Prediction Model and Risk Stratification Tool for Survival in Patients With CKD

Chronic kidney Disease overall survival prediction model based on frailty index score: construction and validation using NHANES data

BACKGROUND

Frailty predicts poor outcomes in chronic kidney disease (CKD) patients. This study compared frailty's predictive power with other factors and aimed to develop a model for predicting overall survival (OS) in CKD patients.

METHODS

The study included 3,714 CKD participants from the National Health and Nutrition Examination Survey 2005-2018. The death data were updated to December 31, 2019. Lasso-Cox regression identified significant predictors among 42 factors, resulting in a prognostic nomogram using 11 key variables. Subsequent evaluation of the nomogram involved the C-index, the Areas Under Time-dependent Receiver Operating Characteristic Curves (AUC) and calibration curves.

RESULTS

Over a median follow-up of 5.92 years, 1,234 deaths occurred. The final predictors of OS in CKD patients included age, ethnicity, smoking status, estimated pulse wave velocity, body fat percentage, blood uric acid concentration, blood urea nitrogen concentration, and albumin concentration, neutrophil-to-lymphocyte ratio, urine albumin-to-creatinine ratio level, and frailty index (FI) score. The FI score was the strongest predictor with an HR of 76.54 (95% CI: 42.93, 136.46, p < 0.0001). In the training set, the AUC values were 80.11% for 1-year, 79.90% for 3-year, 79.53% for 5-year, and 81.34% for 10-year follow-ups. In the internal validation set, AUC values were 78.66%, 77.78%, 77.56%, and 79.54%, respectively. The nomogram's corrected C-index was 0.76 (95% CI: 0.75 - 0.78), and calibration curves showed satisfactory accuracy.

CONCLUSIONS

The FI score is a significant predictor of CKD OS. The developed nomogram based on the FI score is a promising tool for predicting the OS of CKD patients.

The prognostic reasoning system for chronic kidney disease progression (PROGRES-CKD) may help improve waiting list management for outpatient nephrology services in a second-level public hospital in Italy

BACKGROUND

The management of patients with non-dialysis dependent chronic kidney disease (NDD-CKD) is challenging due to coexisting diseases, competing risks and uncertainties around optimal transition planning. Such clinical challenges are further exacerbated by physician shortage, coupled with rising service demands, which may hinder timely medical access due to long waiting times. Accurate progression risk assessment may help optimize resource allocation and adapting care based on individual patients' needs. This study validated the Prognostic Reasoning System for Chronic Kidney Disease Progression (PROGRES-CKD) in an Italian public hospital and compared its potential impact on waiting list optimization against physician-based protocols.

METHODS

First we first validated PROGRES-CKD by assessing its accuracy in predicting kidney replacement therapy (KRT) initiation within 6 months and 24 months in a historical cohort of patients treated at the San Gerardo Hospital (Italy) between 01-01-2015 and 31-12-2019. In a second study we compared PROGRES-CKD to attending nephrologists' prognostic ratings and simulated their potential impact on a waiting list management protocol.

RESULTS

We included 2005 patients who underwent 11,757 outpatient nephrology visits in 4 years. Most visits occurred for NDD-CKD stage 4 patients; the incidence of KRT onset was 10.8 and 9.32/100 patient-years at the 6 and 24-month prediction horizon cohorts, respectively. PROGRES-CKD demonstrated high accuracy in predicting KRT initiation at 6 and 24 months (AUROC = 0.88 and AUROC = 0.85, respectively). Nephrologists' prognostic performance was highly operator-dependent, albeit always significantly lower than PROGRES-CKD. In the simulation exercise, allocation based on PROGRES-CKD resulted in more follow-up visits for patients progressing to end-stage kidney disease (ESKD) and fewer visits for non-progressing patients, compared to allocation determined by nephrologists' prognosis.

CONCLUSIONS

PROGRES-CKD showed high accuracy in a real-world application. Waiting list simulation suggests that PROGRES-CKD may enable more efficient allocation of resources.

A decision-making algorithm proposal for PICCs and midlines insertion in patients with advanced kidney disease: A pilot study

INTRODUCTION

Kidney Disease Outcomes Quality Initiative clinical practice guidelines recommend avoiding placement of peripherally inserted vascular access devices in patients with an estimated glomerular filtration rate (eGFR) <45 ml/min. On the other hand, many patients with severe chronic kidney disease (CKD) have poor prognosis.This study carried out a global assessment of mortality at 2 years through Charlson Comorbidity Index (CCI) and Beclap score in patients with PICCs or Midlines, assuming that in those with an estimated high mortality rate at 2 years, it could be acceptable to implant a peripheral vascular access device (PVAD) despite the presence of CKD.

METHODS

We analyzed data on patients with PICCs or Midlines inserted from October 2018 to November 2019. CCI, Beclap score, and eGFR were calculated for each patient at the time of the catheter insertion. We then followed patients for 2 years to assess 2-year mortality for each.

RESULTS

One hundred and thirty-one patients were enrolled, 49 (37.4%) had eGFR<45 ml. The 2-year mortality rate was 57.3%. The cut off derived from ROC curve analysis of 15 for Beclap score and 5 for CCI, showed good sensitivity and specificity in predicting mortality of the total population, patients without an oncological disease and patients with eGFR<45 ml/min.

CONCLUSION

CCI and Beclap score are good predictors of mortality at 2 years.Physicians and nurses can use these tools in the evaluation of patients at risk for future dialysis, instead of relying exclusively on renal function to decide whether implanting PICCs, Midlines, or other vascular access devices.

Frailty, but not cognitive impairment, improves mortality risk prediction among those with chronic kidney disease-a nationally representative study

BACKGROUND

Though older adults with chronic kidney disease (CKD) have a greater mortality risk than those without CKD, traditional risk factors poorly predict mortality in this population. Therefore, we tested our hypothesis that two common geriatric risk factors, frailty and cognitive impairment, and their co-occurrence, might improve mortality risk prediction in CKD.

METHODS

Among participants aged ≥ 60 years from National Health and Nutrition Examination Survey (2011-2014), we quantified associations between frailty (physical frailty phenotype) and global/domain-specific cognitive function (immediate-recall [CERAD-WL], delayed-recall [CERAD-DL], verbal fluency [AF], executive function/processing speed [DSST], and global [standardized-average of 4 domain-specific tests]) using linear regression, and tested whether associations differed by CKD using a Wald test. We then tested whether frailty, global cognitive impairment (1.5SD below the mean), or their combination improved prediction of mortality (Cox models, c-statistics) compared to base models (likelihood-ratios) among those with and without CKD.

RESULTS

Among 3,211 participants, 1.4% were cognitively impaired, and 10.0% were frail; frailty and cognitive impairment co-occurrence was greater among those with CKD versus those without (1.2%vs.0.1%). Frailty was associated with worse global cognitive function (Cohen's d = -0.26SD,95%CI -0.36,-0.17), and worse cognitive function across all domains; these associations did not differ by CKD (pinteractions > 0.05). Mortality risk prediction improved only among those with CKD when accounting for frailty (p[likelihood ratio test] < 0.001) but not cognitive impairment.

CONCLUSIONS

Frailty is associated with worse cognitive function regardless of CKD status. While CKD and frailty improved mortality prediction, cognitive impairment did not. Risk prediction tools should incorporate frailty to improve mortality prediction among those with CKD.

Albuminuria improves R2CHA2DS2-VASc score in predicting mortality in high cardiovascular risk population

BACKGROUND AND AIMS

The CHA₂DS₂-VASc score estimates the risk of cardioembolism in patients with atrial fibrillation (AF). It also predicts vascular events and death in different clinical settings, even in the absence of AF. The R₂CHA₂DS₂-VASc score, obtained by adding the glomerular filtration rate to CHA₂DS₂-VASc, shows a higher prediction ability for new events and all-cause mortality. The present study aims to assess whether the addition of albuminuria to R₂CHA₂DS₂-VASc score further improves its discrimination ability in predicting all-cause mortality in a sample of high cardiovascular risk population.

METHODS AND RESULTS

Prospective, monocentric, observational study, evaluating a subset of 737 subjects consecutively undergoing to coronary angiography at Coronary Unit of Scientific Institute "Casa Sollievo della Sofferenza" from June 2016 to December 2018. The presence of albuminuria was significantly associated with all-cause mortality (p < 0.0001). Any one-point increase of Alb-R₂CHA₂DS₂-VASc score increased mortality of about 1.5-fold (adjusted HR 1.49; 95%CI: 1.37-1.63; p < 0.0001). Considering tertiles of Alb-R₂CHA₂DS₂-VASc, the third tertile showed a 9.5-fold increased risk of mortality (HR 9.52; 95% CI: 5.15-17.60, p < 0.001). Comparing the two scores, the Alb-R₂CHA₂DS₂-VASc score (C-statistic = 0.751; 95%CI: 0.69-0.81) outperformed the R₂-CHA₂DS₂-VASc score (C-statistic = 0.736; 95%CI: 0.68-0.961) in predicting mortality (delta C-statistic = 0.015; 95%CI: 0.001-0.029). The better prediction ability of the Alb-R₂CHA₂DS₂-VASc score was also proven by an IDI of 0.024 (p < 0.0001) and a relative IDI of 24.11% (p < 0.0001), with an NRI = 0.608 (p < 0.00001).

CONCLUSIONS

The addition of albuminuria to R₂CHA₂DS₂-VASc significantly and independently predicts the risk of all-cause mortality in a sample of high CV risk patients. Moreover, Alb-R₂CHA₂DS₂-VASc outperforms R₂CHA₂DS₂-VASc.

Development of clinical prediction models for renal and cardiovascular outcomes and mortality in patients with type 2 diabetes and chronic kidney disease using time-varying predictors

AIMS

To develop a set of prediction models for end-stage kidney disease (ESKD), cardiovascular outcomes, and mortality in patients with type 2 diabetes (T2D) and chronic kidney disease (CKD) using commonly measured clinical variables.

METHODS

We studied 1432 participants with T2D and CKD enrolled in the Chronic Renal Insufficiency Cohort, followed for a median period of 7 years. We used Cox proportional-hazards models to model the six outcomes (ESKD, stroke, myocardial infarction (MI), congestive heart failure (CHF), death before ESKD, and all-cause mortality). We internally evaluated these models using concordance and calibration measures.

RESULTS

The newly developed six prediction models included 15 predictors: age at diabetes diagnosis, sex, blood pressure, body mass index, hemoglobin A1c, high density lipoprotein cholesterol, urine protein-to-creatinine ratio, estimated glomerular filtration rate, smoking status, and history of stroke, MI, CHF, ESKD, and amputation. The resulting models demonstrated good/strong discrimination (cross-validation C-index range: 0.70 to 0.90) and calibration.

CONCLUSIONS

This study provided an internally validated and useful tool for predicting individual adverse outcomes and mortality in patients with T2D and CKD. These models may inform optimal use of targeted health interventions.

Identifying high-risk hospitalised chronic kidney disease patient using electronic health records for serious illness conversation

INTRODUCTION

This study aimed to identify risk factors that are associated with increased mortality that could prompt a serious illness conversation (SIC) among patients with chronic kidney disease (CKD).

METHODS

The electronic health records of adult CKD patients admitted between August 2018 and February 2020 were retrospectively reviewed to identify CKD patients with >1 hospitalisation and length of hospital stay ≥4 days. Outcome measures were mortality and the duration of hospitalisation. We also assessed the utility of the Cohen's model to predict 6-month mortality among CKD patients.

RESULTS

A total of 442 patients (mean age 68.6 years) with median follow-up of 15.3 months were identified. The mean (standard deviation) Charlson Comorbidity Index [CCI] was 6.8±2.0 with 48.4% on chronic dialysis. The overall mortality rate until August 2020 was 36.7%. Mortality was associated with age (hazard ratio [HR] 1.51, 95% confidence interval [CI] 1.29-1.77), CCI≥7 (1.58, 1.08-2.30), lower serum albumin (1.09, 1.06-1.11), readmission within 30-day (1.96, 1.43-2.68) and CKD non-dialysis (1.52, 1.04-2.17). Subgroup analysis of the patients within first 6-month from index admission revealed longer hospitalisation stay for those who died (CKD-non dialysis: 5.5; CKD-dialysis: 8.0 versus 4 days for those survived, P<0.001). The Cohen's model demonstrated reasonable predictive ability to discriminate 6-month mortality (area under the curve 0.81, 95% CI 0.75-0.87). Only 24 (5.4%) CKD patients completed advanced care planning.

CONCLUSION

CCI, serum albumin and recent hospital readmission could identify CKD patients at higher risk of mortality who could benefit from a serious illness conversation.

Predictive Risk Models to Identify Patients at High-Risk for Severe Clinical Outcomes With Chronic Kidney Disease and Type 2 Diabetes

Introduction/Objective:

Predictive risk models identifying patients at high risk for specific outcomes may provide valuable insights to providers and payers regarding points of intervention and modifiable factors. The goal of our study was to build predictive risk models to identify patients with chronic kidney disease (CKD) and type 2 diabetes (T2D) at high risk for progression to end stage kidney disease (ESKD), mortality, and hospitalization for cardiovascular disease (CVD), cerebrovascular disease (CeVD), and heart failure (HF).

Methods:

This was a retrospective observational cohort study utilizing administrative claims data in patients with CKD (stage 3-4) and T2D aged 65 to 89 years enrolled in a Medicare Advantage Drug Prescription plan offered by Humana Inc. between 1/1/2012 and 12/31/2017. Patients were enrolled ≥1 year pre-index and followed for outcomes, including hospitalization for CVD, CeVD and HF, ESKD, and mortality, 2 years post-index. Pre-index characteristics comprising demographic, comorbidities, laboratory values, and treatment (T2D and cardiovascular) were evaluated and included in the models. LASSO technique was used to identify predictors to be retained in the final models followed by logistic regression to generate parameter estimates and model performance statistics. Inverse probability censoring weighting was used to account for varying follow-up time.

Results:

We identified 169 876 patients for inclusion. Declining estimated glomerular filtration rate (eGFR) increased the risk of hospitalization for CVD (38.6%-61.8%) and HF (2-3 times) for patients with eGFR 15 to 29 mL/min/1.73 m² compared to patients with eGFR 50 to 59 mL/min/1.73 m². Patients with urine albumin-to-creatinine ratio (UACR) ≥300 mg/g had greater chance for hospitalization for CVD (2.0 times) and HF (4.9 times), progression to ESKD (2.9 times) and all-cause mortality (2.4 times) than patients with UACR <30 mg/g. Elevated hemoglobin A1c (≥8%) increased the chances for hospitalization for CVD (21.3%), CeVD (45.4%), and death (20.6%). Among comorbidities, history of HF increased the risk for ESKD, mortality, and hospitalization for CVD, CeVD, and HF.

Conclusions:

The predictive models developed in this study could potentially be used as decision support tools for physicians and payers, and the risk scores from these models can be applied to future outcomes studies focused on patients with T2D and CKD.

Low performance of prognostic tools for predicting death before dialysis in older patients with advanced CKD

INTRODUCTION

Chronic kidney disease (CKD) is a disease which is spreading worldwide, especially among older patients. Several prognostic scores have been developed to predict death in older CKD patients, but they have not been validated. We aimed to evaluate the existing risk scores for predicting death before dialysis start, identified via an in-depth review, in a cohort of elderly patients with advanced CKD.

METHODS

We performed a review to identify scores predicting death, developed in and applicable to CKD patients. Each score was evaluated with an absolute risk calculation from the patients' baseline characteristics. We used a French prospective multicentre cohort of elderly patients (> 75 years) with advanced CKD [estimated glomerular filtration rate (eGFR) < 20 mL/min/1.73 m²], recruited from nephrological centres, with a 5-year follow-up. The outcome considered was death before initiating dialysis. Discrimination [area under curve (AUC)], calibration and Brier score were calculated for each score at its time frame.

RESULTS

Our review found 6 equations predicting death before dialysis in CKD patients. Four of these (GOLDFARB, BANSAL, GRAMS 2 and 4 years) were evaluated. The validation cohort (Parcours de Soins des Personnes Âgées Parcours de Soins des Personnes Âgées, PSPA) included 573 patients, with a median age of 82 years and a median eGFR of 13 mL/min/1.73 m². At the end of follow-up, 287 (50%) patients had started dialysis and 238 (41%) patients had died before dialysis. The four equations evaluated showed average discrimination (AUC 0.61-0.70) and, concerning calibration, a global overestimation of the risk of death.

DISCUSSION

The available scores predicting death before dialysis showed low performance among older patients with advanced CKD in a French multicentre cohort, indicating the need to upgrade them or develop new scores for this population.

Serious Illness Conversations in advanced kidney disease: a mixed-methods implementation study

OBJECTIVE

Advanced kidney disease is associated with a high risk of morbidity and mortality. Consequently, invasive treatments such as dialysis may not yield survival benefits. Advance care planning has been encouraged. However, whether such discussions are acceptable when done earlier, before end-stage kidney treatment decision-making occurs, is unclear. This pilot study aimed to explore whether use of the Serious Illness Conversation Guide to aid early advance care planning is acceptable, and to evaluate the information gained from these conversations.

METHODS

Patients with advanced kidney disease (stage 3B and above) and high mortality risk at 2 years were enrolled in this mixed-methods study from an academic nephrology clinic. Semi-structured interviews were conducted using the adapted Serious Illness Conversation Guide. Thematic analysis was used to assess patients' perceptions of the conversation. Participants completed a questionnaire assessing conversation acceptability.

RESULTS

Twenty-six patients participated, 50% were female. Participants felt that the conversation guide helped them reflect on their prognosis, goals of care and treatment preferences. Most did not feel that the conversation provoked anxiety (23/26, 88%) nor that it decreased hopefulness (24/26, 92%). Some challenges were elicited; patients expressed cognitive dissonance with the kidney disease severity due to lack of symptoms; had difficulty conceptualising their goals of care; and vocalised fear of personal failure without attempting dialysis.

CONCLUSIONS

Patients in this pilot study found the adapted Serious Illness Conversation Guide acceptable. This guide may be used with patients early in the course of advanced kidney disease to gather information for future advanced care planning.

Low performance of prognostic tools for predicting dialysis in elderly people with advanced CKD

INTRODUCTION

Clinical decision-making about care plans can be difficult for very elderly people with advanced chronic kidney disease (CKD). Current guidelines propose the use of prognostic tools predicting end stage renal disease (ESRD) to assist in a patient-centered shared decision-making approach. Our objective was to evaluate the existing risk model scores predicting ESRD, from data collected for a French prospective multicenter cohort of mainly octogenarians with advanced CKD.

METHODS

We performed a rapid review to identify the risk model scores predicting ESRD developed from CKD patient cohorts and evaluated them with data from a prospective multicenter French cohort of elderly (> 75 years) patients with advanced CKD (estimated glomerular filtration rate [eGFR] < 20 mL/min/1.75m²), followed up for 5 years. We evaluated these scores (in absolute risk) for discrimination, calibration and the Brier score. For scores using the same time frame, we made a joint calibration curve and compared areas under the curve (AUCs).

RESULTS

The PSPA cohort included 573 patients; their mean age was 83 years and their median eGFR was 13 mL/min/1.73 m². At the end of follow-up, 414 had died and 287 had started renal replacement therapy (RRT). Our rapid review found 12 scores that predicted renal replacement therapy. Five were evaluated: the TANGRI 4-variable, DRAWZ, MARKS, GRAMS, and LANDRAY scores. No score performed well in the PSPA cohort: AUCs ranged from 0.57 to 0.65, and Briers scores from 0.18 to 0.25.

CONCLUSIONS

The low predictiveness for ESRD of the scores tested in a cohort of octogenarian patients with advanced CKD underlines the need to develop new tools for this population.

Evaluation of administrative case definitions for chronic kidney disease in children

INTRODUCTION

Administrative data is increasingly used for chronic disease surveillance; however, its validity to define cases of chronic kidney disease (CKD) in children is unknown. We sought to evaluate the performance of case definitions for CKD in children.

METHODS

We utilized population-based administrative data from the Manitoba Center for Health Policy to evaluate the validity of algorithms based on a combination of hospital claims, outpatient physician visits, and pharmaceutical use over 1-3 years in children <18 years of age. Algorithms were compared with a laboratory-based definition (estimated glomerular filtration rate < 90 ml/min/1.73 m² and/or presence of proteinuria).

RESULTS

All algorithms evaluated had very low sensitivity (0.20-0.39) and moderate positive predictive value (0.52-0.68). Algorithms had excellent specificity (0.98-0.99) and negative predictive value (0.96-0.97). Receiver operating characteristic (ROC) curves indicate fair accuracy (0.60-0.68). Sensitivity improved with increasing years of data. One or more physician claims and one or more prescriptions over 3 years had the highest sensitivity and ROC.

CONCLUSIONS

The sensitivity of administrative data algorithms for CKD is unacceptably low for a screening test. Specificity is excellent; therefore, children without CKD are correctly identified. Alternate data sources are required for population-based surveillance of this important chronic disease.

Developing and testing models to predict mortality in the general population

We have previously proposed an approach using information collected from published reports to generate prediction models. The goal of this project was to validate this technique to develop and test various prediction models. A risk indicator (R) is calculated as a linear combination of the hazard ratios for the following predictors: age, male gender, diabetes, albuminuria, and either CKD, CVD or both. We developed a linear and two exponential expressions to predict the probability of the outcome of 2-year mortality and compared to actual outcome in the target dataset from NHANES. The risk indicator demonstrated good performance with area under ROC curve of 0.84. The linear and two exponential expressions generated similar predictions in the lower categories of risk indicator (R ≤ 6). However, in the groups with higher R value, the linear expression tends to predict lower, and the exponential expressions higher, probabilities than the observed outcome. A Combined model which averaged the linear and logistic expressions was shown to approximate the actual outcome data the best. A simple technique (named Woodpecker™) allows derivation functional prediction models and risk stratification tools from reports of clinical outcome studies and their application to new populations by using only summary statistics of the new population.

Prediction model for cardiovascular events or all-cause mortality in incident dialysis patients

Some variables including age, comorbidity of diabetes, and so on at dialysis initiation are associated with patient prognosis. Cardiovascular (CV) events are a major cause of death, and adequate models that predict prognosis in dialysis patients are warranted. Therefore, we created models using some variables at dialysis initiation. We used a database of 1,520 consecutive dialysis patients (median age, 70 years; 492 women [32.4%]) from a multicenter prospective cohort study. We established the primary endpoint as a composite of the incidence of first CV events or all-cause death. A multivariable Cox proportional hazard regression model was used to construct a model. We considered a complex and a simple model. We used area under the receiver operating characteristic curve (AUROC) to assess and compare the predictive performances of the prediction models and evaluated the improvement in discrimination using the complex model versus the simple model using net reclassification improvement (NRI). We then assessed integrated discrimination improvement (IDI) to evaluate improvements in average sensitivity and specificity. Of 392 deaths, 152 were CV-related. Totally, 506 CV events occurred during the follow-up period (median 1,285 days). Finally, 692 patients reached the primary endpoint. Baseline data were set at dialysis initiation. AUROC for the primary endpoint was 0.737 (95% confidence interval [CI], 0.712–0.761) in the simple model and 0.765 (95% CI, 0.741–0.788) in the complex model. There were significant intergroup differences in NRI (0.44; 95% CI, 0.34–0.53; p < 0.001) and IDI (0.02; 95% CI, 0.02–0.03; p < 0.001). We prepared a Shiny R application for each model to automatically calculate the predicted occurrence probability (https://statacademy.shinyapps.io/App_inaguma_20190717/). The complex model made more accurate predictions than the simple model. However, the intergroup difference was not significant. Hence, the simple model was more useful than the complex model. The tool was useful in a real-world clinical setting because it required only routinely available variables. Moreover, we emphasized that the tool could predict the incidence of CV events or all-cause mortality for individual patients. In the future, we must confirm its external validity in other prospective cohorts.