| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ann Thorac Surg 2001;72:1118-1124
© 2001 The Society of Thoracic Surgeons
a Division of General Thoracic Surgery, Boston, Massachusetts, USA
b Clinical Research Program, Massachusetts General Hospital, Boston, Massachusetts, USA
Address reprint requests to Dr Wright, Division of General Thoracic Surgery, Massachusetts General Hospital, 55 Fruit St, Blake 1570, Boston, MA 02114
e-mail: wright.cameron{at}mgh.harvard.edu
Presented at the Thirty-seventh Annual Meeting of The Society of Thoracic Surgeons, New Orleans, LA, Jan 29–31, 2001.
 |
Abstract |
|---|
 Top Abstract IntroductionPatients and methodsResultsCommentDiscussionReferences |
|---|
Methods. Data were obtained from the Massachusetts Health Data Consortium on discharge information for all acute care hospitals in Massachusetts regardless of payer from 1992 to 2000. The influence of hospital volume was related to days in the intensive care unit, length of stay, discharge disposition, hospital mortality, and total cost. Hospitals were stratified to low volume hospitals (< 6 cases per year) and high volume hospitals (> 6 cases per year).
Results. One thousand one hundred ninety-three patients underwent esophagectomy during this 8-year study period in Massachusetts. Three high volume hospitals performed 56.5% of all resections (674 of 1,193). Sixty-one low volume hospitals performed 43.5% of the resections (519 of 1,193) with an average volume of only 1 case of esophagectomy per year. High volume hospitals were associated with a 2-day decrease in median length of stay (p < 0.001), a 3-day reduction in median intensive care unit stay (p < 0.001), an increased rate of home discharges (as opposed to rehabilitation hospitals) (p < 0.001), and a 3.7-fold decrease in hospital mortality (9.2% vs 2.5%; p < 0.001). The odds ratio of death at a low volume hospital was 4.3 (95% confidence interval, 2.3 to 7.7; p < 0.001). The median cost was $755 dollars greater at high volume hospitals (p = not significant).
Conclusions. Hospitals that perform a high volume of esophagectomies have better results with early clinical outcomes and marked reductions in mortality compared with low volume hospitals.
|
Introduction |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentDiscussionReferences |
|---|
Recent studies examining the relationship of volume and outcome for esophagectomies have demonstrated a consistent improvement in clinical outcomes with increased hospital volume [7, 8]. However, these studies have been limited to either Medicare patients or California patients. Massachusetts has a unique hospital environment of several large academic health care providers with access to discharge information on all patients in the state of Massachusetts. The present study analyzes all esophagectomies over an 8-year period in the state of Massachusetts. We hypothesized that the relationship of volume and outcome as evident in Medicare patients and in California patients would also be evident in our study group.
|
Patients and methods |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentDiscussionReferences |
|---|
The study population comprised of all patients discharged from a nonfederal acute care hospital in Massachusetts who had an esophagectomy (International Classification of Disease Procedure code 42.4x and 42.52 or 42.62 or 43.5) performed for esophageal cancer (International Classification of Disease [ICD-9] diagnosis code 150.x) from January 1992 through December 1999.
A hospital was included in the study if at least one esophagectomy was performed during the study period. The hospitals were categorized into high volume and low volume based on whether or not they averaged more or less than six procedures per year.
Other independent variables examined in the analysis were patient age, gender, race, place of residence, urgency of admission, source of admission, year of admission, primary payer type, and secondary payer type. A comorbidity score was calculated for each patient using the Dartmouth-Manitoba adaptation of the Charlson comorbidity index [9, 10]. This comorbidity system only accounts for the most severe types of conditions before admission. Malignancy was not included in the calculation because it was already an inclusion criterion for the study.
Patient outcomes
Primary quality outcomes studied were mean and median length of hospital stay, length of intensive care unit (ICU) stay, in-hospital mortality, and discharge destination (home, secondary care facility, transfer to acute care facility, death). Intensive care unit stay was chosen as a potential marker for quality with the thought that early surgical morbidity would be reflected by a longer ICU stay. This reasoning might be confounded by a tendency for low volume hospitals to keep complex surgical patients in the ICU longer than really necessary. Both mean and median analyses were performed because patients who had adverse outcomes significantly skewed the mean. Hospital charge data from the Massachusetts Health Data Consortium were converted to cost data using ratios provided by Healthshare, Inc (Boston, MA). Using Medicare cost reporting data from the Health Care Financing Administration (Healthshare, Inc, Boston, MA), which calculates a relative cost conversion ratio for each hospital in Massachusetts every year, these ratios were used to convert charge data to cost data. Hospital costs were adjusted for inflation based on the appropriate market basket index from the Medicare Payment Advisory Commission’s (MedPAC) biannual reports to Congress [11]. Costs are presented in constant 1999 dollars.
Statistical analysis
The distributions of patient characteristics between high volume hospitals and low volume hospitals were compared using two sample t-tests for the continuous variables and 
-square tests for categorical variables. Any imbalance between the high volume hospitals and low volume hospitals was included in the multiple regression models.
In the univariate analysis, -squared tests were used to compare the unadjusted mortality and the distribution of discharge disposition, whereas nonparametric tests were used to compare the median length of stay and cost outcomes to reduce the influence of outliers. Multiple linear and logistic regression models were used to assess the independent effect of hospital volumes. Because length of hospital stay and cost outcomes were skewed to the right, natural log transformation was used to minimize the effect of outliers to achieve a more normal distribution. Adjustments for volume, age, race, comorbidity score, urgency of admission, source of admission, year, payer type, and residence were performed for the final regression models. Cluster analysis was also performed on the data [12] because data from patients within the same hospital might not be completely independent. Adjusting for the clustering reduces the effective sample size.
All p values were based on two-sided tests of significance. Data management and analysis were performed using Excel (Microsoft, Redwood, WA) and SAS (SAS Inc, Cary, NC), respectively.
|
Results |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentDiscussionReferences |
|---|
|
|
|
|
|
The median length of ICU stay was significantly lower at the high volume hospitals (2 days) compared with the low volume hospitals (5 days) (p < 0.001). After adjusting for confounding factors, this difference was close to significant (p = 0.08). A significantly higher percentage of patients were home discharged from the high volume hospitals (78.5%) when compared with low volume hospitals (68.0%) (p < 0.001).
Hospital costs and charges
Hospital charges and total costs were analyzed. The median statewide charge was $46,871 (mean, $67,592). High volume hospitals charged $48,764 (mean, $72,060) and low volume hospitals charged $42,357 (mean, $61,790) (p < 0.001). After covariate adjustment and cluster analysis were performed, none of the differences in cost were significant. Only a $755 dollar difference in median cost was noted between high and low volume hospitals after adjustment (p = 0.33). In the patients that died, the median cost was 1.8 times higher at high volume hospitals ($103,296) compared with low volume hospitals ($57,057) (p < 0.001).
|
Comment |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentDiscussionReferences |
|---|
|
In 1998, Sutton examined the performance of a single, well-trained surgeon in 150 consecutive esophagectomies over a 7-year period in England. He demonstrated a significant improvement with reduced single-lung operating time, reduced blood loss, reduced transfusion requirement, reduced ICU stay, and increased yield of lymph nodes [16].
Three studies have examined the relationship between the number of esophagectomies performed yearly by individual surgeons and mortality rates in England, Canada, and the Netherlands. These studies all reported significant decreases in hospital mortality with surgeons performing greater than six esophagectomies per year [17–19] (Table 5). Individual surgeon experience was unavailable in our database, thus it was impossible to analyze the influence of this factor on the overall results.
|
Quality and outcome are clearly influenced by the knowledge, skill, and experience of the whole surgical team. Studies on pancreatic cancer have demonstrated that hospital volume is a stronger determinant of outcome than individual surgeon volume. Sosa and colleagues [4] demonstrated that in Maryland low volume surgeons at high volume hospitals had similar outcomes to the high volume surgeons, whereas low volume surgeons at low volume hospitals had the worst outcomes. A recent study by Swisher and colleagues [8] of Medicare-reimbursed esophagectomies from 1994 to 1996 in 13 national cancer institutions and 88 community hospitals demonstrated that volume of esophagectomies performed and not the number of nonesophageal operations, hospital size, or cancer specialization was the independent risk factor for operative mortality. Characteristics of high volume hospitals and their staff are thought to contribute to the observed improved quality outcomes. A high volume hospital provides an opportunity for a small number of surgeons, anesthesiologists, intensivists, house staff and nursing staff to gain expertise through experience at managing and performing this specific complex operation. This expertise can result in improved patient selection, operating skills, and postoperative management. In addition, care pathways, case managers, specialized facilities, equipment, and 24-hour ICU teams are more likely to be present at high volume hospitals. Increased experience can result in early detection and standardized treatment of complications. During the time period of this study, the high volume hospitals were able to decrease their median length of stay more than their low volume counterparts (4 days vs 2 days). This suggests a greater improvement in the overall system of care at high volume hospitals.
It has been argued that selective referral patterns lead to high volume hospitals having patients that are healthier than patients at low volume hospitals. In this study, low volume hospitals saw slightly more patients that were greater than 70 years of age and admitted on an urgent or emergency basis. These factors can contribute to increased surgical risk and affect quality outcomes. However, after adjusting for differences in patient demographics and comorbidity, there still remained a significant association of improved quality outcomes with increased hospital volume. Thus it is unlikely that referral pattern bias is responsible for our results.
Previous studies analyzing the relationship between economic outcomes and hospital volume have all focused on charge data. Swisher and colleagues [8] noted a $22,227 decrease in charges ($39,867 vs $62,094) for hospitals performing a large number of esophagectomies in their study of 340 Medicare patients. The high volume hospitals were noted to have a median charge of $6,407 more than their low volume counterparts. All three high volume hospitals in the study were academic teaching institutions in Boston with a relatively high cost profile compared with rural or suburban hospitals. Teaching hospitals have the additional economic burden of training house staff that is not a factor at smaller community hospitals. However, there was only a $755 difference when the median costs were compared between the high volume and low volume hospitals. Patients with complications or prolonged hospital courses can cost significantly more than other patients; high volume hospitals had almost twice the median total cost of low volume hospitals for the patients who died. Patients who had adverse outcomes were kept alive a median of 7.5 days longer in high volume hospitals, and this greatly skewed the cost data against these institutions.
There are several potential limitations to this study. First, there are potential errors in the coding of diagnosis and procedures in the Massachusetts Health Data Consortium database. The coding is performed at each individual hospital and then reported to the state and extracted for this database. Extensive efforts were used in this study to limit this source of error and ensure that the cases included in the study fit the study criteria. In addition, coding of major procedures like this one tend to be very reliable. Second, because of the limitations of statewide reporting of discharge data, the stage of the cancer, preoperative radiation or chemotherapy, and specific provider information were not available. Third, although it would be valuable to explore other quality outcomes, such as complications, functional status, quality of life and long-term survival, these outcome measures were also not available in the statewide database. Fourth, the generalizability of this study is uncertain. In the state of Massachusetts, three hospitals account for more than 50% of the cases performed in the state. In the greater Boston area, there is a unique concentration of academic, tertiary care institutions. This data may not be generalized to states with different demographics or health care resources, however it does approximate other published research from the United States and Europe.
In the current medical environment, application of the data in this study can be approached from two separate directions or from both. It can be used for quality improvement programs or it can be used to influence referral patterns, or both. By studying the relationship between volume and outcome, important links between process and outcome can be discovered. This knowledge can improve the overall outcome of patients at all hospitals.
Recently, several initiatives have been aimed at using the purchasing power of the private sector to encourage managed care organizations to contract only with high volume hospitals or provide incentives for patients to seek care at such institutions. The statewide data in this study suggest that patients with esophageal cancer in need of an esophagectomy would benefit from referral to a high volume hospital with experienced staff, facilities, and equipment. This could lead to a substantial decrease in statewide morbidity and mortality associated with this procedure. Currently, 43.5% of cases are performed at low volume hospitals that perform fewer than 6 cases a year. Over two thirds of the hospitals in this study (44 of 64) perform less than one esophagectomy a year. Esophagectomy for esophageal cancer is rarely an emergent operation and the potential quality benefit of referring these patients to a high volume hospital is substantial. Theoretically, if all patients in the low volume group were referred to high volume hospitals, 35 lives (4.4 per year) would have been saved over the 8-year period at a cost of only $755 per patient. In addition, the increased experience and volume as a result of these referrals would likely result in continued improvements of quality outcome measures for this procedure.
|
Discussion |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentDiscussionReferences |
|---|
The authors used the Massachusetts Health Data Consortium Database. Like our South Carolina Lung Cancer Outcome Study, this paper uses a preoperative comorbidity score. The Charlson Comorbidity Index is an old and simple method. It would not include such factors as induction therapy. We do not know whether mortality was the result of increased preoperative risk or postoperative complications. Since the analysis is based on an administrative database, completeness of the chart could be assured, but accuracy could not. One also cannot model for selection bias. The authors appropriately acknowledge limitations of the study. To solve these problems, either an extensive chart audit could be undertaken or, preferably, a prospective cohort study of outcomes for esophagectomy patients needs to be performed. Such a prospective study would allow adequate risk stratification and allow development of a clinical prediction rule.
I have a few specific comments and questions for the authors. I think you do a slight disservice to the low volume hospitals because the text itself does not state that after adjustment the mortality is 4.3%. Considering that low volume hospitals averaged one case per year, this is actually remarkably good. Twenty-four point five percent of low volume hospital cases were either emergent or urgent. How do you account for this? Possibilities include lack of coding experience or overcoding to maximize reimbursement. Although your median values are significant, adjusted hospital mean length of stay, adjusted mean length of intensive care unit (ICU) stay and cost did not reach statistical significance. Your contention is that outliers account for this, but how can you be sure? A mean ICU stay of 4.6 days, even a median of 2 days, is long. In the last year, our average length of ICU stay is less than 24 hours. Also, ICU stay is probably not a good outcome indicator, as low volume hospitals probably keep patients in the ICU for noncritical reasons, such as inadequate nursing on the floors.
In summary, I accept and I endorse the "take-home" message. This paper significantly adds to the outcome literature and emphasizes the need for a prospective study.
I enjoyed your report and I thank the Society for the invitation to discuss this paper. Thank you.
DR WRIGHT: Characteristically, Dr Reed hit the nail on the head and pointed out, as we did in our own paper, the weaknesses of this particular database. The strength, of course, is that it is prospective and completely unbiased. The weakness is that it is not clinically driven and calls for the need, such as we heard yesterday from Dr. Matloff and others, for a national thoracic surgery database. We need to know how we are doing so we can tell our insurers how we are doing and our patients how we are doing.
The mortality, even after adjustment, was still threefold higher, and this was still statistically significant at a 0.001 level. The length of stay indeed was essentially the same, just one day better in the mean data; but that mean data, as we pointed out in the paper, is greatly skewed by outliers, which we think takes away from what, perhaps, the average patient would experience. We did not understand why so many urgent patients were seen in the low volume hospitals. As you all know, except for a perforation during an endoscopy, there is essentially no reason to do an esophagectomy under an emergency circumstance. Patients can always be transferred if appropriate. So we don’t understand that, and it is a potential source of nonunderstood bias. Thank you.
|
References |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  R. Jeganathan, H. Kinnear, J. Campbell, S. Jordan, A. Graham, A. Gavin, K. McManus, and J. McGuigan A surgeon's case volume of oesophagectomy for cancer does not influence patient outcome in a high volume hospital Interactive CardioVascular and Thoracic Surgery, July 1, 2009; 9(1): 66 - 69. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. C. Fernando, S. C. Murthy, W. Hofstetter, J. B. Shrager, C. Bridges, J. D. Mitchell, R. J. Landreneau, E. R. Clough, and T. J. Watson The Society of Thoracic Surgeons Practice Guideline Series: Guidelines for the Management of Barrett's Esophagus With High-Grade Dysplasia. Ann. Thorac. Surg., June 1, 2009; 87(6): 1993 - 2002. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. L. Gruen, V. Pitt, S. Green, A. Parkhill, D. Campbell, and D. Jolley The Effect of Provider Case Volume on Cancer Mortality: Systematic Review and Meta-Analysis CA Cancer J Clin, May 1, 2009; 59(3): 192 - 211. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Kronebusch Assessing Changes in High-Volume Hospital Use: Hospitals, Payers, and Aggregate Volume Trends Med Care Res Rev, April 1, 2009; 66(2): 197 - 218. [Abstract] [PDF]
|
|
|
|
 |
|
 R. A. Meguid, E. S. Weiss, D. C. Chang, M. V. Brock, and S. C. Yang The effect of volume on esophageal cancer resections: what constitutes acceptable resection volumes for centers of excellence? J. Thorac. Cardiovasc. Surg., January 1, 2009; 137(1): 23 - 29. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. W. Wouters, B. P. Wijnhoven, H. E. Karim-Kos, H. G. Blaauwgeers, L. P. Stassen, W.-H. Steup, H. W.Tilanus, and R. A. Tollenaar High-Volume versus Low-Volume for Esophageal Resections for Cancer: The Essential Role of Case-Mix Adjustments based on Clinical Data Ann. Surg. Oncol., January 1, 2008; 15(1): 80 - 87. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Badruddoja Results of Esophagectomy in High- and Low-Volume Centers Arch Surg, November 1, 2007; 142(11): 1112 - 1113. [Full Text] [PDF]
|
|
|
|
|
|
S. Sundaresan, B. Langer, T. Oliver, F. Schwartz, M. Brouwers, H. Stern, and Expert Panel on Thoracic Surgical Oncology Standards for Thoracic Surgical Oncology in a Single-Payer Healthcare System Ann. Thorac. Surg., August 1, 2007; 84(2): 693 - 701. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Migliore, C. K. Choong, E. Lim, K. A. Goldsmith, A. Ritchie, and F. C. Wells A surgeon's case volume of oesophagectomy for cancer strongly influences the operative mortality rate Eur. J. Cardiothorac. Surg., August 1, 2007; 32(2): 375 - 380. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. F. Reed, G. Tolis Jr, B. H. Edil, J. S. Allan, D. M. Donahue, H. A. Gaissert, A. C. Moncure, J. C. Wain, C. D. Wright, and D. J. Mathisen Surgical Treatment of Esophageal High-Grade Dysplasia Ann. Thorac. Surg., April 1, 2005; 79(4): 1110 - 1115. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. B. Dimick, J. A. Cowan Jr, G. Ailawadi, R. M. Wainess, and G. R. Upchurch Jr National Variation in Operative Mortality Rates for Esophageal Resection and the Need for Quality Improvement Arch Surg, December 1, 2003; 138(12): 1305 - 1309. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. M. Shahian and S.-L. T. Normand The volume-outcome relationship: from Luft to Leapfrog Ann. Thorac. Surg., March 1, 2003; 75(3): 1048 - 1058. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. B. Dimick, P. J. Pronovost, J. A. Cowan, and P. A. Lipsett Surgical volume and quality of care for esophageal resection: do high-volume hospitals have fewer complications? Ann. Thorac. Surg., February 1, 2003; 75(2): 337 - 341. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. C. Wolfsen, T. A. Woodward, and M. Raimondo Photodynamic Therapy for Dysplastic Barrett Esophagus and Early Esophageal Adenocarcinoma Mayo Clin. Proc., November 1, 2002; 77(11): 1176 - 1181. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 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 |
