrial fibrillation and severe sepsis. They should be commended for that initiative.

Nonetheless, I don't think readers should walk away from this thinking that patients with severe sepsis are at increased risk for atrial fibrillation, and if they get it, that they are at increased risk for an embolic event, which can be associated with mortality issues. Therefore, if a patient has new-onset atrial fibrillation, the patient should [undergo anticoagulation] sooner than he or she might have otherwise. These data are not saying that. The data are only generating hypotheses.

More importantly, these data need to be confirmed to determine whether this association is truly causal by looking at large randomized controlled trials of septic shock and seeing how often atrial fibrillation occurs and what happens afterward. Observational data cannot ever show causation and, of course, we are never going to be able to do a randomized controlled trial of early intervention in atrial fibrillation vs late intervention in atrial fibrillation and severe sepsis because it is hard enough just to do intervention studies for the treatment of severe sepsis itself. However, there are other ways to look at the data to help us decide whether this relationship truly exists and whether we need to be worried about it.

Biological plausibility certainly exists. Patients with severe sepsis are probably procoagulable or hypercoagulable in some way and are therefore more likely to have a clot develop in their left atrial appendage. Similarly, with fluid shifts, their left atrial appendage may stretch more, and so clots may be predisposed to form and, because of endothelial dysfunction, they may be at greater risk for embolic events related to that clot if small pieces of it break away. I think we need to look at other data when we are trying to do causation studies before we rely purely on observational datasets to do anything that might actually change our practice.

This is a very interesting article in the November 23rd issue of JAMA. It goes through some methodologic approaches that I think are crucial if we are going to understand what is going on in critical care from a health services research perspective. At its root, however, the observations need to be taken with caution and be viewed as only hypothesis-generating, rather than as something that needs to change our management at this point. This is the first step in that process.

This is Andy Shorr from Washington, DC.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: