Nephrology Provider Prognostic Perceptions and Care Delivered to Older Adults with Advanced Kidney Disease

Abstract

Background and objectives 

Prognostic uncertainty is one barrier that impedes providers in engaging patients with CKD in shared decision making and advance care planning. The surprise question has been shown to identify patients at increased risk of dying.

Design, setting, participants, & measurements 

In our prospective observational study, 488 patients ≥60 years of age with CKD stage 4 or 5 were enrolled. Binary surprise question (i.e., “Would you be surprised if this patient died in the next 12 months?”) responses were recorded, and dialysis planning preferences, presence of advance care planning documentation, and care preceding death were abstracted.

Results 

The median patient age was 71 (65–77) years old. Providers responded no and yes to the surprise question for 171 (35%) and 317 (65%) patients, respectively. Median follow-up was 1.9 (1.5–2.1) years, during which 18% of patients died (33% of surprise question no, 10% of surprise question yes; P<0.001). In patients with a known RRT preference (58%), 13% of surprise question no participants had a preference for conservative management (versus 2% of yes counterparts; P<0.001). A medical order (i.e., physician order for life-sustaining treatment) was documented in 13% of surprise question no patients versus 5% of yes patients (P=0.004). Among surprise question no decedents, 41% died at home or hospice, 38% used hospice services, and 54% were hospitalized in the month before death. In surprise question yes decedents, 39% died at home or hospice (P=0.90 versus no), 26% used hospice services (P=0.50 versus no), and 67% were hospitalized in the month before death (P=0.40 versus surprise question no).

Conclusions 

Nephrologists’ prognostic perceptions were associated with modest changes in care, highlighting a critical gap in conservative management discussions, advance care planning, and end of life care among older adults with CKD stages 4 and 5 and high-risk clinical characteristics.

Podcast 

This article contains a podcast at https://www.asn-online.org/media/podcast/CJASN/2017_09_18_CJASNPodcast_17_11.mp3

Introduction

The prevalence of CKD among the elderly is high (¹,²). Older adults with CKD are more likely to die than to progress to ESRD (²,³). However, nephrology care for many of these patients is focused on slowing CKD progression and in those who progress, preparing for RRT. For this latter group, less time is dedicated to engaging patients in shared decision making about the tradeoffs of dialysis versus conservative management or in advance care planning discussions (⁴,⁵). Although such conversations could help align care delivery with patients’ values and preferences (⁶,⁷), it is unclear how often providers alter care for patients with recognizably limited prognoses. These observations are of particular importance, because patients with advanced CKD report limited discussions of treatment options and prognosis (⁶).

The surprise question (SQ; i.e., “Would you be surprised if this patient died in the next 12 months?”) has been proposed as a pragmatic tool to help providers identify patients who could benefit from more prudent consideration of the value of interventions and timely advance care planning (^8–10). In a previous interim analysis, we have shown that the SQ seems to be a useful and reliable tool for identifying older adults with nondialysis-dependent CKD stages 4 and 5 who are at increased risk of dying (¹¹). We hypothesized that surprise question no (SQ-No) patients would be more likely to receive conservative management, have an advance directive or other evidence of advance care planning, and die at home or inpatient hospice and with hospice services than their surprise question yes (SQ-Yes) counterparts. To test these hypotheses, we collected SQ responses from nephrology providers and examined how subsequent usual standard of care differed among older adults with nondialysis-dependent CKD stages 4 and 5 attending an outpatient nephrology clinic.

Materials and Methods

Study Setting and Participants

As part of an ongoing, prospective observational study examining mortality in older (age ≥60 years old) adults with nondialysis-dependent CKD stages 4 and 5, we enrolled 488 participants (i.e., all eligible patients) from the nephrology clinic at Vanderbilt University Medical Center between June of 2014 and February of 2016 and followed them through November of 2016 (¹¹). The clinic has 13 attending physicians and one advanced practitioner, and it serves over 3000 unique patients with CKD per year. As described in a prior interim analysis (¹¹), patients were excluded for dialysis dependence, prior kidney transplant, initial visit with a provider, or possible AKI. Because the study included a provider-based subjective health measure, we excluded first visits to allow providers to gain familiarity with their patients.

We obtained written informed consent from all nephrology providers with a nephrology clinic at the local site to participate in the study. The Vanderbilt University Medical Center Institutional Review Board approved this study, including a waiver of patient informed consent (140402).

SQ Assessment

We captured provider responses to the SQ immediately after each eligible patient visit using a paper form (¹¹). SQ responses were stored outside the legal medical record and were not shared with other providers, and there was no intervention associated with SQ responses.

Covariates

We performed manual chart reviews and electronic health record (EHR) structured data abstractions to obtain baseline variables, including sociodemographics (i.e., age, sex, race, marital status, and insurance status), comorbidities (i.e., diabetes, hypertension, coronary artery disease, congestive heart failure, peripheral vascular disease, and malignancy), and common clinical measurements (i.e., body mass index, serum creatinine, eGFR, serum albumin, hemoglobin, and urine protein-to-creatinine ratio) within the prior 12 months. The local EHR also allows for key word searches, which we used to augment our findings using a published list of search terms (¹²) that we supplemented (¹¹).

Outcomes

Two physicians manually abstracted each chart to gather data and adjudicate outcomes of interest. Disagreements were discussed and adjudicated by consensus with the help of a third physician. If outcome information was missing, it was recorded as unknown and reported as such. Our primary outcomes were RRT planning, advance care planning, and end of life care characteristics.

RRT Planning.

We searched for documentation of RRT preferences in the EHR and abstracted the earliest documented modality preference and date of documentation. RRT planning options included conservative management, hemodialysis, peritoneal dialysis, other, and unknown (i.e., no clear preference documented and/or patient undecided). Conservative management was defined as a documented wish to forego dialytic therapies while pursuing other treatments (e.g., pharmacotherapy, dietary restriction, symptom relief, etc.) for uremia when the need arose (¹³). Only two patients had RRT preferences of other, and they were excluded from the RRT planning analyses. We continued to update patients’ research records every 4 months to capture updates and changes to RRT preferences. For this study, we only considered changes in RRT preferences from peritoneal dialysis/hemodialysis to conservative management or vice versa. No such changes were documented during the follow-up reported here.

We captured the presence and date of dialysis access placement on the basis of recurring EHR reviews. We recorded the earliest access placed. In addition, we recorded RRT initiation on the basis of EHR review and recorded the modality and date of initiation.

Advance Care Planning.

During each nephrology office visit, clinic intake personnel routinely asked the patient whether she/he had an advance directive, including documentation of a medical order (i.e., physician orders for scope of treatment or similar), durable power of attorney for health care (DPOA), or living will, or if she/he would like to update any of those respective forms. Clinic personnel recorded patient responses in the EHR, and the respective forms were scanned and stored in a dedicated section of the EHR. Study team members performed manual chart reviews to capture patient responses and examine the contents of any scanned advance care planning documents. To improve ascertainment, we also searched routine clinical documentation (e.g., outpatient notes, inpatient progress notes, etc.) for evidence that advance care planning forms existed.

For medical orders, we recorded the specialty of the documenting provider, whether the form was completed in an ambulatory or inpatient setting, the presence of modifications over time, and the preferences documented (e.g., comfort care, do not resuscitate, do not intubate, etc.). Because of the small number of medical orders and preferences, we collapsed values into the following categories: treatment limitations (i.e., do not resuscitate, etc.) and nonlimited/aggressive care (i.e., no limitation specified). For living wills/DPOA, we recorded their presence and their date of completion.

Death and End of Life Care.

As previously described (¹¹), vital status and date of death were determined through a multistep process. First, we linked with the Social Security Death Master File. Second, we performed EHR chart review and obituary searches. Third, we linked to the Tennessee Department of Health Vital Records Office (https://www.tn.gov/assets/entities/health/attachments/ph-3055.pdf). For decedents, we reviewed obituaries and EHR documentation to determine location of death (i.e., home, hospice, hospital/rehabilitation center, or nursing home), hospitalization within 30 days of death, use and timing of hospice services, and use and timing of nursing home or skilled nursing facility services (Supplemental Material).

Statistical Analyses

Patient characteristics were summarized by SQ-Yes, SQ-No, and the combined cohort. Continuous variables were summarized as medians with interquartile ranges, and categorical variables were summarized as frequencies and percentages. Univariate comparisons of categorical baseline variables between patients with SQ-Yes and patients with SQ-No were calculated with a Pearson chi-squared test, whereas comparison of continuous variables was calculated with the Wilcoxon rank sum test. Analyses related to RRT were repeated while restricting to patients with a Modification of Diet in Renal Disease eGFR ≤20 ml/min per 1.73 m². Logistic regression was used to model presence of a documented advance care plan (i.e., medical order, living will, or DPOA) with the following a priori variables of interest: age, race, sex, and SQ response (^14–16). Age was included in the model as a restricted cubic spline to allow for potential nonlinear effects. The association of the SQ response and death was examined using Cox proportional hazards regression analysis to obtain the hazard ratio adjusted for age, sex, and race at the end of the follow-up period (November of 2016) (¹¹). To allow for a potentially nonlinear effect of age, the variable was included in the model as a restricted cubic spline with three knots. Statistical analysis was performed using R, version 3.3.0 (http://www.r-project.org), and two-sided P values <0.05 were considered statistically significant.

Results

Among 488 participants, providers responded SQ-No and SQ-Yes for 171 (35%) and 317 (65%) patients, respectively. The median patient age was 71 (65–77) years old, 51% were women, and 17% were black. As expected, the cohort had a heavy comorbidity burden, with a median Charlson Comorbidity Index score of five (four to seven), and the median eGFR was 22 (16–26) ml/min per 1.73 m² (Table 1). Generally, patients with SQ-No were older, had higher comorbidity scores, and had lower eGFR, serum albumin, and hemoglobin values (Table 1). During a median follow-up of 1.9 years (interquartile range, 1.5–2.1), 87 (18%) patients died, including 33% of patients with SQ-No versus 10% of patients with SQ-Yes (P<0.001). In the age, sex, and race Cox proportional hazards–adjusted analysis, an SQ-No response was associated with a hazard ratio for death of 3.7 (95% confidence interval [95% CI], 2.4 to 5.8).

Table 1. - Baseline characteristics

Characteristic	Total Cohort, n=488	SQ Yes, n=317	SQ No, n=171	P Value
Age, yr	71 (65–77)	69 (65–75)	74 (68–80)	<0.001
Women	249 (51)	164 (52)	85 (50)	0.70
Race				0.80
White	404 (83)	261 (83)	143 (84)
Black	82 (17)	54 (17)	28 (16)
Marital status				0.07
Married	301 (63)	197 (63)	104 (62)
Divorced/separated	53 (11)	42 (13)	11 (7)
Widow/widower	105 (22)	61 (19)	44 (26)
Single	21 (4)	13 (4)	8 (5)
Insurance				0.06
Medicare	397 (81)	248 (78)	149 (87)
Private	87 (18)	66 (21)	21 (12)
Medicaid/self-pay	4 (1)	3 (1)	1 (1)
Time since first visit with provider, mo	22(7–57)	27 (9–62)	14 (5–45)	0.002
Length of follow-up, yr	1.9 (1.5–2.1)	2.0 (1.6–2.2)	1.9 (1.2–2.1)	<0.001
CCI	5 (4–7)	5 (4–6)	6 (5–8)	<0.001
Comorbidities
Diabetes	268 (55)	167 (53)	101 (59)	0.20
Hypertension	480 (98)	312 (98)	168 (98)	0.90
Coronary artery disease	197 (40)	109 (34)	88 (51)	<0.001
Congestive heart failure	154 (32)	80 (25)	74 (43)	<0.001
Peripheral vascular disease	104 (21)	59 (19)	45 (26)	0.05
Active cancer	81 (17)	45 (14)	36 (21)	0.05
CKD etiology
Diabetes	234 (48)	144 (45)	90 (53)	0.10
Hypertension	366 (75)	231 (73)	135 (79)	0.10
Analgesic nephropathy	41 (8)	30 (9)	11 (6)	0.30
Cardiorenal	44 (9)	14 (4)	30 (18)	<0.001
Renal artery stenosis	39 (8)	22 (7)	17 (10)	0.20
AKI	85 (17)	47 (15)	38 (22)	0.04
Glomerular disease	34 (7)	26 (8)	8 (5)	0.10
BMI	29 (25–34)	30 (26–35)	28(24–32)	0.001
Laboratory tests
eGFR, ml/min per 1.73 m2	22 (16–26)	23 (17–27)	20 (15–25)	0.001
Serum creatinine, mg/dl	2.5 (2.1–3.2)	2.5 (2.1–3.1)	2.7 (2.2–3.6)	0.01
Serum albumin, g/dl	4.0 (3.7–4.2)	4.0 (3.8–4.2)	3.9 (3.6–4.1)	<0.001
Hemoglobin, g/dl	11.5 (10.4–12.7)	11.8 (10.9–12.8)	11.0 (9.8–12.3)	<0.001
Urine protein-to-creatinine ratio, mg/g	41 (12–1560)	29 (10–96)	65 (15–2300)	<0.001

Continuous variables are presented as medians (interquartile ranges), and categorical variables are presented as N (%). SQ, surprise question; CCI, Charlson Comorbidity Index; BMI, body mass index.

RRT Planning

A documented preference for conservative management was uncommon in both groups: 13% among patients with SQ-No and 2% among patients with SQ-Yes (P<0.001) (Figure 1A, Table 2). Most had a documented preference for some form of dialysis, including 53% of patients with SQ-No and 51% of patients with SQ-Yes (Figure 1A, Table 2).

Figure 1.: Prognostic perceptions and differences in RRT preferences, advance care planning (ACP), and end of life (EOL) care. Point estimates and 95% confidence intervals are shown. (A) RRT preference: conservative management (CM), P<0.001. (B) ACP: any ACP form, P=0.03; medical order (MO), P=0.01. (C) Place of death, P=0.90. (D) EOL processes of care: use of hospice, P=0.50; use of hospital (i.e., hospitalization within 30 days of death), P=0.40. HD, hemodialysis; NH, nursing home; PD, peritoneal dialysis; SQ, surprise question. **Analyses restricted to decedents.

Table 2. - RRT planning, vascular access, and dialysis initiation

Care Characteristic	Total, n=488	SQ Yes, n=317	SQ No, n=171	P Value
Dialysis modality preference a				<0.001
Conservative management	29 (6)	6 (2)	23 (13)
Maintenance hemodialysis	147 (30)	94 (30)	53 (31)
Maintenance peritoneal dialysis	107 (22)	69 (22)	38 (22)
No preference specified	205 (42)	148 (47)	57 (33)
Dialysis access present	109 (22)	71 (22)	38 (22)
Arteriovenous fistula b	93 (85)	63 (89)	30 (79)	0.01
Arteriovenous graft b	7 (6)	1 (1)	6 (16)
Tunneled catheter b	9 (8)	7 (10)	2 (5)
Dialysis modality initiated	61 (13)	36 (11)	25 (15)	0.40
Hemodialysis c	41 (67)	24 (67)	17 (68)
Peritoneal dialysis c	20 (33)	12 (33)	8 (32)
Duration between SQ date and dialysis initiation	255 (127–412)	266 (178–426)	179 (94–383)	0.10

Continuous variables are presented as medians (interquartile ranges), and categorical variables are presented as N (%). SQ, surprise question.

^aSurprise question yes and surprise question no do not equal 100% due to rounding.

^bLimited to patients who had access placed.

^cLimited to patients who were initiated on dialysis

Advance Care Planning

Documented advance care plans were present in only a minority of each group: 34% of patients with SQ-No and 24% of patients with SQ-Yes (P=0.02) (Figure 1B, Table 3). A medical order was present in 13% of patients with SQ-No and 5% of patients with SQ-Yes (P=0.004) (Figure 1B, Table 3). Most participants with a medical order had specified a desire for treatment limitations, including 77% of patients with SQ-No and 59% of patients with SQ-Yes (P=0.20). During follow-up, six patients (all SQ-No) had a documented change in their medical order preferences (four from treatment limitations to aggressive care and two from aggressive care to treatment limitations). Among the 39 patients who had a medical order, the largest percentage was completed by emergency department providers (n=9; 23%). In contrast, very few were completed by a nephrology provider (n=4; 10%) or primary care provider (n=5; 13%) (Table 3).

Table 3. - Advance care planning characteristics

ACP Characteristic	Total, n=488	SQ Yes, n=317	SQ No, n=171	P Value
Advance care plan documentation	135 (28)	77 (24)	58 (34)	0.02
Medical order (i.e., POST or similar)	39 (8)	17 (5)	22 (13)	0.004
Treatment limitations (DNR/DNI/CMO) a	27 (69)	10 (59)	17 (77)	0.20
Aggressive care (full code) a	12 (31)	7 (41)	5 (23)	0.20
Provider completing medical order				0.40
Nephrologist b	4 (10)	1 (6)	3 (14)
Primary care physician b	5 (13)	1 (6)	4 (18)
Palliative care b	4 (10)	1 (6)	3 (14)
Emergency department physician b	9 (23)	4 (24)	5 (23)
Other b	17 (44)	10 (59)	7 (32)
Duration between SQ date and ACP form completion date, d c	−53 (−366–173)	100 (−201–209)	−221 (−546–43)	0.10

Continuous variables are presented as medians (interquartile ranges), and categorical variables are presented as N (%). SQ, surprise question; POST, physician orders for scope of treatment; DNR, do not resuscitate; DNI, do not intubate; CMO, comfort measures only; ACP, advance care planning.

^aLimited to patients with advance directives.

^bLimited to patients who had completed an advance directive.

^cRange includes negative values when event occurred before SQ capture date.

End of Life Care

Among 87 decedents, 41% of patients with SQ-No and 39% of patients with SQ-Yes died at home or hospice. An additional 39% of patients with SQ-No and 42% of patients with SQ-Yes died in the hospital (P=0.90) (Figure 1C, Table 4). Among SQ-No decedents, 38% used hospice care, with a median length of use of 6 (3–27) days. Among SQ-Yes decedents, 26% used hospice (P=0.50 versus SQ-No), with a median length of use of 8 (2–12) days (P=0.40 versus SQ-No). A majority of patients in each group were hospitalized in the month before death (SQ-No: 54% versus SQ-Yes: 67%; P=0.40) (Table 4). Only a minority of decedents in each group had documented advance care planning forms (39% SQ-No versus 32% SQ-Yes; P=0.50) (Supplemental Table 1).

Table 4. - End of life care characteristics in decedents

Care Characteristic	Total, n=87	SQ Yes, n=31	SQ No, n=56	P Value
Place of death				>0.99
Home	24 (28)	8 (26)	16 (29)
Hospice	11 (13)	4 (13)	7 (12)
Nursing home	5 (6)	2 (6)	3 (5)
Hospital	35 (40)	13 (42)	22 (39)
Unknown	12 (14)	4 (13)	8 (14)
Hospice use by the patient				0.50
Yes	29 (33)	8 (26)	21 (38)
No	45 (52)	18 (58)	27 (48)
Unknown	13 (15)	5 (16)	8 (14)
Duration of hospice use a	6 (3–25)	8 (2–12)	6 (3–27)	0.40
Hospitalization within 30 d of death				0.40
Yes	51 (59)	21 (68)	30 (54)
No	26 (30)	7 (23)	19 (34)
Unknown	10 (11)	3 (10)	7 (12)
Length of last hospitalization within 30 d of death, a d	11 (6–21)	8 (4–16)	13 (9–21)	0.30
Dialysis initiated	16 (18)	8 (26)	8 (14)
Hemodialysis b	11 (69)	5 (62)	6 (75)	0.60
Peritoneal dialysis b	5 (31)	3 (38)	2 (25)
Duration between RRT initiation and death, d	169 (60–306)	134 (57–289)	184 (135–376)	0.30
Duration between SQ date and death, d	351 (173–581)	380 (230–590)	336 (120–546)	0.10

Continuous variables are presented as medians (interquartile ranges), and categorical variables are presented as N (%). SQ, surprise question.

^aRestricted to known users.

^bLimited to those patients who were initiated on dialysis.

Supplemental Analyses

In the prespecified subgroup analysis examining patients with an eGFR≤20 ml/min per 1.73 m², results were similar (Supplemental Table 2). A documented preference for conservative management was present in a minority of patients, despite 78% of patients having a documented RRT preference. Similarly, use of advance care planning forms remained infrequent. Permanent hemodialysis access was present in a larger proportion of patients in the eGFR≤20 groups (SQ-No: 32% arteriovenous fistula or graft; SQ-Yes: 43% arteriovenous fistula or graft; P=0.90) (Supplemental Table 2).

In the multivariate model examining presence of advance care planning documentation in the entire cohort, nonblack race (odds ratio, 4.28; 95% CI, 1.90 to 9.64; white versus black) and older age (odds ratio, 1.86; 95% CI, 1.38 to 2.45; 80 versus 70 years old) but not SQ response were associated with a higher likelihood of advance care planning documentation (Supplemental Table 3). Additional adjustment for length of follow-up did not alter these findings.

Discussion

Despite a 2-year mortality rate of approximately 33%, we found that advance care planning documentation, a recorded preference for conservative management, and markers of high-quality end of life care were present in a minority of patients who were identified as high risk for clinical deterioration. Nephrologists’ prognostic perceptions were associated with only modest increases in preferences for conservative management and advance care planning documentation. Similarly, modest differences in hospice use and hospitalization within 30 days of death were noted, although these were not statistically significant. In adjusted analyses, SQ response was not associated with presence of an advance care planning form.

Our findings highlight a critical gap in advance care planning and end of life care among older adults with advanced nondialysis-dependent CKD and clinically recognizable high-risk features, in whom prognosis is similar to some advanced cancers (¹⁷,¹⁸). Health care delivered near the end of life is often incongruent with patients’ stated desires (^19–21). This is particularly evident among patients with advanced CKD, who tend to experience less high-quality end of life care and receive more aggressive care compared with patients with heart failure, cancer, or dementia (^22–24). Many patients with advanced nondialysis-dependent CKD report a lack of advance care planning discussions with their providers, a factor that may contribute to low rates of conservative management as well as suboptimal end of life care (⁶,²⁵,²⁶). Our results parallel those in patients with ESRD (^22–24). Unfortunately, temporal trends in aggressive care near the end of life suggest limited progress has been made in patients with advanced CKD over the last decade (²⁷).

Our findings indicate that timely advance care planning depends on key factors that extend beyond prognostic perceptions. These potentially include providers’ communication skills, time availability, health system resources to support patient-provider engagement (e.g., decision aids) and facilitate timely communication between team members (e.g., shared EHRs), clarity on provider roles in the process, patient/surrogate disease insight, patient/surrogate trust, and patient/surrogate communication skills (²¹,^28–30). Studies are needed to prioritize these barriers and discover effective strategies to overcome them in advanced CKD, because timely advance care planning and high-quality end of life care improve patient and bereaved caregiver quality of life outcomes (^31–34).

We noted low rates of conservative management use in this study, similar to recent findings in a Veterans Affairs cohort (³⁵). Use of conservative management in older adults with advanced CKD in other industrialized nations seems to be higher (20%–50%), with substantial heterogeneity depending on patient age, case mix, and location (^36–39). Our study design precluded us from discerning whether patients were informed about conservative management and its outcomes in patients with heavy comorbidity burdens, advanced age, or functional impairments (³⁶,⁴⁰,⁴¹). Provider uncertainty regarding prognosis with conservative management (⁴²) and how RRT decision making connects with the overarching need for advance care planning may have impeded these conversations (⁴³). Enhancing the level of evidence regarding the utility of conservative management in the United States is one step to improving provider comfort with shared decision making in this setting (¹³,^42–45). Improving patient engagement in these decisions is necessary to avoid unwanted dialysis initiations that may reduce quality of life, independence, and psychosocial functioning (⁴¹,⁴⁶,⁴⁷). Improvements in shared decision making in advanced stage CKD may also help patients exert control over their future and empower them to engage in advance care planning.

Notably, the date of SQ capture is unlikely to be the first occasion that providers considered patients’ prognoses. Indeed, over one half of the advance care planning forms documented in patients with SQ-No were completed >6 months before SQ capture (Table 3). Hence, providers had more time and opportunities to engage in advance care planning or enlist support to help deliver high-quality end of life care in patients with SQ-No responses than we are estimating. Our findings suggest that providers caring for an older adult with nondialysis-dependent CKD stage 4 or 5 should reflect on their patient’s clinical status using the SQ. In individuals who elicit an SQ-No response, providers should consider whether they have the skills and resources available to engage in advance care planning, including a meaningful discussion of conservative management. In the instances where barriers prevent this, providers should consider enlisting additional services (e.g., palliative care or trained advance care planning facilitator) with the requisite resources to address their patient’s and surrogate’s needs (⁴⁸).

We recognize that advance care planning documents do not ensure that patients and their surrogates have discussed patient care preferences, an important step to facilitate future decision making, reduce guilt, and improve caregivers’ outcomes. Hence, advance care planning forms do not guarantee patient-centered care; however, we believe that their absence is a useful indicator of deficits in high-quality advance care planning. In our cohort, age and race were significant predictors of advance care planning documentation, findings mirrored in other chronic diseases (⁴⁹). Given the elevated rates of nondialysis-dependent CKD in blacks, further work to characterize barriers and facilitators of advance care planning in this population is vital to help inform the development of effective interventions.

Our study has several strengths. We used the SQ to prospectively identify high-risk patients and compare their care with that of lower-risk patients. Similarly, we examined care preceding death in patients who were prospectively identified as higher risk to die. This approach is less subject to bias than retrospective approaches examining decedents, some of whom may have suffered unexpected deaths that would limit a provider’s ability to deliver high-quality end of life care (⁵⁰). In addition, our patients were systematically queried about the presence of an advance care plan at every outpatient visit as part of the routine clinic intake procedure. Our study also has several limitations. First, this was a single-center study performed at an academic nephrology practice. However, we suspect that care patterns in most busy community-based practices are unlikely to be better. Second, the number of patients with a documented medical order was limited, which precluded meaningful comparisons regarding differences in care. Third, we had a relatively small number of deaths in our cohort. Fourth, we did not have information on other salient outcomes, including congruence between patient preferences and end of life care delivery as well as patient and caregiver quality of life and symptom burdens. Fifth, some patients, even those with an eGFR≤20 ml/min per 1.73 m², were missing a documented RRT preference. This may lead us to underestimate actual conservative management use. Conversely, some patients who reported a preference for conservative management may eventually choose to start dialysis, leading us to overestimate preferences for conservative management. Sixth, our study was observational in nature and should not be interpreted as a failure of the SQ. Rather, we believe that our findings highlight the need to test multifaceted, systematic approaches to patient-centered end of life care in those with advanced CKD and high-risk clinical characteristics.

In conclusion, although patients with SQ-No responses from their providers and CKD stages 4–5 were at markedly increased risk of dying, conservative management adoption and advance care planning documentation rates were low. Markers of high-quality end of life care were modestly improved in higher-risk patients compared with their lower-risk counterparts, although these differences were generally not statistically significant. Future work will need to elucidate patient-, provider-, and system-level facilitators for the use of conservative management and advance care planning in the United States. Rigorous trials are needed to identify generalizable interventions that can expedite timely advance care planning and high-quality end of life care in this setting.

Disclosures

None.