The prevalence of kidney disease has reached alarming levels globally, with 850 million people affected, as highlighted by statements from the American Society of Nephrology, European Renal Association, and International Society of Nephrology.1 Chronic kidney disease (CKD) has a global prevalence of around 9.1–13%, underscoring the urgent need for effective management strategies.1–3 In Portugal, CKD prevalence is notably high at 20.9%, double the global average.4

Nearly four million individuals worldwide are undergoing kidney replacement therapy (KRT), with hemodialysis (HD) standing out as the leading modality.5 In Portugal, of the 2705 patients starting KRT in 2022, 82.7% opted for HD.6 Incident hemodialysis patients face a limitation of life expectancy of 5–10 years, with major causes of mortality including infection and cardiovascular disease.7 Furthermore, the mortality rate within the first year of starting HD is around 15.2%.8 To accurately interpret this, consideration must be given to the baseline characteristics of patients at HD start, as these significantly impact treatment outcomes and mortality rates. Prognosis assessment after starting HD is challenging and risk prediction scores can be useful in planning treatment.

In 2023, Clark et al. developed a risk prediction model to effectively assess the one-year likelihood of kidney recovery and mortality in patients starting HD.9 This was a population-based registry study with a derivation and an external validation cohort of patients starting HD during hospitalization from which they were discharged to continue outpatient HD. The risk model is available as an online tool and considers variables such as age, Charlson comorbidities index, cancer, length of hospital admission, intensive care status, discharge disposition, prehospital admission estimated glomerular filtration rate (eGFR), and random urine albumin-to-creatinine ratio, which are routinely available at discharge.

Given the notable number of incident HD patients in Portugal, there is a critical need for a mortality risk predictor. Such a tool is essential for optimizing therapeutic interventions and refining management strategies in this specific patient population.

Acknowledging the potential variability in the accuracy of prediction models across diverse populations, our study aims to validate the Death Outcome risk score in a Portuguese cohort.

This study is a retrospective analysis of patients who initiated hemodialysis during hospitalization in Unidade Local de Saúde Santa Maria (ULS-SM) in Lisbon, Portugal. The Ethical Committee approved of this study, in agreement with institutional guidelines. Informed consent was waived, given the retrospective and non-interventional nature of the study.

Our study focused on a cohort of 369 patients with a mean age of 71.1±14.3 years. The majority were male (59.9%, n=221) and 87% were Caucasian (n=321). Baseline characteristics are described in Table 1.

Eighty six percent of patients had hypertension (n=319), 73.3% had CKD (n=272) and 55.8% were followed in nephrology consults, 46.1% had diabetes mellitus (n=170) and 43.6% had heart failure (n=161). Twenty percent (n=74) of patients had been diagnosed with malignancy in the 5 years prior to this study. Median Charlson score of this population was 7±3.

Mean baseline SCr was 3.7±1.5mg/dL, eGFR was 28±21.8ml/min/1.73m2 and urine ACR was 387.6±1324.2mg/g. Mean time from baseline data collection to hemodialysis initiation was 123.1±101.2 days. Concerning laboratory at hemodialysis start, mean hemoglobin was 9.4±1.7g/dL, albumin was 3.3±0.6g/dL, creatinine was 6.3 (3.01)mg/dL with an eGFR of 8.2±4.4ml/min/1.73m2.

The time from admission to dialysis start was 5.1±8.9 days and from the start of dialysis to hospital discharge 16.3±17.4 days. Most patients started hemodialysis with a central venous catheter (78.6%, n=290). Dialysis was started in the intensive care unit in 14 patients (3.8%).

At discharge all patients were on hemodialysis and 65% were independent (n=240), 22% had home care (n=81), and 13% were discharged to a nursing home (n=48).

The mortality rate within one year after discharge was 22.2% (n=82). Patients who died within the first year after hemodialysis start were older (79.4±8.7 vs 68.7±14.7 years, p<0.001), had a lower prevalence of hypertension (79.3 vs 88.5%, p=0.031), and a higher median Charlson score (8±2 vs 7±3, p<0.001), with more prevalence of heart failure (59.8 vs 39.0%, p=0.001), ischemic cardiomyopathy (36.6 vs 23.0%, p=0.013) and dementia (11.0 vs 4.9%, p=0.044). These patients also had significantly lower serum albumin (3.1±0.6 vs 3.4±0.6g/dL, p<0.001), a longer time from dialysis start to discharge, and were more frequently discharged to home care or nursing home (p<0.001).

The one-year death outcome (DO) risk of the cohort was 29.7±14.0%, and it was significantly higher in the subgroup of patients who died (38.8±10.3 vs 27.1±13.9%, p<0.001).

Concerning the DO risk, patients were divided in four groups, from lowest score or probability of death to highest, from D1 to D4, as shown in Table 2. Thirty percent of patients were D1 (n=112), 21.7% were D2 (n=80), 19.2% were D3 (n=71) and 28.7% were D4 (n=106). The one-year survival was significantly lower in patients with the highest probability of death (D4=61.3% vs D3=67.6% vs D2=81.3% vs D4=97.3%, p<0.001) – Fig. 1. The Kaplan–Meier plot for death of each DO group is displayed in Fig. 2.

Fig. 1.

Cumulative probability of death curves according to ReDO Death score groups. Log-rank test p<0.001.

(0.19MB).

Fig. 2.

Area under the curve of the REDO risk model for the prediction of death within the first year (auROC 0.741, p<0.001).

(0.16MB).

Patients in D4 were older (p<0.001), had a higher median Charlson score (p<0.001), higher prevalence of active malignancy (p<0.001), longer time from dialysis start to discharge (p<0.001), and were more often discharged with home care or to a nursing home (p<0.001) (Table 3).

The ReDO Death score accurately predicted the one-year risk of mortality, with a hazard ratio of 1.2 [95% CI (1.15–1.31), p<0.001]. The Hosmer–Lemeshow test indicated good fit of this model (p=0.167).

The ReDO Death score predicted the one-year risk of mortality with an auROC of 0.741 [95% CI (0.687–0.794), p<0.001] (Fig. 2), with a sensitivity of 78.9% and specificity of 60.6% (Table 4). The optimal ReDO Death risk cut-off was >30%, with a hazard ratio of 6.57 [95% CI (3.48–12.2), p<0.001] for one year risk of death.

Mortality was significantly higher in older patients (HR 1.1 [95% CI (1.05–1.11), p<0.001]), Caucasians (HR 2.7 [95% CI (1.04–7.09), p=0.042]), patients with heart failure (HR 2.3 [95% CI (1.40–3.83), p=0.001]), patients with ischemic cardiomyopathy (HR 1.9 [95% CI (1.14–3.27), p=0.014]) and higher Charlson score (HR 1.3 [95% CI (1.13–1.39), p<0.001]). Hypertension was associated with lower risk of death (HR 0.49 [95% CI (0.26–0.95), p=0.034]), and so were higher albumin levels at hemodialysis start (HR 0.4 [95% CI (0.27–0.66), p<0.001]).

On a multivariate analysis, older age (adjusted HR 1.08 [95% CI (1.05–1.12), p<0.001]) was the only significant predictor of one-year mortality. Hypertension (adjusted HR 0.29 [95% CI (0.13–0.62), p=0.001]) and higher albumin levels (adjusted HR 0.46 [95% CI (0.28–0.75), p=0.002]) were protective factors for one-year mortality.

In this cohort of patients who initiated hemodialysis during hospitalization, the mortality rate within one year of discharge was 22.2% and the ReDO Death score accurately predicted the one-year risk of mortality. We also identified that a ReDO Death risk >30% was a significant risk predictor for one-year mortality.

Data from the United States Renal Data System shows that although mortality in the first year after dialysis initiation has been decreasing, it remains high (217.3 per 1000 person-years), with over half being due to cardiovascular disease.15 The ERA Registry Annual Report 2021 demonstrated a one-year survival of 87%, corresponding to a 13% mortality rate, which is lower than what we found in our cohort, but still significant.16

As such, it is important to assess patients’ prognosis after hemodialysis start. This is fundamental to tailoring medical care, facilitating shared decision-making, and providing appropriate support and resources. In this setting of significant early mortality risk, understanding and managing patient trajectories enables healthcare professionals to deliver patient-centered care that aligns with the individual needs and goals of each patient while optimizing resource use.

The original ReDO score included a derivation and validation cohort of a total of 9160 Canadian patients, which started dialysis during hospitalization and were discharged to continue outpatient dialysis. The authors created a model to predict kidney recovery to dialysis independence and death (all-cause mortality) within 1 year of hospital discharge, generating the ReDO-Kidney Recovery Score and ReDO Death Outcome Score, respectively. Predictive variables included age, comorbidities, health burden (including the Charlson score, and new cancer diagnosed in previous 5 years), length of hospital admission, intensive care status, discharge disposition, prehospital admission eGFR and random urine albumin-to-creatinine ratio. These are easily attained variables at discharge from hospitalization or at patient admission in a Dialysis Unit and using the online tool can easily provide useful information. Compared to the original derivation cohort, our cohort is similar in terms of age, if slightly older (71.1 years vs. 67 years), and of male preponderance (59.9% vs 62%). However, our Portuguese cohort had lower prevalence of diabetes (46.1% vs. 67%) and of heart failure (43.6% vs. 52%), but higher prevalence of active malignancy (20.1% vs. 13%) and ischemic cardiomyopathy (26% vs. 15%). Median Charlson score was higher in our cohort than in the derivation cohort (7 vs. 5).

The ReDO Death Outcome also accounted for a significant interaction between age and Charlson score, and for a lower predicted probability of death with lower baseline GFR. In the derivation cohort, the c-statistic of the DO risk score was 0.65 (95% CI, 0.60–0.69) and although in the validation group model discrimination was modest (c-statistics [95% CI] 0.66 [0.62–0.69]), calibration was good (integrated calibration index [95% CI] was 4% [2–6%] for death). These results were similar in our cohort, in which the DO risk score had an adequate auROC of 0.741, which corroborates the accuracy of this risk score.

To our knowledge, the ReDO Death Outcome Score has not been studied in other populations. Nevertheless, other studies have looked at predicting early mortality after dialysis initiation. Thamer et al. presented a comprehensive risk model (auROC=0.72) to predict all-cause mortality in the first 3 and 6 months after hemodialysis start in patients aged 67 years or older, that assigned points for age, sex, race, serum creatinine, hypoalbuminemia, catheter use, Nephrology referral, functional status and comorbidities as predictors of mortality.17

We found that patients who died within the first year were significantly older, with more prevalent comorbidities such as heart failure, ischemic cardiomyopathy, and dementia, and with a higher median Charlson score.18 They were also more frequently discharged to home care or nursing home settings, had lower serum albumin and longer hospital stays. Indeed, the fact that the Death Outcome risk score incorporates many of these variables render it effective in assessing for early mortality after starting hemodialysis.

A systematic review and meta-analysis of validated risk scores predicting death after dialysis start evaluated 36 studies (of which 31 on hemodialysis and 5 on peritoneal dialysis), and most studies used prognostic variables such as age and comorbidities; five studies also used functional status and one evaluated length of hospital stay.19,20 The Charlson comorbidity index presented the most consistent discrimination performance (auROC=0.74), which is also included in the DO risk score. In a third of the cases no external validation was conducted, contrasting with the ReDO Death score original study.

The ReDO Death score is a simple tool that can be used at the time of hospital discharge, to better tailor medical care to the patients’ needs. An individualized care plan should be discussed with the patient or with their healthcare proxy throughout their entire hospital stay. And when the discharge date is presumed to be near, the subject should be revisited and discussed to ensure a smooth transition to out-of-hospital care. The patients’ needs, goals and medical trajectory may need to be reassessed, and decisions such as transition of care to a hemodialysis facility versus best supportive care can be discussed.

When the decision to remain on hemodialysis is made, vascular access planning must also be considered. In most of our cohort, a catheter was used at the start of hemodialysis. Although more than half the patients had previous Nephrology follow-up, the lack of functional fistula or grafts might be in some cases explained by unexpected acute-on-chronic kidney injury, patient option and/or frailty. The most recent guidelines recognize the complexity of decisions regarding hemodialysis access, and promote individualized decisions, taking into account the patient's foreseeable lifespan, functional status, social support, preferences and life goals.21 In this regard, the ReDO Death score might be an additional tool in deciding whether to create a fistula or graft.

As limitations of this study, we recognize its retrospective and single center nature, which limit the generalization of these results. Secondly, the moderate size of this sample also limits further data generalization. Thirdly, although there was a high prevalence of prior CKD in our cohort, we acknowledge that CKD cause was not ascertained which could have implication on patient mortality (e.g. patients may have been under prior immunosuppression). Fourthly, mortality causes were not assessed which could also be important to interpret our results. Additionally, mortality within the first week post-discharge was excluded due to challenges in accurate ascertainment. During the transition to outpatient care, patients may not yet be captured in the receiving dialysis clinic's mortality records. To account for this potential data gap, these cases were excluded from the study population.

Finally, in contrast to the original study, we did not evaluate kidney function recovery to hemodialysis independence.

Future research on outcomes in patients initiating hemodialysis may consider comparing the discriminatory performance of the Charlson comorbidity index to the ReDO Death score.

Nevertheless, our study has several strengths. This is the first study to validate the ReDO Death Outcome risk score as a prognostic tool in a Portuguese population. The necessary variables for its application are routinely recorded information, which allow for important multivariate analysis. In addition, the good discriminatory ability and high sensitivity of this model suggest that many patients may be identified using this tool.

In conclusion, we validated the ReDO Death Outcome risk score for one-year mortality prediction after hemodialysis start in a Portuguese population. This score should be used as a tool to inform goals-of-care discussion at the time of transition to out-of-hospital care, involving the in-hospital nephrology care team, the patient, and, if applicable, the future care team, as it can enlighten clinical decisions and, in some cases, lead to better end-of-life planning.

JG was responsible for study conception and design. TP, BT, MC, JAF, CO, SJ and CR collected the data. TP, BT and JG contributed to data analysis and interpretation and draft elaboration. JAL, CO, SJ, CR, JAL and JG provided significant intellectual content. All authors contributed to revision of the manuscript. All authors have approved the final version of the article.

None to report.

On behalf of all authors, the corresponding author states that there is no conflict of interest.