Timing of patient-reported renal replacement therapy planning discussions by disease severity among children and young adults with chronic kidney disease

Abstract

Background

Preparing children with chronic kidney disease (CKD) for renal replacement therapy (RRT) begins with a discussion about transplant and dialysis, but its typical timing in the course of CKD management is unclear. We aimed to describe participant-reported RRT planning discussions by CKD stage, clinical and sociodemographic characteristics, in the Chronic Kidney Disease in Children (CKiD) cohort.

Methods

Participants responded to the question “In the past year, have you discussed renal replacement therapy with your doctor or health care provider?” at annual study visits. Responses were linked to the previous year CKD risk stage based on GFR and proteinuria. Repeated measures logistic models estimated the proportion discussing RRT by stage, with modification by sex, age, race, socioeconomic status and CKD diagnosis (glomerular vs. non-glomerular).

Results

721 CKiD participants (median age= 12, 62% boys) contributed 2856 person-visits. Proportions of person-visits reporting RRT discussions increased as CKD severity increased (10% at the lowest disease stage and 87% at the highest disease stage). After controlling for CKD risk stage, rates of RRT discussions did not differ by sex, age, race and socioeconomic status.

Conclusions

Despite participant-reported RRT discussions being strongly associated with CKD severity, a substantial proportion with advanced CKD reported no discussion. While recall bias may lead to underreporting, it is still meaningful that some participants with severe CKD did not report or remember discussing RRT. Initiating RRT discussions early in the CKD course should be encouraged to foster comprehensive preparation and to align RRT selection for optimal health and patient preferences.

Introduction

Preparation for renal replacement therapy (RRT) is a necessary part of managing pediatric chronic kidney disease (CKD) as it progresses to end stage kidney disease (ESKD). Kidney transplant is the preferred treatment modality for pediatric ESKD, but dialysis is common when organs are not available or if transplant is temporarily contraindicated [1]. While effective, dialysis is associated with very low quality of life for children, near or worse than that experienced by pediatric cancer or cystic fibrosis patients [2]. Dialysis patients also have worse outcomes compared to children who receive a kidney transplant. Importantly, children have had improved access to organs since the early 2000s [2, 3]. Therefore, appropriate planning for RRT that focuses on the benefits of kidney transplantation may optimize outcomes, especially since pediatric patients receive organ priority in the US [4]. Indeed, new policy goals proposed by the National Kidney Foundation aim to improve selection of RRT to align with “patient lifestyle preferences, values and goals”[5]. While there are many steps leading to successful kidney transplantation of the patient with ESKD, the first in this process is a discussion about RRT choices with their nephrologist.

The timing and frequency of when nephrologists or healthcare providers have RRT discussions in the management of the pediatric CKD patient is unclear. Studies in adult CKD populations have highlighted the importance of early RRT planning to improve ESKD outcomes, and have documented problems related to perceived knowledge, education, complexity of CKD, lower health literacy and lack of readiness to learn. At the same time, the efficacy of interventional educational efforts have been documented [6–9]. It is likely that children with kidney diseases and their families face similar obstacles and, at least in adults, nephrologists are most well suited to correct knowledge deficits [8]. While Ricardo et al. [10] observed overall high parental health literacy among families with children with CKD, lower health literacy was associated with accelerated disease progression. In addition, RRT discussions occurring late in the course of pediatric CKD progression may contribute to a decreased likelihood of a preemptive transplant [11]. This is significant since dialysis while awaiting transplant is associated with an increased incidence and severity of comorbid conditions [12–14]. Sociodemographic differences in the incidence of RRT [15, 16] and subsequent mortality [17] have been documented (e.g., socioeconomic status [SES], race and sex) and may be due to differences in RRT planning. Whereas adult guidelines recommend referral for RRT education when the glomerular filtration rate (GFR) falls to 30 ml/min|1.73m², the extent to which these discussions occur in clinical pediatric practice across the spectrum of CKD is not well described [18].

Using data from the Chronic Kidney Disease in Children (CKiD) study, we aimed to describe the patterns of participant-reported RRT discussions by CKD severity and by clinical and sociodemographic variables to identify patient groups that may benefit from earlier RRT planning. The longstanding and longitudinal nature of CKiD provides a unique opportunity to prospectively describe patient-reported preparatory RRT discussions at annual study visits by linking disease stage at the year prior to question administration.

Methods

Study participants and data collection

The CKiD study is a prospective cohort of children enrolled between 6 months and 16 years of age with a diagnosis of CKD, a GFR < 90 ml/min|1.73m² and without a history of ESKD from 56 sites in the US and Canada. Longitudinal data were collected at annual visits to describe the participant’s CKD severity, general medical history, cardiovascular health, growth and neurocognitive development. Standardized questionnaires were administered to obtain information about their clinical experience (e.g., discussion with healthcare providers, health insurance use, etc.). Initial enrollment was between April 2005 and August 2008 (n= 586), followed by a second enrollment from February 2011 to March 2014 (n= 305), and a third enrollment wave started in October 2016 (n= 184 as of September 2019). Details of the study design and methods have been previously published [19]. All CKiD protocols and sites were approved by local Institutional Review Boards and all parents/participants provided informed consent/assent.

Measurements and data collection

CKiD participants were asked “In the past year, have you discussed renal replacement therapy with your doctor or health care provider?” at each annual visit starting in July 2008. If they responded yes, they were asked to respond whether they discussed transplant, dialysis, both, and were allowed to answer “don’t know”.

Since the outcome referred to discussions that had taken place within the previous year, the pediatric CKD risk stage at the previous visit, approximately one year before, was used as the primary exposure and this was considered the baseline visit for this analysis. The pediatric risk stage was jointly developed in the CKiD and ESCAPE cohorts [20]. Risk stages were six mutually exclusive categories (Stages A to F of increasing severity; Stage A is the least severe while Stage F is the most severe) based on estimated GFR and urine protein-creatinine ratio (UPCR). Serum creatinine-based estimated GFR was categorized as ≥90, ≥60 to <90, ≥45 to <60, ≥30 to <45, and ≥15 to <30 ml/min|1.73m². UPCR was categorized as <0.5, 0.5 to 2.0, and >2.0mg/mgCr. The specific criteria of each risk stage has been previously described (13) and is outlined in Figure 1 along with the distribution of person-visits within each category. For participants ≥ 18 years of age, the combined serum creatinine and cystatin-based estimated GFR was used since creatinine-only GFR in this population is biased towards lower values (i.e., underestimated GFR) [21].

Figure 1.

Distribution of person-visits by GFR and proteinuria (UPCR) categories, colored by risk stage (A to F), kidney disease diagnosis (dx) and proportion engaging in a renal replacement therapy (RRT) discussion with their health care provider in the past year.

Note: Distribution of person-visits, chronic kidney disease (CKD) diagnosis (dx), and proportions of engaging in a renal replacement therapy (RRT) discussion, colored by CKD risk stage (A to F). Cell coloring defined the six risk stages ordered from the least to the most severe as follows: dark green (risk stage A), light green (risk stage B), gold (risk stage C), tan (risk stage D), salmon (risk stage E), and red (risk stage F). Adapted from Furth SL, Pierce C, Hui WF, White CA, Wong CS, Schaefer F, Wühl E, Abraham AG, Warady BA, Samuels J, Furth S, Neuhaus TJ et al. Estimating Time to ESRD in Children With CKD. American Journal of Kidney Diseases 2018; 71: 783–792.

Additionally, we investigated whether selected sociodemographic or clinical variables at the previous annual visit modified the relationship between disease severity and RRT discussions. Sociodemographic variables included sex (male vs. female), age (< 18 and ≥18 years), self-reported race (African American vs. non-African American), country of residence (United States vs. Canada), and SES (low vs. high). Age at the time of question administration was dichotomized at <18 years to classify RRT discussions that occurred prior to the participant’s 18^th birthday, since transplant allocation is prioritized for patients listed prior to 18 years of age according to the Kidney Allocation System mandated by the Organ Procurement and Transplantation Network [22]. While young adults represented a small proportion of our study population, this age stratification was epidemiologically meaningful because of the health policy related to transplantation. Low SES was defined as the presence of household income < $36000, maternal education less than college, or the presence of any public insurance (among those living in the US). Age and SES were treated as time-varying variables. Since glomerular diagnoses are known to have faster disease progression than non-glomerular diagnoses [20, 23, 24], we tested whether diagnosis modified the association between CKD severity and RRT discussions.

Statistical analysis

The primary unit of analysis was person-visit in which the exposures (and modifiers) were measured one year prior to the outcome (participant’s report on RRT discussion in the past year).

Repeated measures logistic regression was used to estimate the proportion engaging in an RRT discussion in the past year within each risk stage (i.e., risk stage as a categorical predictor) by converting the log odds of the outcome to proportions, with corresponding 95% confidence intervals (95%CI). Six separate models additionally included sex, age, site, race, SES and CKD diagnosis as separate modifiers of risk stage, with interaction terms such that each variable modified the odds of having an RRT discussion within each CKD risk stage. For all models, generalized estimating equations were used to account for within-person correlation and estimate appropriate standard errors. Pairwise Wald test comparisons within each risk stage category for each sociodemographic and clinical variable were conducted with statistical significance defined as p<0.05.

To assess the association between the participant’s response at the prior visit and reporting an RRT discussion in the following visit, similar repeated measures logistic regression models were fit with the presence of an RRT discussion in the visit prior to the question as a predictor, and adjusting for risk stage. This odds ratio represented how having a prior discussion was associated with reporting an RRT discussion in the next year, and is a measure of within-person correlation. Lastly, we explored whether reporting an RRT discussion at the previous visit was different by risk stage and diagnosis.

All analyses were conducted in SAS 9.4 (SAS Institute, Cary NC) with graphs constructed using R 3.5.1 (Vienna, Austria).

Results

A total of 721 CKiD participants responded to the RRT discussion questionnaire, with a total of 2856 person-visits from July 2008 to December 2018. Table 1 describes the sociodemographic and clinical characteristics at analytic baseline which corresponded to the year prior to answering the questionnaire about RRT discussions. The study cohort comprised 61.9% boys with a median age of 12 [IQR: 8, 15] years. Most were diagnosed with non-glomerular kidney disease (72.4%), which was nearly all congenital or diagnosed at a very early age. At analytic entry, the median GFR was 56 ml/min|1.73m² [IQR: 42, 73] and median urine protein was 0.30 mg/mg Creatinine [IQR: 0.12, 0.89].

Table 1.

Baseline sociodemographic, clinical and longitudinal data characteristics of participants at the visit prior to responding to question of engaging in renal replacement therapy (RRT) discussions. Median [interquartile range] and % (n).

	Overall CKiD Cohort (n=721)
Demographic characteristics
Age	12 [8, 15]
Age < 18 years	91.1% (657)
Male	61.9% (446)
African American race	21.9% (158)
Hispanic ethnicity	13.9% (100)
Socioeconomic characteristics
Income < $36 000	23.5% (169)
Any public insurance	26.6% (191)
Maternal education < College	58.6% (420)
Attending Canadian clinical site	12.1% (87)
Kidney disease characteristics
Non-glomerular CKD diagnosis	72.4% (522)
Age at CKD onset
< 1 year	70.7% (504)
1 to 5 years	7.9% (56)
5 to 10 years	8.8% (63)
10+ years	12.6% (90)
eGFR, ml/min|1.73m2	56 [42, 73]
uPrCr, mg/mgCr	0.3 [0.12, 0.89]
Longitudinal data
Person-visits	2856
Risk stage A (lowest)	50.4% (1439)
Risk stage B	22.2% (633)
Risk stage C	10.3% (295)
Risk stage D	5.3% (151)
Risk stage E	8.7% (248)
Risk stage F (highest)	3.2% (90)

RRT discussions by CKD risk stages

Figure 1 displays the proportion of person-visits reporting an RRT discussion with their health care provider in the previous year by GFR and UPCR categories, and categories of the same color correspond to the staging used in subsequent analyses. In general, the proportion of visits with RRT discussions increased with a lower GFR, but only marginally so for increasing UPCR. Figure 2 summarizes the overall proportions of visits in which an RRT discussion took place in the previous year within each risk stage. Starting with 10% in the lowest risk group (risk stage A), those at risk stage B reported an RRT discussion for 23% of person-visits. Furthermore, 32% of person-visits reporting a discussion in the past year for risk stage C and about half (49%) reporting in risk stage D. The largest increase was transitioning from risk stage E to risk stage F: a discussion was reported to have taken place at 56% of visits in risk stage E, while 87% of participants reported RRT discussions in risk stage F (GFR between 15 and 30 ml/min|1.73m² with nephrotic range proteinuria) in which ESKD was imminent. Of note, 67% of those participants with a GFR between 15 and 30 ml/min|1.73m², regardless of proteinuria level, reported having an RRT discussion (Figure 1).

Figure 2.

Percent of person-visits (n= 2856 person-visits from 721 participants) in which participant reported a renal replacement therapy discussion with their healthcare provider in the previous year, stratified by chronic kidney disease risk stage at the previous study visit. Proportions were estimated by a repeated measures logistic regression model and corresponding 95% confidence intervals were calculated using generalized estimating equations to account for within-individual correlation of responses.

Modifiers of prevalence of RRT discussions by CKD risk stages

Figure 3 describes the proportions of person-visits in which an RRT discussion took place within each CKD risk stage by the following six characteristics: sex, age, site location, race, SES and glomerular diagnosis. There were no consistent significant differences by sex, age, race and SES. Canadian participants were more likely to have had RRT discussions for more severe CKD (risk stages C to E) compared to US participants, and this was most pronounced for risk stage C (52% vs. 30%, p= 0.017). Participants with non-glomerular diagnoses were more likely to have had a RRT discussion at risk stage C (36% vs. 19%, p= 0.015), and risk stage D (55% vs. 20% p= 0.011).

Figure 3.

Percent of person-visits in which participant reported a renal replacement therapy discussion with their healthcare provider in the previous year, stratified by chronic kidney disease risk stage at the previous study visit, and sex (A), age (B), US and Canada (C), race (D), socioeconomic status (E) and kidney disease diagnosis (F). A total of 2856 person-visits were contributed by 721 participants. Proportions and within-risk stage pairwise comparisons were estimated by a repeated measures logistic regression model and corresponding 95% confidence intervals were calculated using generalized estimating equations to account for within-individual correlation of responses.

We investigated how prior responses were related to having an RRT discussion in the past year. An RRT discussion at the previous visit was associated with a much higher odds of reporting a discussion at the following visit (OR: 7.96, 95%CI: 5.99, 10.56), after adjustment for disease stage. This association did not significantly differ by disease stage (p for interaction= 0.495) or diagnosis (p for interaction= 0.309).

Discussions of treatment modalities among those who reported an RRT discussion

Figure 4 describes type of modality discussed among those participants who reported having an RRT discussion, stratified by risk stage. The width of the color bars are proportional to those who reported having a discussion, and the stacked color bars show the proportions discussing both transplant and dialysis (purple), transplant only (pink), dialysis only (blue) and don’t know (gray). As CKD severity increased, the proportion of person-visits in which the RRT discussion addressed both transplant and dialysis increased from 39% in risk stage A to 76% in risk stage F. The proportion of participants reporting a discussion of only dialysis was less than 10% for all risk groups and ranged from 3% to 8%. At lower risk stages, the proportion not knowing what type of modality was discussed ranged from 15% to 20% for risk stages A to C and was about 1% for risk stage F.

Figure 4.

Distributions of renal replacement therapy (RRT) modalities considered among person-visits reporting discussing RRT in the previous year, stratified by chronic kidney disease risk stage (n= 2856 person-visits from 721 participants). Each risk group rectangle represents the number of person-visits responding the question, “In the past year, have you discussed renal replacement therapy with your doctor or health care provider?” and the width of the colored bars is proportional to those responding yes to that question. Stacked colors represent the proportion discussing both transplant and dialysis (purple), transplant only (pink), dialysis only (blue) and “don’t know” (gray).

Discussion

The present study sought to describe how RRT discussions occurred in the context of disease severity in this representative clinical cohort of pediatric CKD, and to identify factors that potentially modify this relationship. This study demonstrated that the likelihood of patient-reported RRT discussions with a clinician was commensurate with CKD severity. In addition, previously identified risk factors for differences between preemptive transplant or dialysis as the initial ESKD therapy, including SES [15, 25], race [16] and sex [17], were not associated with differences in the timing of patient-reported RRT discussions. It was especially encouraging that differences by race and SES were not observed in RRT discussion planning since these variables are strongly linked to differential progression and incident RRT [16, 25]. In light of the present study, these factors may be more important after planning initiation. Discussions alone are likely insufficient to minimize putative RRT differences.

This analysis showed that there were no consistent differences in the timing of RRT discussions by age. Ages less than 18 were of interest since pediatric patients registering for transplant receive priority for deceased organ allocation according to US health policy. While younger participants had a slightly higher proportion of reported discussions than older participants, these differences were relatively small and non-significant. This was particularly encouraging since those ≥18 years may have been in the process of transitioning from pediatric to adult care, with the potential for disruption in disease management and planning. It should be noted that the majority of person-visits ≥18 years (72%) were participant self-reported without parental assistance, compared to person-visits <18 years (20%). Since this may lead to reporting bias, within each risk group the proportions of parental-reported RRT planning discussions were compared to participant self-report and we observed no evidence of differences by reporting source (see Supplemental Material).

The proportions of person-visits reporting RRT discussions were higher among those with a non-glomerular diagnosis, compared to those with a glomerular diagnosis in risk stages C and D. Since non-glomerular diagnoses are largely congenital and are slower to progress than glomerular diseases, it is possible that the higher proportion of person-visits reporting RRT discussions was due to a longer duration of CKD. Supplemental Table 2 presents the distribution of years with CKD stratified by diagnoses and shows that glomerular diagnoses had a shorter disease duration than those with non-glomerular diseases. However, the average duration was still 5 to 10 years, which should provide ample time for an RRT discussion to occur. Upon investigation of the distribution of diagnoses, the overall differences by diagnosis described in Figure 2F are most likely explained by confounding. Specifically, diagnoses are related to distinct GFR and proteinuria patterns even within risk stages, and GFR is a major factor for engaging in RRT discussions (Supplemental Table 1). For example, stage C is defined by GFR between ≥30 and <45ml/min|1.73m² and proteinuria between 0.5 and 2.0 mg/mgCr, or by GFR ≥60 to <90ml/min|1.73m² and proteinuria >2.0 mg/mgCr. In this cohort, 84% of person-visits in the former group had non-glomerular diagnoses, while 79% in the latter group had glomerular diagnoses. Similarly, risk stage D comprised those with GFR ≥15 to <30ml/min|1.73m² and proteinuria <0.5 mg/mgCr (of these person-visits, 90% were non-glomerular and 10% were glomerular) and GFR ≥45 to <60ml/min|1.73m² with proteinuria >2.0 mg/mgCr (67% were non-glomerular and 33% were glomerular). The disparate disease profiles within risk stages C and D related to diagnosis suggest that those with glomerular diagnoses and higher GFR were less likely to have a discussion than non-glomerular diagnoses with lower GFR, despite being in the same risk stage.

In this analysis, participant-reported outcomes reflect what is remembered at the time of data collection. While the study did not collect physician-reported data on the same question, Salter et al. reported that among adults initiating dialysis, patient reporting bias underestimated the true prevalence of these discussions [26]. This may explain why of those in the most severe disease category, for whom ESKD was imminent, only 87% (compared to an expected 100%) reported having had a discussion regarding the planning for RRT. On the other hand, that these participants did not recall having an RRT planning discussion is troubling, as all would be expected to remember at least some planning with their clinician by this severe disease stage. While the estimates of proportions discussing RRT are potentially underestimated due to recall bias in the sense that participants did not remember a conversation that actually occurred, these estimates may be more clinically valid as they most likely reflect discussions that occurred and were remembered. Future interventional efforts should address this deficit in clinical communication and patient knowledge.

There is currently no consensus on when RRT discussions should occur according to risk stage, but our results show that in practice, the likelihood of RRT discussions occurring is commensurate to disease severity. Disease progression is heterogeneous in children with kidney diseases, but pediatric calculators designed to estimate times to ESKD in children may help guide decisions for optimal timing of RRT discussions and planning. These calculators range from estimating times to ESKD based on GFR, proteinuria and diagnosis [20], to a comprehensive calculator based on a larger panel of clinical data and longitudinal profiles [23]. Future research should also investigate the impact of the regular integration of these calculators in pediatric CKD management on preferential preemptive transplantation.

The analytic design of this study used disease severity at the prior year (i.e., longitudinally) for risk stratification since the questionnaire referred to RRT discussions that occurred during the previous year. Treating the risk as cross-sectional would lead to underestimation bias since physicians would not necessarily have had the opportunity to discuss RRT options based on their current disease severity. While our analysis was more valid to estimate how disease severity was related to RRT discussions, this meant that baseline visits were excluded from analysis since previous CKD risk stage was missing (i.e., prior to enrollment). We therefore assessed whether the cross-sectional relationship between disease severity and the prevalence of RRT discussions differed between person-visits excluded (n person-visits= 876 with no risk stage data prior to responding to RRT questionnaire) and those included in our analysis (n= 2856 person-visits). There was no statistical difference between those visits that were excluded and those that were included in our analysis (p for main effect= 0.442 and p for interaction= 0.215). The estimated proportions of visits engaging in RRT discussions from cross-sectional data were indeed lower than those estimated from longitudinal data, but the former demonstrated the same increasing prevalence of RRT discussions with higher disease severity.

The main study outcome in this analysis was based on a broad question that described one dimension of patient and family knowledge about RRT through their doctor or health care provider. While we believe this question is valuable to broadly understand the role of doctors and providers in discussing RRT, a limitation to note is that there are other sources of education and information about RRT that can be part of the planning process, including other health professionals, written materials (books or the internet) and discussions with other families dealing with pediatric kidney diseases. Future research may seek to evaluate these other sources, particularly those that are patient-initiated, as a more comprehensive overview of RRT planning in this high risk population.

There were several other limitations to this analysis. The observations from participants in the CKiD cohort study may not be generalizable to the experience of all pediatric CKD patients since the RRT discussion question was first administered about 3 years after study initiation. Participants who joined the cohort study early and progressed rapidly were not represented here. Another limitation was that only six variables were selected a priori as putative modifiers of the likelihood of having an RRT discussion and CKD severity. Other characteristics, such as participant non-adherence, may potentially be related to differences in the timing of RRT planning discussions, but were not considered here. Importantly, the targeted clinical and sociodemographic characteristics in the present analysis have been previously reported as strong modifiers of RRT modality types and RRT outcomes. We also note that all modifiers were dichotomized and this may not fully capture heterogeneity within these binary categories. We investigated the impact of dichotomizing SES category by at least two indicators of low SES, or all three indicators of low SES, and found that inferences did not differ in either case.

While it was notable that Canadian sites were identified as having a higher proportion of person-visits with RRT planning discussions within each risk stage, these inferences were from only three Canadian clinics. The observed differences may reflect site-specific practices rather than specific policy and health care differences between Canada and the US. The data were also not sufficient to investigate whether discussions at earlier disease stages were associated with a higher incidence of preemptive kidney transplant compared to dialysis. To address this, we restricted our analysis to those who received preemptive transplant or transitioned to dialysis, but this constituted only 25% of the full cohort (183/721). There was suggestive evidence that those with a preemptive transplant were more likely to have had an RRT discussion at earlier risk stages (Stage A and D; Supplemental Figure 1), but additional data are clearly needed. For the remaining 75% who have not yet transitioned to RRT, we were unable to determine how the timing of RRT planning may be related to transplant or dialysis.

For pediatric patients with CKD, the transition to RRT begins with a discussion about RRT choices. This study demonstrated that clinicians are engaging in RRT planning efforts commensurate with risk stage, although increased efforts in the highest risk stages are warranted. In this cohort, there were mostly no differences in the timing of RRT discussions by key variables typically associated with disease progression, RRT modality, and outcomes including mortality after ESKD. Clinicians should consider discussions regarding RRT planning early in the disease course. Future research should investigate patient- and family-centered preferences for kidney transplantation and dialysis, and also empirically investigate optimal timing for RRT planning discussions in terms of their potential to promote preemptive transplant as an initial RRT modality.