Development and Validation of Prediction Scores for Early Mortality at Transition to Dialysis

Development of a prognostic risk score to predict early mortality in incident elderly Japanese hemodialysis patients

BACKGROUND

Information of short-term prognosis after hemodialysis (HD) introduction is important for elderly patients with chronic kidney disease (CKD) and their families choosing a modality of renal replacement therapy. Therefore, we developed a risk score to predict early mortality in incident elderly Japanese hemodialysis patients.

MATERIALS AND METHODS

We analyzed data of incident elderly HD patients from a nationwide cohort study of the Japanese Society for Dialysis Therapy Renal Data Registry (JRDR) to develop a prognostic risk score. Candidate risk factors for early death within 1 year was evaluated using multivariate logistic regression analysis. The risk score was developed by summing up points derived from parameter estimate values of independent risk factors. The association between risk score and early death was tested using Cox proportional hazards models. This risk score was validated twice by using an internal validation cohort derived from the JRDR and an external validation cohort collected for this study.

RESULTS

Using the development cohort (n = 2,000), nine risk factors were retained in the risk score: older age (>85), yes = 2, no = 0; sex, male = 2, female = 0; lower body mass index (<20), yes = 2, no = 0; cancer, yes = 1, no = 0; dementia, yes = 3, no = 0; lower creatinine (<6.5 mg/dL), yes = 1, no = 0; lower albumin (<3.0 g/dL), yes = 3, no = 0; normal or high calcium (≥8.5 mg/dL), yes = 1, no = 0; and higher C reactive protein (>2.0 mg/dL), yes = 2, no = 0. In the internal and external validation cohorts (n = 739, 140, respectively), the medium- and high-risk groups (total score, 6 to 10 and 11 or more, respectively) showed significantly higher risk of early death than the low-risk group (total score, 0 to 5) (p<0.001).

CONCLUSION

We developed a prognostic risk score predicting early death within 1 year in incident elderly Japanese HD patients, which may help detect elderly patients with a high-risk of early death after HD introduction.

Predicting outcomes in chronic kidney disease: needs and preferences of patients and nephrologists

INTRODUCTION

Guidelines on chronic kidney disease (CKD) recommend that nephrologists use clinical prediction models (CPMs). However, the actual use of CPMs seems limited in clinical practice. We conducted a national survey study to evaluate: 1) to what extent CPMs are used in Dutch CKD practice, 2) patients' and nephrologists' needs and preferences regarding predictions in CKD, and 3) determinants that may affect the adoption of CPMs in clinical practice.

METHODS

We conducted semi-structured interviews with CKD patients to inform the development of two online surveys; one for CKD patients and one for nephrologists. Survey participants were recruited through the Dutch Kidney Patient Association and the Dutch Federation of Nephrology.

RESULTS

A total of 126 patients and 50 nephrologists responded to the surveys. Most patients (89%) reported they had discussed predictions with their nephrologists. They most frequently discussed predictions regarded CKD progression: when they were expected to need kidney replacement therapy (KRT) (n = 81), and how rapidly their kidney function was expected to decline (n = 68). Half of the nephrologists (52%) reported to use CPMs in clinical practice, in particular CPMs predicting the risk of cardiovascular disease. Almost all nephrologists (98%) reported discussing expected CKD trajectories with their patients; even those that did not use CPMs (42%). The majority of patients (61%) and nephrologists (84%) chose a CPM predicting when patients would need KRT in the future as the most important prediction. However, a small portion of patients indicated they did not want to be informed on predictions regarding CKD progression at all (10-15%). Nephrologists not using CPMs (42%) reported they did not know CPMs they could use or felt that they had insufficient knowledge regarding CPMs. According to the nephrologists, the most important determinants for the adoption of CPMs in clinical practice were: 1) understandability for patients, 2) integration as standard of care, 3) the clinical relevance.

CONCLUSION

Even though the majority of patients in Dutch CKD practice reported discussing predictions with their nephrologists, CPMs are infrequently used for this purpose. Both patients and nephrologists considered a CPM predicting CKD progression most important to discuss. Increasing awareness about existing CPMs that predict CKD progression may result in increased adoption in clinical practice. When using CPMs regarding CKD progression, nephrologists should ask whether patients want to hear predictions beforehand, since individual patients' preferences vary.

A Machine Learning Model for Predicting Mortality within 90 Days of Dialysis Initiation

Background

The first 90 days after dialysis initiation are associated with high morbidity and mortality in end-stage kidney disease (ESKD) patients. A machine learning–based tool for predicting mortality could inform patient-clinician shared decision making on whether to initiate dialysis or pursue medical management. We used the eXtreme Gradient Boosting (XGBoost) algorithm to predict mortality in the first 90 days after dialysis initiation in a nationally representative population from the United States Renal Data System.

Methods

A cohort of adults initiating dialysis between 2008–2017 were studied for outcome of death within 90 days of dialysis initiation. The study dataset included 188 candidate predictors prognostic of early mortality that were known on or before the first day of dialysis and was partitioned into training (70%) and testing (30%) subsets. XGBoost modeling used a complete-case set and a dataset obtained from multiple imputation. Model performance was evaluated by c-statistics overall and stratified by subgroups of age, sex, race, and dialysis modality.

Results

The analysis included 1,150,195 patients with ESKD, of whom 86,083 (8%) died in the first 90 days after dialysis initiation. The XGBoost models discriminated mortality risk in the nonimputed (c=0.826, 95% CI, 0.823 to 0.828) and imputed (c=0.827, 95% CI, 0.823 to 0.827) models and performed well across nearly every subgroup (race, age, sex, and dialysis modality) evaluated (c>0.75). Across predicted risk thresholds of 10%–50%, higher risk thresholds showed declining sensitivity (0.69–0.04) with improving specificity (0.79–0.99); similarly, positive likelihood ratio was highest at the 40% threshold, whereas the negative likelihood ratio was lowest at the 10% threshold. After calibration using isotonic regression, the model accurately estimated the probability of mortality across all ranges of predicted risk.

Conclusions

The XGBoost-based model developed in this study discriminated risk of early mortality after dialysis initiation with excellent calibration and performed well across key subgroups.

Shared decision-making in advanced kidney disease: a scoping review

OBJECTIVES

To provide a comprehensive overview of interventions that support shared decision-making (SDM) for treatment modality decisions in advanced kidney disease (AKD). To provide summarised information on their content, use and reported results. To provide an overview of interventions currently under development or investigation.

DESIGN

The JBI methodology for scoping reviews was followed. This review conforms to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA-ScR) checklist.

DATA SOURCES

MEDLINE, Embase, Web of Science, Cochrane Library, Emcare, PsycINFO, PROSPERO and Academic Search Premier for peer-reviewed literature. Other online databases (eg, clinicaltrials.gov, OpenGrey) for grey literature.

ELIGIBILITY FOR INCLUSION

Records in English with a study population of patients >18 years of age with an estimated glomerular filtration rate <30 mL/min/1.73 m². Records had to be on the subject of SDM, or explicitly mention that the intervention reported on could be used to support SDM for treatment modality decisions in AKD.

DATA EXTRACTION AND SYNTHESIS

Two reviewers independently screened and selected records for data extraction. Interventions were categorised as prognostic tools (PTs), educational programmes (EPs), patient decision aids (PtDAs) or multicomponent initiatives (MIs). Interventions were subsequently categorised based on the decisions they were developed to support.

RESULTS

One hundred forty-five interventions were identified in a total of 158 included records: 52 PTs, 51 EPs, 29 PtDAs and 13 MIs. Sixteen (n=16, 11%) were novel interventions currently under investigation. Forty-six (n=46, 35.7%) were reported to have been implemented in clinical practice. Sixty-seven (n=67, 51.9%) were evaluated for their effects on outcomes in the intended users.

CONCLUSION

There is no conclusive evidence on which intervention is the most efficacious in supporting SDM for treatment modality decisions in AKD. There is a lot of variation in selected outcomes, and the body of evidence is largely based on observational research. In addition, the effects of these interventions on SDM are under-reported.

Development and validation of the MMCD score to predict kidney replacement therapy in COVID-19 patients

BACKGROUND

Acute kidney injury (AKI) is frequently associated with COVID-19, and the need for kidney replacement therapy (KRT) is considered an indicator of disease severity. This study aimed to develop a prognostic score for predicting the need for KRT in hospitalised COVID-19 patients, and to assess the incidence of AKI and KRT requirement.

METHODS

This study is part of a multicentre cohort, the Brazilian COVID-19 Registry. A total of 5212 adult COVID-19 patients were included between March/2020 and September/2020. Variable selection was performed using generalised additive models (GAM), and least absolute shrinkage and selection operator (LASSO) regression was used for score derivation. Accuracy was assessed using the area under the receiver operating characteristic curve (AUC-ROC).

RESULTS

The median age of the model-derivation cohort was 59 (IQR 47-70) years, 54.5% were men, 34.3% required ICU admission, 20.9% evolved with AKI, 9.3% required KRT, and 15.1% died during hospitalisation. The temporal validation cohort had similar age, sex, ICU admission, AKI, required KRT distribution and in-hospital mortality. The geographic validation cohort had similar age and sex; however, this cohort had higher rates of ICU admission, AKI, need for KRT and in-hospital mortality. Four predictors of the need for KRT were identified using GAM: need for mechanical ventilation, male sex, higher creatinine at hospital presentation and diabetes. The MMCD score had excellent discrimination in derivation (AUROC 0.929, 95% CI 0.918-0.939) and validation (temporal AUROC 0.927, 95% CI 0.911-0.941; geographic AUROC 0.819, 95% CI 0.792-0.845) cohorts and good overall performance (Brier score: 0.057, 0.056 and 0.122, respectively). The score is implemented in a freely available online risk calculator ( https://www.mmcdscore.com/ ).

CONCLUSIONS

The use of the MMCD score to predict the need for KRT may assist healthcare workers in identifying hospitalised COVID-19 patients who may require more intensive monitoring, and can be useful for resource allocation.

Smoothing transition to dialysis to improve early outcomes after dialysis initiation among old and frail adults-a narrative review

The number of patients ≥65 years of age suffering from advanced chronic kidney disease and transitioning to end-stage kidney disease (ESKD) is increasing. However, elderly patients often have poor outcomes once haemodialysis is initiated, including high mortality within the first year as well as fast cognitive and functional decline and diminished quality of life. The question is how we can smooth this transition to ESKD in older patients who also exhibit much higher proportions of frailty when compared with community-dwelling non-dialysis older adults and who are generally more vulnerable to invasive treatment such as kidney replacement therapy. To avoid early death and poor quality of life, a carefully prepared smooth transition should precede the initiation of treatment. This involves pre-dialysis physical and educational care, as well as mental and psychosocial preparedness of the patient to enable an informed and shared decision about the individual choice of treatment modality. Communication between a healthcare professional and patient plays a pivotal role but can be challenging given the high rate of cognitive impairment in this particular population. In order to practise patient-centred care, adapting treatment tailored to the individual patient should include comprehensive conservative care. However, structured treatment pathways including multidisciplinary teams for such conservative care are still rare and may be difficult to establish outside of large cities. Generally, geriatric nephrology misses data on the comparative effectiveness of different treatment modalities in this population of old and very old age on which to base recommendations and decisions.

Chronic kidney disease

Chronic kidney disease is a progressive disease with no cure and high morbidity and mortality that occurs commonly in the general adult population, especially in people with diabetes and hypertension. Preservation of kidney function can improve outcomes and can be achieved through non-pharmacological strategies (eg, dietary and lifestyle adjustments) and chronic kidney disease-targeted and kidney disease-specific pharmacological interventions. A plant-dominant, low-protein, and low-salt diet might help to mitigate glomerular hyperfiltration and preserve renal function for longer, possibly while also leading to favourable alterations in acid-base homoeostasis and in the gut microbiome. Pharmacotherapies that alter intrarenal haemodynamics (eg, renin-angiotensin-aldosterone pathway modulators and SGLT2 [SLC5A2] inhibitors) can preserve kidney function by reducing intraglomerular pressure independently of blood pressure and glucose control, whereas other novel agents (eg, non-steroidal mineralocorticoid receptor antagonists) might protect the kidney through anti-inflammatory or antifibrotic mechanisms. Some glomerular and cystic kidney diseases might benefit from disease-specific therapies. Managing chronic kidney disease-associated cardiovascular risk, minimising the risk of infection, and preventing acute kidney injury are crucial interventions for these patients, given the high burden of complications, associated morbidity and mortality, and the role of non-conventional risk factors in chronic kidney disease. When renal replacement therapy becomes inevitable, an incremental transition to dialysis can be considered and has been proposed to possibly preserve residual kidney function longer. There are similarities and distinctions between kidney-preserving care and supportive care. Additional studies of dietary and pharmacological interventions and development of innovative strategies are necessary to ensure optimal kidney-preserving care and to achieve greater longevity and better health-related quality of life for these patients.

[Treatment of geriatric patients with end-stage renal failure : Dialysis or conservative procedure?]

Compared to younger patients, an acceptable state of health in older patients with various comorbidities is rarely achieved by the initiation of dialysis. Despite dialysis treatment, further functional and cognitive impairments often rapidly occur in geriatric patients. Thus, newer studies are concerned with the quality of life of this patient group after initiation of dialysis as well as with palliative treatment strategies as alternatives. A structured clarification for the patients on all possibilities with mediation of all necessary information is a prerequisite for a shared decision-making. To assess life expectancy after dialysis initiation, various scores have been developed but the sensitivity could not fulfil the expectations. In the case of renal replacement, chronic intermittent hemodialysis is the treatment form most frequently performed in geriatric patients. The main concern of conservative palliative treatment is the quality of life and the management of uremic symptoms, which have to be addressed by a multidisciplinary team.

Predialysis Potassium Variability and Postdialysis Mortality in Patients With Advanced CKD

Introduction

Patients with advanced non–dialysis-dependent chronic kidney disease (NDD-CKD) are prone to potassium (K) imbalances due to reduced kidney function. Both hypo- and hyperkalemia are associated with increased mortality; however, it is unclear if K variability before dialysis initiation is associated with outcomes after dialysis initiation.

Methods

We identified 34,167 US veterans with advanced NDD-CKD transitioning to dialysis between October 1, 2007, through March 31, 2015, who had at least 1 K measurement each year over a 3-year period before transition (3-year prelude). For each patient, a linear mixed-effects model was used to regress K over time (in years) over the 3-year prelude to derive K variability (square root of the average squared distance between the observed and estimated K). The main outcomes of interest were 6-month all-cause and cardiovascular mortality after dialysis initiation. Multivariable Cox and Fine-Gray competing risk regression adjusted for 3-year prelude K intercept, K slope (per year), demographics, smoking status, comorbidities, length of hospitalizations, body mass index, vascular access type, medications, average estimated glomerular filtration rate, and number of K measurements over the 3-year prelude were used to assess the association of K variability (expressed as quartiles) with all-cause and cardiovascular mortality, respectively.

Results

Higher prelude K variability was associated with higher multivariable-adjusted risk of all-cause mortality but not cardiovascular mortality (adjusted hazard/subhazard ratios [95% confidence interval] for highest quartile [vs. lowest] of K variability, 1.14 [1.03–1.25] and 0.99 [0.85–1.16] for all-cause and cardiovascular mortality, respectively).

Conclusion

Higher K variability is associated with higher all-cause mortality after dialysis initiation.

The quest for cardiovascular disease risk prediction models in patients with nondialysis chronic kidney disease

PURPOSE OF REVIEW

Cardiovascular disease (CVD) is the leading cause of death in patients with chronic kidney disease (CKD). However, traditional CVD risk prediction equations do not work well in patients with CKD, and inclusion of kidney disease metrics such as albuminuria and estimated glomerular filtration rate have a modest to no benefit in improving prediction.

RECENT FINDINGS

As CKD progresses, the strength of traditional CVD risk factors in predicting clinical outcomes weakens. A pooled cohort equation used for CVD risk prediction is a useful tool for guiding clinicians on management of patients with CVD risk, but these equations do not calibrate well in patients with CKD, although a number of studies have developed modifications of the traditional equations to improve risk prediction. The reason for the poor calibration may be related to the fact that as CKD progresses, associations of traditional risk factors such as BMI, lipids and blood pressure with CVD outcomes are attenuated or reverse, and other risk factors may become more important.

SUMMARY

Large national cohorts such as the US Veteran cohort with many patients with evolving CKD may be useful resources for the developing CVD prediction models; however, additional considerations are needed for the unique composition of patients receiving care in these healthcare systems, including those with multiple comorbidities, as well as mental health issues, homelessness, posttraumatic stress disorders, frailty, malnutrition and polypharmacy. Machine learning over conventional risk prediction models may be better suited to handle the complexity needed for these CVD prediction models.

Mortality Risk in Chronic Kidney Disease Patients Transitioning to Dialysis: Impact of Opiate and Non-Opiate Use

BACKGROUND

Population-based studies show there is a high prevalence of chronic kidney disease (CKD) patients suffering from chronic pain. While opiates are frequently prescribed in non-dialysis-dependent CKD (NDD-CKD) patients, there may be toxic accumulation of metabolites, particularly among those progressing to end-stage renal disease (ESRD). We examined the association of opiate versus other analgesic use during the pre-ESRD period with post-ESRD mortality among NDD-CKD patients transitioning to dialysis.

METHODS

We examined a national cohort of US Veterans with NDD-CKD who transitioned to dialysis over 2007-14. Among patients who received ≥1 prescription(s) in the Veterans Affairs (VA) Healthcare System within 1 year of transitioning to dialysis, we examined associations of pre-ESRD analgesic status, defined as opiate, gabapentin/pregabalin, other non-opiate analgesic, versus no analgesic use, with post-ESRD mortality using multivariable Cox models.

RESULTS

Among 57,764 patients who met eligibility criteria, pre-ESRD opiate and gabapentin/pregabalin use were each associated with higher post-ESRD mortality (ref: no analgesic use), whereas non-opiate analgesic use was not associated with higher mortality in expanded case-mix analyses: HRs (95% CIs) 1.07 (1.05-1.10), 1.07 (1.01-1.13), and 1.00 (0.94-1.06), respectively. In secondary analyses, increasing frequency of opiate prescriptions exceeding 1 opiate prescription in the 1-year pre-ESRD period was associated with incrementally higher post-ESRD mortality (ref: no analgesic use).

CONCLUSIONS

In NDD-CKD patients transitioning to dialysis, pre-ESRD opiate and gabapentin/pregabalin use were associated with higher post-ESRD mortality, whereas non-opiate analgesic use was not associated with death. There was a graded association between increasing frequency of pre-ESRD opiate use and incrementally higher mortality.

Prediction of Risk of Death for Patients Starting Dialysis: A Systematic Review and Meta-Analysis

BACKGROUND AND OBJECTIVES

Dialysis is a preference-sensitive decision where prognosis may play an important role. Although patients desire risk prediction, nephrologists are wary of sharing this information. We reviewed the performance of prognostic indices for patients starting dialysis to facilitate bedside translation.

DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS

Systematic review and meta-analysis following the PRISMA guidelines. We searched Ovid MEDLINE, Ovid Embase, Ovid Central Register of Controlled Trials, Ovid Cochrane Database of Systematic Reviews, and Scopus for eligible studies of patients starting dialysis published from inception to December 31, 2018.

SELECTION CRITERIA

Articles describing validated prognostic indices predicting mortality at the start of dialysis. We excluded studies limited to prevalent dialysis patients, AKI and studies excluding mortality in the first 1-3 months. Two reviewers independently screened abstracts, performed full text assessment of inclusion criteria and extracted: study design, setting, population demographics, index performance and risk of bias. Pre-planned random effects meta-analysis was performed stratified by index and predictive window to reduce heterogeneity.

RESULTS

Of 12,132 articles screened and 214 reviewed in full text, 36 studies were included describing 32 prognostic indices. Predictive windows ranged from 3 months to 10 years, cohort sizes from 46 to 52,796. Meta-analysis showed discrimination area under the curve (AUC) of 0.71 (95% confidence interval, 0.69 to 073) with high heterogeneity (I2=99.12). Meta-analysis by index showed highest AUC for The Obi, Ivory, and Charlson comorbidity index (CCI)=0.74, also CCI was the most commonly used (ten studies). Other commonly used indices were Kahn-Wright index (eight studies, AUC 0.68), Hemmelgarn modification of the CCI (six studies, AUC 0.66) and REIN index (five studies, AUC 0.69). Of the indices, ten have been validated externally, 16 internally and nine were pre-existing validated indices. Limitations include heterogeneity and exclusion of large cohort studies in prevalent patients.

CONCLUSIONS

Several well validated indices with good discrimination are available for predicting survival at dialysis start.

Machine Learning to Identify Dialysis Patients at High Death Risk

Introduction

Given the high mortality rate within the first year of dialysis initiation, an accurate estimation of postdialysis mortality could help patients and clinicians in decision making about initiation of dialysis. We aimed to use machine learning (ML) by incorporating complex information from electronic health records to predict patients at risk for postdialysis short-term mortality.

Methods

This study was carried out on a contemporary cohort of 27,615 US veterans with incident end-stage renal disease (ESRD). We implemented a random forest method on 49 variables obtained before dialysis transition to predict outcomes of 30-, 90-, 180-, and 365-day all-cause mortality after dialysis initiation.

Results

The mean (±SD) age of our cohort was 68.7 ± 11.2 years, 98.1% of patients were men, 29.4% were African American, and 71.4% were diabetic. The final random forest model provided C-statistics (95% confidence intervals) of 0.7185 (0.6994–0.7377), 0.7446 (0.7346–0.7546), 0.7504 (0.7425–0.7583), and 0.7488 (0.7421–0.7554) for predicting risk of death within the 4 different time windows. The models showed good internal validity and replicated well in patients with various demographic and clinical characteristics and provided similar or better performance compared with other ML algorithms. Results may not be generalizable to non-veterans. Use of predictors available in electronic medical records has limited the assessment of number of predictors.

Conclusion

We implemented and ML-based method to accurately predict short-term postdialysis mortality in patients with incident ESRD. Our models could aid patients and clinicians in better decision making about the best course of action in patients approaching ESRD.