Volume Thresholds And Hospital Characteristics In The United States

Transcription

1 Volume Thresholds And Hospital Characteristics In The United States Nationwide evidence that skill and experience of staff are part of the volume-outcome link for certain surgical procedures. by Anne Elixhauser, Claudia Steiner, and Irene Fraser ABSTRACT: Procedure volume has been used as a proxy for quality and recommended as a basis for hospital referrals. We studied the volume, mortality, and associated hospital and staffing characteristics of ten complex procedures in U.S. hospitals using the 2000 HCUP Nationwide Inpatient Sample. Although the majority of patients had their procedures performed in high-volume hospitals, for seven procedures, more than three-fourths of hospitals would be considered low-volume. Unadjusted mortality rates were significantly higher at low-volume hospitals for five procedures. Low-volume hospitals also tended to have lower mean numbers of residents and RNs. However, for two procedures, low-volume hospitals had RN and resident staffing equal to or higher than those of high-volume hospitals, and the unadjusted mortality rates were no different. Increased concern about patient safety in the hospital has brought renewed research and policy attention to a body of research linking mortality to the volume of inpatient procedures performed. Two recent literature reviews concluded that for certain complex and high-technology procedures, hospitals that perform more than a threshold number have significantly lower mortality than do hospitals where fewer procedures are performed. 1 Volume threshold levels have been recommended as the basis for evidence-based referrals, and a recent study concluded that in the absence of other information on quality of care, volume of procedures is a reasonable standard for selecting hospitals. 2 It is not clear how many people or hospitals would be affected by volume-based referrals, because much of the research has not provided national baseline information. Adams Dudley and colleagues used California data to estimate the impact of volume-based referrals. 3 John Birkmeyer and colleagues focused on Medicare patients. 4 In addition, past research does not provide much insight on why the link between volume and outcomes exists. Given that some low-volume hospitals Anne Elixhauser is a senior research scientist in the Center for Organization and Delivery Studies at the Agency for Healthcare Research and Quality in Rockville, Maryland. Claudia Steiner is a senior research physician there, and Irene Fraser is the center s director. HEALTH AFFAIRS ~ Volume 22, Number Project HOPE The People-to-People Health Foundation, Inc.

2 Research Challenge have outcomes comparable to those of high-volume hospitals and some highvolume hospitals have poor outcomes, it would follow that hospital volume is not the sole explanatory variable. 5 Volume may be a proxy for other explanatory variables, such as processes of care, expertise of staff, and ratios of staff to patients. 6 Without any underlying explanation for any observed volume-outcome link, it is difficult to recommend strategies for improvement. This study seeks to provide two important additions to the volume-outcome literature and policy debate. First, we provide nationwide baseline estimates on the number of high- and low-volume hospitals and the number of procedures they perform. Second, we present information on staffing and other hospital characteristics that are associated with high- or low-volume hospitals. These characteristics can contribute to an improved understanding of the volume-outcome link, from which strategies for improvement could be derived. Data And Methods The analysis is based on hospital inpatient data from the Healthcare Cost and Utilization Project (HCUP), a federal-state-industry partnership in health care data. 7 One of its components the Nationwide Inpatient Sample (NIS) contains information found in a typical hospital discharge or billing record. NIS 2000 includes hospital discharge data from twenty-eight states (see acknowledgment). Discharges in these states account for more than 80 percent of all U.S. hospital discharges and constitute the NIS sampling frame. The NIS is designed to be a 20 percent sample of U.S. community hospitals, defined by the American Hospital Association (AHA) as all nonfederal, short-term, general and other specialty hospitals. 8 The NIS is a stratified probability sample of hospitals in the frame, with sampling probabilities proportional to the number of U.S. community hospitals in the universe (defined using the AHA Annual Survey of Hospitals) in each stratum. This universe is divided into strata using five hospital characteristics: ownership/control, bed size, teaching status, rural/urban location, and geographic region. Hospitals from the HCUP-participating states are selected to represent these strata. All discharges from sampled hospitals are included in the database; thus, it is possible to assess the volume of procedures performed in each sampled hospital. Two types of weights are available for the NIS. Discharge weights indicate the number of discharges that the sample discharge represents in the universe of discharges from U.S. hospitals for that year in that stratum. Hospital weights indicate the number of hospitals that the sample hospital represents in the universe of U.S. hospitals in that stratum. We used hospital weights because the unit of analysis is the hospital; however, analyses using discharge weights gave comparable findings. The entire sample of NIS hospitals (N = 4) and discharges (N = 7,450,2) for 2000 were used for this study. These discharges and hospitals were weighted to obtain estimates that are representative of U.S. inpatient discharges and hospitals. 168 March/April 2003

3 The estimated total number of U.S. discharges based on the NIS is 36,400,000, which compares favorably with the estimate of 35,800,000 discharges (including newborns) in 1 based on the National Hospital Discharge Survey. Weighted analyses were performed using SAS, and standard errors for all estimates were calculated using SUDAAN to account for the complex sampling design, which clusters discharges within hospitals. Z-tests were used to assess statistically significant differences between high- and low-volume hospitals. Dudley and colleagues identified ten procedures with a strong volumeoutcome relationship in the literature and defined volume thresholds for each. 10 Our study uses these definitions of procedures and thresholds (Exhibit 1). Only hospitals that performed at least one of each specified procedure were included in analyses for that procedure. We determined the mean number of each procedure performed in low- and high-volume hospitals, and the overall percentage of low-volume hospitals performing each one. We calculated the number of all procedures performed in low-volume hospitals. We also calculated the unadjusted in-hospital mortality, and the mean patient age and number of comorbidities, for each procedure in high- versus low-volume hospitals. We used our comorbidity method, which flags thirty unique coexisting illnesses that affect length-of-stay, charges, and mortality. 11 EXHIBIT 1 Nationwide Comparison Of Low- And High-Volume Hospitals, By Number Of Hospitals, Mean Number Of Procedures, Percentage Of Hospitals That Are Low-Volume, And Percentage Of Procedures Performed At Low-Volume Hospitals, 2000 Procedure Abdominal aortic aneurysm repair Carotid endarterectomy Mean number of procedures performed b Number of U.S. hospitals that are high- and low-volume Volume thresholds a High Low High Low ,44 (88.3%) 2,088 (82.1%) Procedures performed at low-volume hospitals 54.1% 45.3 Lower-extremity arterial bypass CABG , ,178 (43.6%) 783 (75.1%) Coronary angioplasty Heart transplantation (52.2%) 58 (44.4%) Pediatric heart surgery Pancreatic cancer surgery (80.4%) 1,204 (0.1%) Esophageal cancer surgery Cerebral aneurysm surgery (5.%) 1,015 (2.4%) SOURCE: 2000 Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample, unless otherwise indicated. NOTES: Standard errors (available from the authors on request) for all estimates were calculated using SUDAAN to account for the complex sampling design, which clusters discharges within hospitals. CABG is coronary artery bypass graft. a R.A. Dudley et al., Selective Referral to High-Volume Hospitals: Estimating Potentially Avoidable Deaths, Journal of the American Medical Association 283, no. (2000): b p <.001 for all of the procedures listed for the observed differences between high- and low-volume hospitals. HEALTH AFFAIRS ~ Volume 22, Number 2 16

4 Research Challenge We linked the NIS database to the AHA Annual Survey of Hospitals, which provides information on an array of hospital characteristics. We performed unadjusted comparisons of low- to high-volume hospitals for the following hospital characteristics: mean number of residents and interns per 100 beds; mean number of full-time-equivalent (FTE) registered nurses (RNs) per 100 beds; owner status (investor-owned, private, not-for-profit, and public); location/teaching status (rural, urban teaching, and urban nonteaching), and region (Northeast, South, Midwest, and West). A hospital is considered to be a teaching hospital if it has an American Medical Association (AMA) approved residency program, is a member of the Council of Teaching Hospitals (COTH), or has a ratio of FTE interns and residents to beds of 0.25 or higher. Study Results Exhibit 1 provides information on the percentage of U.S. hospitals that met or exceeded the volume thresholds in For seven of the ten procedures examined, more than 75 percent of hospitals would be considered low-volume institutions. For only two procedures do at least half of the hospitals reach the highvolume threshold: heart transplantation and lower-extremity arterial bypass. In 2000, 1.4 million procedures were performed in the ten categories examined here. As one would expect, the mean number of procedures performed in highvolume institutions was significantly higher than in low-volume institutions. For example, hospitals designated as high-volume for coronary artery bypass graft (CABG) performed, on average, 863 of these procedures over the year, compared with 218 in low-volume hospitals (Exhibit 1). Similarly, hospitals designated as high-volume for esophageal cancer surgery performed an average of nine of these procedures, compared with only two in low-volume hospitals. For most of these procedures, low-volume hospitals treat fewer than half of all patients receiving these procedures. Ten percent or less of patients with lowerextremity arterial bypass and heart transplant had their procedure performed at a low-volume hospital, and only about 15 percent of patients receiving coronary angioplasty did so. Fewer than half of all patients receiving CABG, carotid endarterectomy, and pediatric heart surgery had these procedures performed in a low-volume hospital. However, for esophageal cancer surgery, cerebral aneurysm repair, pancreatic cancer surgery, and abdominal aortic aneurysm repair, most patients had their procedures performed at low-volume hospitals. Across all ten procedures, about 27 percent were performed in low-volume institutions. Unadjusted mortality rates for half of these procedures were significantly higher in low-volume than in high-volume hospitals (Exhibit 2). For example, low-volume hospitals for CABG had in-hospital mortality rates that were, on average, 18 percent higher than those in high-volume hospitals. Low-volume hospitals for esophageal cancer surgery had mortality rates that were on average 1.5 times higher than those in high-volume hospitals. Unadjusted mortality rates 170 March/April 2003

5 EXHIBIT 2 Nationwide Comparison of Low- And High-Volume Hospitals, By Mean In-Hospital Mortality, Mean Number Of Comorbidities, And Mean Age Of Patients Undergoing Each Procedure, 2000 Mean in-hospital mortality, percent unadjusted Mean number of comorbidities Mean age (years) Procedure High Low High Low High Low Abdominal aortic aneurysm repair Carotid endarterectomy Lower-extremity arterial bypass CABG Coronary angioplasty Heart transplantation Pediatric heart surgery Pancreatic cancer surgery Esophageal cancer surgery Cerebral aneurysm surgery ** *** *** *** *** *** 65.6*** * 11.5*** * SOURCE: 2000 Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample. NOTES: Standard errors for all estimates (available from the authors on request) were calculated using SUDAAN to account for the complex sampling design, which clusters discharges within hospitals. Mortality is the mean of the percentage of patients that died and received the procedure. CABG is coronary artery bypass graft. *p <.05 **p <.01 ***p <.001 could be higher in low-volume hospitals if the patient population was generally more ill. For pancreatic cancer surgery, whose unadjusted mortality is higher in low-volume hospitals, patients at those hospitals also appeared to be older and bore a slightly higher comorbidity burden. However, for the remaining procedures, mean age and number of comorbidities differed little between high- and low-volume hospitals. Exhibit 3 provides information on two measures of staffing intensity and expertise: the mean number of residents and interns per 100 beds and the mean number of FTE RNs per 100 beds. For four procedures, low-volume hospitals had significantly lower mean numbers of residents and interns per bed. In addition, there were significantly fewer RNs per 100 beds in low-volume hospitals for eight procedures. Of note, low-volume hospitals performing heart transplants and pediatric heart surgeries had as many or significantly more residents/interns and RNs than their high-volume counterparts did. Although low-volume hospitals on average did significantly fewer heart transplants and pediatric heart surgeries than their high-volume counterparts did, their overall mortality rates were no different. We then compared high- and low-volume hospitals along other characteristics: teaching status, number of beds, hospital ownership or control, and region (Exhibits 4 and 5). Low-volume hospitals tended to be small, urban nonteaching, or rural institutions that had for-profit ownership or were located in the South. HEALTH AFFAIRS ~ Volume 22, Number 2 171

6 Research Challenge EXHIBIT 3 Nationwide Comparison of Low- And High-Volume Hospitals, By Mean Number Of Residents And Interns Per 100 Beds And Mean Number Of Full-Time-Equivalent (FTE) Registered Nurses (RNs) Per 100 Beds, 2000 Mean number of residents and interns Mean number of FTE RNs Procedure High Low High Low Abdominal aortic aneurysm repair Carotid endarterectomy *** *** 11*** Lower extremity arterial bypass CABG 14 2*** *** 133* Coronary angioplasty Heart transplantation ** 182 Pediatric heart surgery Pancreatic cancer surgery 36 3*** 7*** *** Esophageal cancer surgery Cerebral aneurysm surgery *** *** 134** SOURCE: 2000 Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample; and American Hospital Association Annual Survey of Hospitals. NOTES: Standard errors for all estimates (available from the authors on request) were calculated using SUDAAN to account for the complex sampling design, which clusters discharges within hospitals. CABG is coronary artery bypass graft. *p <.05 **p <.01 ***p <.001 Discussion And Policy Implications This study provides a nationwide picture of ten procedures for which a volume-outcome relationship has been established in U.S. hospitals. For many of these volume-sensitive procedures, the majority of patients are treated in highvolume hospitals. To the extent that volume is considered a proxy for quality, these national figures provide good news and a more optimistic picture than one would deduce from some smaller-scale studies. Only patients receiving esophageal cancer surgery were predominantly treated in low-volume hospitals. For most categories, the percentage of U.S. patients treated in high-volume hospitals greatly exceeds the percentage that Dudley and colleagues found in California. Rates of procedures performed in low-volume hospitals there were 61 percent for lower-extremity arterial bypass (compared with 2 percent nationally), 64 percent for coronary angioplasty (86 percent nationally), 34 percent for CABG (57 percent nationally), 37 percent for carotid endarterectomy (55 percent nationally), and 36 percent for elective abdominal aortic aneurysm repair (46 percent nationally). Only for pediatric heart surgery did the California high-volume hospital rate (7 percent) greatly exceed the national rate (5 percent). This study also shows that any effort to raise the share of procedures performed by high-volume hospitals could have a negative impact on many hospitals. For eight procedures, fewer than half of U.S. hospitals that performed at least one pro- 172 March/April 2003

7 EXHIBIT 4 Percentage Of Hospitals Classified As Low-Volume Hospitals, By Region, 2000 Procedure Northeast South Midwest West Abdominal aortic aneurysm repair a,b,c 86.5% Carotid endarterectomy a,b,c,d,e 83.8 Lower-extremity arterial bypass a,d,f 33.4 CABG a,c,d,e,f 42.3 Coronary angioplasty a,d,f 32.5 Heart transplantation a,d,f % % % Pediatric heart surgery 87.3 Pancreatic cancer surgery b,e 87.0 Esophageal cancer surgery e 0.1 Cerebral aneurysm surgery a,d,f SOURCES: 2000 Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample; and American Hospital Association Annual Survey of Hospitals. NOTES: Significant differences between regions (p <.05) are as follows. CABG is coronary artery bypass graft. a Northeast significantly different from South. b South significantly different from Midwest. c South significantly different from West. d Northeast significantly different from West. e Midwest significantly different from West. f Northeast significantly different from Midwest. cedure met or exceeded the high-volume threshold for that procedure. Only for heart transplantation and lower extremity arterial bypass surgery could at least half of U.S. hospitals be considered high-volume institutions. This finding suggests that any systematic effort to encourage use of high-volume hospitals for some procedures (whether through education, selective contracting, or other means) could divert patients from up to half of all hospitals, thereby affecting their financial picture. 12 However, because many of these hospitals perform relatively few of these procedures, the direct financial impact for the lowest-volume hospitals would be small. Although this is difficult to assess, other related lines of service also could be adversely affected. Hospitals that perform a higher volume of procedures but are still below the threshold could be more directly adversely affected. On the other hand, patients who are required to travel longer distances to high-volume centers would bear an additional burden in terms of travel costs, inconvenience, and increased distance from family and support networks. These issues should be explored in future studies and should be balanced against improved medical outcomes. The explanation for this combined finding most procedures are done in highvolume hospitals, but most hospitals perform them at low-volume levels lies in the wide differences in the number of procedures performed by high- as compared with low-volume hospitals. This study shows that low-volume hospitals often perform very small numbers of procedures, whereas many high-volume hospitals HEALTH AFFAIRS ~ Volume 22, Number 2 173

8 Research Challenge EXHIBIT 5 Percentage Of Hospitals Classified As Low-Volume, By Urban/Rural Status, Teaching Status, And Ownership, 2000 Procedure Urban, teaching Urban, nonteaching Rural Private, for-profit Private, not-for-profit Public Abdominal aortic aneurysm repair a,b,c,d Carotid endarterectomy a,b,c,e 75.% % % % % % 85.8 Lower-extremity arterial bypass a,b,c,e,f CABG a,c,f Coronary angioplasty a,b,c Heart transplantation g g g Pediatric heart surgery c,f Pancreatic cancer surgery a,b,c,d g Esophageal cancer surgery a,b,c Cerebral aneurysm surgery a,b SOURCES: 2000 Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample; and American Hospital Association Annual Survey of Hospitals. NOTES: Significant differences between location/teaching status and ownership types (p <.05) are as follows. CABG is coronary artery bypass graft. a Urban, teaching significantly different from urban, nonteaching. b Urban, teaching significantly different from rural. c Private, for-profit significantly different from private, not-for-profit. d Private, for-profit significantly different from public. e Urban, nonteaching significantly different from rural. f Private, not-for-profit significantly different from public. g Too few cases to make reliable estimates. perform far more than the threshold number. For most of the ten procedures, the mean number performed by the low-volume hospitals was one-third the threshold level or lower. This suggests that policymakers and others interested in an incremental approach to volume-based referrals (or concerned that methodological problems create uncertainty about how to handle hospitals near the threshold levels) might be able to identify some very-low-volume hospitals as a starting point for changing referral patterns. One caveat of this study lies in the precise thresholds for these procedures that are based upon the results of a single literature review that surveyed numerous studies and determined the threshold based on the highest-quality studies. Future analyses and policies based on the volumeoutcome literature should explore alternative thresholds to assess whether different thresholds would result in much different definitions of high- and low-volume hospitals. This study examines unadjusted mortality rates between high- and low-volume hospitals and finds that for five of the ten procedures, mortality was significantly higher at the low-volume hospitals. Mean age and mean number of comorbidities were generally no different between high- and low-volume hospitals, which suggests that the higher mortality rates at low-volume hospitals are not explained by 174 March/April 2003

9 Our study provides further evidence of a link between staffing/ expertise and improved outcomes. the presence of more severely ill patients. The exception appears to be low-volume hospitals performing pancreatic cancer surgery, where patients appear to be somewhat older and have a slightly higher comorbidity burden. Our study also provides further insight into some associated characteristics of low-volume as compared with high-volume hospitals. Low-volume hospitals tend to be located in the South or in rural areas; are nonteaching institutions; and hold private, for-profit investor status. Furthermore, low-volume hospitals tend to have lower fewer residents and interns and fewer RNs per 100 beds. Higher levels of nurse staffing have also been associated with improved quality of care in hospitals. 13 Teaching status and its relationship to quality of care and outcomes have been examined across illnesses and procedures. 14 Many of these studies suggest that teaching hospitals have more favorable clinical outcomes. Levels of expertise and staffing may be an underlying explanation for the observed volume-outcome link; this could be modified to improve outcomes in low-volume hospitals. In contrast to five other complex procedures (CABG, coronary angioplasty, pancreatic and esophageal cancer surgery, and cerebral aneurysm surgery), lowvolume hospitals performing pediatric heart surgery and heart transplants had mortality rates that were comparable to those of high-volume hospitals. Furthermore, for pediatric heart surgery and heart transplants, the mean numbers of RNs and residents/interns per 100 beds were not much higher at the high-volume hospitals, in contrast to other procedures. Thus, our study provides further evidence of a link between staffing/expertise and improved outcomes. Although these findings cannot provide a specific explanation for the association between volume and outcomes (for example, are patients selectively referred to institutions, resulting in higher volumes?), the findings do add further evidence of the relationship between hospital staffing and patient outcomes. Although procedure volume may be a convenient proxy for quality of care, questions have been raised about the ramifications of policy making based on volume. Although there appears to be a statistical link between volume and quality of care, the nature of this link is still poorly understood. For example, recent studies have compared morbidity and mortality at low- and high-volume centers for esophagectomy, pancreatic resection, and carotid endarterectomy and suggest that volume alone is not a sufficient signal of quality. 15 These studies point to at least two additional factors that influence outcomes: surgeons skill and experience, and the presence of an organizational structure for assuring high quality of care, such as treatment protocols. In this study, heart transplants and pediatric heart surgery were performed at relatively few hospitals compared with the other procedures. It is possible that for the most complex procedures, whether frequent HEALTH AFFAIRS ~ Volume 22, Number 2 175

10 Research Challenge or infrequent, hospitals providing them must maintain a certain level of staffing and technology to support even one procedure. In addition, a few specialized surgeons might perform these procedures at more than one hospital, so an individual hospital s volume, whether high or low, is a poor proxy for outcomes. Individual surgeon volume, staffing, or measures of the presence of key technologies or practices, such as protocols, may be better measures. Such factors should be further examined so that we can better understand how to improve quality and to provide the basis for quality improvement initiatives. This study focuses on high-technology, complex procedures for which a link between volume and mortality has been suggested. This link is likely to be indirect and complex, reflecting at least the organization of health care services and the skill and experience of staff. Our study illustrates the value of administrative data in providing insights into the delivery and outcomes of medical care, as a tool for hypothesis generation, and as a step in understanding complex relationships that require further in-depth analysis before decisions are made. The authors gratefully acknowledge the programming support provided by Suzanne Worth and Shiqiang Li of Social and Scientific Systems. The authors also acknowledge the state data organizations that participate in the HCUP Nationwide Inpatient Sample (NIS) 2000: Arizona Department of Health Services; California Office of Statewide Health Planning and Development; Colorado Health and Hospital Association; Connecticut Hospital Association (CHIME); Florida Agency for Health Care Administration; Georgia Hospital Association; Hawaii Health Information Corporation; Iowa Hospital Association; Illinois Health Care Cost Containment Council; Kansas Hospital Association; Kentucky Department for Public Health; Maine Health Data Organization; Massachusetts Division of Health Care Finance and Policy; Maryland Health Services Cost Review Commission; Missouri Hospital Industry Data Institute; New Jersey Department of Health and Senior Services; New York State Department of Health; North Carolina Department of Health and Human Services; Oregon Association of Hospitals and Health Systems and Office for Oregon Health Plan Policy and Research; Pennsylvania Health Care Cost Containment Council; South Carolina State Budget and Control Board; Tennessee Hospital Association; Texas Health Care Information Council; Utah Department of Health; Virginia Health Information; Washington State Department of Health; West Virginia Health Care Authority; and Wisconsin Department of Health and Family Services. The views expressed in this paper are those of the authors and do not necessarily represent those of the Agency for Healthcare Research and Quality or the U.S. Department of Health and Human Services. 176 March/April 2003

11 NOTES 1. E.A. Halm, C. Lee, and M.R. Chassin, How Is Volume Related to Quality in Health Care? A Systematic Review of the Research Literature, in Institute of Medicine, Interpreting the Volume-Outcome Relationship in the Context of Health Care Quality: Workshop Summary (Washington: National Academy Press, 2000), 4 ; and R.A. Dudley et al., Selective Referral to High-Volume Hospitals: Estimating Potentially Avoidable Deaths, Journal of the American Medical Association 283, no. (2000): Leapfrog Group, Fact Sheet: Evidence Based Hospital Referral, November 2000, www. leapfroggroup.org/factsheets/ehr_factsheet.pdf ( December 2002); and J.D. Birkmeyer et al., Hospital Volume and Surgical Mortality in the United States, New England Journal of Medicine 346, no. 15 (2002): Dudley et al., Selective Referral to High-Volume Hospitals. 4. Birkmeyer et al, Hospital Volume and Surgical Mortality. 5. E.L. Hannan, The Relation between Volume and Outcome in Health Care, New England Journal of Medicine 340, no. 21 (1): Ibid.; Halm et al., How Is Volume Related to Quality in Health Care? ; and K.A. Phillips and H.S. Luft, The Policy Implications of Using Hospital and Physician Volumes as Indicators of Quality Care in a Changing Health Care Environment, International Journal for Quality in Health Care, no. 5 (17): Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project (HCUP), : A Federal-State-Industry Partnership in Health Data, November 2001, hcup-pkt.htm (8 November 2002). 8. American Hospital Association, Hospital Statistics, 2002 ed. (Chicago: AHA, 2002).. J.R. Popovic, 1 National Hospital Discharge Survey: Annual Summary with Detailed Diagnosis and Procedure Data, Vital and Health Statistics no. 151 (Hyattsville, Md.: National Center for Health Statistics, 2001), International Classification of Diseases, Ninth Revision, Clinical Modification (ICD--CM) code definitions for all procedures can be found in Dudley et al., Selective Referral to High-Volume Hospitals. 11. A. Elixhauser et al., Comorbidity Measures for Use with Administrative Data, Medical Care 36, no. 1 (18): Birkmeyer et al., Hospital Volume and Surgical Mortality ; and J.D. Birkmeyer, C.M. Birkmeyer, and J.S. Skinner, Leapfrog Patient Safety Standards: Economic Implications (Washington: Leapfrog Group, 2001). 13. J. Needleman, Nurse-Staffing Levels and the Quality of Care in Hospitals, New England Journal of Medicine 346, no. 22 (2002): ; and J. Needleman et al., Nurse Staffing and Patient Outcomes in Hospitals, Executive Summary of Final Report to U.S. Department of Health and Human Services, Health Resources and Services Administration, ftp://ftp.hrsa.gov/bhpr/nursing/staffstudy/staffexecsum.pdf ( December 2002); and C. Kovner and P.J. Gergen, Nurse Staffing Levels and Adverse Events following Surgery in U.S. Hospitals, Image The Journal of Nursing Scholarship 30, no. 4 (18): J.J. Allison et al., Relationship of Hospital Teaching Status with Quality of Care and Mortality for Medicare Patients with Acute MI, Journal of the American Medical Association 284, no. 22 (2000): ; M.M. Hutter, R.E. Glasgow, and S.J. Mulvihill, Does the Participation of a Surgical Trainee Adversely Impact Patient Outcomes? A Study of Major Pancreatic Resections in California, Surgery 128, no. 2 (2000): ; S.F. Khuri et al., Comparison of Surgical Outcomes between Teaching and Nonteaching Hospitals in the Department of Veterans Affairs, Annals of Surgery 234, no. 3 (2001): ; G.E. Rosenthal et al., Severity-Adjusted Mortality and Length of Stay in Teaching and Nonteaching Hospitals, Results of a Regional Study, Journal of the American Medical Association 278, no. 6 (17): ; and Z. Yuan et al., The Association between Hospital Type and Mortality and Length of Stay: A Study of 16. Million Hospitalized Medicare Beneficiaries, Medical Care 38, no. 2 (2000): R.S. Padmanabhan et al., Should Esophagectomy Be Performed in a Low-Volume Center? American Surgeon 68, no. 4 (2002): ; A. Afsari et al., Outcome Analysis of Pancreaticoduodenectomy at a Community Hospital, American Surgeon 68, no. 3 (2002): ; and C. Peck, J. Peck, and A. Peck, Comparison of Carotid Endarterectomy at High- and Low-Volume Hospitals, American Journal of Surgery 181, no. 5 (2001): HEALTH AFFAIRS ~ Volume 22, Number 2 177