Replacement of the native aortic valve with a mechanical

Transcription

1 Mortality After Aortic Valve Replacement: Results From a Nationally Representative Database Brad C. Astor, MPH, Ronald G. Kaczmarek, MD, Brockton Hefflin, MD, and W. Randolph Daley, DVM Center for Devices and Radiological Health, Food and Drug Administration, Rockville, Maryland Background. Nationally representative estimates of inhospital mortality after aortic valve replacement are needed to evaluate whether results from The Society of Thoracic Surgeons National Cardiac Surgery Database are applicable to other institutions in the United States performing these procedures. Methods. Data from the 1994 Nationwide Inpatient Sample were used to estimate the patient characteristics and in-hospital mortality rates associated with aortic valve replacements performed in nonfederal hospitals in the United States. Procedural and hospital characteristics were examined for possible associations with in-hospital mortality. Results. An estimated 46,397 aortic valve replacements were performed. In-hospital mortality occurred in 4.3% of first-time isolated aortic valve replacements and 6.4% overall. The highest quartile of procedure-specific hospital volume, compared with the lowest quartile, was associated with lower in-hospital mortality (adjusted odds ratio, 0.58; 95% confidence interval, 0.42 to 0.81). Conclusions. The in-hospital mortality rates observed in this study are very similar to those reported from The Society of Thoracic Surgeons database. These data provide substantial evidence that results from The Society of Thoracic Surgeons database are representative of those achieved at other institutions. However, procedurespecific hospital volume must be considered in applying these results to individual institutions. (Ann Thorac Surg 2000;70: ) 2000 by The Society of Thoracic Surgeons Accepted for publication April 26, Address reprint requests to Mr Astor, Welch Center for Prevention, Epidemiology and Clinical Research, The Johns Hopkins University, 2024 East Monument St, Baltimore, MD 21205; bastor@jhsph.edu. Replacement of the native aortic valve with a mechanical or bioprosthetic valve has been performed for four decades and is the treatment of choice for most patients with symptomatic aortic valve disease in the United States. As the US population ages, the incidence of aortic valve disease, and thus the number of prosthetic aortic valve implantations, is expected to increase [1]. Concomitantly, changes are expected in the distribution of clinical and demographic characteristics of prosthetic aortic valve recipients, including increases in the mean age of patients and the proportion of patients undergoing concurrent coronary artery bypass graft (CABG) or replacement of a previously implanted prosthetic valve [2, 3]. These changes in patient characteristics may have a significant impact on the rates of perioperative mortality after aortic valve replacements observed by clinical centers in the United States. Published studies of mortality after aortic valve replacement have important limitations. Many studies report a single large center s experience over an extended period of time, often more than 10 years [4 9]. Although these reports are necessary to assess long-term valve survival and incidence of valve-related complications, the perioperative results reported may not be representative of current practice or be generalized to other centers. Also, many reports do not include results from operations to replace existing prosthetic valves [4, 10]. In contrast, The Society of Thoracic Surgeons (STS) National Cardiac Surgery Database [11] reports results from multiple institutions and includes data on repeat heart valve replacements and valve procedures combined with CABG. However, these data are also from a self-selected subset of the institutions performing these procedures, and how representative these participating hospitals and patient population are has been questioned [12]. The effects of hospital characteristics on mortality after aortic valve replacements that may bias the results of this database are not known. The present study characterizes the patient population and examines risk factors for in-hospital mortality in a nationally representative sample of both first-time and repeat aortic valve replacement operations in the United States. These results provide a benchmark useful for tracking future changes in demographic and clinical characteristics of patients and rates of in-hospital mortality. Hospital characteristics that may have an impact on in-hospital mortality after aortic valve replacement are examined, and the patient characteristics and inhospital mortality rates observed in this sample are compared to those reported from the STS database. These comparisons allow evaluation of whether results from the STS database can be generalized to other institutions in the United States by The Society of Thoracic Surgeons /00/$20.00 Published by Elsevier Science Inc PII S (00)

2 1940 ASTOR ET AL Ann Thorac Surg MORTALITY IN AORTIC VALVE REPLACEMENT 2000;70: Material and Methods Data Source Data were obtained from the 1994 Nationwide Inpatient Sample (NIS), a part of the Healthcare Cost and Utilization Project (HCUP-3) which was sponsored by the Agency for Healthcare Research and Quality (AHRQ) [13]. The NIS is a stratified probability sample of more than 900 nonfederal hospitals in 17 states participating in HCUP-3 and is designed to approximate a 20% national sample. Administrative data were collected from all discharges by each participating hospital, compiled by state data organizations, and edited and checked for invalid entries by the AHRQ. The sampling frame consisted of all nonfederal hospitals in the participating states as identified by the American Hospital Association Annual Survey of Hospitals. Sampling probabilities were proportional to the number of community hospitals in each stratum based on geographical region of the United States (northeast, north central, south, or west), location (urban or rural), teaching status, ownership/control (government controlled, private nonprofit, or private forprofit), and bed size (small, medium, or large, the definitions of which varied by location and teaching status). Due to the inconsistent definition and for computational efficiency, strata were collapsed on bed size. Weights based on the sampling probabilities for each stratum in the dataset were assigned to each sampled hospital to allow projection to the universe of all nonfederal hospitals in the United States [14]. All hospital discharges with a procedure code indicating implantation of a bioprosthetic (code 35.21) or mechanical (code 35.22) aortic valve, as defined in International Classification of Diseases, 9th revision, Clinical Modification (ICD-9) [15], were extracted from the NIS. Selected additional procedures, including replacements of other valve positions (codes to 35.28) and coronary artery bypass graft (codes to 36.19), occurring during the same hospital stay were also identified. Patients less than 18 years of age were excluded. We divided the patient population into quartiles (ie, low, medium-low, medium-high, and high volume) based on the unweighted volume of aortic valve replacements performed during 1994 at the hospitals from which the patients were discharged. Diagnoses were identified by ICD-9 diagnosis codes. The following diagnoses were specifically identified, with the ICD-9 codes in parentheses: diabetes mellitus (250), rheumatic valve disease (394 to 397), hypertensive disease (401 to 405), ischemic heart disease (410 to 414), cardiomyopathy (425), cardiac conduction disorders (426), cardiac dysrhythmias (427), heart failure (428), chronic renal failure (581 to 583 and 585 to 589), cerebrovascular disease (430 to 438), and failure of a previously implanted heart valve (996.02, , and ). Up to 15 procedures and 15 diagnoses were coded for each stay. Hospital admission status was coded as emergent, urgent, or elective. A dummy variable was used to code missing values of hospital admission status (n 164 records). The results were unchanged in analyses that excluded records with missing data. Data Analyses All analyses were performed by weighting the response variables by the sampling weights. Age- and sex-specific as well as total estimates of the number of prosthetic aortic valve implantations and in-hospital mortality rates were calculated, as were descriptive statistics of patient and hospital characteristics. Age- and sex-specific population projections for 1994 based on data from the 1990 US Census were used to estimate rates of implantation [1]. Bivariate tests of independence were performed by corrected 2 tests [16]. Multivariate logistic regression models were used to identify independent predictors of in-hospital mortality. Those comorbid conditions found to be associated with increased mortality in the bivariate analyses were included in the multivariate models. We constructed separate models that included all procedures, only procedures with concurrent CABG, only procedures without concurrent CABG, and only firsttime isolated aortic valve replacements without concurrent CABG. Similar models were constructed after stratification by hospital teaching status. The reported confidence intervals (CI) are based on weighted estimates of variance adjusted for the correlation of responses within each hospital (ie, clustering by hospital). All CIs reported are 95% confidence. Data extraction from the NIS database was performed using SAS (Cary, NC) [17], and analysis was performed using Stata (College Station, TX) [18]. Results Patient and Procedural Characteristics Among the 6,385,011 hospital discharges in the NIS database, 8,741 were identified as including replacement of an aortic valve in an adult ( 18 years) patient. These procedures were performed at 176 hospitals. Of these hospitals, 173 (98.3%) were in an urban (as opposed to a rural) location, 103 (58.7%) were teaching hospitals, 140 (79.7%) were private nonprofit, 24 (13.4%) were government-controlled, and 12 (7.0%) were private for-profit. Hospital procedure-specific volume during 1994 was categorized as low (60 or fewer procedures), medium-low (61 to 100 procedures), medium-high (101 to 180 procedures), or high (more than 180 procedures). Weighted projection to all nonfederal hospitals in the United States resulted in an estimated total of 46,397 (95% CI 40,005 to 52,789) aortic valve replacements in adult patients during All subsequent results are based on weighted estimates. The rate of aortic valve replacement increased significantly with age up to 75 to 79 years, peaking at 0.95 and 1.80/1,000 for women and men, respectively (Fig 1). The rate of replacement decreased significantly for ages 80 years and older. The mean age of aortic valve recipients was 67.3 years (Table 1). This varied significantly by the type of aortic valve received (72.5 years for those receiving a tissue valve versus 66.1 years for those receiv-

3 Ann Thorac Surg ASTOR ET AL 2000;70: MORTALITY IN AORTIC VALVE REPLACEMENT 1941 (6.3%) versus public for-profit (4.9%; p 0.11), and procedure-specific volume (7.2% at low volume hospitals, 6.2% at medium-low volume hospitals, 7.2% at mediumhigh volume hospitals, and 5.1% at high volume hospitals; p 0.09). Fig 1. Estimated rate of aortic valve replacement (per 1,000) from the 1994 Nationwide Inpatient Sample, by age and sex. ing a mechanical valve; p 0.001). An estimated 5.9% of patients underwent reoperations with replacement of a previously implanted prosthetic heart valve. Mechanical failure was noted in 63% of these reoperations, and infection was noted in 15%. Urban hospitals performed 98.5% of the aortic valve replacements, and one-half were in teaching hospitals. Private nonprofit hospitals accounted for 82.2% of the hospitalizations, whereas 11.6% were in government-controlled hospitals and 6.2% were in private for-profit hospitals. Diagnoses and comorbid conditions coded for these hospitalizations included ischemic heart disease (51.3%), cardiac dysrhythmia (46.0%), heart failure (37.8%), hypertensive disease (32.7%), rheumatic valve disease (21.6%), diabetes mellitus (12.3%), cardiac conduction disorders (10.3%), cerebrovascular disease (6.8%), cardiomyopathy (2.9%), and chronic renal failure (2.0%). In-Hospital Mortality There were an estimated 2,988 in-hospital deaths in aortic valve recipients, resulting in an overall in-hospital mortality rate of 6.4% (95% CI 5.8 to 7.1). In-hospital mortality was significantly higher for women than men (8.8 versus 5.6; p 0.001) and in patients undergoing CABG as compared to others (8.3% versus 5.0%; p 0.001) in unadjusted analyses. A higher proportion of men underwent concurrent CABG than did women (46.5% versus 37.2%; p 0.001). The rate of in-hospital mortality increased significantly with age for both sexes (p 0.001) (Fig 2), and was higher in patients receiving a tissue valve than in those receiving a mechanical valve (8.3% versus 6.0%; p 0.002). Diagnoses that were significantly associated (p 0.001) with higher inhospital mortality in unadjusted analyses were rheumatic valve disease (8.1% versus 6.0%), heart failure (9.1% versus 4.8%), chronic renal failure (16.8% versus 6.2%), and cerebrovascular disease (13.5% versus 5.9%). In-Hospital Mortality by Hospital Characteristics In-hospital mortality did not differ significantly by hospital characteristics in unadjusted analyses, including teaching (6.8%) versus nonteaching status (6.1%; p 0.32), rural (8.5%) versus urban location (6.4%; p 0.27), government-controlled (8.4%) versus public nonprofit In-Hospital Mortality by Procedure In-hospital mortality for procedures without concurrent CABG varied from 4.3% for isolated replacements of a native aortic valve to 15.2% for reoperations involving multiple valves (Table 2). The rates for procedures involving concurrent CABG were higher for every category, ranging from 7.6% (95% CI 6.7 to 8.6) for first-time Table 1. Demographic and Procedural Characteristics of Aortic Heart Valve Recipients in the 1994 Nationwide Inpatient Sample (Estimated From Weighted Responses) Characteristic No. (95% CI) % (95% CI) Age 65 yr 31, (27,372 36,542) ( ) Female 17, (15,176 20,221) ( ) Mechanical valve a 37, (32,232 42,052) ( ) Bioprosthetic valve a 9, (6,658 11,936) ( ) Number of valve positions replaced 1 42, (36,510 47,795) ( ) 2 4, (3,269 5,099) ( ) (26 93) ( ) Replacement of previously 2, implanted prosthetic heart valve b (2,229 3,277) (5.2 67) Mechanical failure 1, (1,358 2,126) ( ) Infection ( ) ( ) Other failure ( ) ( ) Concurrent CABG 19, (16,888 22,951) ( ) Hospital admission status c Emergent 8, (6,984 10,598) ( ) Urgent 13, (10,471 16,369) ( ) Elective 23, (19,291 27,328) ( ) n 46,397 (95% Confidence interval 40,005 to 52,789). a An estimated 42 patients (7 records) reportedly received both a mechanical and bioprosthetic aortic valve. b An estimated 115 patients (36 records) had diagnoses of more than one heart valve failure mode. c An estimated 876 patients (164 records) did not specify hospital admission status and are excluded from this analysis. CABG coronary artery bypass graft; CI confidence interval.

4 1942 ASTOR ET AL Ann Thorac Surg MORTALITY IN AORTIC VALVE REPLACEMENT 2000;70: Multivariate Associations With In-Hospital Mortality A multivariate logistic regression model including all procedure types found each procedural characteristic (ie, number of valves replaced, repeat versus first-time procedure, with versus without concurrent CABG) to be significantly associated with in-hospital mortality (Table 3). Higher age, female sex, implantation of a bioprosthetic versus a mechanical valve, and emergent or urgent (versus elective) hospital admission status were independently associated with higher odds of in-hospital mortality. Heart failure, chronic renal failure, and cerebrovascular disease remained significantly associated with mortality after adjustment (p 0.001), whereas rheumatic valve disease did not. The estimates of the odds ratios remained relatively unchanged after stratification by the presence or absence of concurrent CABG. In a model including only first-time isolated aortic valve replacements without concurrent CABG, the association of implantation of a bioprosthetic valve with in-hospital mortality was no longer significant (adjusted odds ratio, 1.17; 95% CI 0.76 to 1.81). The other estimates were relatively unchanged in this analysis. Fig 2. Estimated in-hospital mortality after aortic valve replacement in the 1994 Nationwide Inpatient Sample, by age and sex. aortic valve replacements to 25.3% (95% CI 8.2 to 42.3) for reoperations involving multiple valves. Multivariate Associations Between Hospital Characteristics and In-Hospital Mortality Higher hospital volume was independently associated with lower odds of in-hospital mortality (p for trend 0.002), with the highest volume hospitals, as compared to the lowest, being associated with an adjusted odds ratio of 0.58 (95% CI 0.42 to 0.81). The second and third higher quartiles of hospital volume were not significantly different than the lowest quartile. There was a trend toward higher odds of in-hospital mortality in teaching hospitals as compared to nonteaching hospitals (odds ratio, 1.26; p 0.06) in the adjusted analyses of all procedures. In a model limited to first-time isolated aortic valve replacements without concurrent CABG, the adjusted odds ratio for teaching versus nonteaching status increased to 1.45 (95% CI 1.02 to 2.06), and the adjusted odds ratio for high volume versus low volume hospitals decreased to 0.39 (95% CI 0.23 to 0.64). Higher hospital volume remained significantly associated with lower in-hospital mortality in models including only teaching hospitals (p 0.02) and in models including only nonteaching hospitals (p 0.01). Table 2. Estimated Number and In-Hospital Mortality by Procedure Among Aortic Heart Valve Recipients in the 1994 Nationwide Inpatient Sample (Estimated From Weighted Responses) Procedure Without Concurrent CABG In-Hospital Mortality (95% CI) No. (95% CI) With Concurrent CABG In-Hospital Mortality (95% CI) No. (95% CI) Isolated aortic valve replacement First-time 4.3% 21, % 18,213 ( ) (18,868 24,793) ( ) (15,490 20,935) Repeat 8.2% 1, % 621 ( ) (1,194 1,784) ( ) ( ) Overall 4.5% 23, % 18,834 ( ) (18,806 27,831) ( ) (14,972 22,696) Multiple valve replacement First-time 7.5% 2, % 955 ( ) (2,039 3,252) ( ) (686 1,224) Repeat 15.2% % 131 ( ) ( ) ( ) (66 196) Overall 8.8% 3, % 1,086 ( ) (2,334 3,982) ( ) (752 1,420) n 46,397 (95% Confidence interval 40,005 to 52,789). CABG coronary artery bypass graft; CI confidence interval.

5 Ann Thorac Surg ASTOR ET AL 2000;70: MORTALITY IN AORTIC VALVE REPLACEMENT 1943 Table 3. Adjusted Odds Ratio for In-Hospital Mortality Associated With Demographic, Clinical, and Procedural Characteristics of Aortic Heart Valve Recipients in the 1994 Nationwide Inpatient Sample (Estimated From Weighted Responses) Characteristic Adjusted Odds Ratio for In-Hospital Mortality 95% Confidence Interval p Value Age (per 10 yr) Female (versus male) sex Bioprosthetic (versus mechanical) valve a Multiple valve positions replaced Replacement of previously implanted prosthetic heart valve b Concurrent CABG Hospital admission status c (versus elective status) Emergent Urgent Comorbid conditions Rheumatic valve disease Heart failure Chronic renal failure Cerebrovascular disease Hospital characteristics Urban (versus rural) location Teaching (versus nonteaching) Hospital control (versus government control) Private nonprofit Private for-profit Hospital volume (versus low volume) Medium-low ( procedures) Medium-high ( procedures) High ( 180 procedures) n 46,397 (95% Confidence interval 40,005 to 52,789). a An estimated 42 patients (7 records) reportedly received both a mechanical and bioprosthetic aortic valve. b An estimated 115 paitents (36 records) had diagnoses of more than one heart valve failure mode. c An estimated 876 patients (164 records) did not specify hospital admission status and were assigned a dummy variable representing missing admission status. CABG coronary artery bypass graft. Comment These results provide the first nationally representative estimates of the number and characteristics of adult patients undergoing aortic valve replacement in the United States. The aging of the US population is expected to have a significant impact on the number and distribution of demographic, clinical, and procedural characteristics of aortic valve recipients in the future. Some of these changes have already been observed in large registries [3, 19] and specific institutions [20]. The distribution of patient characteristics observed in this study will be useful as a comparison for future data to track expected changes in these characteristics among prosthetic aortic valve recipients in the United States. This study also provides nationally representative estimates of in-hospital mortality associated with aortic valve replacement in the United States. Direct comparison of these results to previous reports is clouded by the difference in the definition of risk used in the present study (ie, in-hospital mortality) and the definition used elsewhere (ie, operative mortality, which includes both in-hospital deaths and postdischarge deaths occurring within 30 days of the procedure) [21]. This difference would tend to bias the comparisons toward slightly lower rates in the present study. Nonetheless, the mortality rates observed in the present study are comparable to those reported recently by the STS National Cardiac Surgery Database. For example, Jamieson and colleagues [19] report a 30-day mortality rate of 4.3% in patients undergoing an isolated aortic valve replacement from 1986 through 1995 in the STS database, based on 26,317 procedures. The present study estimates the in-hospital mortality rate at 4.5% (95% CI 3.8 to 5.4) for the same procedure. Combined CABG and isolated aortic heart valve replacement increased the 30-day mortality to 8.0% in the STS database (n 22,713), similar to the 8.0% (95% CI 7.0 to 9.0) in-hospital mortality rate estimated for the same procedure in the present study. For procedures involving multiple valves, the STS data report 9.6% operative mortality for procedures without concurrent CABG (n 3,840), compared to 8.8% (95% CI 6.6 to 11.5) in-hospital mortality in multiple valve procedures among

6 1944 ASTOR ET AL Ann Thorac Surg MORTALITY IN AORTIC VALVE REPLACEMENT 2000;70: aortic valve recipients in this report. The present study limits this category to those procedures involving the aortic position in addition to any other position, whereas the STS results do not exclude those procedures involving only other (ie, nonaortic) multiple valve positions. Procedures involving multiple nonaortic valve positions are likely associated with significantly higher mortality. Patient populations and hospital characteristics may be quite different between institutions that voluntarily participate in a professional organization s database and the remaining community hospitals performing specific procedures. In addition, as the STS operates the National Cardiac Surgery Database as a voluntary system, there is no systematic assurance that all eligible patients undergoing aortic valve replacement at participating centers are entered. Even with these potential differences, the results reported from the NIS and the STS database are very similar. This suggests that the institutions participating in the STS database are representative of all institutions performing aortic valve replacements in the United States, and that the patients entered into the database are representative of all patients undergoing these procedures. Thus, the results from the STS database can be applied to other institutions in the United States that perform these procedures but do not participate in the STS database. It should be noted, however, that the results referenced from the STS database include procedures from 1986 through 1995, whereas the present study used data from In the STS database, significant declines in perioperative mortality among patients undergoing CABG procedures were observed in this same time period [12], and declines may also be expected to have occurred in other procedures, including aortic valve replacement. However, with the rapid growth in the number of surgeons and institutions participating in the STS database during this same time, a 12-fold increase occurred in the number of patient records entered between 1990 and 1994 [12], weighting the results heavily toward the later years of this interval. The relevance of results from the STS database to nonparticipating institutions is further evidenced by comparison of the present study population with the 11,987 aortic valve recipients entered into the STS database in 1994 in terms of several demographic and procedural characteristics, including age (67.3 versus 67.6 years), female sex (38.1 versus 39.2), and concurrent CABG (42.9 versus 44.6) [22]. More detailed comparisons of clinical characteristics are encumbered by the present study s reliance on coded diagnoses to capture comorbidities, whereas the STS data collection forms specifically request this information. We found a significant association between higher procedure-specific hospital volume and lower in-hospital mortality in adjusted analyses. When limited to first-time isolated aortic valve replacements without concurrent CABG, the highest quartile of hospital volume, as compared to the lowest quartile, was associated with an adjusted odds ratio of 0.39 (95% CI 0.23 to 0.64). This association was found among both teaching and nonteaching hospitals. Although there was a significant trend of decreasing in-hospital mortality across increasing quartiles of hospital volume, the second and third quartiles of hospital volume did not differ significantly from the lowest quartile. Higher hospital volume has been associated with lower mortality in other surgical procedures [23 26]. This association may be due to the experience of the physicians and staff, differences in case-mix based on characteristics not included in our analyses, or referral of patients to centers with better results. Due to the lack of detailed clinical, procedural, and institutional data, the NIS database is not well suited for investigating possible causes for this relation. However, the present study demonstrates the need to explore the association between hospital volume of aortic valve replacements and lower mortality in datasets with more detailed information. Surgical and postoperative procedures practiced by hospitals with more experience may be instructional for other institutions. Policies regarding regional resources may also be significantly affected by results of such studies. The results of the present study also point out that hospital volume must be considered in applying results from multicenter databases to individual institutions. There was a trend toward higher in-hospital mortality in teaching hospital as compared to nonteaching hospitals in the present study. Because of assumed differences in case-mix between teaching and nonteaching hospitals, a more detailed risk stratification algorithm than was available in this database may explain part or all of this association. In the present study, receipt of a bioprosthesis, as compared to a mechanical valve, was associated with a higher rate of in-hospital mortality in unadjusted analyses. A weaker association was found after adjustment for other factors, and no significant association was found in a model limited to first-time isolated aortic valve procedures without concurrent CABG (adjusted odds ratio, 1.16; 95% CI 0.76 to 1.81). These results suggest that differences in patient characteristics, such as comorbid conditions that we did not include in our models, may explain the nonsignificant increase in odds of in-hospital mortality among recipients of bioprosthetic valves. Most previous studies have found no difference in acute mortality results between patients receiving bioprosthetic and mechanical valves after adjustment for clinical characteristics [9, 27]. Female sex was a significant predictor of in-hospital mortality (adjusted odds ratio, 1.33; 95% CI 1.12 to 1.59). This result is similar to that reported by the STS National Database for operative mortality, which found adjusted odds ratios associated with female sex of 1.25 in isolated aortic valve replacements and 1.61 in combined aortic valve replacement plus CABG [19]. Several other studies have found similar results [2, 3, 28, 29], although some smaller studies found female sex to be associated with reduced mortality [9]. Several studies have examined the higher periprocedural and late mortality in women after CABG and other cardiac procedures [30, 31], and numerous reasons for this disparity have been proposed (eg, referral bias, smaller vessel diameter) with no obvious conclusions. Unfortunately, the current study can only

7 Ann Thorac Surg ASTOR ET AL 2000;70: MORTALITY IN AORTIC VALVE REPLACEMENT 1945 provide additional evidence of the existence of this disparity following aortic valve replacement. The large size and representative nature of the data are obvious strengths of this study. However, the study is limited by the lack of detailed clinical information in the NIS. This precludes adjustment for measures of disease severity (eg, cardiac functional class) and some significant comorbid conditions. As all of the diagnoses are abstracted from discharge summaries, it cannot be determined whether a specific condition was present before the valve replacement or was a result of such operations. Thus, conditions that may be the result of cardiac operations cannot be analyzed as possible preoperative predictors of in-hospital mortality. We thank Josef Coresh, MD, PhD, for his insight and advice on the preparation of this manuscript. References 1. Day JC. Population projections of the United States by age, sex, race, and Hispanic origin: 1995 to Washington, DC: US Government Printing Office, US Bureau of the Census, Aranki SF, Rizzo RJ, Couper GS, et al. Aortic valve replacement in the elderly. Effect of gender and coronary artery disease on operative mortality. Circulation 1993;88:II Taylor KM, Gray SA, Livingstone S, Brannan JJ. The United Kingdom Heart Valve Registry. J Heart Valve Dis 1992;1: Bessell JR, Gower G, Craddock DR, Stubberfield J, Maddern GJ. Thirty years experience with heart valve surgery: isolated aortic valve replacement. Aust N Z J Surg 1996;66: Bortolotti U, Milano A, Mossuto E, Mazzaro E, Thiene G, Casarotto D. Early and late outcome after reoperation for prosthetic valve dysfunction: analysis of 549 patients during a 26-year period. J Heart Valve Dis 1994;3: Cohn LH, Allred EN, Cohn LA, et al. Early and late risk of mitral valve replacement: a 12 year concomitant comparison of the porcine bioprosthetic and prosthetic disc mitral valves. J Thorac Cardiovasc Surg 1985;90: Fiane AE, Geiran OR, Svennevig JL. Up to eight years follow-up of 997 patients receiving the CarboMedics prosthetic heart valve. Ann Thorac Surg 1998;66: Lund O, Chandrasekaran V, Grocott-Mason R, et al. Primary aortic valve replacement with allografts over twenty-five years: valve-related and procedure-related determinants of outcome. J Thorac Cardiovasc Surg 1999;117: Asimakopoulos G, Edwards MB, Taylor KM. Aortic valve replacement in patients 80 years of age and older: survival and cause of death based on 1100 cases: collective results from the UK Heart Valve Registry. Circulation 1997;96: Rabago G, Corbi P, Tedy G, et al. Five-year experience with the Medtronic Hall prosthesis in isolated aortic valve replacement. J Card Surg 1993;8: Clark RE. The Society of Thoracic Surgeons National Database status report (comment). Ann Thorac Surg 1994;57: Clark RE. The STS Cardiac Surgery National Database: an update. Ann Thorac Surg 1995;59: Healthcare Cost, and Utilization Project (HCUP-3). Nationwide Inpatient Sample, Release 3. Rockville, MD: Agency for Health Care Policy, and Research, Cochran WG. Sampling techniques. New York: John Wiley & Sons, Public Health Service, Health Care Financing Adminstration. International Classification of Diseases, 9th revision, Clinical Modification (ICD-9-CM). Washington, DC: US Department of Health and Human Services, Rao JNK, Scott AJ. The analysis of categorical data from complex sample surveys: chi-squared tests for goodness of fit and independence in two-way tables. J Am Stat Soc 1981; 76: SAS Institute. SAS user s guide: statistics. Cary, NC: SAS Institute, StataCorp. Stata statistical software. College Station, TX: Stata Corporation, Jamieson WR, Edwards FH, Schwartz M, Bero JW, Clark RE, Grover FL. Risk stratification for cardiac valve replacement. National Cardiac Surgery Database. Database Committee of The Society of Thoracic Surgeons. Ann Thorac Surg 1999;67: Rao V, Christakis GT, Weisel RD, et al. Changing pattern of valve surgery. Circulation 1996;94:II The Society of Thoracic Surgeons National Cardiac Surgery Database. Manual for data managers. Minneapolis: Summit Medical Systems, The Society of Thoracic Surgeons. Data analyses of the STS National Cardiac Surgery Database. Chicago: The Society of Thoracic Surgeons, Luft HS, Bunker JP, Enthoven AC. Should operations be regionalized? The empirical relation between surgical volume and mortality. N Engl J Med 1979;301: Showstack JA, Rosenfeld KE, Garnick DW, Luft HS, Schaffarzick RW, Fowles J. Association of volume with outcome of coronary artery bypass graft surgery. Scheduled vs nonscheduled operations [published erratum appears in JAMA 1987 May 8;257(18):2438]. JAMA 1987:257: Manheim LM, Sohn MW, Feinglass J, Ujiki M, Parker MA, Pearce WH. Hospital vascular surgery volume and procedure mortality rates in California, J Vasc Surg 1998:28: Begg CB, Cramer LD, Hoskins WJ, Brennan MF. Impact of hospital volume on operative mortality for major cancer surgery. JAMA 1998;280: Holper K, Wottke M, Lewe T, et al. Bioprosthetic and mechanical valves in the elderly: benefits and risks. Ann Thorac Surg 1995;60:S Gehlot A, Mullany CJ, Ilstrup D, et al. Aortic valve replacement in patients aged eighty years and older: early and long-term results. J Thorac Cardiovasc Surg 1996;111: Lytle BW, Cosgrove DM, Loop FD, et al. Replacement of aortic valve combined with myocardial revascularization: determinants of early and late risk for 500 patients, Circulation 1983:68: Edwards FH, Carey JS, Grover FL, Bero JW, Hartz RS. Impact of gender on coronary bypass operative mortality. Ann Thorac Surg 1998;66: O Connor GT, Morton JR, Diehl MJ, et al. Differences between men and women in hospital mortality associated with coronary artery bypass graft surgery. The Northern New England Cardiovascular Disease Study Group. Circulation 1993;88: