| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J Thorac Cardiovasc Surg 2009;137:28-29
© 2009 The American Association for Thoracic Surgery
Invited Commentary |
Several large single-institution series have been published with remarkably low mortality rates, including your institution; ours, where in a series of 222 consecutive esophagectomies had a mortality rate of 1.4%; and excellent results from other larger series, such as those from Drs Orringer, Altorki, and Swanson, to name a few.
However, this relationship between esophagectomy volume and outcome is complex, with several factors playing a role. These factors include the surgeon volume, specialty training of the surgeon, comorbidities in the patient, and provision for critical care services. Adding to these factors is the case mix seen at a particular hospital, which may contribute. For example, the referral pattern of a private hospital may be much different than that of an inner-city hospital.
So my first question is, given the complexities of the volume–outcome relationship, do you think it is possible that you can reduce this to a single number across all hospitals in the United States without taking into consideration other important factors, such as surgeon volume, expertise, and patient population?
Dr Meguid. The impetus for our study was the curiosity to see how the seemingly arbitrary cutoff of 13 fared against other volume cutoffs. We expected to see a dramatic difference in mortality rates, for instance, but were, quite frankly, shocked by the apparent lack of the difference between choosing 13 and any other volume cutoff.
As you point out, the relationship between individual and hospital operative volume and the processes of care and outcome is very complex. Unfortunately, it's difficult to study the effect of different processes of care because of lack of information available in these multi-institutional databases. Using a patient-focused database, such as the Society of Thoracic Surgeons' database, would be ideal; however, at the present time data are lacking for such analysis, and unfortunately, use of all of our own single-institution databases lends to bias.
Dr Pennathur. The second question is, did you first conduct an analysis of all the data in this particular cohort of patients to establish a functional relationship between mortality and volume in this cohort before dichotomizing these patients?
Dr Meguid. Yes, sir. Before dichotomizing the data at the different cutoff points, we did examine the unadjusted relationship between volume and in-hospital death and found this to be an inversely linear relationship. When we adjusted for age, patient gender, patient race, and patient comorbidities, we also saw that that persisted.
Dr Pennathur. The next thing is, using mortality as an outcome variable, how do you do a risk-adjusted mortality rate? For example, from your article, when the volume was greater than 10, the mortality rate was 5.3%. However, when the volume was greater than or equal to 29, the mortality was actually higher; it was 8.6%. Is this because sicker patients are going to high-volume hospitals? Perhaps a more useful reporting might be a risk-adjusted mortality rate.
Dr Meguid. Yes, I fully agree with you. In dichotomizing a continuous variable, one is combining all of the values below and above that cutoff into 2 values. As a result, the lowest mortality rate was observed at a volume cutoff of 10, but when the volume cutoff was raised, a lower mortality rate was not observed. Again, that's one of the prime complications of using a dichotomous model for continuous variables.
Dr Pennathur. Along the same lines, are you going to attempt to take other variables into consideration, such as nutritional status of patients, which has been shown to be important; socioeconomic status; specialty training of the surgeon, which has been show to be important; and elective versus emergent procedures, all of which are going to have an impact on mortality?
Dr Meguid. Yes, I fully agree with you regarding the importance of these factors. Unfortunately, because of the administrative nature of the NIS database, one is unable to account for many germane factors, such as cancer staging, preoperative nutritional status, and neoadjuvant or postoperative chemo- and radiotherapy. One can control for gender, some patient demographics, which aren't necessarily specific to operative mortality, the Charlson Comorbidity Index, and some hospital demographics. This is, again, a big limitation with these administrative datasets.
Dr M. Jaklitsch (Boston, Massachusetts). I'm just deeply concerned that your highest output hospital was 29 cases per year, and we all know of excellent academic centers of excellence that do more than 29 per year. So either they are not in the database, the most sterling outcomes are not in this database, or they are in this database but with incomplete data. I don't know how you can draw conclusions from this if either 1 of those 2 cases is true.
Dr Meguid. Dr Jaklitsch, that's an excellent point. In fact, when we looked at a similar example using pancreatic resections, we found a similarly small range, and that motivated us at Johns Hopkins to look into why we don't find institutions with 100 resections per year. Subsequently, in this nationally representative sample, we find that a lot of the hospitals are lower-volume hospitals and we don't see a lot of the larger academic centers included every year in this database. So that is, again, a limitation specific to the NIS database.
Dr Jaklitsch. Can I ask, how many cases a year do you do at Hopkins?
Dr Meguid. I'm not sure, I believe it's in the 50s. Dr Yang?
Dr Yang. It's about 75.
Dr Jaklitsch. So Hopkins' data, for instance, is not in this?
Dr Meguid. It's not in the NIS, no, sir.
Dr T. Grodzki (Szczecin, Poland). Did you make a differentiation between hospital volume and surgeon volume? Because it's not the same. And did you analyze the mortality reasons? Were they due to technical failure or the imperfections in postoperative care?
Dr Meguid. Those are very good points. We chose to use hospital volume, although we could have examined physician volume. We chose to use hospital volume because that is what has been used in other models, including the Leapfrog Group, and we wanted to analyze these cutoffs in particular. Unfortunately, we can't tease out what postoperative complications occur in these patients because of the limitations of this dataset.
Dr D. Wood (Seattle, Wash). I have the same concern that Dr Jaklitsch expressed, but also a concern about how you have represented the conclusion. It would seem that volume is an important surrogate for mortality outcomes, yet you have found that there is not a good cutoff for volume. That does not mean that volume isn't important, which it sounds like in the conclusion. Rather, volume is very important; we just cannot create a cutoff to define an "adequate" volume. So I think that it is very important to refine the message, because policymakers, like Leapfrog, need a clear message that volume is important in terms of quality of outcomes, unless you think that this research disputes that premise.
Dr Meguid. Dr Wood, you have made an excellent point. I don't want to misrepresent our findings. Our findings are that a specific cutoff is an inappropriate way to determine centers of excellence. However, in this example, increased volume is correlated with decreased complications and mortality, and that should not be overlooked.
| ||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
