Temporal Trends in Decisions to Forgo KRT

Visual Abstract

Abstract

Key Points

• Approximately 6% of patients approaching kidney failure forgo dialysis, a proportion that has remained unchanged over the recent decade.

• Treatment with dialysis is more common in the United States than in other developed countries and even in not-for-profit US health systems.

• There is need to develop services, and the nephrology workforce to ensure conservative management is a viable treatment option.

Background

After decades of rise, dialysis initiation rates are now declining. In parallel, conservative kidney management has emerged as an important therapeutic alternative to dialysis. However, contemporary trends in patients opting to forgo dialysis are poorly understood.

Methods

This retrospective cohort study included adults with advanced kidney disease as defined by ≥2 outpatient measures of eGFR ≤15 ml/min per 1.73 m² for ≥90 days, who were receiving care in a large US health system between 2012 and 2020. Patients were followed through death or December 31, 2022. Administrative codes were used to identify patients who received KRT during follow-up. For remaining patients, documentation in their clinical progress notes was reviewed to determine the KRT decision made closest to the end of follow-up. Descriptive analyses evaluated trends in the distribution of treatment decisions made in 3-year periods (2012–2014, 2015–2017, and 2018–2020) and stratified by age group (<65, 65–74, 75–85, and ≥85 years).

Results

Among the 12,678 patients in the cohort (age 66±15 years, 49% non-Hispanic White, 52% male), 90% received KRT on the basis of either administrative codes or chart documentation, 4% had been discussing and/or preparing for KRT, and 6% decided against KRT over a median follow-up of 3 (1–5) years. The distribution of KRT decisions did not meaningfully change over time. The proportion of patients in 2012–2014, 2015–2017, and 2018–2020 who went on to be treated with KRT was 90%, 91%, and 90%, respectively. Across the same periods, 7%, 6%, and 6% of patients had forgone KRT, respectively. The distribution of KRT decisions remained stable for all age groups over time.

Conclusions

Forgoing KRT is uncommon. Over the recent decade, most patients approaching kidney failure receive or are preparing to receive KRT, underscoring its continued use as a default approach to kidney failure care.

Introduction

The United States National Kidney Registry indicates that, after decades of rise, the incidence rate of patients starting maintenance dialysis has been decreasing.¹ This decline in the incidence rate of dialysis has been greatest among adults aged 75 years or older in whom the benefits of dialysis toward prolonging life and maintaining function are most controversial.¹ Initiation of dialysis in older patients is associated with increased health care utilization and institutionalization and loss of independence.^2–4 Significant comorbid burden and frailty also severely diminish the survival benefits of dialysis initiation.^5,6 In light of the substantial trade-offs in longevity and quality of life with undergoing dialysis, it is increasingly recognized that not all patients who develop kidney failure choose to undergo dialysis or benefit from starting this treatment. Concurrently, dedicated conservative kidney management services have emerged through nephrology training programs and/or specialized clinicians with expertise in caring for patients with kidney failure without dialysis to enable a decision to forgo dialysis as a viable option.^7–15

Because our country's national kidney registry captures only patients treated with KRT (i.e., dialysis and kidney transplantation), little is known about the frequency with which patients forgo KRT. Understanding of patients with kidney failure who forgo dialysis has been largely limited to clinician self-reported practices. National survey studies of clinicians suggest approximately 5% of patients with kidney failure forgo dialysis^16,17; however, estimates are subject to recall bias. In a national study conducted in the Department of Veterans Affairs (VA) that used clinician documentation in clinical progress notes to identify treatment decisions made for kidney failure, it was found that a decision not to pursue dialysis was made in one in seven patients who had reached an eGFR <15 ml/min per 1.73 m². This study was restricted to veterans who are predominantly male, White, and older adults, and may not reflect KRT preferences in the general population. No prior studies have examined temporal trends in treatment preferences toward dialysis.

Methods

Study Design

This retrospective cohort study included adults with advanced kidney disease from the Center for Kidney Disease Research, Education, and Hope (CURE-CKD) Registry. The CURE-CKD Registry contains clinical data derived from electronic health records at Providence and University of California Los Angeles health systems serving patients in rural and urban areas in five western US states (Washington, Alaska, Montana, Oregon, and California). Development of the CURE-CKD Registry has been previously described in detail.^18,19 Briefly, the CURE-CKD Registry was established in 2015 as a collaboration between Providence and University of California Los Angeles health systems to form an observational real-world cohort for study of patients with or at risk for CKD on the basis of a combination of laboratory data, administrative encounter data, vital signs, and medication records from inpatient, outpatient, and ambulatory settings. The CURE-CKD Registry contains electronic health record data that is updated annually to provide ongoing contemporaneous data for longitudinal studies.

This study is limited to patients treated at Providence health system for whom clinical chart notes were available for the current analysis. This study was approved by the Providence Institutional Review Board, which determined that written informed consent was not required for this limited dataset.

Study Population

Adults aged 18 years or older with advanced kidney disease as defined as ≥2 outpatient measures of eGFR <15 ml/min per 1.73 m² for ≥90 days between 2012 and 2020 were included. Cohort entry was the date of the second eGFR <15 ml/min per 1.73 m². Patients were followed through death or December 31, 2022, whichever came first. Exclusion criteria were an administrative code for KRT before cohort entry (Supplemental Table 1) or insufficient documentation in clinical progress notes to determine the KRT decision (Figure 1).

Figure 1.

Study design and cohort selection of patients with advanced kidney disease for the determination of kidney failure treatment decisions.

KRT Decisions

Using previously described methods,^20,21 we used a combination of diagnostic code search of administrative files and deductive content analysis of documentation in each patient's electronic health records to determine the treatment decision for their kidney failure during follow-up. First, all patients treated with KRT during follow-up on the basis of administrative code search for KRT were identified. For remaining patients, clinical progress notes from cohort entry through end of follow-up were reviewed to classify patients into three predefined, mutually exclusive treatment groups reflecting the KRT decision made closest to the end of follow-up: (1) Received KRT: Patients had documentation in notes that they received KRT; (2) Discussing/Preparing for KRT: Patients appeared in notes to be discussing and/or preparing for KRT; and, (3) Decision against KRT: Note documentation indicated that patients, families, and/or their care team made a decision to forgo KRT.

Lucene text-searching software^22–24 was used to facilitate chart review and categorization of patients into the respective treatment groups. A combination of inclusive search terms (such as “dialysis” and “kidney failure”) were used to ensure completeness of the chart review. Specific search terms and phrases targeting patients who receive dialysis and those who forgo dialysis assembled from prior related studies were used to augment identification of patients in each treatment group (Supplemental Table 2).^20,25

Patient Characteristics

Demographic information, including age, sex, race and ethnicity, and insurance status at cohort entry, were obtained from administrative records. Clinical information from laboratory and vital sign records (eGFR, outpatient systolic BP, body mass indices, and hemoglobin A1c measures) and health care utilization at Providence from administrative records (hospitalizations, outpatient visits, and outpatient nephrology care) were aggregated over 1 year before cohort entry. Comorbidities at baseline were identified using diagnostic codes at any encounter before cohort entry. Identification of individuals with hypertension and diabetes was supplemented by using a combination of patient-level laboratory values, vital signs, and prescription medications before cohort entry.¹⁸ Prescribed classes of medications were identified for any encounter before cohort entry with a concurrent prescription record.

Data Analysis

Characteristics of patients were reported at cohort entry. Frequency distributions were reported for categorical variables, the mean and SD were reported for normally distributed continuous variables, and the median and interquartile range were reported for skewed continuous variables. The distribution of KRT decisions were calculated during follow-up and stratified by 3-year periods (2012–2014, 2015–2017, and 2018–2020). Because decisions to forgo dialysis are more common at older ages,²⁰ analyses were further stratified by age group (<65, 65–74, 75–85, and ≥85 years). Only small temporal differences in KRT treatment decisions that were unlikely to be clinically significant were found upon preliminary inspection of descriptive statistics. Thus, to avoid type 1 errors, preplanned statistical analyses of differences were not performed.²⁶ All analyses were performed using R version 4.1.3 (R core team, 2022).

Results

A total of 12,678 individuals with advanced kidney failure were included in the analysis. For baseline characteristics, the largest proportions were patients of non-Hispanic White race (49%, n=6236) and male sex (52%, n=6533) and those with common comorbidities of diabetes (56%, n=7136) and hypertension (57%, n=7165; Table 1). The mean age of the cohort was 66±15 years. Only 21% (n=2685) of patients were receiving nephrology care. Hospitalization in the year before cohort entry was common (48%, n=6060).

Table 1.

Baseline characteristics of those with advanced kidney disease by time

Patient Characteristics	2012–2020 (n=12,678)	2012–2014 (n=2905)	2015–2017 (n=4625)	2018–2020 (n=5148)
Demographics
Sex, n (%)
Male	6533 (52)	1413 (49)	2402 (52)	2718 (53)
Female	6145 (48)	1492 (51)	2223 (48)	2430 (47)
Race and ethnicity, n (%)
American Indian or Alaska native	146 (1)	33 (1)	56 (1)	57 (1)
Asian	1018 (8)	244 (8)	363 (8)	411 (8)
Black or African American	1093 (9)	284 (10)	378 (8)	431 (8)
Hispanic or Latino/a	1635 (13)	199 (7)	673 (15)	763 (15)
Native Hawaiian or pacific Islander	224 (2)	55 (2)	79 (2)	90 (2)
Other a	2326 (18)	508 (18)	829 (18)	989 (19)
White	6236 (49)	1582 (55)	2247 (49)	2407 (47)
Primary health insurance, n (%)
Commercial	1137 (9)	267 (9)	414 (9)	456 (9)
Medicaid	365 (3)	88 (3)	120 (3)	157 (3)
Medicare	4212 (33)	1109 (38)	1491 (32)	1612 (31)
Unknown	6964 (55)	1441 (50)	2600 (56)	2923 (57)
Age, mean±SD, yr	66±15	66±16	65±16	66±15
Age category, yr, n (%)
<65	5644 (45)	1264 (44)	2106 (46)	2274 (44)
65–74	3187 (25)	723 (25)	1138 (25)	1326 (26)
75–84	2615 (21)	606 (21)	918 (20)	1091 (21)
≥85	1232 (10)	312 (11)	463 (10)	457 (9)
Clinical
eGFR, n (%)	12,678 (100)	2905 (100)	4625 (100)	5148 (100)
mean ± SD, ml/min /1.73 m 2	13±6	13±6	12±7	13±6
Systolic BP, outpatient, n (%)	8906 (70)	1593 (55)	3351 (73)	3962 (77)
mean±SD, mm Hg	140±20	139±21	139±20	141±20
BMI, n (%)	7970 (63)	1336 (46)	3122 (68)	3512 (68)
mean±SD, kg/m 2	30±7	30±7	30±7	30±7
HbA1c, n (%)	5322 (42)	1082 (37)	1995 (43)	2245 (44)
mean±SD, %	6.8 ± 1.6	6.8 ± 2	6.8 ± 2	6.9 ± 2
Medications, n (%)
ACE inhibitor or ARB	4264 (34)	852 (29)	1461 (32)	1951 (38)
Antiplatelets	1075 (8)	198 (7)	329 (7)	548 (11)
β blockers	5148 (41)	1105 (38)	1781 (39)	2262 (44)
Cholesterol	4467 (35)	907 (31)	1507 (33)	2053 (40)
Diabetes	4014 (32)	789 (27)	1368 (30)	1857 (36)
Diuretic	4616 (36)	986 (34)	1595 (35)	2035 (40)
GLP-1 receptor agonists	205 (2)	22 (1)	39 (1)	144 (3)
Potassium resin	1042 (8)	179 (6)	379 (8)	484 (9)
Opioids	4 (0)	0 (0)	3 (0.1)	1 (0.02)
SGLT-2 inhibitors	23 (0)	0 (0)	2 (0.04)	21 (0.4)
Comorbidities, n (%)
Alcohol or drug abuse	354 (3)	39 (1)	119 (3)	196 (4)
Autoimmune conditions	307 (2)	65 (2)	101 (2)	141 (3)
Cancer	853 (7)	172 (6)	304 (7)	377 (7)
Cerebrovascular disease	793 (6)	139 (5)	284 (6)	370 (7)
Chronic pulmonary disease	856 (7)	166 (6)	303 (7)	387 (8)
Dementia	240 (2)	56 (2)	89 (2)	95 (2)
Depression	777 (6)	136 (5)	280 (6)	361 (7)
Diabetes mellitus	7136 (56)	1583 (55)	2419 (52)	3134 (61)
Heart failure	1603 (13)	330 (11)	567 (12)	706 (14)
Hypertension	7165 (57)	1296 (45)	2593 (56)	3276 (64)
Liver disease	336 (3)	42 (1)	124 (3)	170 (3)
Neurological disorders	665 (5)	99 (3)	234 (5)	332 (6)
Obesity	880 (7)	138 (5)	290 (6)	452 (9)
Pulmonary circulation disease	268 (2)	38 (1)	99 (2)	131 (3)
Valvular disease	619 (5)	119 (4)	209 (5)	291 (6)
Health care utilization
Hospitalization, n (%)	6060 (48)	1365 (47)	2201 (48)	2494 (48)
Outpatient visits, median (IQR)	10 (4–21)	8 (4–18)	12 (5–21)	10 (4–22)
Nephrology care, n (%)	2685 (21)	546 (19)	910 (20)	1229 (24)
Follow-up
Yr, median (IQR)	3 (1–5)	4 (1–7)	4 (1–5)	2 (1–3)
Deaths, n (%)	4684 (37)	1347 (46)	1805 (39)	1532 (30)

ACE, angiotensin-converting enzyme; ARB, angiotensin receptor blocker; BMI, body mass index; GLP-1, glucagon-like peptide 1; Hb, hemoglobin; IQR, interquartile range; SGLT-2, sodium-glucose cotransporter 2.

^aIncludes persons who did not identify with main census categories or whose race is unknown.

Although the number of patients approaching kidney failure increased between 2012 and 2020, the distribution of baseline demographic characteristics remained similar over the same period. There were modest increases (<10%) in the receipt of nephrology care, use of prescription medications, and diagnosis of other comorbidities over study periods, except for the diagnosis of hypertension, which changed from 45% (n=1296) in 2012–2014 to 64% (n=3276) in 2018–2020. The degree of control of major risk factors for kidney failure, including mean outpatient systolic BP, body mass index, and hemoglobin A1c, was largely unchanged over time.

Over a median follow-up of 3 (1–5) years, 90% (n=11,398) received KRT based on either administrative codes (55%, n=6931) or chart documentation (35%, n=4467). An additional 4% (n=505) had been discussing/preparing for KRT at their latest follow-up. Only 6% (n=775) of patients decided against KRT, although this decision was more common among patients between age 75 and 84 years (10%, n=264) and 85 years of older (27%, n=336). Example documentation used to categorize patients into their respective treatment groups is provided in Supplemental Table 3.

The distribution of KRT decisions did not meaningfully change over time. The proportion of patients who had advanced kidney disease in 2012–2014 and went on to be treated with KRT during follow-up was 89% (n=2599), 90% (n=4184) in 2015–2017, and 90% (n=4615) in 2018–2020. The proportion of patients who had forgone KRT was 7% (n=189), 6% (n=268), and 6% (n=318) across the same periods. The distribution of KRT decisions remained stable for all age groups over time (Figure 2).

Figure 2.

Distribution of treatment decisions for KRT over time and by age group. Blue represents patients who received KRT, yellow represents patients who were discussion or preparing for KRT, and gray represents patients who decided against KRT.

Discussion

In this large, not-for-profit US health system, the overwhelming majority were treated with or preparing for KRT, signaling that KRT is the default approach to treating kidney failure. Just one of 17 people with advanced kidney disease decided to forgo KRT, irrespective of age. The distribution of decisions with respect to KRT also did not change between 2012 and 2020, despite increase in the number of patients developing kidney failure, their comorbidity burden and their access to nephrology care over time.

The decision to forgo dialysis among this cohort (6%) was much less common than that reported in the US VA (15%) and by cohorts in other high-income countries, such as the United Kingdom (18%) and Australia (14%).^20,27–29 These differences are more striking at older ages (Figure 3). In the current cohort, 16% of patients aged 75 years or older had forgone KRT as compared with 28% of similarly aged patients in the VA and 40% in the UK. Among patients aged 85 years or older in the present cohort, 27% had forgone KRT versus 47% in the VA. These findings are consistent with the observation that dialysis practices are more liberal in the United States than in other developed countries,²⁰ private versus government-run health systems,^30,31 and even for not-for-profit health systems.

Figure 3.

Decisions to forgo maintenance dialysis among patients with eGFR <15 ml/min per 1.73 m² across the United States, United Kingdom, and Australia. The light blue bars represent the Providence health system data described in this study, the yellow bars represent the VA population, the green bars represent the United Kingdom population, and the dark blue bars represent the Australian population. AUS, Australia; UK, United Kingdom; VA, Department of Veterans Affairs.

Although financial incentives may play a role,³² the reasons for the observed differences in treatment rates with KRT across these settings are most likely multifactorial. Although Providence and VA share characteristics as large, nonprofit health systems, the VA's integrated health services and prioritization of in-home and palliative care services through all of its medical centers may reduce barriers to conservative management of kidney failure as compared with non-VA settings.^33–35 Furthermore, there is a relative paucity of conservative management pathways to care for patients who forgo KRT in the US as compared with some international settings, positioning dialysis initiation as often the sole treatment choice for kidney failure. In addition to system-level differences, there is likely variation in clinician practices and patient support. Conservative kidney management is not consistently presented as an option to patients by clinicians nor is it a regular topic in kidney disease patient education.^36–38 Nephrology clinicians receive minimal training and have very little professional experience with caring for patients who forgo KRT to confidently offer this therapeutic approach to patients or support patients who make this choice.^39–41 Thus, it is not surprising that clinicians are often reluctant to honor patients' wishes to forgo KRT and manage their patients with the expectation that patients will change their minds and start dialysis when patients become sick.^24,41 Similarly, patients are least familiar with conservative management of all the treatment options for kidney failure.⁴² Decisions to forgo dialysis are also stigmatized as giving up or no care.^24,36 Clinicians who proactively support patients in this decision are made to feel as though they are practicing outside of medical norms,^9,24,36 and patients are made to feel as though they might be abandoned by their health care professionals if they were to make this choice.^24,43 Collectively, our findings underscore the need to build the cultural and clinical infrastructure to proactively support conservative kidney management as a viable therapeutic option for patients in this country.

Although the prevalence of decisions to forgo KRT did not change over time for patients in this health system, there was a 59% increase in the number of prevalent cases of patients who had forgone KRT from 189 cases in 2012–2014 to 318 cases in 2018–2020. This finding highlights how health systems must not only contend with the growing needs of patients with advanced kidney care but also ensure a workforce and clinical services capable of caring for patients who opt out of KRT. This likely means improving training of nephrology clinicians in the care of patients who forgo KRT and the availability of dedicated conservative management pathways to support this treatment choice and optimize outcomes. Available data from other countries indicate that patients who forgo dialysis and are cared for by nephrology clinicians might survive longer than those cared for in non-nephrology settings.⁴⁴ Patients cared for in dedicated conservative management pathways or by dedicated clinicians delivering this care might also experience fewer hospitalizations and spend fewer days in a hospital than those cared for in usual nephrology settings.⁴⁵

Despite endorsement of conservative kidney management by professional societies and clinical practice guidelines,^46,47 there is a striking lack of information systems to monitor its practice, quality, costs, and outcomes. Similar to other countries,⁴⁸ our national kidney failure registry does not systematically collect information on patients with kidney failure who forgo KRT, limiting efforts to support delivery of high-quality conservative kidney management and shared decision-making about KRT. While a number of systematic reviews and meta-analyses have been attempted to synthesize evidence on conservative kidney management,^49–52 the absence of clinical trial data, heterogeneity of populations and approaches to treatment, and residual confounding and bias have limited the medical community in drawing robust conclusions about how best to approach caring for these patients. Electronic health records coupled with text search software, such as those used in this study, offer a comprehensive means of capturing population-level and system-level information on treatment practices for kidney failure,²⁰ addressing gaps left by a lack of a unified approach to epidemiological surveillance. The findings of other studies suggest substantial disparities in access to quality end-of-life care among patients who forgo dialysis as compared with other patients with serious illness.^49,53 Developing lasting, population-wide and system-wide data collection systems on conservative kidney management that are comparable with those for KRT is a step forward in driving development of quality metrics, practice improvement, and sound policy.

This study has several limitations. First, although the study population spanned a large geographic area in the United States, it was limited to a single health system and may not be generalizable to those receiving care in other health systems. Second, differences in study design argue for caution in comparing rates of decisions to forgo KRT in this study with those reported in prior studies conducted in and outside the United States. The proportion of patients not treated with KRT is likely sensitive to differences in exclusion and inclusion criteria across studies. While we adopted a rigorous approach to capture episodes of KRT using administrative data, linkage to national kidney registry data was not performed, and thus, our results may underestimate the proportion of the cohort who received dialysis. In a related point, although much less uncommon,¹ decisions to start or forgo KRT can occur before patients reach an eGFR <15 ml/min per 1.73 m² and were not accounted for in the current analysis. Third, categorization of patients into treatment groups was based on documentation in the Providence electronic health record and might not fully or accurately capture treatment decisions occurring in other health systems. Documentation is also based on clinicians' perception of patients' treatment preferences, which might not fully or precisely capture patients' attitudes toward dialysis, the adequacy of patients' understanding of their treatment options, and their engagement in decision making. Fourth, preferences surrounding KRT evolve and can vacillate over time.²⁵ Thus, categorization of treatment decisions for KRT was based on decisions documented at most recent follow-up and may not capture final decisions or future changes in treatment preferences of some patients.

Among this large cohort of US patients with advanced kidney disease receiving care in a not-for-profit health system, treatment decisions for kidney failure with respect to KRT did not change substantially over time. It was much less common for patients in this cohort to forgo KRT as compared with similar patients in government-run health systems in the United States and abroad. Nonetheless, the number of incident cases of patients who forgo KRT has been increasing over time and highlights a need for further research on population-level health initiatives targeted at developing care pathways and the nephrology workforce to meet the needs of this population.

Supplementary Material

kidney360-7-2008-s001.pdf ^{(2.7MB, pdf)}

kidney360-7-2008-s002.pdf ^{(197.7KB, pdf)}

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/KN9/B676.

Author Contributions

Conceptualization: Kenn B. Daratha, Cami R. Jones, Lindsey M. Kornowske, Christina L. Reynolds, Katherine R. Tuttle, Susan P.Y. Wong.

Data curation: Cami R. Jones, Lindsey M. Kornowske, Christina L. Reynolds.

Formal analysis: Kenn B. Daratha, Cami R. Jones, Lindsey M. Kornowske, Ryan Laundry, Christina L. Reynolds, Katherine R. Tuttle, Susan P.Y. Wong.

Funding acquisition: Susan P.Y. Wong.

Investigation: Cami R. Jones, Lindsey M. Kornowske, Christina L. Reynolds, Katherine R. Tuttle, Susan P.Y. Wong.

Methodology: Cami R. Jones, Lindsey M. Kornowske, Ryan Laundry, Christina L. Reynolds.

Project administration: Cami R. Jones.

Resources: Cami R. Jones.

Software: Lindsey M. Kornowske, Ryan Laundry, Christina L. Reynolds.

Supervision: Kenn B. Daratha, Katherine R. Tuttle.

Writing – original draft: Susan P.Y. Wong.

Writing – review & editing: Kenn B. Daratha, Cami R. Jones, Lindsey M. Kornowske, Katherine R. Tuttle, Susan P.Y. Wong.

Funding

S.P.Y. Wong: Doris Duke Charitable Foundation (Clinical Scientist Development Award).

Declarative Statements

This study is exempt from the Institutional Review Board or Ethics Committee approval, and I have provided the reason for exemption. IRB deemed not necessary to obtain informed consent for analysis of limited dataset.

Data Availability Statements

Original data generated for the study will be made available upon reasonable request to the corresponding author. Data Type: Aggregated Data. Reason for Restricted Access: The data that support the findings of this study are available upon request from the corresponding author. These data are not publicly available due to privacy or ethical restrictions and would require a data use agreement.

Supplemental Material

This article contains supplemental material online, published as provided by the authors, at http://links.lww.com/KN9/B677.

Supplemental Table 1. International Classification of Diseases-10 administrative codes for the identification of KRT.

Supplemental Table 2. Search terms used for chart review.

Supplemental Table 3. Exemplar quotation from the clinical progress notes of patients in each treatment decision group.