Wednesday, January 30, 2008

GIGO and the CABG

Dontcha just love it when you take your car to the shop for one problem, and they call you back about 7 hours later to tell you that, well, there's maybe a few more? When I was younger, this phone call always involved some problem with the carburetor; now, it's always some electronic gizmo gone haywire.

Something like that happens not infrequently when folks go to the hospital for emergent or elective surgery. "Well, sir, your appendicitis is treated, but we found out that you have diabetes." "The surgery to take out your ruptured spleen went great. The thing is, you aspirated and now have pneumonia, which is going to set back your recovery a few weeks." "Looks like you have had yourself a little heart attack while you were stressed in here. Must have happened when your mother in law came to visit." And so on.

Most of the time, these types of occurrences are pretty easily explained -- wound infections, perioperative pneumonia or cardiac problems, previously undiagnosed diabetes or hypertension. But there is a group of patients that have problems that are harder to understand; these are folks that have GI complications following cardiac surgery. For a variety of reasons, patients undergoing cardiac surgery (with coronary artery bypass grafting being the most common) can develop a range of complications that are physically remote and seemingly unrelated to the heart operation. This isn't common, but according to a recently published study, may be more frequent than we have previously understood.

Gastrointestinal Complications after Coronary Artery Bypass Grafting: A National Study of Morbidiy and Mortality Predictors is more than just a mouthful, it's the title of a review published in December's JACS. From the abstract:


Previous single-institution studies have documented a 0.6% to 2.4% incidence of gastrointestinal (GI) complications after coronary artery bypass grafting (CABG), with an associated 14% to 63% mortality rate. To better determine the incidence and impact of GI complications after CABG, national outcomes for CABG were examined from 1998 to 2002.

Study Design

The Nationwide Inpatient Sample was queried for all patients undergoing CABG (ICD9 procedure codes 36.10 to 36.16). Two cohorts were compared: CABGs with and without GI complications. Both demographic and outcomes variables were compared by either t-test or chi-square analysis. Logistic regression analyses indicated potential predictors of CABG inpatient mortality and GI complications after CABG.


The incidence of GI complications among 2.7 million CABGs identified was 4.1%. Total hospital length of stay (19.3 versus 8.8 days) and inpatient mortality (12.0% versus 2.5%, both p < style="font-weight: bold;">Factors associated with increased risk of GI complications included: age greater than 65 years (odds ratio [OR], 2.1); hemodialysis (OR, 3.4); intraaortic balloon pump (OR, 1.6); concomitant valve procedure (OR, 1.5); and procedure urgency (OR, 1.22). Use of an internal mammary graft was protective (OR, 0.5), but GI complications increased inpatient mortality risk (OR, 2.6).


This national population-based study indicates that GI complications after CABG occur at a higher rate than previously described, leading to increased hospital length of stay and mortality.


"Oh, BS," you say. "DocSurg, you've lost it, if you ever actually had it. How can you say this stinking pile of statistics is interesting?" Well, for starters, given my chosen profession, I can pretty much be assured of being asked to see this very type of patient every once in a while --- and I can be pretty confident in telling patients and families that if they are seeing my smiling face at some point after a CABG, they need to plan on a more than double the normal postop hospital stay, as well as a not inconsiderable increase in the risk of death . And those numbers come with or without my charming bedside manner.

Looking a bit deeper into the paper, there is a laundry list of problems that were teased out of the 2.7 million patient database and labeled as GI complications:
  • Intraabdominal Abscess 28.1%
  • Ileus 23.9%
  • GI ulcer (perforated or bleeding) 16.9%
  • Diverticulitis and Diverticular bleeding 14.4%
  • Pancreatitis 5.5%
  • Liver failure/hepatitis 5.1%
  • Mesenteric ischemia 4.4%
  • Esophageal ulceration or perforation 4%
  • Bowel obstruction 3.2%
  • Biliary tract complication/gallbladder disease 2.9%
  • Clostridium difficile colitis 1.5%
  • Appendicitis 0.2%

What jumps out at me are a few numbers --- the rate of ileus (about 24%) is much higher than I would have anticipated. Part of the problem with labeling an ileus as a complication is that it is a diagnosis that increases the reimbursement a hospital receives from Medicare; in my experience, that leads to overdiagnosis by hospital coders (and I have indeed had this very problem at a local hospital; when I would write in the chart that the patient's expected postoperative ileus had not resolved --- in other words, after a major bowel operation, it is expected that it will take a few days for GI tract function to recover --- I was labeled as having an inordinately high complication rate, because the hospital coders listed ileus as a complication. Needless to say, that word has not been written by my hand in a patient's chart since). So, for argument's sake, let's just throw that out of the list.

Even with "ileus" on the trash heap, intraabdominal and hepatic abscesses remain the most frequent post-CABG complication in the study. Though I am not a cardiac surgeon, I cannot recall ever having been asked to see a post-CABG patient with a hepatic abscess. Once again, this makes me a bit suspicious of how this study got its results.

Of the remaining complications on the list, there is one statistical oddity --- namely, gallbladder problems are relatively infrequent compared to my own personal (and highly subjective) experience. Actually, I'd say the top three reasons I get asked to see a post-CABG patient are cholecystitis, mesenteric ischemia (or the possibility of the same), and bowel obstruction, with pancreatitis in the near vicinity.

The article goes on to nicely summarize the costs and mortality associated with these postop problems, and also describes the current thinking about why some of these complications arise. Without going down a long, winding road with many forks that have to be fully explored, suffice it to say that a lot of this can result from poor gut perfusion --- there's a problem with delivering enough little red boxcars with their important load to the tissues in the abdomen.

OK. Cool. Nice paper, summarizing a pile of data (even though my eyes started to bleed when I read that this came from a database of 2,732,158 patients). But, since I'm particularly curmudgeonly today, and because I don't like to commit things to memory that may not be useful, let me throw a few tomatoes at this paper.

First of all, this study encompassed the period from 1998 through 2002; I guess I would ask, where's the more recent data, at least through 2005? This stuff is, I hope, available on a computer, and database searches for diagnosis codes are pretty doggone easy. But in asking that question, we really get to the heart of my worries about this type of data review ---- we are extrapolating data from diagnosis coding, done by thousands of hospital employees, using a set of diagnosis codes that are so inadequate, incomplete and confusing that they resemble a train schedule from pre-war Italy. I am not criticizing the authors for their choice of data --- it's what is available --- but I am not convinced that we should draw sweeping conclusions from this type of data analysis.

"Well, DocSurg," you say, you are an arrogant @#^%&. Give us a better alternative!"

OK. Here it is.

You see, the above study relied on the National Inpatient Study, data collected by the Healthcare Cost and Utilization Project (HCUP), which is described as a "Federal-State-Industry partnership sponsored by the Agency for Healthcare Research and Quality." Once again, this data is pretty much generated by and collected by non-clinical personnel (I'm quite sure that there are clinicians involved along the way). The Society of Thoracic Surgeons database, however, has a far greater degree of physician input; the surgeon involved with the case fills out the form, which has places for patient risk data, complications, etc. The data is entered online, and there is even a really cool risk calculator that allows real-time feedback of what a patient's risk-adjusted expected morbidity and mortality rates should be.

So what? Is there a difference between what a surgeon considers risks and complications and what a hospital coder does? You bet our sweet patootie there is. As I alluded to above, hospital coding is all geared towards one thing --- getting paid. And, quite frankly, they are getting shafted when it comes to reimbursement for the sickest patients (not to mention outright fraud on the part of some insurers), and they need to fight like the Hansen brothers for every penny they can. Because of our convoluted payment system, they get paid more when an "ileus" is coded, for example. So, that leaves me a bit suspicious of the kind of studies generated from sifting through data from hospital diagnosis codes. Or, to put it another way, GIGO -- garbage in, garbage out.