Skip Navigation


NDT Advance Access originally published online on February 13, 2008
Nephrology Dialysis Transplantation 2008 23(5):1690-1696; doi:10.1093/ndt/gfm728
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
23/5/1690    most recent
gfm728v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Mauri, J. M
Right arrow Articles by Registry, C. R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Mauri, J. M
Right arrow Articles by Registry, C. R.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© The Author [2008]. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org


Design and validation of a model to predict early mortality in haemodialysis patients

Joan M Mauri1, Montse Clèries2, Emili Vela2 and Catalan Renal Registry2

1 Nephrology Department, Hospital Universitari de Girona Dr Josep Trueta, Girona, Spain 2 Catalan Health Service, RMRC, Barcelona, Spain

Correspondence and offprint requests to: Dr Montse Clèries, Catalan Health Service, Travessera de Les Corts, 131-159, 08021-Barcelona, Spain. Email: mcleries{at}catsalut.net



   Abstract
Top
 Abstract
Introduction
 Patients and methods
 Statistical analysis
 Results
 Discussion
 References
 
Background. Mortality and morbidity rates are higher in patients receiving haemodialysis therapy than in the general population. Detection of risk factors related to early death in these patients could be of aid for clinical and administrative decision making.

Objectives. The aims of this study were (1) to identify risk factors (comorbidity and variables specific to haemodialysis) associated with death in the first year following the start of haemodialysis and (2) to design and validate a prognostic model to quantify the probability of death for each patient.

Methods. An analysis was carried out on all patients starting haemodialysis treatment in Catalonia during the period 1997–2003 (n = 5738). The data source was the Renal Registry of Catalonia, a mandatory population registry. Patients were randomly divided into two samples: 60% (n = 3455) of the total were used to develop the prognostic model and the remaining 40% (n = 2283) to validate the model. Logistic regression analysis was used to construct the model.

Results. One-year mortality in the total study population was 16.5%. The predictive model included the following variables: age, sex, primary renal disease, grade of functional autonomy, chronic obstructive pulmonary disease, malignant processes, chronic liver disease, cardiovascular disease, initial vascular access and malnutrition. The analyses showed adequate calibration for both the sample to develop the model and the validation sample (Hosmer-Lemeshow statistic 0.97 and P = 0.49, respectively) as well as adequate discrimination (ROC curve 0.78 in both cases).

Conclusions. Risk factors implicated in mortality at one year following the start of haemodialysis have been determined and a prognostic model designed. The validated, easy-to-apply model quantifies individual patient risk attributable to various factors, some of them amenable to correction by directed interventions.

Keywords: early mortality; epidemiology; haemodialysis; predictive model; registry



   Introduction
Top
 Abstract
 Introduction
Patients and methods
 Statistical analysis
 Results
 Discussion
 References
 
Overall mortality in haemodialysis (HD) patients is 6.8 times greater than in the general population. This higher mortality is associated with, and may in part be explained by, higher morbidity, which is either pre-existent [1] or produced during the period on renal replacement therapy (RRT) [2]. It would be highly desirable to have a tool that could predict early death in these patients as an aid in decision-making related to RRT, to provide objective information to patients and their families, and to perform health planning analyses and comparative studies. In addition, the early identification of risk factors would allow the implementation of corrective measures in factors that are amenable to some type of intervention.

The aims of this study were to identify risk factors associated with death in the first year following the start of HD, and to design and validate an easily applied prognostic model to determine the specific risk status of each patient.



   Patients and methods
Top
 Abstract
 Introduction
 Patients and methods
Statistical analysis
 Results
 Discussion
 References
 
The data source used for this study is the Catalan Renal Registry (RMRC, Registre de Malalts Renals de Catalunya), created in 1984 by the Health Department of the Government of Catalonia. Notification to the RMRC is mandatory for all centres, both public and private, that offer this treatment in Catalonia, an autonomous region of Spain. Information for the Registry is collected for each patient at the initiation of RRT and yearly thereafter. Data on mortality (date and cause) were obtained from the online notification system used by the centres and from the annual Registry update. In cases lost to follow-up, other sources of population data were consulted. Since 1984, information on morbidity has been recorded for the Registry [3] at the start of RRT, and since 1990 the data have been updated at each yearly follow-up.

This study includes all patients older than 17 years and residing in Catalonia who started haemodialysis treatment during the period 1997–2003 (n = 5738) and had a minimum follow-up of 1 year.

The risk factors studied are easily obtained by the attending physician: age, sex, type of primary renal disease, presentation form of ESRD (acute is attributed to a lack of predialysis nephrologic care, whereas subacute or normal progression implies predialysis nephrologic care during at least 6 months), functional autonomy degree (FAD), cardiovascular disease, peripheral vascular disease, malignant processes (diagnosis of any malignant disease before starting HD, whether active or not), chronic liver disease, chronic obstructive pulmonary disease (COPD), arthropathy, intestinal disease, esophageal, stomach and duodenal diseases, malnutrition (BMI < 20), and first vascular access.

Age was taken as a continuous variable and, following analysis of the linear pattern, was divided into 10-year units for the logistic regression analysis. Primary renal disease was divided into three groups: diabetes (EDTA codes 80 and 81), systemic (EDTA codes 73, 74 and 82–89) and standard disease (the remaining EDTA codes) (Appendix A). Functional autonomy degree was assessed with Gutman's modification of the Karnofsky index for renal patients [4]. Morbid conditions were grouped according to the International Classification of Diseases (ICD-9) coding (Appendix B). The type of vascular access used at the start of dialysis was divided into two categories: catheter, or arteriovenous fistula (AV fistula) and graft.

The total population was randomly divided into two samples, one with 60% of the patients (n = 3455), which was used to develop the model, and the other with the remaining 40% (n = 2283), which was used to validate the model.



   Statistical analysis
Top
 Abstract
 Introduction
 Patients and methods
 Statistical analysis
Results
 Discussion
 References
 
The prognostic model was constructed using a logistic regression analysis, in which the dependent variable was mortality at 1 year following the start of haemodialysis. The variables were entered in the model one by one and retained when their significance was <0.10, with the exception of the variable sex, which was forced into the model for demographic purposes even though it was not significant. The main effect variables were sex, age, primary renal disease, functional autonomy degree, cardiovascular disease, COPD, malignant processes, chronic liver disease and malnutrition.

Possible interactions between the variables in the model were tested two-by-two, retaining those that were significant, that had sufficient events in their categories (>1%), and that were clinically plausible. An examination of all possible two-way interactions among these main effects determined that three of them satisfied all our criteria for inclusion in the model.

Calibration of the models was assessed with the Hosmer–Lemeshow goodness-of-fit test, which determines the accuracy of the statistical probabilities generated by the model studied; i.e. it compares the predicted number of patient deaths with the actual number observed for each decile of risk [5]. Discrimination was assessed using the area under the receiver operating characteristic (ROC) curve to evaluate how well the model recognized patients who had died by 12 months [6]. Statistical analyses were performed using SPSS software, version 12.01.



   Results
Top
 Abstract
 Introduction
 Patients and methods
 Statistical analysis
 Results
Discussion
 References
 
Description of the population
The demographic and clinical characteristics of the 5738 patients studied are shown in Table 1. Among the total, 16.5% (n = 946) had died by 1 year following the start of HD treatment. The mean age was 64.6 ± 14.4 years and 62.2% were men. The etiology of renal failure was diabetic nephropathy in 20% of the cases and systemic disease in 5%. Ten percent of the patients had not been seen by a nephrologist; hence the debut of end-stage renal disease (ESRD) was acute in these cases. Thirteen percent required special care for daily living and 29% had somewhat limited functional autonomy. More than half of the patients initiated dialysis with a catheter as the first vascular access. The most frequent comorbid conditions were cardiovascular disease in more than 40%, peripheral vascular disease 27%, arthropathy 23% and COPD 17%.


View this table:
[in this window]
[in a new window]

  		Table 1 Description of the study population. New cases on haemodialysis, 1997–2003

 
Development of the model
In the sample used to develop the model (n = 3445), first-year mortality was 16.2% (n = 558). Information on all the variables studied was available in 3326 cases (96.3%).

The results of the predictive model for mortality in the first year of HD treatment are shown in Table 2. The table contains ß coefficients with the standard error and OR values with the 95% confidence interval for all the risk factors in the model, except the variables age and malignant processes. The coefficients of these two variables cannot be interpreted separately; they should be interpreted together as an interaction. Since age is a continuous variable, ORs are shown as a graph (Figure 1) rather than in a table.


View this table:
[in this window]
[in a new window]

  		Table 2 Prognostic model for mortality at 1 year following initiation of haemodialysis treatment

 

Figure 1
View larger version (15K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 1 Odds ratio of the risk of death one year after starting hemodialysis according to age and neoplastic disease diagnosis.

 
The model shows that the mortality risk of patients with systemic primary renal disease is 2.7 times greater than that of patients with standard primary renal disease, after adjusting for the other factors (Table 2). Patients who require special care for daily living at the start of HD have a 4-fold greater risk of death at 1 year than those with normal functional independence, and patients with COPD have a 1.3-fold higher risk than those without this comorbid condition.

The interaction between first vascular access and cardiovascular disease should be interpreted by taking as the reference patients without cardiovascular disease who started HD with an arteriovenous fistula. Thus, the fact of presenting cardiovascular disease before HD implied a 3.2-fold increased risk, the fact of having a catheter as first access without cardiovascular disease implied a nearly 4-fold increased risk and the fact of starting with a catheter and having cardiovascular disease increased the risk by 5.6-fold as compared to the reference population. Examining a specific example, in two patients with cardiovascular disease differentiated only by the vascular access at the start of HD, the patient starting with a catheter will have a 1.7-fold higher risk (5.63/3.24) of 1-year mortality. In another scenario, the risk in a patient without cardiovascular disease starting HD with a catheter (OR 3.8) is higher than that of a patient starting with an AV fistula (OR 3.2).

The interpretation of the other interaction between two dichotomous variables, first vascular access and nutrition, should be done in a similar manner. As was seen from the above interaction, the risk of starting HD with a catheter and normal nutritional status is greater than starting with an AV fistula and malnutrition (OR 3.8 versus 3.4).

Interactions between age and the presence or absence of a malignant process are shown in Figure 1. Using 20-year-old patients with no malignancy as the reference, mortality increases with age in patients without neoplastic disease, but not in patients with neoplastic disease. The increased risk of death in patients with a malignant process relative to those without is higher in younger patients than in older ones; nevertheless, patients with a malignant disease are always at a higher risk than those without, whatever the age. Between two 20-year-old patients, differentiated only by the presence of a malignant process, the risk of death is almost 15 times higher in the patient with a neoplasm.

The Hosmer and Lemeshow test results (P = 0.97) demonstrated good calibration and goodness-of-fit for the model studied (Figure 2). The area under the ROC curve for the model was 0.78, indicating appropriate discrimination among patients who died by 1 year after starting treatment and those who did not.


Figure 2
View larger version (27K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 2 Expected and observed mortality by deciles risk group in the development and the validation populations (Hosmer-Lemeshow test).

 
Validation of the model
The model was validated in the remaining 40% of the patients inscribed in the Registry during the study period (n = 2283 cases). One-year mortality was 17% (n = 388) in these patients and information on all the variables studied was available in 2192 cases (96.0%).

The coefficients of the prognostic model were applied to this population according to the following formula:


Formula

which yielded the specific probability of death at 1 year for each individual patient. These probabilities were distributed into 10 groups (deciles of risk), allowing estimation of the expected number of survivors and non-survivors in the first year of treatment for each of the 10 groups. These results were compared with the actual number of survivors and non-survivors at 1 year in each of the groups (Figure 2). There were no statistically significant differences between the data estimated by the model and the true data on 1-year mortality in the validation population. The Hosmer and Lemeshow test results (P = 0.49) again demonstrated good calibration and goodness-of-fit for the model studied. The area under the ROC curve for the model was 0.78, indicating appropriate discrimination between patients who died by 1 year after starting treatment and those who did not.



   Discussion
Top
 Abstract
 Introduction
 Patients and methods
 Statistical analysis
 Results
 Discussion
References
 
Identification of risk factors and development of indicators predictive of mortality are important for improving the outcome of medical care from both the clinical and administrative standpoints. In the present study, a prognostic indicator of death at 1 year following the start of HD was formulated to aid in clinical decision-making and management of nephrological resources.

It is known that the presence of ischaemic heart disease, cardiac or respiratory failure, or neoplastic disease is associated with early death due to any cause in renal patients [1,7]. In addition, certain clinical factors that are not specifically linked to renal disease, such as age, type of vascular access and degree of functional dependence, are also related to early death in this population [8,9]. To our knowledge, there is only one (non-validated) study in the literature undertaken in patients starting dialysis in a hospital, in which the severity of the comorbid conditions and the grade of functional autonomy were found to be more important than the age for predicting morbidity and mortality, and late referral was also identified as an adverse factor [10]. Despite the recognized transcendence of these risk factors, they have been rarely assessed in population settings.

Our results on morbidity concur with the findings of other studies that have compiled and characterized the variables in a similar manner [11–14]. Specifically, the presence of heart disease, malnutrition, neoplasm or liver disease before starting RRT were all associated with early death. Nevertheless, our findings on predialysis risk factors apart from morbidity show some differences as compared to those reported in the literature. In the present study, the association between late referral to the nephrologist and early mortality unexpectedly disappeared when the model was adjusted for the type of vascular access used to start dialysis. We believe this may have occurred because of the link between these two factors: the small number of patients in the cohort with late referral (10.4% of the total) [15–17] and the elevated number of patients (around 50%) starting dialysis with a catheter, mainly because of organizational problems in the health system (coordination among specialists and lengthy waiting lists in vascular surgery). In addition, our model highlighted another variable, the grade of functional autonomy derived from the Karnofsky index [4], which proved to be a good predictor of mortality in several analyses with our data [1].

One unexpected finding from the present study was that patients with systemic renal disease, but not those with diabetic disease, showed a higher risk of death as compared to patients with standard primary renal disease. Nevertheless, diabetic nephropathy is a well-recognized risk factor for mortality [18]. According to data from the RMRC, long-term mortality in diabetic patients is also higher in our dialysis population [1]; however, in the first year of dialysis, the risk decreased after adjusting for factors such as functional autonomy and presence of cardiovascular disease, which carried a great deal of weight in the analysis in this group.

Examination of the various interactions between variables also provided relevant information, although this type of analysis was infrequent in the articles consulted. It is usually assumed that the risk is constant for different combinations of the variables; hence this aspect is not analysed. As a result of this practice, the models obtained are simpler, but may not be able to explain more complex realities, and as a consequence, the predictions offered by the prognostic indicators may be less accurate. In the present study, the added risk of death implicated in starting HD with a catheter was not the same for all patients. For example, for patients who do not have cardiovascular disease at initiation of HD, initial catheter access implies a 4-fold higher risk of death than initial AV fistula access. However, for patients who already have cardiovascular disease, the increased risk associated with catheter access is ‘only’ double that of AV fistula access. A similar pattern was observed with the interaction between vascular access and malnutrition.

Along the same line, we found that the risk of death associated with the presence of a malignant process was not constant at all ages. The younger age groups had a higher associated risk. This may be because neoplasms can be more aggressive in this population or because the time interval between the diagnosis and treatment of the malignant process, and the start of dialysis is likely to be longer in elderly than in young patients; hence the older group would have a greater probability of having overcome the disease when dialysis was started.

Several general indicators of comorbidity are in use, such as the Charlson [19] and Elixhauser [20] indexes; the former has been adapted for renal patients [21]. Several studies comparing various comorbiditiy indexes for their capacity to predict survival of renal patients have concluded that all are appropriate for analysing the impact of comorbidity on mortality [9,22] and that comorbid conditions explain a higher percentage of the variations in heath status when they are analysed individually than when they are analysed as a group by means of the different indicators [20,23].

Other predictive models of mortality specifically designed for patients initiating dialysis and including a large number of relevant laboratory determinations have attributed a higher predictive value to comorbidity indexes than to laboratory findings [24]. Application of these models tends to be complex and most of them have not as yet been validated [24,25], with the exception of one published recently [26]. This last study presents a model developed from a retrospective analysis of a cohort of patients from different countries who died during a specific period of time, and was validated with a cohort of patients who started HD treatment in the same time period. Our model was developed and validated in a large number of incident cases in a population context that avoids the selection bias that may have occurred in the aforementioned study.

The limitations of this study are those inherent to any study performed with data extracted from a population registry, in which the variables utilized are limited in number and have a relatively low clinical specificity, but are highly robust. Another limitation related to the data source is the absence of certain parameters that might be relevant for predicting mortality in this population, such as haemoglobin determination, residual renal function and date of the first visit to the nephrologist. These factors were not available for the cases analysed during the study period, and retrospective data collection was excluded because of the large number of cases involved and the low reliability of data compiled in this manner.

The fact that the model was developed and validated in the same geographic setting may pose a problem for applying it to other patient populations on haemodialysis, in which the health care organization may imply substantial modifications in the non-morbid determinants of nephrological care (e.g. the priority of starting with AV fistulas). Nonetheless, the findings obtained are similar to other recently reported results [26], which leads us to believe that it could be feasible to apply our model in other settings. To this purpose, it would be interesting to validate it in other populations.

Despite the limitations mentioned, studies carried out with population registries can be more reliable than those performed with hospital data or in samples because the information is collected prospectively, there is no selection bias, a large number of cases are available and loss of patients to follow-up in the period studied is virtually non-existent. The Renal Registry of Catalonia, a mandatory population registry that records information on all patients receiving RRT in the Mediterranean region of Catalonia, is well suited for this purpose. In 1988, the Registry underwent an external validation process, which showed exhaustive notification of relevant variables and excellent agreement. These findings verify the validity of the data for the use in clinical and epidemiological studies, and indicate the Registry's proper functioning. In 1990, the RMRC became a local registry of the European Dialysis and Transplant Association (EDTA) [27].

Another advantage of this study is the fact that it was conducted in a relatively small community (7 million inhabitants) with a registry that has been fully functioning for 23 years, which facilitates homogeneity in the system for data notification and clinical management of these patients, thereby conferring a high degree of robustness to the model.

Finally, we mention that the variables used in the model are easily compiled by the attending physician, even at the first visit with the patient, and that analytical and biological determinations are not required.

The availability of an easy-to-apply risk indicator when the indication for RRT is being assessed can provide an orientation regarding interventions focussed on correcting modifiable risk factors, which should be implemented during the predialysis period whenever possible. It is important for the clinician to have a reliable estimation of the probability of death for each individual patient. Despite the fact that nephrologists can predict early death with reasonable certitude based on their experience, acceptance or rejection of RRT depends on the objective information offered to patients and their family. In other words, the availability of valid, easily applicable prognostic indicators adds an objective element to the evaluation of indicating or discontinuing renal replacement therapy. The study also provides the possibility for objective criteria to be remitted to the public health system to improve the establishment of priorities in the care provided. Lastly, the scores obtained with this model can be used as an adjusting variable when comparing survival results from different populations.

In conclusion, the model presented possesses good predictive capacity, as was seen in the validation study, and the variables used can be easily obtained in daily clinical practice. We believe the simplicity of the system is a decisive advantage when considering its applicability and utility in nephrological patient care.



   Acknowledgments
 
The authors thank Celine Cavallo for English language editing.

Conflict of interest statement. None declared.



   References
Top
 Abstract
 Introduction
 Patients and methods
 Statistical analysis
 Results
 Discussion
 References
 

  1. Catalan Renal Registry 21st Report (2004). Catalan Health Department, Catalan Government, Barcelona. (2006) http://www10.gencat.net/catsalut/ocatt/ca/htm/est_pub_trans_renal.htm.
  2. Van Dijk PC, Zwinderman AH, Dekker FW, et al. Effect of general population mortality on the north-south mortality gradient in patients on renal replacement therapy in Europe. Kidney Int (2007) 71:53–59.[CrossRef][Web of Science][Medline]
  3. Catalan Renal Registry 1st Report (1984). Catalan Health Department, Catalan Government, Barcelona. (1985).
  4. Gutman RA, Stead WW, Robinson RR. Physical activity and employment status of patients on maintenance dialysis. New England J Medicine (1981) 304:309–313.[Abstract]
  5. Hosmer DW, Lemeshow S. Applied Logistic Regression (1989) New York: Wiley.
  6. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristics (ROC) curve. Radiology (1982) 153:29–36.
  7. Davies SJ, Phillips L, Naish PF, et al. Quantifying comorbidity in peritoneal patients and its relationship to other predictors of survival. Nephrol Dial Trasplant (2002) 17:1085–1099.[CrossRef]
  8. Xue JL, Dahl D, Ebben JB, et al. The association of initial hemodialysis access type with mortality. Outcomes in elderly. Medicare ESRD patients. Am J Kidney Dis (2003) 42:1013–1019.[Web of Science][Medline]
  9. Hemmelgam BR, Manns BJ, Quan H, et al. Adapting the Charlson Comorbidity Index for use in patients with ESRD. Am J Kidney Dis (2003) 42:125–132.[CrossRef][Web of Science][Medline]
  10. Chandna SM, Schulz J, Lawrence C, et al. Is there a rationale for rationing chronic dialysis? A hospital based cohort study of factors affecting survival and morbidity. BMJ (1999) 318:217–223.[Abstract/Free Full Text]
  11. U.S. Renal Data System. (2005) USRDS 2005 Annual Data Report: Atlas of end-stage renal disease in the United States, National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD.
  12. ANZDATA 28th Report 2005. Australia and New Zealand Dialysis and Transplant Registry, Adelaide, South Australia. http://www.anzdata.org.au/ANZDATA/AnzdataReport/download.htm.
  13. Van Manen JG, van Dijk PC, Stel VS, et al. Confounding effect of co-morbidity in survival studies in patients on RRT. Nephrol Dial Trasplant (2007) 22:187–195.
  14. UK Renal Registry Report 2004. UK Renal Registry, Bristol, UK.
  15. Kazmi WH, Obrador GT, Khan SS, et al. Late nephrology referral and mortality among patients with end-stage renal disease: a propensity score analysis. Nephrol Dial Transplant (2004) 19:1671–1673.[Free Full Text]
  16. Lameire N, Van Biesem W. The pattern of referral of patients with end-stage renal disease to the nephrologist—a Europe survey. Nephrol Dial Transplant (1999) 14:16–23.[Abstract/Free Full Text]
  17. Lorenzo V, Martin M, Rufino M, et al. Predialysis nephrologic care and functioning arteriovenous fistula at entry are associated with better survival in incident hemodialysis patients: an observational cohort study. Am J Kidney Dis (2004) 43:999–1007.[CrossRef][Web of Science][Medline]
  18. Johnson JG, Gore SM, Firth J. The effect of age, diabetes, and other comorbidity on the survival of patients on dialysis: a systematic quantitative overview of the literature. Nephrol Dial Trasplant (1999) 14:2156–2164.[CrossRef]
  19. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chron Dis (1987) 40:373–383.[CrossRef][Web of Science][Medline]
  20. Elixhauser A, Steiner C, Harris R, et al. Comorbidity measures for use with administrative data. Med Care (1998) 36:8–27.[CrossRef][Web of Science][Medline]
  21. Miskulin DC, Martin AA, Brown R, et al. Predicting one-year mortality in an outpatient haemodialysis population; a comparison of co-morbidity instruments. Nephrol Dial Transplant (2004) 19:413–420.[Abstract/Free Full Text]
  22. Postorino M, Marino C., Tripepi G, Zoccali C. Calabrian Registry of Dialysis and Transplantation. Prognostic value of the New York Heart Association classification in end-stage renal disease. Nephrol Dial Transplant (2007) 25(5):1377–82.
  23. Van Manen JG, Korevaar JC, Dekker FW, et al. For the NECOSAD-Study group. Adjustment for comorbidity in studies on health status in ESRD patients: which comorbidity index to use? J Am Soc Nephrol (2003) 14:148–485.[Abstract/Free Full Text]
  24. Miskulin DC, Meyer KB, Martin AA, et al. Comorbidity and its change predict survival in incident dialysis patients. Am J Kidney Dis (2003) 41:149–161.[CrossRef][Web of Science][Medline]
  25. Khan I. Co-morbidity: the major challenge for survival and quality of life in end-stage renal disease. Nephrol Dial Transplant (1998) 13:76–79.[Free Full Text]
  26. Geddes CC, van Dijk PCW, McArthur S, et al. The ERA_EDTA cohort study-comparison of methods to predict survival on renal replacement therapy. Nephrol Dial Transplant (2006) 21:945–956.[Abstract/Free Full Text]
  27. Aubia J, Puig JM, Clèries M, et al. Cómo validar un registro de enfermos renales: Consideraciones a propósito del audit del Registro de enfermos renales de Catalunya. Nefrologia (XI) (1991) 4:332–339.
Received for publication: 27. 4.07
Accepted in revised form: 18. 9.07


Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?


This article has been cited by other articles:


Home page
 Nephrol Dial TransplantHome page
R. Fiedler, P. M. Jehle, B. Osten, O. Dorligschaw, and M. Girndt
Clinical nutrition scores are superior for the prognosis of haemodialysis patients compared to lab markers and bioelectrical impedance
Nephrol. Dial. Transplant., July 15, 2009; (2009) gfp346v1.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
23/5/1690    most recent
gfm728v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Mauri, J. M
Right arrow Articles by Registry, C. R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Mauri, J. M
Right arrow Articles by Registry, C. R.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?