| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Click here to read this article as a CME activity
a Division of Cardiac Surgery, Department of Surgery, The Johns Hopkins Medical Institutions, Baltimore, Maryland
b Division of Pulmonology and Critical Care Medicine, Department of Medicine, and the Bloomberg School of Public Health, The Johns Hopkins Medical Institutions, Baltimore, Maryland
Accepted for publication June 1, 2009.
* Address correspondence to Dr Shah, Department of Surgery, Division of Cardiac Surgery, The Johns Hopkins Hospital, Blalock 618, 600 N. Wolfe St, Baltimore, MD 21287 (Email: ashah29{at}jhmi.edu).
Presented at the Forty-fifth Annual Meeting of The Society of Thoracic Surgeons, San Francisco, CA, Jan 26–28, 2009.
| GENERAL THORACIC SURGERY:
The Annals of Thoracic Surgery CME Program is located online at http://cme.ctsnetjournals.org. To take the CME activity related to this article, you must have either an STS member or an individual non-member subscription to the journal.
|
 |
Abstract |
|---|
 Top Abstract IntroductionMaterial and MethodsResultsCommentDiscussionAcknowledgmentsReferences |
|---|
Methods: We reviewed United Network for Organ Sharing data (1998 through 2007) to identify 10,496 first-time adult lung transplantation recipients at 79 centers. Centers were stratified by quartiles of mean annual volume. Risk of 30-day mortality and 1- and 5-year mortality (censored for 30-day death) were assessed by multivariable Cox proportional hazards regression.
Results: Mean center volume ranged from less than 1 to 58.2 (median, 9.4 cases/year; volume quartiles: 0 to 2.1, 2.2 to 9.4, 9.5 to 19.9, and 20 to 58.2 cases). Each 1 case/year decrease led to a 2% increase in 30-day mortality (hazard ratio, 1.02; 95% confidence interval, 1.01 to 1.02; p < 0.001). Centers of lowest quartile (performing 
2.1 lung transplantations/year) had a 30-day cumulative mortality of 9.6% or 89% increase in the risk of death (hazard ratio, 1.89; 95% confidence interval, 1.01 to 3.44; p = 0.05) compared with the highest quartile centers despite fewer idiopathic pulmonary fibrosis patients (15.6% versus 25.8%; p < 0.001) and younger age (40.9 versus 51.5 years; p < 0.001). Low-volume centers had double the risk of 30-day censored 1-year mortality (hazard ratio, 1.95; 95% confidence interval, 1.30 to 2.92; p = 0.001). High-volume centers (
20 lung transplantations/year) had the lowest 30-day mortality (4.1%).
Conclusions: We provide an initial examination of the relationship of volume and lung allocation score to outcomes for lung transplantation. Low center volume is associated with increased short-term and cumulative mortality despite fewer idiopathic pulmonary fibrosis patients and younger patients.
For two decades, lung transplantation (LTx) has been an important treatment for end-stage lung disease. Despite its acceptance in the medical and surgical community, LTx is a complex procedure with a high risk of mortality, requiring experienced providers and specialized ancillary staff.
With more than 79 U.S. centers performing LTx during the past decade, there exists a wide discrepancy in the average annual number of cases performed. The Centers for Medicare and Medicaid Services mandates an institutional volume of 10 cases per year to qualify for funding [1]. Despite this mandate, there surprisingly exists few reports examining volume-based outcomes for LTx patients.
Both hospital and provider volume have been shown to influence outcomes for a variety of surgical procedures. Several studies have now demonstrated increased hospital volume to be associated with improved survival [2–10], shorter length of stay [11], decreased readmissions [12], and decreased hospital costs [11, 13, 14]. These effects have been observed for procedures within the realm of general surgery, thoracic surgery [2–10], urology [15], gynecology [16, 17], and vascular [18] and cardiac surgery [19–21].
The relationship between center volume and outcomes would seem to assume increasing importance for complex surgical procedures such as LTx. Studies examining orthotopic heart transplantation (OHT; a similar procedure of complexity requiring specialized providers and support systems) have identified a center effect whereby increasing volume leads to decreased mortality [22–24]. Adding to the growing body of literature on this subject, our group recently completed an investigation demonstrating increased short-term mortality for centers failing to meet sufficient volume standards in OHT [25].
Given the importance that volume holds for complex surgical procedures, we sought to provide an initial examination of the impact of hospital volume on short-term mortality in LTx. We used United Network for Organ Sharing data to examine an open cohort of U.S. patients receiving LTx during the last decade. We hypothesize that similar to OHT, increased institutional volume will be associated with decreased short-term mortality in LTx.
|
Material and Methods |
|---|
|
TopAbstractIntroductionMaterial and MethodsResultsCommentDiscussionAcknowledgmentsReferences |
|---|
Study Design
We retrospectively examined a cohort of first-time adult LTx patients receiving either single or double LTx (patients receiving heart-lung transplantation were excluded) during the past 10 years of U.S. LTx (January 1998 through December 2007). The variable annual institutional volume was derived from the existing center identifier codes present in the data set. Using this primary variable, we examined volume as a continuous variable and also divided the cohort by quartiles of annual center volume.
Variables Examined and Outcome Measures
The data set contains more than 400 unique preoperative, intraoperative, and postoperative variables. Pertinent variables examined included demographic factors (age, sex, race, education level, and insurance type), markers of pulmonary status (oxygen requirement, 6-minute walking distance, forced expiratory volume at 1 second, forced vital capacity, forced expiratory volume at 1 second to forced vital capacity ratio, mechanical ventilation before LTx, and intensive care unit care before transplant), comorbidities (diabetes mellitus, body mass index, preoperative creatinine levels, and hypertension), and transplant variables (ischemic time, HLA mismatch, panel-reactive antibody level, year of transplant, and wait-list times). We also examined donor variables including donor age, race, sex, and body mass index.
The primary end point was the incidence of mortality within 30 days. We also examined mortality at 1 and 5 years after LTx conditional on surviving 30 days to examine long-term survival independent of short-term mortality.
Analysis
We compared baseline characteristics among the four center volume strata by one-way analysis of variance (continuous variables) and the 
2 test (categorical variables). For significant associations, post hoc pair-wise comparisons between strata were performed with the Tukey-Kramer method (continuous variables) and by univariate logistic regression (categorical variables).
Cumulative survival was estimated using the Kaplan-Meier product-limit estimator focused on time intervals with adequate follow-up. Survival estimates by group were compared using the log-rank test. Center-specific mortality rates were plotted by means of two-way scatter plots with linear best-fit regression lines to assess the effect of center volume on mortality. Influence (leverage x residual) was assessed by examining the scaled differences between predicted mortality outcomes with each single observation left out, and points of high influence were excluded in sensitivity analysis.
Risk of death was estimated using a multivariable Cox proportional hazards regression model with censoring occurring for lost to follow-up and for those whose follow-up time ended at the end of the study period. Independent covariates with potential for confounding and with minimal missing data (<10% missing) were first tested in a univariate model. Those reaching statistical significance (p < 0.05) or known confounders based on literature were incorporated into the multivariable model in a stepwise fashion using the likelihood ratio test to examine nested models. The final model incorporated the following covariates: age 65 years, year of transplant, donor and recipient race, creatinine before transplant, history of malignancy, body mass index, history of diabetes, intensive care unit before transplant, disease type (idiopathic pulmonary fibrosis, chronic obstructive pulmonary disease, or other), and ischemic time. Model goodness of fit was assessed by calculation of the Harrell's concordance statistic (C-statistic) [26], which estimates how well the Cox model correctly identifies the order of survival times between pairs of patients.
For all analyses, a probability value of less than 0.05 (two-tailed) was considered significant. Means are presented with standard deviations. Medians are presented with interquartile ranges. All hazard ratios (HR) are presented with 95% confidence intervals (CI). Statistical analyses were performed with the aid of STATA software (version 9.2 Special Edition; StataCorp LP, College Station, TX).
|
Results |
|---|
|
TopAbstractIntroductionMaterial and MethodsResultsCommentDiscussionAcknowledgmentsReferences |
|---|
Center Volume
The 10,496 LTx cases occurred in 79 distinct centers (Fig 1). Forty-eight centers (60.1%) performed at least one LTx in each year of the study. The median annual institutional volume was 9.4 LTx/year (range, 0.1 to 58.2 LTx/year). Stratification by volume quartile revealed the following ranges: quartile 1, 0 to 2.1 LTx/year, n = 160; quartile 2, 2.2 to 9.4 LTx/year, n = 1,060; quartile 3, 9.5 to 19.9 LTx/year, n = 2,836; and quartile 4, 20 to 58.2 LTx/year, n = 6,440.
|
|
|
|
|
|
When the multivariable analysis was repeated with the lowest volume quartile as the reference, only the highest volume quartile showed a protective effect at both 30 days (HR, 0.53; 95% CI, 0.29 to 0.99; p = 0.05 for 30-day mortality) and 1 year (censored for 30-day death; HR, 0.51; 95% CI, 0.34 to 0.77; p = 0.001 for 1-year censored mortality), indicating that it is the high-volume centers that lead to the greatest decrease in adjusted mortality.
|
Comment |
|---|
|
TopAbstractIntroductionMaterial and MethodsResultsCommentDiscussionAcknowledgmentsReferences |
|---|
It is important to note that high-volume centers achieved superior outcomes despite transplanting older patients with a greater proportion of idiopathic pulmonary fibrosis. Several studies have now demonstrated the substantial acuity associated with LTx patients with idiopathic pulmonary fibrosis [27, 28], and it is interesting, that better outcomes persisted at high-volume centers despite this demographic.
Although a clear trend existed pointing toward lower mortality at high-volume centers, it is important to mention that not all low-volume centers had high rates of mortality. Specifically, four centers with mean volumes of less than 2 LTx/year did not have any deaths in the first 30 days of transplant. However six centers in this group averaged more than 25% cumulative mortality within the first 30 days of LTx, and the group as a whole averaged 5.5% lower cumulative survival at 30 days.
Examination of the Harrell's C-statistic in the unadjusted multivariable analysis further revealed the minimal effect that volume had on the explanatory power of the model. The C-statistic indicates the percentage of times that the model predicts the order of survival between random pairs of subjects. A C-statistic of 50% indicates no prediction. In the model incorporating volume alone, the C-statistic was only 54%, indicating that the predictive power of 30-day and 1- and 5-year censored mortality was not substantially influenced by volume alone.
This fact coupled with the wide variance in survival at low-volume centers underscores the fact that volume is insufficient to explain mortality in LTx. It should be mentioned that the C-statistic is unlikely to provide a strongly predictive result with any one covariate as evidenced by the low predictive power of additional covariates such as diagnosis and ischemic time. Despite this, it is clear that the complexity of modeling both short- and long-term mortality after LTx exceeds that able to be explained by volume alone. What this analysis demonstrates is that although volume by itself is insufficient to explain outcomes, high-volume centers have lower overall mortality rates and less variation among centers. It is clear that further research is needed to identify those factors which led to decreased variance and superior outcomes in the high-volume centers.
Relationship Between Volume and Mortality
Ample evidence now exists to confirm a center effect for a variety of surgical procedures [4, 10, 25, 29]. This relationship becomes especially important for complex procedures such as transplantation [22, 25]. Although the reasons for this are not entirely clear and not measurable using this data set, they likely involve processes of care delivered effectively at high-volume centers. Recent substantive work has focused on identifying and disseminating these processes of care to help improve outcomes broadly. Specifically studied have been the presence of dedicated intensive care unit providers [30, 31], patient safety measure implementation [32], and the importance of multidisciplinary teams and of critical pathways [5, 33, 34]. It is our belief that these processes of care are especially important for solid organ transplantation.
Although volume outcome relationships in LTx have not been thoroughly examined, there have now been several studies examining volume outcome relationship in heart transplantation. Heck and colleagues [22] and Evans and associates [24] in two separate studies used International Society of Heart and Lung Transplantation data to show increased mortality at low-volume OHT centers. After this investigation, an important study was undertaken by Hosenpud and coworkers [23] using United Network for Organ Sharing data to demonstrate lower 30-day and 1-year survival in low-volume institutions.
As these studies were reported more than 14 years earlier, our group used United Network for Organ Sharing data to examine volume outcome effects in a modern cohort of OHT patients [25]. Similar to the results of the current study, we found a doubling of the risk of death at 30 days in centers falling below Centers for Medicare and Medicaid Services standards of 10 OHTs/year and a 2% increase in the odds of 30-day mortality for each decrease of 1 case/year.
Our current study contributes to this literature by examining volume mortality relationships for LTx. Similar to heart transplantation, we have found that certain low-volume centers have alarmingly high mortality rates, and centers of high volume achieve consistently superior results.
Restricting Lung Transplantation to Centers of Expertise
Although based on this data alone, the authors do not recommend closing all low-volume centers based purely on the number of cases performed, potential benefits of regionalization must be considered. The data presented show an approximate 12% difference in absolute risk between those centers averaging greater than 20 LTx/year and those less than 20. Based on number needed to treat calculations (1/absolute risk reduction), 8.3 patients would need to be transplanted at high-volume centers to save 1 life at 5 years. With 4,056 patients undergoing transplantation at centers averaging less than 20 cases per year, shifting LTx patients to high-volume centers would have saved approximately 489 lives or 13% of the 3,818 patients who died within 5 years during the past decade of LTx.
This calculation does not account for additional processes of care influencing outcomes after LTx, and regionalization does pose unique challenges for patients, families, and payers. However, as our group has previously discussed with OHT [25], regionalization does offer distinct advantages infrequently identified in the literature. Specifically, high-risk patients (who are generally not accepted as candidates in low-volume centers) may gain greater access to LTx through restricting LTx to centers of expertise. Furthermore, donor utilization might be improved with this approach. High-volume centers are generally more likely to use marginal donors (including donation after cardiac death), allowing increased organ utilization. Modeling the potential nationwide regionalization of LTx was beyond the scope of this investigation. However, these data highlight the need to further discuss implications of this type of policy decision.
Benchmark for Low-Volume Centers
That some low-volume centers produce excellent outcomes calls attention to the fact that factors influencing mortality after LTx are far more complex than can be explained by experience alone. This is highlighted by the fact that on multivariable analysis, only 4% of the variance in the model is explained by volume. Volume outcome analyses such as this one provide important early insight, but must be accompanied by detailed multiinstitutional data to examine those factors for which volume is a surrogate and which truly drive outcomes. Our intention is not only to examine the relationship between volume and mortality but also to provide a benchmark for low-volume centers to strive for. Clearly centers with superior outcomes are equipped with skilled staff and integrated patient support systems, and a culture of excellence. These characteristics should be quantified and provided to all centers large and small.
Limitations
The study is limited by its retrospective nature. We acknowledge a lack of control over all potential confounders. Although we use multivariable statistics to control for patient level covariates, we acknowledge the potential for unmeasured confounders. United Network for Organ Sharing records provide limited information on transplant issues such as immunosuppressive regimens and secondary outcomes such as quality of life. Furthermore, we believe that volume may be serving as a surrogate for unknown factors influencing mortality in this data set. We acknowledge that a limitation of the study is the inability to identify those factors. Finally, we cannot confirm the accuracy of coding, although we generally assume equal distribution of coding errors.
Conclusions
In this analysis, we have provided an initial examination of the effect of volume on mortality for LTx. High-volume centers achieve consistently superior results whereas low-volume centers are marked by variable rates of mortality. Multiinstitutional data are needed to identify those factors tied to institutional volume that portend improved outcomes in LTx.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterial and MethodsResultsCommentDiscussionAcknowledgmentsReferences |
|---|
Thank you.
DR WEISS: Thank you Dr Bousamra for your comments. Certainly if you look at the scatterplot I presented, there clearly are centers performing a low number of lung transplantations that do just as well as the high-volume centers. The point of the study, however, is that on the aggregate, the low-volume centers tend to fall short. We believe that processes of care, which likely exist to a greater extent at the high-volume centers, contribute to superior outcomes. These processes need to be investigated and disseminated to low-volume centers so that all centers can achieve superior results. I want to make the point that we are not saying that centers with low volume should be shut down or that the cutoff we presented should be strictly enforced. We are identifying a benchmark that low-volume centers can strive to achieve. The data show that centers performing 20 lung transplants a year have superior outcomes. This is something that low-volume centers can strive to achieve.
DR CERFOLIO: But if the government and Obama get your paper, they are going to be shutting programs down. Yes or no? Possibly?
DR WEISS: That is difficult to comment on.
DR BOUSAMRA: But I want to make the point that low-volume transplant centers shouldn't be striving to do 20 transplants if their results are good. You don't want to encourage transplant centers to do transplants they don't need to be doing.
DR CERFOLIO: We understand.
DR MALCOLM M. DECAMP (Boston, MA): You are not suggesting that transplant volume is an independent predictor of outcome. You had two statements in one slide that said it's a surrogate for things that might be a lesson to programs that are underperforming, if I understand your paper correctly.
The comment I had is, one of the issues in lung transplantation outcomes is that we have been very focused on operative mortality and 1-, 2-, and 5-year survival, without paying much attention to deaths while waiting on the list. That is another way that's underrepresented or underreported in terms of potentially gaming transplant outcomes. The lung allocation score is supposed to get around that, but it remains an imperfect science. I would be interested in how you might incorporate into your analysis deaths while waiting on the list.
DR WEISS: Well first of all, thank you Dr DeCamp for your comments. Yes, I agree with you, volume is most likely a surrogate for other processes of care, which cannot be identified from this analysis. What we have presented here is the raw data which shows that volume matters. Why it matters is less clear and needs to be the subject of future investigations. I agree with you in that volume by itself is certainly not everything and I think it would be incorrect to assume that. There are clearly additional processes of care likely driving outcomes.
In terms of incorporating deaths on the waiting list, I think that is an interesting point. It's a difficult analysis to conduct and not something that we did in this paper. Therefore I can't comment on wait-list mortality for low- and high-volume centers. I think that type of analysis would make an interesting and important next step for this project.
DR JESSICA S. DONINGTON (New York, NY): Were you able to identify any basic processes of care, like larger centers allocating lungs differently, the use of singles versus double lungs, or other factors like that, from your analysis?
DR WEISS: Specific processes of care are hard to identify using the United Network for Organ Sharing data set. The data set is comprehensive and provides a real-world snapshot of United States practice patterns, but like all retrospective data sets, it has limitations. We have limited information on the processes of care you mentioned. It is noteworthy that the use of bilateral lung transplantation was fairly uniform across the different volume quartiles in this analysis.
|
Acknowledgments |
|---|
|
TopAbstractIntroductionMaterial and MethodsResultsCommentDiscussionAcknowledgmentsReferences |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterial and MethodsResultsCommentDiscussionAcknowledgmentsReferences |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 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 |
