Saturday, April 4, 2009

They write letters....

For those of you who didn't see it, Dick Wenzel, Mike Edmond, and Gonzalo Bearman published a piece in ICHE last November pointing out the flaws in MRSA screening and isolation programs, arguing instead for a broad-based, "horizontal" approach to infection prevention--one designed to reduce rates of all infections, those due to MRSA included.

Barry Farr and Bill Jarvis don't agree. They say that, and more, in a lively letter to ICHE published in the May issue. The letter, and the response by Wenzel, et al., are good reading, so take a look. I recently published my views on this issue in a JAMA editorial, so I won't rehash them here.

We can establish a few facts, though: (1) MRSA accounts for fewer than 1 in 10 hospital acquired bacterial infections, (2) MRSA bloodstream infection rates in US ICUs are declining significantly, without any widespread MRSA screening, (3) MRSA screening programs are resource intensive and pose risks to patients, and (4) resource limitations are becoming increasingly urgent in our cash-strapped hospitals. So knowing all of this, is it wise to demand that all hospitals begin MRSA screening programs? (I report, you decide)

I also note that Farr and Jarvis now routinely denigrate calls for well-designed, controlled trials to address the impact of MRSA screening, arguing instead that "over 100" observational, quasi-experimental, multiple intervention, or otherwise flawed studies should be dispositive.

In that spirit, let me propose just such an observational study. Let's choose as our "intervention" hospital (Hospital A) a large tertiary care center where MRSA screening has been performed for almost 3 decades (Dr. Farr's hospital). The control hospital shall be a similarly large, tertiary-care hospital where no MRSA screening is performed among adult inpatients, and in which a broad-based, "horizontal" approach to infection prevention is preferred (Dr. Wenzel's hospital, "Hospital B"). Hospital B has already reported their ICU MRSA infection rates for the past several years (decreasing significantly). We now only need Hospital A to provide their data, in order to make the comparison. The publicly available data we do have finds Hospital A to have an overall ICU bloodstream infection rate that is more than double that of Hospital B. Food for thought....


  1. April 10, 2009
    Dr. Diekema says that MRSA CR-BSI rates in ICUs are declining in hospitals not using active detection and isolation (ADI), but fails to mention that this reduction represents a tiny fraction of hospital-wide MRSA HAIs.
    He says that we regularly denigrate calls for randomized controlled trials (RCTs) to evaluate the efficacy of ADI. What we actually oppose is the erection of a double standard for studies of MRSA control by Diekema and others who choose to ignore the fact that there have been no RCTs to evaluate the efficacy of isolation for any other hospital pathogens.
    RCTs are valuable for preventing selection bias in studies of pharmacological agents in which individual patients are the unit of randomization. It is not so clear what bias RCTs of ADI efficacy would seek to prevent when a hospital unit, or perhaps more appropriately, a national healthcare system, would be the unit of randomization [since spread occurs throughout the healthcare system, randomization of individual patients would be not only ineffective but counterproductive]. Moreover, it is also unclear how well RCTs would work to prevent this unspecified bias.
    Importantly, Dr. Diekema doesn’t mention that the only RCT attempted of hospital isolation so far [focusing on MRSA and VRE ADI] was too poorly conducted to provide accurate data and that it seemed to introduce more biases than its randomization of study units could have prevented [e.g., the study protocol provided a phase-in period for staff to adjust to trial study measures in one study limb but provided no phase-in period to adjust to study measures in the other study limb]. The interval from “admission screening sample,” which staff were allowed up to 48 hours to collect, to implementation of isolation was reportedly usually 6 to 7 days (i.e., the patient usually had already been discharged from the unit before the result returned), biasing the study result toward the null.
    After implying that only randomized comparisons would be valid, Diekema then seems to contradict himself by proposing what we estimate to be about the 190th study of ADI efficacy without randomization [over 180 studies having yielded positive results]. He also compares crude rates of CR-BSI from two different hospitals [the University of Virginia (UVA) and the Medical College of Virginia (MCV)] without the benefit of multivariate adjustment for such variables as case mix index—something most hospital epidemiologists have long opposed because patients with a higher case mix index will inevitably have higher infection rates even if the same preventive measures are employed. It is unclear why Dr. Diekema believes such a casual, unstudied comparison should be “dispositive.”
    Dr. Jarvis has never worked at UVA and Dr. Farr retired there 5 years ago, so current UVA rates seem irrelevant, but if Dr. Diekema insists on seeing a comparison of hospital-wide MRSA and VRE primary BSI frequencies for UVA and MCV from a year in which Dr. Farr and Dr. Edmond were both working at the respective hospitals, we can provide them.
    The information was obtained almost a decade ago because Dr. Farr had created a Problem Pathogen Partnership and asked Virginia hospitals to help control the MRSA and VRE spread they were facilitating. Dr. Edmond said MCV couldn’t afford an ADI program. Because a cost comparison conducted at UVA suggested that other university hospitals of comparable size and complexity might be losing more money than UVA was spending on ADI (due to allowing higher rates of more expensive MRSA and VRE infections)[ICHE 2002; 23:407-410.], we asked Dr. Edmond how many nosocomial MRSA and VRE primary BSIs MCV had the year before. He said 60 and 35; UVA had 13 and 6. Some of this difference could have been because MCV was operating more beds that year [MCV was approved for 14% more beds], but a 1.14-fold difference in hospital size wouldn’t explain 4.6-fold more MRSA BSIs or 5.8-fold more VRE BSIs.
    Finally, Dr. Diekema may be interested to learn that the UVA Hospital has often had the highest case-mix index in Virginia [and did at the time of the comparison mentioned in the paragraph above] and was one of only 15 major teaching hospitals to be listed among America’s Top 100 hospitals by the Thomson Reuters 2008 National Benchmarks Study, which reportedly evaluated such things as baseline acuity-adjusted rates of mortality, medical complications, and patient safety measures.


    Barry M. Farr MD, MSc
    William R. Jarvis, MD

  2. For my reply, see this post:

  3. As Dr. Farr points out, the data on infections at my hospital that he cites are a decade old and antedate our aggressive agenda of interventions which began in 2004. Over the last five years we have reduced HAIs overall by 73%, CLABSIs by 76% and VAP by 90%. It is important to note that these results were achieved without any of the added costs of of a MSRSA active surveillance program. And yes, careful stewardship of resources is especially important at safety net hospitals such as VCU Medical Center. We have published our MRSA reduction results in the peer-reviewed literature (AJIC 2008;26:461-463). Since UVA has a longer experience with active surveillance than any other hospital in the country that I’m aware of, it would be interesting to see what has happened to MRSA infections there. But it is also important to see what has happened to all HAIs. As Dick Wenzel has pointed out, not one of the studies cited by Farr and Jarvis that showed reduction in MRSA using active surveillance have shown a decrease in overall infection rates. We must remember that the goal of the patient is not to avoid acquiring a MRSA infection, but rather to avoid any infection. For healthcare consumers the current infection rates will be much more valuable than those from a decade ago. And although Dr. Farr criticizes Dan Diekema’s tongue-in-cheek comparison of CLABSI rates at UVA and VCU, Dr. Farr actually published such a study (ICHE 2002;23:429-35) where he compared VRE bloodstream infection rates at UVA and the University of Maryland (and he didn’t mention the case mix index). At the end of the day, the goal should be to decrease infection rates to the irreducible minimum, and I believe there is no one solution. If hospitals have been successful using active surveillance to reduce MRSA AND have reduced all infections AND can demonstrate they are not harming patients with contact precautions, then I would not attempt to persuade them to do otherwise. Likewise, for hospitals like mine that have been successful in reducing infections without active surveillance, there’s no reason to begin. Lastly, according to the American Hospital Directory (, in 2008 the case mix index at UVA Medical Center was 1.95 and at VCU Medical Center 1.85.

    Mike Edmond