Saturday, February 25, 2012

Attention, Pennsylvania IPs and hospital epidemiologists!

I share Mike’s amazement at the low overall prevalence of healthcare-associated infections (HAIs) in recent reports from Pennsylvania (1% of all admitted patients for the years 2009 and 2010). Especially since the 5-10% estimates aren’t just from older studies. Recent studies-- in fact, virtually every study performed outside of high-stakes public reporting mandates--report similar estimates (see this WHO summary, pages 12-15, from which the I pasted part of a Figure to the right, with PA added in red). As Mike points out, an amazing achievement.

But I have to stay true to the name of this blog, so for the sake of argument I will now assume a highly impolite and cynical stance regarding this report. Here goes: I don’t believe it.

I think we have reason to assume that a healthy proportion of CLABSIs are “adjudicated away” by many centers, and lord only knows how many other HAIs (VAP, anyone?) vanish into the ether of consensus review, clinician veto, administrative adjustment, or just plain old bad surveillance. Recall also the SHEA 2011 presentation reporting evidence for gaming in the earlier PA data (using methods that can also be criticized, but the “difference-in-differences” analysis was persuasive enough for me to accept that there was both a real reduction in HAI rates, and evidence for gaming, since the PA public reporting mandate).

Now that I’ve thrown down the gauntlet, Pennsylvania, please convince me that these data are accurate. The report references validation—how was it done? Was your own hospital subject to audit of HAI rates? How many records were reviewed, if chart review was performed? What were the credentials of the auditors? As a “consumer” of health care, would you base your decisions about where to receive care on facility level data from your state reporting program?

4 comments:

  1. I think the problem of comparing the PA data with the data in the WHO report is that the WHO report includes studies as far back as 1995, which is well before there was any real push to reduce infections. In my own hospital, our infections peaked in 2003. In 2004, we became very aggressive about reducing infections, and have seen significant reductions since. Unfortunately, I can't compare our overall infection rate with the PA data since our scope of surveillance is more limited. However, I can say that our hospital-wide device associated infection rate is <1%. So that is why the PA 1.1% metric seemed reasonable to me. Nonetheless, as you point out, some hospitals are probably cheating. But even if they intentionally or unintentionally missed half of the infections, the rate would only be 2%, and I'd still say that's pretty damn good!

    I also agree that validation can be problematic. When our CLABSI data were validated by the state health department, the validator concluded that we falsely elevated our infection rate by over-calling infections. We appealed each discordant case and the validator concluded that we were correct.

    So who knows what the true incidence of infection is or was in hospitals? I think it's fair to day that an enormous amount of work has been invested in reducing infections and I would be amazed if that did not result in a significant reduction.

    ReplyDelete
  2. Hi Mike,

    It's true that some of the studies in the WHO report go back to the mid-90s, but many of them are more recent, up to the mid-2000s, and in EU countries that I believe in have been good about HAI prevention. And this isn't a single hospital, remember, it's an entire state (a state I will soon move to, or at least move to before I get sick, if these numbers are correct).

    I also agree that we've made tremendous progress. I just think that in the current era, with incentives aligned as they are, we are simply counting different things, for different reasons, than we were 10-15 years ago. When HAI surveillance was done solely for internal quality improvement and patient safety, the incentive was to be overly inclusive, not miss any infections that might be hospital acquired, preventable or not. Things have changed enormously, not just in how we prevent infections, but also in how we count them. Our new benchmarks will be problematic if the pendulum ever swings back to center (i.e. if the creative surveillance approaches taken at many hospitals are inhibited by expanded validation, or if manual surveillance is replaced by purely electronic measure reporting).

    Dan

    ReplyDelete
    Replies
    1. I was a Pa IP and was subject to audit. They had a trained IP do the audit. They asked for every positive blood culture in inpatients in a year. They then took a sampling of reported CLABSI and picked other blood cultures randomly (from what I can tell) They disagreed with me on a case and I was required to report it despite the fact I did not believe it was related to the central line. Not sure if you know but in Pa hospitals are required to send a letter to the patient telling them that they acquired an infection in the hospital, even though in my professional judgment it was transient bacteremia.

      The state auditors are now threatening fining hospitals for not reporting when the cultures are positive, hospitals are required to report "serious events" within 24 hours of confirmation and the state auditors are interpreting confirmation as the date of the positive cultures.

      Delete
  3. Thanks for your comment, PA ICP, that's extremely helpful information! I tried to find published descriptions of the PA audit procedure but couldn't (maybe something is out there and I just missed it). But you are describing a serious audit, glad to hear the auditor had IP training. If you can indulge me, did the auditor review very many of the CLABSIs you called (say, more than half), and if so did they find any "overcalls" (where they felt it shouldn't have been reported)? How many of the other "random" blood cultures did they review, and what proportion of those reviewed had central lines? Probably way too much detail to expect you to recall, so feel free to ignore.

    I had heard a bit about the rapid reporting requirements. I really have no idea how a hospital could adhere to those requirements, given the time constraints and the fact that a secondary source may not be known at the time the first positive culture is called.

    So I'm still on the fence about whether to move to PA--if the HAI rate really is 1% statewide, that's pretty cool, but those reporting requirements are insane. Plus, too, the fracking.

    ReplyDelete