M uch has been made of the excessive mortality experienced by dialysis patients in the United States. Even when adjusted for the severity of associ-

Dialysis Survival in a Large Inner-City Facility: A Comparison to National Rates1 John C. Stivelman,2 J. Michael Soucie, Elizabeth S. Hall, and Edwin J. Macon J.C. Stivelman. ES. Hall. E.J. Macon, Department of Medicine, Renal Division, Emory University School of Medicine, Atlanta, GA. J.M. Soucie, Department of Family and Preventive Medicine. Emory University School of Medicine. Atlanta. GA (J. Am. Soc. Nephrol. 1995; 6:125#{243}-1261) ABSTRACT Impoverished patients may represent a high-risk population with poor survival. With 1993 U.S. Renal Data System survival tables (to adjust the risk of death for differences in age, race, and ESRD diagnosis), the mortality rates of patients over 3 yr in a large inner-city dialysis facility using high-flux technique were compared with national averages. At least 93.7% of patients were African-American, 50% had incomes below $7,000 per year, and employment was 5% or less. Observed and expected deaths (the latter derived from the U.S. Renal Data System tables) were used to calculate a standardized mortality ratio (observed deaths/expected deaths); the U.S. average is 1.0. The standardized mortality ratio at this facility for each year was <0.600 and was significantly lower than the U.S. average in 1991, In 1992 (P < 0.05), and for all 3 yr (P <.001). Over all 3 yr. It was lower for females (0.540, P < 0.05), males (0.620, P < 0.05), patients with diabetes (0.593, P < 0.05), and glomerulonephritis (0.318, P < 0.05). For the 3 yr. a Cox regression analysis revealed independent associations between mortality and age (P = 0.004), serum albumin (P = 0.02), Kt/V (P = 0.02), and dialysis for more than 2 yr (P = 0.01). Patients with economic hardship can attain survival significantly better than the national average with the provision of adequate dialysis, nutrition, and support services. Key Words: Hemodialysis, economic hardship, dialysis adequacy, nutritional status. ESRD M uch has been made of the excessive mortality experienced by dialysis patients in the United States. Even when adjusted for the severity of associ- Received september 26, 1994. Accepted Aprii 12, 1995. 2Correspondence to Dr. J.C. Stivelman, Renal Division. Emory University School of Medicine, 69 Butler Street, SE., Atlanta, GA 30303. 1046.6673/0604-1256$03.O0/O Journal of the American Society of Nephroiogy copyright 1995 by the American Society of Nephroiogy ated diseases, the relative risk of death for Americans treated with hemodialysis for 5 yr approaches 1.5 times that of other industrialized countries (1,2). In recent years, attempts to explain this striking disparity have centered on the increasing age and heterogeneity of patients in the U.S. dialysis population, and the delivery of inadequate dialytic therapy compared with those of countries with better survival (3). This troubling issue is magnified among Americans living in inner cities, for whom poverty, homelessness, and violence aggravate the difficulties associated with living with ESRD. Although it has been suggested that these patients might comprise a high-risk population with survival below the national average, surprisingly few investigators have addressed this issue (4,5). Standardized statistics yielding risks of death of dialysis patients in United States are found in the U.S. Renal Data System (USRDS) Survival Tables (6,7). These tables not only provide a simple mechanism for determining the risk of death of patients throughout the United States, but they also establish the means by which survival in a population can be normalized to account for differences in age, race, and ESRD diagnosis. Moreover, the USRDS data can be used to compare mortality rates in specific patient subgroups either within a single facility, between facilities, on with all U.S. dialysis patients. These tables adjust the risk of death for differences in age, race, and ESRD diagnosis, thereby permitting the attribution of differences in death rate to other factors. Stimulated by the controversy surrounding the excessive mortality rate of dialysis patients in the United States and the unusual characteristics of patients treated in the large inner-city dialysis unit staffed by this Renal Division, we used the USRDS tables to examine their overall, sex-, age- and ESRD diagnosisspecific survival rates. We also evaluated parameters of dialytic adequacy and nutritional status to determine their relative contribution to the risk of death. Additionally, specific subgroups were studied to identi1 any increased risk of death and to determine the relative contribution of Kt/V and nutritional variables to that risk. METHODS All patients in this analysis received ESRD care under the supervision of Emory University Renal Division faculty in a 38-station inner-city facifity using all high-efficiency/highflux technique. Patients were assessed during dialysis treatments by physicians twice weekly and by a physician s assistant once weekly. All patients received physical assessments quarterly. Patients requiring hospitalization were admitted to Emory University Hospital or Grady Memorial Hospital, where their care was supervised by faculty of the Emory Renal Division. 1256 Volume 6 Number 4. 1995

Stivelman et al Patients were dialyzed three times weekly on Fresenius Model D machines with either F80 or F60 dialyzers (a few non-reuse patients used Gambro G6N dialyzers.). The reprocessing of polysulfone dialyzers was accomplished with Seratronics DRS-4 automated equipment: dialyzers were rinsed with bleach, packed in 1.5% formaldehyde, and incubated overnight at 40#{176}C.The dialysis prescription was adjusted to deliver a minimum Kt/V of 1.2 and was not changed during the study. The dialysate flow rate was 800 ml/min for F60 and F80 dialyzers. Urea kinetic modeling was performed monthly by use of the two-point method (third point extrapolated) arid PACKH (Fresenius) software (8). The second BUN was sampled 3 to 5 mm postdialysis to correct for cardiopulmonary and dialyzer recirculation. Residual renal function was assessed yearly and used in determining the dialysis prescription. Dietary counseling was provided monthly to all patients, with a goal of at least 1.0 g of protein/kg per day. Inclusion Criteria Patients eligible for this analysis satisfied USRDS Mortality Table eligibilityrequirements, which include: (1) enrollment at the facility by January 1, 1990, 1991, or 1992; (2) on dialysis 90 days after beginning renal replacement therapy; and (3) no prior renal transplant (6,7). Transient patients were not included. Patients birth dates, dates of initiation of hemodialysis (to determine length of treatment), and ESRD diagnoses were obtained from hospital discharge summaries, medical records, patient dialysis records, or contact with the patients physicians. Age was calculated as of December 31 of the year before the first year of USRDS eligibility and for the purposes of risk analysis was categorized as 45, 46 to 60, 61 to 75; or >75 yr. Time on dialysis before eligibility was categorized as more than 2 yr or 2 yr or less. Income data were obtained by social workers at the facifity. Use of USRDS Survival Tables Primary ESRD diagnoses were those specified by 1993 USRDS Survival Tables: hypertension, diabetes, glomerulonephritis, and other. Patients were designated as hypertensive Ifthey were hypertensive and a diagnosis of hypertensive nephropathy was noted in the chart, if a biopsy revealed nephrosclerosis, if extrarenal end-organ hypertensive disease was demonstrable, and if there was no other renalrelated diagnosis. Patients were classified as diabetic if there were no other renal-related diagnoses and the patients were diabetic. Non-nephrotic diabetic patients with few or no complications of diabetes but severe hypertension requiring treatment with several drugs were designated as hypertensives. If diabetes significantly antedated hypertension, the patient was designated as diabetic. Patients were classified as having glomerulonephritis if there was a diagnostic biopsy, if the patient had AIDS-associated nephropathy, or a history of long-standing intravenous drug abuse and the nephrotic syndrome. Patients were monitored to December 31 of each calendar year. unless death, transplantation, or transfer to another facility occurred. The expected death rate for each patient based on his/her age, race, and ESRD diagnosis was obtained from the 1993 USRDS Tables (7). This rate was multiplied by each patient s follow-up time to arrive at an expected probability of death. The sum of all probabilities in each year was the expected number of deaths for the facility (ED). This ED for the facility was then compared with the actual number of deaths (OD). The observed deaths (OD) have been expressed as a fraction of the expected deaths (OD/ED) to provide a standardized mortality ratio (SMR) in order to compare the yearly mortality data with the USRDS national average. For these analyses, we calculated both yearly and 3-yr cumulative mortality for each diagnosis, gender, and the group as a whole. Dialysis Adequacy and Nutritional Status Urea kinetic modeling was used to determine delivered Kt/V and protein catabolic rate normalized for body weight (npcr). These values and monthly serum albumin measurements were averaged for the year for each patient. A weighted average of these yearly values was used in the 3-yr multivanate analysis. Statistical Analysis Values are presented as mean ± standard deviation. Expected and observed deaths were compared by the use of a f test. The survivor function was estimated for all USRDSeligible patients by use of the actuarial life table method (9). Differences in survival between subgroups of the study population were assessed by use of the nonparametnic linear rank test of Tarone and Ware (10). The proportional hazards model developed by Cox (11) was used to measure the simultaneous effects of multiple factors on the risk of death. RESULTS Patients ProfIles of the patients for each calendar year are presented in Table 1. At the beginning of 1991, 53% of patients had annual incomes below $7,000; in 1992, 54% of patients had annual incomes below $7,000. The percentage of patients employed was between 3 and 5% In each year. Overall Mortality and Diagnosis-Related Mortality The ratio of observed to expected deaths (SMR) was calculated for all patients in each calendar year (Table 2). Expected deaths, adjusted for age, race, and ESRD diagnosis, were determined from the 1993 USRDS survival tables (7). The SMR for 1991 and 1992 individually, and for the entire 3-yr period as well, were significantly different from the national average. In Table 3, similar calculations for all 3 yrs were performed for each of the four USRDS diagnostic groups. Significant differences in SMR were seen among patients with diabetes (P < 0.05) and glomerulonephnitis TABLE 1. USRDS-eligible patients by year#{176} Year N M/F Age (yr) Race (t M F Black White 1990 95 59/36 46.6 ± 15.0 58.0 ± 13.5 93 2 1991 142 87/55 49.3 ± 15.2 58.4 ± 13.6 133 9 1992 162 97/65 49.1 ± 13.9 58.9 ± 13.3 157 5 #{176}Vaiues ore mean ± SD. The number of patients eligible(f, the ratio of males to females CM/F), age (inyears),and race are presented. Journal of the American Society of Nephrology 1257

Dialysis Mortality in an Inner-City Facility TABLE 2. Mortality by year#{176} Year. N OD ED SMR P 1990 95 8 13.812 0.579 NS 1991 142 13 22.713 0.572 <0.05 1992 162 15 25.374 0.591 <0.05 Cumulative 36 61.898 0.582 <0.001 #{176}Yearlyand cumuiative mortally. GD and ED caicuiated from 1993 USPDS Survival Tables (adjusting for age, race, and ESRD diagnosis) are presented, as is the ratio of observed to expected deaths (SMR). NS, not significant. TABLE 3. Cumulative mortality by USRDS diagnostic group#{176} Diagnosis N OD ED SMR P DM 67 15 25.316 0.593 <0.05 GN 34 3 9.446 0.318 <0.05 HTN 93 16 23.296 0.687 NS 0TH 15 2 3.841 0.521 NS O Diagnosis-related mortality. OD. ED calculated from the 1993 IJSRDS Survival Tables adjusting for age, race, and ESRD diagnosis, and SMR for all 3 yr are presented for each of the LJSRDS-designated four diagnostic groups. DM. diabetic nephropathy; GN, glomerulonephritis; HIN, hypertensive renal disease; 0TH, an other diagnoses; NS. not significant. (P < 0.05), but not for the diagnoses of hypertension or other. Because USRDS eligibility criteria exclude some patients from analysis, crude mortality for all hemodialysis patients at this facility was also calculated. Crude mortality was 9.1% in 1990, 10.7% in 1991, and 8.8% in 1992. These values are comparable to the mortality we observed for USRDS-eligible patients in each year (8.4, 9.2, and 9.3%, respectively). SMR were calculated separately for men and women, as displayed in Table 4. Over the 3 yr, mortality was significantly lower than the national average for both genders. Urea Kinetic Data and Serum Albumin Each patient s data for Kt/V, npcr, and serum albumin concentrations, from all available years, were combined to provide a time-weighted mean value for his or her entire period of study. Overall, the mean Kt/V was 1.24 ± 0.18, the mean npcr was 0.98 ± 0.17 g of protein/kg per day, and the mean serum albumin concentration was 3.73 ± 0.34 g/dl. Survival Analysis We initially derived the crude mortality rates by year and found no differences from year to year (results not TABLE 4. Gender and cumulative mortality Gender N OD ED SMR P Men 130 20 32.263 0.620 <0.05 Women 79 16 29.635 0.540 <0.05 shown). Data from all 3 yr were, therefore, combined and analyzed as a single period. This analysis included 209 patients monitored for between 4 and 1,096 days. Total follow-up was 130,104 patient-days (356.2 patient-years). The length of dialysis treatment before the patient became eligible for this analysis ranged from 91 to 8,157 days, with an average of 730 days. A stepwise Cox regression analysis Identified the following factors to be independently associated with mortality: serum albumin, age, Kt/V, and time on dialysis before entering the study (Table 5). Race and gender were not significantly associated with mortality in this model and are not displayed. A separate Cox analysis by diabetic status showed that the risk of death for diabetic patients at lower Kt/V (<1.0) was much greater than for nondiabetic patients when compared with a standard Kt/V of 1.0 to 1.2. AssociatIons between advanced age and albumin <3.5 g/dl were statistically significant in diabetics only. In Figure 1, survival among patients in this study is compared with the estimated survival of all African- Americans on hemodialysis on the basis of the yearly risk of death from data in the USRDS 1993 Annual Report (7). Survival curves for the four ESRD diagnostic groups are shown in Figure 2. No significant differences in survival between groups were seen. In all diagnostic groups, survival was better than that reported by the USRDS. DISCUSSION Using 1993 USRDS Survival tables, we adjusted mortality rates at a large inner-city dialysis facility for age, race, and diagnosis and found a mortality rate significantly below the national average despite in- TABLE 5. Stratified Cox regression analysis results#{176} Patients Covariate Relative Risk P N % Age (yr) 45 72 34 0.15(0.04-0.55) 0.004 46-60 75 36 0.61 (0.26-1.41) 0.24 61-75 52 25 Reference 75+ 10 5 1.63 (0.49-5.47) 0.4,3 Albumin >3.5 gil 150 72 Reference 3.5g/L 56 27 2.43(1.12-5.27) 0.02 KtIV <1.0 17 8 5.05(1.30-19.6) 0.02 1.0-<1.2 60 29 Reference 1.2-<1.4 84 40 0.77(0.29-2.07) 0.60 1.4+ 38 18 0.81 (0.26-2.60) 0.73 Time on Dialysis z2 yr 165 79 Reference >2 yr 44 21 2.60 (1.22-5.58) 0.01 O 95% confidence interval in parentheses. Note that all percents do not total to 100 because of missing data. 1258 Volume 6 Number 4 1995

Stivelman et al 1 0.8 0.6 0.4 0.2 Cumulative Proportion Surviving 0300 600 900 1200 -Present Series Days - USRDS Figure 1. Survival curve for all eligible patients on the basis of OD (solid line) and the estimated survival curve for all African-Americans on hemodialysis for more than 60 days and never transplanted, from USRDS (hatched line). The curve for all African-Americans assumes a constant rate of death over the 3-yr period on the basis of a rate of 198.9 deaths/i,000 patient-years (7). 0.8 0.6 0.4 0.2 1 Cumulative Proportion Surviving 0 300 600 900 1200 Days -HTN DM --GN 0TH Figure 2. Survival curves for OD In each USRDS diagnostic category. The USRDS i-yr survival for all African-Americans (on hemodialysis for more than 60 days, never transplanted) in each diagnostic category is: hypertensives (HTN). 81%; diabetics (DM), 75.2%; giomerulonephritics (GN), 85.1%; and other (OTN), 80.9% (7). comes in over half the patients of less than $7,000! year and unemployment rates of 95%. Mortality was significantly lower for both men and women, as well as for patients with diabetes and glomerulonephritis, but not for patients with hypertension or other causes of ESRD. Although hypertensive and diabetic African- Americans have been shown to have better survival on hemodialysis than other ethnic groups [12-181. particularly whites, we believe our results cannot be attributed to race or age-related factors because both are subsumed into the analysis when the USRDS tables are used. Few investigators have examined the effect of socioeconomic status on the dialytic mortality of minorities in inner cities. Port et al. (4) examined the incomes of dialysis patients in Michigan and found a direct relationship between survival and income among African- American but not among whites. The adjusted relative risk of death among African-American patients decreased by 3.3% per $1,000 increase in income. In unpublished studies, similar relationships between income and death risk have been suggested for the United States as a whole (L. Agodoa, MD, personal communication). In an additional study analyzing data from the Michigan Kidney Registry, Ferguson et at. found that mortality rate was lower among patients with higher incomes (as estimated by zip code of residence) (5). We found several factors that were independently associated with mortality in this prospectively monitored cohort (Table 5). PatIents whose albumin averaged less than 3.5 gidl. whose delivered dialysis dose was less than 1.0, or who had been receiving dialysis for longer than 2 yr were at increased risk of death. Serum albumin levels have been found to be strong predictors of mortality in dialysis patients (19-21). Collins et at. (21), in a retrospective analysis of 1,773 patients accumulated over a 14-yr period from a regional registry in Minnesota, found that serum albumin concentrations below 3.5 g/dl were significantly associated with increased mortality. Our results confirm the importance of this factor. The fact that 73% of the patients we treated had serum albumiii concentrations above 3.5 g/dl may have contributed to the low mortality we observed. Dialyzer reuse with bleach was recently shown to contribute to albumin depletion (22), but its role cannot be ascertained because we did not measure albumin losses through the dialyzer. Inadequate delivery of dialysis therapy has become increasingly recognized as a major source of increased morbidity and mortality (20,23-25). Hakim et at. (26) analyzed the mortality rate in the Vanderbilt University dialysis facility over a 4-yr period, during which the average, delivered Kt!V was systematically increased from 0.82 to 1.18 (double-pool urea kinetics). They found that the crude mortality rate declined from 22.8 to 9.1% and the SMR fell from 1.03 to 0.61. In another recently published study of dialysis adequacy and mortality, Parker and colleagues (27) collected mortality data on patients treated by Dallas Nephrology Associates facifities as Kt!V was increased from 1.18 in 1989 to 1.46 in 1992 (single-pool urea kinetics). Although the crude mortality rate declined, the decrease observed (22.5 to 18.1%) was not as marked as that observed by Hakim et at., even though the delivered dialysis dose in Dallas was higher. In this study, the delivered Kt!V averaged 1.24 for the 3 yr of study and has been the treatment goal since the time of the facility s opening. The average annual mortality rate experienced by this cohort was similar to that reported by Hakim et at. during the year 1991, when Kt!V averaged 1.18; It was substantially lower than that reported by Parker et at. in their fourth year of Journal of the American Society of Nephrology 1259

Dialysis Mortality in an Inner-City Facility therapy, despite the fact that our Kt/V average was less than either of these when calculated by the use of comparable kinetic formulae. This outcome is more striking In view of the economic disadvantage of this population and an absentee rate of about 5% of all treatments. One potential, albeit speculative, reason for both this discrepancy and the low mortality rate we report may be the near-exclusive use of polysulfone high-flux membranes at this institution for the entire period of study. Hakim et at. increased Kt!V in their patients by conversion from conventional to high-flux treatment, which occurred gradually over a 4-yr period. By the end of their study, almost all of their patients received this treatment modality. In contrast, some centers in the report by Parker et at. had not switched to a high-flux technique even by the completion of their study. The salutary effects of high-flux hemodialysis (28) and blocompatible membranes (29) may be additional independent factors in improving survival, but their specific roles cannot be ascertained from our results. Another important-although poorly quantifiablefactor that could account for survival at this facility is the frequency and quality of the patient-doctor and patient-dialysis staff encounter. Patients were seen at nearly every treatment by a physician or physician s assistant, visited monthly by dietitians, received physical assessments quarterly, and had access to social workers who provided counseling and information regarding insurance, housing, financial support, and food stamps. Although these factors may have enhanced the quality of medical care, their effect on survival, particularly that of the frequency of physician and staff visits during dialysis, Is speculative. It is doubtful that transplantation influenced the survival statistics at this facility, because over the 3-yr period, the number of USRDS-eliglble patients in the 20- to 44-yr age group transplanted (6.5 ± 2.3% per year) was comparable to that reported in the 1993 and 1994 USRDS database (7,30). Further, because our analysis adjusts for age, we do not believe that the persistence of young and healthy patients in this population had a favorable effect on overall survival. The transfer of unstable patients to hospital-based facilities also did not Influence our mortality rates; during the study period, only one USRDS-eligible patient had to be dialyzed at a hospital because of an unstable medical Illness. This patient survived for over 1 yr after transfer. The calculations of crude and adjusted mortality Included all deaths of patients at home or in hospital. Approximately 6.5% of USRDS-eligible patients per year transferred to other facilities. Deaths among these patients occurring at other facilities were not included in our analysis because we did not have complete Information after transfer (6,31). Although no single factor can account for the improved survival of these patients, we find little support for the suggestion that socioeconomic disadvantage per se contributes to increased morbidity in ESRD. On the contrary, economic hardship need not pose an obstacle to good or even excellent survival. The factors apart from adequate dialysis and adequate nutrition responsible for the enhanced survival that we observed will require further elucidation. REFERENCES 1. Charra B, Calemard E, Chazot C, et al: Dose of dialysis: what index? Blood Purif 1992:10:13-21. 2. Held P, Brunner F, Odaka M, Garcia J, Port F, Gaylln D: Five-year survival for end-stage renal disease patients in the United States. Europe. and Japan. 1982-1987. Am J Kidney Dis 1990:15:451-457. 3. Hull A, Parker T: Proceedings from the morbidity. mortality, and prescription of dialysis symposium. Am J Kidney Dis 1990:15:375-383. 4. Port FK, Wolfe RA, Levin NW, Guire KE, Ferguson CW: Income and survival in chronic dialysis patients. Trans ASAIO 1990;36:Ml54-157. 5. Ferguson CW, Levin NW, Port FW, Wolfe RA, Lamphlear D, Hawthorne V: The role of Income on treatment modality and survival in ESRD patients: A multifactorial analysis. Kidney mt 1987;31:231A. 6. Wolfe R, Gaylin D, Port F, Held P. Wood C: Using USRDS generated mortality tables to compare local ESRD mortality rates to national rates. Kidney Int 1992:42:99 1-996. 7. U.S. Renal Data System. USRDS 1993 Annual Data Report. Bethesda, MD: The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases; 1993. 8. PACKH Urea Kinetic Analysis for the PC. Draft Operator s Manual, Version 6.01, Fresenius, Inc. 9. Cutler S,J, Ederer F: Maximum utilization of the life-table method In analyzing survival. J Chronic Dis 1958;8:699-713. 10. Tarone RE, Ware J: On distribution-free tests for equality of survival distributions. Biometrika 1977:1:156-160. 11. Cox DR: Regression models and life tables. J Royal Statistic Soc 1 972;34)Series B): 187-220. 12. Held PJ, Pauly M. Diamond L: Survival analysis of patients undergoing dialysis. JAMA 1987;257:645-650. 13. Tierney WM, Harris LE: Reply letter. Arch Intern Med 1992:152:879-880. 14. U.S. Renal Data System. USRDS 1990 Annual Report. Bethesda, MD: The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases; 1990. 15. Wolfe RA, Port FK, Hawthorn VM, Guire KE: A comparison of survival among dialytic therapies of choice: incenter hemodlalysis vs continuous ambulatory peritoneal dialysis at home. Am Kidney Dis 1990:15:433-40. 16. Lowrie EG, Lew NL, Huang WH: Race and diabetes as death risk predictors in hemodlalysis patients. Kidney mt 1992;38:S22-S31. 17. Bleyer AJ: Race and dialysis survival. Arch Intern Med 1992:152:879. 18. Kasiske BL, Neylan JF. Riggio RR. et at: The effect of race on access and outcome in transplantation. N Engl J Med 1991:324:302-307. 19. Owen WF, Lew ML, Liu Y. Lowrie EG, Lazarus JM: The urea reduction ratio and serum albumin concentration as predictors of mortality in patients undergoing hemodialysis. N Engl J Med 1993:329:1001-1006. 20. Lowrie EG, Lew NL: Death risk in hemodialysis patients: the predictive value of commonly measured variables and an evaluation of death rate differences between facilities. Am J Kidney Dis 1990;15:458-482. 21. Coffins AJ, Ma JZ, Umen A, Keshaviah P: Urea index and other predictors of hemodialysis patient survival. Am J Kidney Dis 1994:23:272-282. 22. Graeber CW, Halley SE. Lapkin RA. Graeber CA, Kaplan 1260 Volume 6 Number 4 1995

Stivelman et a) AA: Protein losses with reused dialyzers. J Am Soc Nephrol 1993:4:349A. 23. Hakim RM, Depner TA, Parker TF: Adequacy of hemodialysis. Am J Kidney Dis 1993:20:107-123. 24. Held PJ, Blagg CR, Liska DW, Port FK, Hakim R, Levin N: The dose of hemodialysis according to dialysis prescription in Europe and the United States. Kidney mt 1992 (suppl 38):42:S16-S21. 25. Held PJ, Levin NW, Bovbjerg RR, Pau MV, Diamond LH: Mortality and duration of hemodialysis treatment. JAMA 199 1:265:871-875. 26. Hakim RM, Breyer J, Ismail N, Schulman G: Effect of dose of dialysis on morbidity and mortality. Am J Kidney Dis 1994:23:661-669. 27. Parker TF, Husni L. Huang W. Lew N, Lowrie EG, Dallas Nephrology Associates: Survival of hemodialysis patients in the United States is Improved with a greater quantity of dialysis. Am J Kidney Dis 1994:23:670-680. 28. Hornberger JC, Chernew M, Peterson J, Garber A: A multivariate analysis of mortality and hospital admissions with high-flux dialysis. J Am Soc Nephrol 1992:3: 1227-1237. 29. Hakim RM: Clinical implications of hemodialysis membrane biocompatibility. Kidney Int 1993:44:484-494. 30. U.S. Renal Data System. USRDS 1994 Annual Data Report. Bethesda, MD: The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases; 1994. 31. Wolfe RA: The standardized mortality ratio revisited: Improvements, innovations, and limitations. Am J Kidney Dis 1994:24:290-297. Journal of the American Society of Nephrology 1261