Suffering and Symptoms At the End of Life in ICU Patients Undergoing Renal Replacement Therapy

Abstract

Background:

We know little about the end-of-life suffering and symptoms of intensive care unit (ICU) decedents in general and those who undergo renal replacement therapy (RRT) in particular.

Objectives:

To examine differences in end-of-life suffering and various symptoms’ contribution to suffering between ICU decedents who did not undergo RRT, those who underwent RRT for end-stage kidney disease (ESKD), and those who underwent RRT for acute kidney injury (AKI).

Methods:

This is a cross-sectional study conducted at a quaternary-level referral hospital September 2015–March 2017. Nurses completed interviews about ICU patients’ suffering and symptoms in their final week. We dichotomized overall suffering into elevated and non-elevated and each symptom as contributing or not to a patient’s suffering.

Results:

Sixty-four nurses completed interviews on 165 patients. Median patient age was 67 years (interquartile range 57, 78); 41% were female. In a multivariable model, undergoing RRT for AKI (odds ratio [OR] 2.95, 95% confidence interval [CI] 1.34–6.49) was significantly associated with elevated suffering compared to no RRT; undergoing RRT for ESKD was not. Adjusting for length of stay, AKI-RRT patients were more likely than non-RRT patients to have fecal incontinence (OR 2.21, 95% CI 1.00–4.93), painful broken skin (OR 2.41, 95% CI 1.14–5.12), and rashes (OR 3.61, 95% CI 1.35–9.67) contributing to their suffering.

Conclusions:

Undergoing RRT for AKI was associated with elevated suffering in the last week of life in ICU decedents. Painful broken skin, rashes, and fecal incontinence were more likely to contribute to suffering in AKI-RRT patients than in non-RRT patients. How to reduce suffering associated with AKI-RRT in ICU patients merits further study.

Introduction

The parties to the process of death and dying—seriously ill individuals, bereaved family members, physicians, others employed in caregiving—agree that freedom from symptoms such as pain, dyspnea, and anxiety is of utmost importance to ameliorate suffering at the end of life.¹ There are, however, limited data on the symptoms that affect people who die in an intensive care unit (ICU) and the contributions of those symptoms to suffering, even though in the United States (US), 20 to 30 percent of people die during or shortly after an ICU stay.^2,3 To date, most of the research on this subject has focused on pain and dyspnea and addressed how well those symptoms are managed rather than how common or severe they are in the context of a patient’s suffering.^4–6 A major barrier to learning about the symptoms and suffering of any ICU decedent is that most ICU patients are unable to self-report as they approach death. To overcome this barrier, studies investigating quality of death in the ICU routinely use nurse ratings alongside ratings from family members and other clinicians.^4,7–12

As little as we know about the end-of-life symptoms of ICU decedents in general, we know even less about the end-of-life symptoms of ICU decedents who dealt with the additional stress of undergoing renal replacement therapy (RRT), either for end-stage kidney disease (ESKD) or acute kidney injury (AKI). The latter portends a particularly poor prognosis, with about 50% of critically ill AKI-RRT patients not surviving to 90 days.^13–17 Patients with ESKD treated with dialysis face a significant burden from chronic symptoms, some of which are secondary to the dialysis itself.^18–23 As nearly 50% of ESKD patients in the US spend time in an ICU in the last month of life,^24,25 and 1 multinational study estimated that over 20% of patients who die in an ICU will have received RRT for AKI,²⁶ recognizing the symptoms of patients undergoing RRT in an ICU prior to death is crucial to lessening their suffering and improving their quality of death. Notably, our group’s previous work found that, of 6 different types of life-prolonging care delivered to ICU decedents, only RRT was associated with nurse-rated patient suffering and loss of dignity, after adjustment for ICU admission diagnosis.²⁷ There is, however, the possibility that patients’ overall severity of illness, which is not necessarily captured in their ICU admission diagnosis, or their preferences for more aggressive treatment could contribute more to their suffering than RRT itself. This is particularly possible in the case of patients receiving RRT for AKI, who are recognized as an especially ill group of patients receiving especially aggressive care.

To advance understanding of the suffering and symptoms at end of life in ICU patients undergoing RRT, based on our prior finding that only undergoing RRT was associated with nurse-rated patient suffering,²⁷ we conducted the current study as a secondary analysis using the aforementioned nurse ratings to examine differences in apparent end-of-life suffering between ICU decedents who did not undergo RRT, those who underwent RRT for ESKD, and those who underwent RRT for AKI. We hypothesized that decedents who underwent RRT for AKI would be rated as experiencing worse suffering at the end of life, given the sudden failure of a particular organ and the use of an invasive extracorporeal therapy to compensate for this organ failure. In an exploratory analysis, we then compared ratings of individual symptoms’ contribution to suffering between non-RRT and AKI-RRT patients.

Methods

This was a cross-sectional study conducted in the medical and cardiac ICUs at New York-Presbyterian Hospital / Weill Cornell Medical Center, a quaternary-level referral hospital, from September 2015 to March 2017. Every week trained research staff identified all patients who had died in these units in the prior week, along with the nurses who had cared for them for at least one 12-hour shift. After obtaining informed consent from the nurses, research staff individually interviewed participating nurses in person. We excluded the deaths of patients who were in the ICU for less than 24 hours, who did not have a single nurse who cared for them for an entire shift, or for whom no eligible nurse was able to complete the interview within 3 weeks of the patient’s death. For completion of an interview outside of regular work hours, a nurse received a $20 gift card. The Weill Cornell Medicine institutional review board approved this study.

Data Collection and Measures

Structured nurse interviews used a 1–10 scale, with 1 representing “not at all” and 10 representing “extremely,” for nurses to rate the patient’s overall suffering in the last week of life (“In the last week of the patient’s life, to what extent did he / she appear to be suffering?”). Using the same scale, nurses also rated 15 different symptoms on how much they contributed to the patient’s suffering in the last week of life (“In the last week of the patient’s life, to what extent did each of the following contribute to his / her suffering?”). When rating the symptoms, nurses were instructed to take into account how well the symptom was managed, e.g. if a patient had severe pain that rapidly decreased after medication administration and stayed at a low level, the nurse’s rating should have reflected that pain control. In our group’s earlier work, nurse responses to the suffering question were correlated in the expected directions with previously validated measures of psychological distress, physical distress, overall quality of death, and peacefulness at the end of life (P < 0.001 for all, data not shown),²⁷ providing evidence of construct validity.

Study staff abstracted demographic and clinical data from patients’ charts, including age, sex, race, ethnicity, religion, length of stay, processes of care (e.g., do-not-resuscitate [DNR] orders), and medical procedures and treatments in the last week of life (e.g., RRT, mechanical ventilation, pressors). A later chart review by the first author (SJR) established whether patients who underwent RRT did so because of ESKD or AKI. Because only charts from New York-Presbyterian Hospital / Weill Cornell Medical Center were available for review, the analyses described in this paper exclude the patients from Brigham and Women’s Hospital who were included in previous analyses.²⁷

Statistical Analyses

We present patients’ demographic and clinical characteristics as medians and interquartile ranges (IQR) for continuous variables and counts with percentages for categorical variables. To examine whether there were significant differences in these characteristics between patients who did not receive RRT, those who received RRT for ESKD, and those who received RRT for AKI, we used the Kruskal-Wallis test for continuous variables (as they were not all normally distributed) and the Fisher’s exact test for categorical variables (as counts for some variables in certain categories were low).

Although nurses originally rated overall suffering on a continuous 1–10 scale, we dichotomized this outcome after graphical examination of its distribution and with the goal of making the data more clinically useful. Finding that the median suffering rating was 7, we classified scores of 7 or greater as representing “elevated suffering” and scores of 6 or less as representing “not elevated suffering.” We evaluated the associations of every variable, including RRT status and indication, with elevated suffering using generalized estimating equations (GEE) with exchangeable correlation structure and logit link to account for within-nurse correlation of ratings. To adjust for the associations of variables aside from RRT status and indication with elevated suffering, we then entered all other variables with P < 0.2 in bivariate analysis into a multivariate model that employed the same regression technique described above.

We also performed an exploratory analysis of individual symptoms’ contributions to suffering in non-RRT, ESKD-RRT, and AKI-RRT patients. Given the small number of ESKD-RRT patients (n = 11), we are omitting those results. Nurses rated 15 individual symptoms on the extent to which each contributed to a patient’s overall suffering. We dichotomized the ratings into “didn’t contribute to suffering at all” (nurse rating of 1) and “contributed to suffering” (nurse rating of 2–10) after graphical examination of the ratings of each symptom revealed that the modal response was 1. We then calculated the percentages of patients in the non-RRT and AKI-RRT groups that had a symptom contributing to their suffering. We modeled the association between AKI-RRT status and each symptom using GEE models with the same specifications as above to account for within-nurse correlation of ratings. For these models we adjusted for length of hospital stay based on our assumption that it might also be associated with symptom contribution to suffering. We performed all analyses in SPSS version 26.

Results

Sixty-four nurses completed 202 interviews pertaining to 170 patients. We excluded 5 patients from this analysis because of missing ratings for overall suffering, leaving a sample size of 165 patients. To simplify analysis, we used data from a single nurse interview for each patient. In 32 cases in which more than one nurse had completed an interview on a particular patient, we used the interview of the nurse who had spent the most shifts with the patient. If more than one nurse had spent an equal number of shifts with the patient, we used the interview of the nurse with less missing data. If both interviews had the same amount of missing data, we used the interview that was entered into the dataset last. The interviewed nurses cared for each patient a median of 2 shifts (IQR 2–3 shifts); a median of 7 days (IQR 5–11 days) elapsed between the patient’s death and the nurse’s interview.

Patients’ demographic and clinical characteristics are displayed in Table 1. Median age in years (IQR) was 67 (57–78); 41% were female; 69% were white, 15% were black, 14% were Asian, and 2% were more than one race; and 12% were of Hispanic ethnicity. Median length of stay in the hospital in days (IQR) was 13 (6–25); in the ICU, it was 7 (4–13). Eighty-five percent had a DNR order at some point before death and 30% received palliative care consults, while in the last week of life, 81% underwent mechanical ventilation, 88% received vasopressors, 35% underwent cardiac resuscitation, 12% underwent a surgical procedure, and 12% received chemotherapy. Comparing patients who did not undergo RRT, those who underwent RRT for ESKD, and those who underwent RRT for AKI, the only significant differences were in length of hospital stay (median 10 days, IQR 4–21 days in no RRT; median 12 days, IQR 7–24 days in ESKD-RRT; median 19 days, IQR 9–29 days in AKI-RRT; P = 0.01) and length of ICU stay (median 5 days, IQR 2–11 days in no RRT; median 8 days, IQR 7–18 days in ESKD-RRT; median 8 days, IQR 5–19 days in AKI-RRT; P = 0.004).

Table 1.

Patients’ Demographic and Clinical Characteristics by RRT Status.^a,b

	Overall (n = 165)	No RRT (n = 109)	RRT for ESKD (n = 11)	RRT for AKI (n = 45)	Pc
Age, years	67 (57, 78)	69 (57, 82)	66 (62, 74)	64 (55, 72)	0.08
Female	68 (41)	47 (43)	5 (46)	16 (36)	0.70
Raced					0.77
White	105 (69)	71 (70)	7 (70)	27 (64)
Black	23 (15)	12 (12)	2 (20)	9 (21)
Asian	22 (14)	16 (16)	1 (10)	5 (12)
More than one	3 (2)	2 (2)	0 (0)	1 (2)
Hispanic ethnicitye	17 (12)	13 (14)	1 (13)	3 (7)	0.54
Total length of stay, days	13 (6, 25)	10 (4, 21)	12 (7, 24)	19 (9, 29)	0.01
ICU length of stay, days	7 (4, 13)	5 (2, 11)	8 (7, 18)	8 (5, 19)	0.004
DNR order present at any time	136 (82)	92 (84)	8 (73)	36 (80)	0.46
Received palliative care consultf	48 (30)	34 (32)	4 (36)	10 (22)	0.40
Procedures / treatments in last week of life
Mechanical ventilation	133 (81)	86 (79)	7 (64)	40 (89)	0.11
Vasopressors	145 (88)	93 (85)	9 (82)	43 (96)	0.11
Cardiac resuscitation	57 (35)	36 (33)	4 (36)	17 (38)	0.83
Surgical procedure	20 (12)	13 (12)	2 (18)	5 (11)	0.79
Chemotherapy	20 (12)	14 (13)	0 (0)	6 (13)	0.58

Abbreviations: DNR, do not resuscitate; RRT, renal replacement therapy; ESKD, end-stage kidney disease; AKI, acute kidney injury.

^aContinuous variables expressed as median (interquartile range); categorical variables expressed as n (%).

^bPercentages in columns may not add up to 100 due to rounding.

^cKruskal-Wallis test for continuous variables; Fisher’s exact test for categorical variables.

^dOverall n = 153, no RRT n = 101, RRT for ESKD n = 10, RRT for AKI n = 42.

^eOverall n = 143, no RRT n = 93, RRT for ESKD n = 8, RRT for AKI n = 42.

^fOverall n = 162, no RRT = 106, RRT for ESKD n = 11, RRT for AKI n = 45.

Among all patients, nurses rated 91 (55%) as having elevated suffering (score ≥7 on 1–10 scale) in the last week of life; by group, it was 53 (49%) of non-RRT patients, 6 (55%) of ESKD-RRT patients, and 32 (71%) of AKI-RRT patients. Table 2 shows unadjusted associations of demographic and clinical characteristics with elevated suffering. Only undergoing AKI for RRT (odds ratio [OR] 2.60, 95% confidence interval [CI] 1.23–5.48) and undergoing mechanical ventilation (OR 0.41, 95% CI 0.18–0.95) were significantly associated with elevated suffering in these unadjusted analyses, though it should be noted that the latter association is a negative one. In the multivariate model, undergoing RRT for AKI (OR 2.95, 95% CI 1.34–6.49) and undergoing mechanical ventilation (OR 0.34, 95% CI 0.14–0.86) both remained significantly associated with elevated suffering (Table 3).

Table 2.

Unadjusted Associations of Demographic and Clinical Characteristics with Elevated Suffering.

	Odds ratio	95% confidence interval	P
Age (per year)	1.00	0.98 −1.02	0.81
Female (vs. male)	1.43	0.76–2.68	0.27
Race (vs. white)
Black	1.04	0.42–2.58	0.93
Asian	0.67	0.27 −1.68	0.39
More than one1	-	—	—
Hispanic ethnicity (vs. non-Hispanic)	0.87	0.32–2.41	0.79
Total length of stay (per day)	1.00	0.99 −1.02	0.80
ICU length of stay (per day)	1.00	0.98 −1.04	0.72
DNR order present at any time (vs. no DNR at any time)	0.59	0.26 −1.37	0.22
Received palliative care consult (vs. no palliative care consult)	1.22	0.62–2.41	0.57
Procedures / treatments in last week of life (vs. not receiving procedure / treatment)
Mechanical ventilation	0.41	0.18–0.95	0.04
Vasopressors	1.01	0.39–2.58	0.99
Cardiac resuscitation	0.95	0.50 −1.82	0.89
Surgical procedure	0.99	0.39–2.54	0.99
Chemotherapy	1.60	0.60–4.23	0.35
RRT status (vs. no RRT)
RRT for ESKD	1.27	0.37–4.40	0.71
RRT for AKI	2.60	1.23–5.48	0.01

Abbreviations: DNR, do not resuscitate; RRT, renal replacement therapy; ESKD, end-stage kidney disease; AKI, acute kidney injury

¹Not calculated due to small number in group.

Table 3.

Adjusted Associations of Demographic and Clinical Characteristics with Elevated Suffering.

	Adjusted odds ratio	95% confidence interval	P
RRT status (vs. no RRT)
RRT for ESKD	1.09	0.31–3.87	0.89
RRT for AKI	2.95	1.34–6.49	0.01
Mechanical ventilation	0.34	0.14–0.86	0.02

Abbreviations: RRT, renal replacement therapy; ESKD, end-stage kidney disease; AKI, acute kidney injury.

Figure 1 demonstrates the frequency of individual symptoms’ contributions to suffering, as determined by nurse ratings of >1 (1=“not at all”) on a 1–10 scale, in the last week of life among patients who did not undergo RRT and those who underwent RRT for AKI. Physical pain (75% non-RRT, 84% AKI-RRT), trouble breathing (83% non-RRT, 89% AKI-RRT), and swelling (69% non-RRT, 84% AKI-RRT) were the most commonly reported symptoms in both groups, while fecal incontinence was also noted in 77% of AKI-RRT patients (Table 4). In exploratory analyses adjusting for length of stay, AKI-RRT patients were more likely than non-RRT patients to have fecal incontinence (OR 2.21, 95% CI 1.00–4.93), painful broken skin (OR 2.41, 95% CI 1.14–5.12), and rashes (OR 3.61, 95% CI 1.35–9.67) rated as contributing to their suffering, while they were less likely to have thirst (OR 0.41, 95% CI 0.17–0.98) rated as contributing.

Figure 1.

Frequency of individual symptoms’ contributions to suffering in non-RRT and AKI-RRT patients.

Table 4.

Individual Symptoms’ Contributions to Suffering in Non-RRT and AKI-RRT Patients.

Symptom	Symptom contributed to suffering		Adjusted odds ratio (RRT for AKI vs. no RRT)b	95% confidence intervalb	Pb
	No RRT n (%)a	RRT for AKI n (%)a
Physical pain	79 (75)	36 (84)	1.41	0.54–3.66	0.48
Trouble breathing	89 (83)	39 (89)	1.48	0.52–4.19	0.46
Constipation and / or diarrhea	44 (41)	21 (47)	1.14	0.55–2.36	0.72
Nausea and / or vomiting	24 (22)	15 (34)	1.88	0.86–4.11	0.11
Urinary incontinence	27 (25)	9 (21)	0.75	0.32 −1.73	0.50
Fecal incontinence	65 (61)	34 (77)	2.21	1.00–4.93	0.05
Thirst	55 (62)	15 (44)	0.41	0.17–0.98	0.04
Hunger	44 (47)	13 (36)	0.56	0.24 −1.27	0.17
Painful broken skin	48 (45)	29 (67)	2.41	1.14–5.12	0.02
Rashes	9 (8)	10 (24)	3.61	1.35–9.67	0.01
Swelling	75 (69)	38 (84)	2.25	0.90–5.67	0.08
Difficulty sleeping	45 (46)	17 (43)	0.93	0.44 −1.98	0.86
Fatigue	56 (58)	22 (58)	0.93	0.43–2.01	0.86
Loss of control of limbs	62 (59)	26 (63)	1.19	0.56–2.53	0.65
Fever and / or chills	34 (32)	16 (38)	1.32	0.63–2.77	0.47

^an for each symptom variable, as not every patient was assessed for every symptom.

^bAdjusted for total length of stay.

Discussion

In this study, using nurse ratings of end-of-life suffering in ICU decedents who did not undergo RRT and those who underwent RRT for ESKD or AKI, we found that among several demographic and clinical characteristics, only undergoing RRT for AKI was independently and positively associated with perceived elevated suffering in the last week of life. Furthermore, nurses also rated significantly more AKI-RRT patients than non-RRT patients as having painful broken skin, rashes, and fecal incontinence contribute to their suffering in the last week of life, even after adjustment for length of stay.

The possible explanations for greater suffering in patients who underwent RRT for AKI include factors directly related to the AKI, factors directly related to the RRT, and / or factors unrelated to either but present more frequently in AKI-RRT patients. One such “unrelated” factor could be a patient or surrogate preference for more aggressive end-of-life care that is reflected in the decision to initiate RRT when AKI develops. Yet what might be the single most significant marker of aggressive end-of-life care, undergoing cardiac resuscitation in the last week of life, was not associated with greater suffering in our study, nor were other aggressive procedures or treatments. The fact that mechanical ventilation was unexpectedly associated with nurse ratings of lower suffering could be the result of mechanically ventilated patients receiving more opioids, benzodiazepines, and other sedatives than non-ventilated patients, or a result of ventilated patients being less able to express their suffering due to inability to speak.

That patient or surrogate preference for more aggressive end-of-life care probably does not by itself account for the greater suffering in AKI-RRT patients suggests that factors directly related to the AKI or RRT might be operative. Symptoms specific to critically ill patients with severe AKI or resulting from the RRT itself have not previously been described. In our exploratory analysis of symptoms contributing to suffering in non-RRT versus AKI-RRT patients, it is notable that symptoms that could be attributed to uremia, namely nausea and / or vomiting and fatigue, and symptoms that could be attributed to volume overload, namely trouble breathing and swelling, did not contribute to suffering in a significantly higher percentage of AKI-RRT patients than in non-RRT patients. Nor did physical pain, which could be attributed to vascular catheter placement and manipulation and body positioning for RRT. That painful broken skin, rashes, and fecal incontinence were the symptoms that contributed to suffering in significantly more patients who underwent RRT for AKI is somewhat surprising, though the cluster of symptoms is logical, with rashes potentially contributing to painful broken skin anywhere on the body and fecal incontinence potentially contributing to rashes and painful broken skin on the buttocks, sacrum, or back. There are no data on how common uremic pruritus is in kidney dysfunction due to AKI, but we also cannot rule out the possibility that AKI-RRT patients experienced uremic pruritus, causing them to scratch their skin and their nurses to judge rashes as contributing to their suffering. While the cluster of symptoms present more often in AKI-RRT patients could be interpreted as a consequence of more aggressive end-of-life care, we observed these differences in symptoms between non-RRT and AKI-RRT patients despite adjustment for length of stay, which was longer in AKI-RRT patients. Due to the relatively small numbers of patients experiencing certain symptoms, we could not adjust for other covariates that might indicate more aggressive end-of-life care.

Regardless of the exact reasons for greater suffering in AKI-RRT patients in the ICU at the end of life, the need for interventions aimed at reducing this suffering is clear. Based on our findings, interventions that improve skin care or fecal incontinence management might be tested to determine if they lessen overall suffering in this population. More broadly, the possible benefit of specialist palliative care for these patients deserves investigation. The minimal research in this area to date demonstrated that AKI-RRT patients in American hospitals infrequently receive specialist palliative care, though the infrequency was similar to that seen in several other life-threatening conditions.²⁸

Our study had some notable strengths. First, in assessing suffering and symptoms in the last week of life in AKI-RRT patients, we have addressed a topic that up until now has received little attention, despite how common a condition AKI-RRT is at the end of life in ICUs. Second, there was considerable racial and ethnic diversity in our sample, expanding the generalizability of our findings. Third, the use of nurse ratings of overall suffering and specific symptoms’ contributions to suffering permitted us to collect data from all ICU patients, not just the ones able to self-report because they are presumably in better condition, thereby reducing selection bias.

We also acknowledge limitations in this study. Given this study’s observational nature, we cannot conclude that the worse suffering or certain symptoms’ contribution to suffering found in AKI-RRT patients was a result of either the AKI or the RRT. Furthermore, we did not have access to all relevant clinical information about the patients, including vital signs, lab values, ventilator settings, pressor doses, fluid balance measurements, complete problem lists, complete medication lists, timing of procedures and DNR status, type of RRT (in particular, intermittent versus continuous), and number or adequacy of RRT sessions. Thus we were unable to examine the role of or statistically adjust for many variables that might have also influenced suffering and symptoms. For the same reason, we were not able to identify the patients who had lesser degrees of AKI or chronic kidney disease to compare them to patients with AKI-RRT and ESKD, respectively. Another limitation is the use of retrospective nurse report for suffering and symptoms, as prior research has suggested that patient-nurse agreement on symptom intensity and distress in the ICU can be low,²⁹ and nurses, like patients, will bring their own biases to this reporting. Even if this were also the case in our study, we have no reason to suspect that nurses systematically overrated suffering and symptoms in one group and underrated them in another. That is, the nurses would have been equally inaccurate for all patients, and our results would remain valid. Studies of quality of death in the ICU routinely use nurse ratings alongside ratings from family members and other clinicians.^4,7–12 We also acknowledge that we do not know what the term “suffering” meant to each individual nurse and that suffering is a multifaceted phenomenon to which symptoms contribute but do not define.

Taken together, our findings that only undergoing RRT for AKI was independently and positively associated with elevated suffering in the last week of life in ICU decedents, and that painful broken skin, rashes, and fecal incontinence were more likely to contribute to suffering in AKI-RRT patients than in non-RRT patients, provide important insights into an understudied area. These findings may also help inform shared decision-making regarding RRT initiation and continuation for AKI patients in the ICU. That these patients face not only high mortality but also high levels of suffering further establishes the need for additional research into how palliative care might help this vulnerable population.