Discussion

Article Outline

• Copyright

Dr Anil Hingorani (Brooklyn, NY). I want to congratulate you on analyzing such a large data set because I think it is the only way we are going to be able to get at some of these questions. However, when you are dealing with these large data sets, one common problem is that you really can't get at some of the details. First, I don't know how you ended up with the data set from California, being from Johns Hopkins, but I congratulate you on acquiring it and analyzing it. One thing I am really concerned about when I read over the abstract is that there were only five hospitals in the last group that met both standards. Whenever you have such a large data set and only five hospitals in one group and you're trying to derive data from there, I think some of your conclusions may not be as solid.

This is still a retrospective database, and the prospective data that has looked at aneurysms has already shown that β-blockers are quite clearly beneficial in perioperative outcome in terms of mortality, and probably long-term survival of these patients, irrespective of whether they have coronary artery disease. What I am actually more interested in seeing is whether or not some of the other retrospective data that have suggested statins may be useful in these patients, which was not addressed in these groups but your group has looked at before.

I want to congratulate you on analyzing such a large data set. I think it was a valiant attempt. But the fact that there were only five hospitals in the last group that met both standards raises some questions in the fact that there are only 25 hospitals in one group, and 32 hospitals in the other group, still leads me to still question some of the conclusions.

Dr Benjamin Brooke. This analysis was focused on California hospitals alone for a number of different reasons. For one, Leapfrog was rolled out in phases, and California was one of the few states targeted in the first rollout period beginning in 2000 to 2001. Second, this state contained the largest number of hospitals targeted by the initiative overall, so it offered us the most robust statewide data set. Moreover, California has a diverse patient population from which to generalize our results to the broader community. The fact that only five hospitals in this state met both Leapfrog criteria should not come as a surprise, given the results of previous studies showing that only a small number of hospitals in California are able to meet Leapfrog's volume criteria. While outcomes at these hospitals were significantly improved compared with control hospitals, you are correct that we need to be careful about extrapolating too much from limited data. To address these limitations, we plan to do future analyses that measure the clinical impact of Leapfrog standards for AAA repair using a large nationwide data set among Medicare beneficiaries.

You raise another excellent point about the fact that we are evaluating the effect of implementing known evidence-based standards such as β-blocker use. But one of the main reasons for doing this analysis is to address the question: If the evidence is based on well-designed randomized trials and is already out there, why aren't more hospitals standardizing β-blocker use during elective AAA repair? Indeed, our analysis was not designed to evaluate the efficacy of Leapfrog standards on elective AAA repair; its primary intent was to look at their effectiveness on a population basis.

Finally, you are absolutely correct about the evidence for statin use, and I agree that Leapfrog should consider this process measure as another evidence-based standard for hospitals performing elective AAA repair.

Dr John Blebea (Philadelphia, Pa). I am a bit concerned about your conclusions. It appears too simplistic to ascribe outcomes of AAA repair to only these two variables. I would propose that the use of β-blockers and volume may be an associated variable but possibly not be a causal one. One could hypothesize that larger hospitals would be more interested in complying with Leapfrog standards compared with smaller hospitals. Additionally, it is possible that hospitals with less prevalent β-blocker usage had a larger percentage of patients with COPD who would therefore not be given β-blockers but would be at higher risk for surgical morbidity and mortality. Therefore, these two variables alone may not be causally related to a better outcome but may only be associated with improved clinical results. Since you have demonstrated that EVAR vs open surgery has a very large positive impact on mortality, have you done a multivariate analysis to better determine possible causal relationships?

Secondly, is there a significant difference in outcome predictability using β-blockade, volume, or a combination of both factors? It appeared that the results were similar or was one single variable more important?

Dr Brooke. You made an excellent point about the different types of patients that may be treated in hospitals meeting Leapfrog standards for AAA repair versus hospitals that don't meet these criteria. Indeed, we considered using a multilevel or hierarchal model for this analysis, which would allow for adjustment of both patient-level and hospital-level confounders and help risk stratify the types of patients that are being treated at different hospital groups. However, a multilevel analysis becomes more complicated to interpret and would require more observations among hospitals meeting both Leapfrog standards. We focused on hospital-level outcomes because we were looking at effectiveness of Leapfrog standards for AAA repair on a population basis, and for process measures like β-blocker usage, you don't have patient-level data; only self-reported hospital data stating whether they comply with that standard or not.

As far as doing a multivariate analysis, we in fact adjusted for hospital level variables using Poisson regression models for each of the Leapfrog hospital groups. Because we stratified outcomes by the type of procedure, an EVAR variable wasn't adjusted for in this regression model. However, we adjusted for all the other hospital-level variables, such as number of floor or ICU admissions, number of beds, teaching status, et cetera, and consistently found in-hospital mortality to be lower for hospitals meeting Leapfrog standards.

In regards to your last comment about measuring the independent effect of β-blocker vs volume in a multivariable analysis, while the reported outcomes were obtained by comparing hospitals meeting individual Leapfrog standards to control hospitals in separate models, similar results were obtained when all Leapfrog standard variables were included in the same model. We chose to develop separate multivariate models for each Leapfrog standard due to concerns about the collinearity of similar variables in a single model.

Dr Ali F. AbuRahma (Charleston, WV). Did you look into the impact of endovascular procedures in these hospitals on the length of stay for the open repair? I just noticed in our institution that the people who do an open surgery start to release their patients even much earlier so they will have a favorable outcome against the endovascular repair. Did you notice the same in your experience, sir?

Dr Brooke. While there was a very slight trend in reduction of length of stay for the hospitals that met both Leapfrog standards, you really couldn't say there was any difference given the variance. So no, we didn't see any temporal changes for length of stay among open AAA cases. It was only for endovascular AAA repair cases in our population.