Background

Over 500,000 individuals receive dialysis for end stage renal disease (ESRD) in the United States [1]. In the absence of clinical emergencies or overt uremia, dialysis is routinely initiated in the outpatient setting. However, many patients initiate therapy during a hospitalization for acute illness. A recent study showed that among veterans this proportion may be as high as 75% [2]. Decisions to initiate dialysis in the hospital often take place in the context of acute kidney injury (AKI) with concurrent critical illness, a situation associated with high mortality.

Studies in the outpatient setting suggest that patients feel unprepared for starting dialysis and may regret the decision [3, 4]. However, the inpatient setting poses unique challenges to informed decision-making, patient autonomy, and shared decision-making. Outpatient dialysis initiation occurs exclusively among patients who have chronic kidney disease and are followed by a nephrologist. However, critical illness leading to severe AKI is restricted to the hospital setting. Therefore, the mechanisms through which patients end up on dialysis are distinct. For patients in intensive care, decisional capacity is often compromised [5] and informed consent for interventions like dialysis may be obtained from a surrogate decision-maker (SDM).

To our knowledge, there are no studies examining how patients or their SDM’s feel about dialysis initiation in the hospital setting. There is also little known about how often hospitalized patients can participate in the decision to initiate dialysis. Therefore, we conducted a prospective, questionnaire-based study of hospitalized patients with AKI requiring dialysis to assess decisional satisfaction, degree of shared decision-making, and patient participation in the decision to initiate dialysis. A short, structured interview was also utilized for exploratory purposes to gain insight into perceptions of the process of informed consent for dialysis initiation.

Methods

Setting and patient recruitment

The study was conducted at a large, academic referral center in the Northeastern United States (Hospital of the University of Pennsylvania, Philadelphia, PA). Patients who consented to dialysis initiation for AKI during a one-month period in 2016 were prospectively approached for enrollment into this study. Dialysis initiation was defined as the first treatment using either intermittent or continuous dialysis during a patient’s lifetime and instances were reported to study authors by consulting nephrology teams. Patients with ESRD or any previous dialysis were excluded. There were no instances of inpatient initiation of peritoneal dialysis during the study period. The participant for this study was the individual who provided consent for dialysis. Thus, in cases where the patient was unable to consent, the patient’s SDM became the study participant. To minimize the effect of outcome, hindsight, and recall bias on responses to study questions, all participants were approached within 72 h of providing written consent for dialysis. All patients had already initiated dialysis when they (or SDM’s) were approached for enrollment and they were always approached during daytime hours. Their medical record was not accessed unless they provided consent to enroll into the study.

Design of the study questionnaire and structured interview

The study questionnaire included previously validated tools. The Decision Attitude Scale (DAS) was used to measure post-decisional satisfaction among patients or their SDM’s. This tool – along with a nearly identical one called Satisfaction With Decision scale – was developed for situations when there are difficult choices and future consequences or outcomes are unknown [6, 7]. Since the clinical outcome is unknown at the time of the decision, the DAS was designed to assess how respondents feel about the process of making their decision. In the clinical context, it therefore provides a way to characterize the informed consent process (Table 1).

Table 1 Decision attitude scale and additional study questions. Responses were a 5-item Likert Scale (Strongly agree, agree, neutral, disagree, strongly disagree)
Full size table

To assess the degree of shared decision-making, participants and the nephrology provider who obtained consent for dialysis were also asked to complete a modified version of the Control Preferences Scale (CPS, Table 2) [8]. The CPS has been used to assess perceptions of how treatment decisions are influenced by patients’ and physicians’ preferences. Decisions are classified based on five choices that reflect a spectrum of autonomy. Fully autonomous decisions are those made exclusively by either the patient or physician. Alternatively, there can be varying degrees of shared decision-making with input from both patient and physician. By design, nephrology providers could see the CPS during the study, but to prevent the study from influencing physician behavior (Hawthorne effect), the remaining contents of the questionnaire were not revealed to them. Lastly, open-ended questions about aspects of the informed consent process for dialysis were asked as part of a structured interview (Additional file 1: Figure S1). Verbal responses were written down by the interviewer. Audio-recording was not used to maximize participant willingness to answer questions.

Table 2 Modified Control Preferences Scale. The original wording of the CPS can be found in reference [8]
Full size table

Notably, references made to the “consent process” in questionnaire items are referencing the consent process for dialysis initiation, not the consent process for this research study. Standard practice at the study institution is to obtain written consent for dialysis initiation from all patients. These consents are typically obtained by the nephrology fellow. Participants were reminded of this distinction as needed.

Ethics

All participants provided written informed consent. The study protocol, consent forms, and items associated with the questionnaire were approved by the Institutional Review Board in the Office of Regulatory Affairs at the University of Pennsylvania. Prior to initiating the study, the overall aims of the project were also discussed with the nephrology department. All nephrology providers active during the study period (both faculty and fellows) agreed to participate in the study.

Data analysis and reporting

Responses from encounters with study participants were initially handwritten onto study sheets and later entered into a de-identified database (Microsoft Excel) and analyzed using Stata. To streamline data reporting, 5-item Likert scale responses were recoded into three possible responses: agree, disagree, and uncertain. Given that the motivation of this work is descriptive and qualitative, statistical analysis is limited to the computation of means and 95% confidence intervals. These calculations were performed in Stata (StataCorp, College Station, TX).

Results

Of the 31 patients and SDM’s who met study criteria, 21 agreed to participate. Table 3 shows baseline demographics for the enrolled patients. Enrolled patients had a mean age of 63 years, one-third were African-American, and most had both pre-existing diabetes and hypertension. Medical history for potential participants who declined to enroll is not available since their records were not accessed to preserve privacy. Most patients were in an ICU and continuous venovenous hemodialysis (CVVHD) was used more often than intermittent dialysis (IHD) as the initial modality. The mean baseline creatinine for enrolled patients was 2.3 mg/dL. For 19 of the 21 patients, this baseline was defined as the immediate prehospitalization value. Study participants were supposed to be approached within 72 h of signing the dialysis consent form and all cases met this requirement. The mean time between dialysis consent and study enrollment was 36 h.

Table 3 Patient demographics
Full size table

Patient participation in the decision to initiate dialysis

Of the 21 patients included in the study, 13 (62%) were not able to participate in the decision to initiate dialysis and relied on an SDM for consent. This lack of decisional capacity is reflective of the illness severity seen among the cohort of patients studied, most of whom were treated in an ICU setting. This finding is supported by the fact that 12 out of 13 patients who initiated CVVHD relied on an SDM.

Decisional attitudes towards consent for dialysis initiation

The mean score from the DAS for all 21 participants was 4.1 (95% CI 3.8–4.3), with a score of 5 implying the highest post-decision satisfaction. In addition, 17 (81%) participants agreed that “going through the process of consent for dialysis was important” to them. All 21 participants disagreed with the statement that “signing the consent form was a waste of time.” Lastly, 19 (90%) participants felt like they “had a choice in making this decision.” In contrast to this sense that participants valued the consent process, 9 (43%) participants agreed that “signing the consent form was mainly to protect the hospital.”

This study was not designed based on thresholds for statistical power to detect differences in mean DAS scores through stratified analyses. However, there was no statistically significant difference in DAS scores when comparing across dialysis modality or the presence of an SDM (Table 4).

Table 4 Mean decisional attitude score stratified by dialysis modality (a) and presence of an SDM (b)
Full size table

Participants’ responses to structured interview questions are shown in the Additional file 2: Table S1. Although a formal thematic analysis was not performed, responses suggest that there were a variety of feelings and emotions about the circumstances of dialysis initiation. 19 (90%) participants mentioned toxin removal, cleaning, purification, or volume removal when asked to summarize what they knew about dialysis. Consistent with their high DAS scores, most participants (11 out of 21, 52%) had feelings of gratitude for care received.

Alternatives to dialysis initiation

Although 19/21 (90%) of participants felt like they “had a choice” in making the decision to initiate dialysis, 17 (81%) answered that they were unaware of any alternatives to dialysis (Additional file 2: Table S2). Two participants, (9%) alluded to waiting a longer time to see if there was renal recovery, and 2 (9%) mentioned either kidney transplant or another dialysis modality. Of the 17 patients who identified no alternatives to dialysis, two specifically mentioned death in their response to the question.

Perceptions regarding shared decision-making

Study participants and their corresponding nephrology providers were both asked to select one of five choices from the modified CPS (Table 2). The distribution of these paired responses is shown in Fig. 1. Most of the responses indicated that dialysis initiation was either an equally shared decision, or one that incorporated some input from both the patients/SDM’s and nephrology providers. Of the 21 paired responses, there were 9 (43%) pairs in which responses were concordant (patient/SDM and provider chose the same item from five choices). Figure 2 shows the distribution of these concordant responses. Among concordant pairs, none felt dialysis initiation was an autonomous decision and most felt that the decision was shared.

Fig. 1
figure 1

Distribution of responses to Control Preferences Scale (21 pairs). All responses to Control Preferences Scale (CPS) by 21 paired sets of study participants and corresponding nephrology providers. The categories on the horizontal axis correspond to selections on the CPS, which is shown in Table 2. Choice A from the CPS corresponds to “Patient autonomous,” while choice E corresponds with “MD autonomous”

Full size image
Fig. 2
figure 2

Distribution of concordant responses to Control Preferences Scale (9 pairs). Concordant responses refer to when a patient/SDM and nephrology provider chose the same response to the Control Preferences Scale. From the 21 paired sets, there were 9 concordant pairs, which are shown here

Full size image

Discussion

Dialysis is often initiated in the hospital when AKI occurs in the setting of critical illness. When AKI is severe enough to require dialysis, there is a high prevalence of conditions like sepsis, the use of vasopressors, and mechanical ventilation [9]. Thus, when compared to individuals who start hemodialysis as outpatients, hospitalized patients face a different clinical scenario for decision-making. This study prospectively approached all hospitalized patients with AKI requiring dialysis at a large, tertiary medical center for 30 consecutive days. There was predominant use of a continuous dialysis modality (67%), a therapy that can only be performed in an intensive care unit (ICU). Consistent with prior reported outcomes, the study cohort had high in-hospital mortality (33%) [10, 11]. One of the main findings is that 62% of patients could not participate in the decision to initiate dialysis and relied on a SDM. This reliance on a SDM is likely a reflection of the general loss of decisional capacity commonly seen in intensive care. Although this study did not assess for patient participation in other decisions like mechanical ventilation, insertion of central venous catheters, or enteral feeding tubes, a high percentage of critically ill patients rely on SDM’s for multiple decisions, especially near the end of life [5].

The fact that more than half of the decisions to initiate dialysis were made by SDMs and not the patients themselves stresses the overall importance of being aware of patients’ values and preferences. Ideally, gaining such insight would take place prior to clinical decompensation and be part of longitudinal advance care planning.

For patients who survive their acute illness and remain dialysis dependent, our study findings suggest that most of them may need to revisit the risks, benefits, and rationale of dialysis since they were not able to participate in the initial discussion. In our experience, due to the structure of inpatient care, which may involve multiple handoffs, this type of conversation often does not take place.

Study participants were essentially unable to identify any alternatives to dialysis. Alternatives could include time-limited trials, conservative care, comfort-directed care, de-escalation of care, palliative care, or hospice. A lack of awareness of alternatives to dialysis is especially noteworthy considering the high mortality seen in the study cohort. Based on the data presented here, it is not possible to extrapolate what drives this inability to articulate alternatives to dialysis. However, we speculate that in the context of acute illness, clinical momentum favoring continuation of life-sustaining therapies likely framed the decision to initiate dialysis as “necessary” and more conservative options were probably not presented. It is also possible that participants were unable to recall a discussion of alternatives to dialysis, but we feel this is less likely.

Acute initiation of dialysis in the United States relies mainly on the hemodialysis modality. Data presented here are consistent with this trend as there were no patients who initiated on peritoneal dialysis. Prior work has shown dialysis modality selection can be influenced by the timing of nephrology referral, with late referral being less likely to be associated with the use of peritoneal dialysis [12, 13]. Therefore, our findings suggest that patients who initiate dialysis in the hospital may need heightened attention to modality education since hemodialysis is essentially used as the “default” therapy in most acute settings.

This work has some limitations. The data are from a single center in the United States and may not be representative of larger trends. Due to resource limitations, prospective enrollment into the study was pre-determined to continue for 1 month. In addition, although the study questionnaire was aimed at assessing aspects of nephrology care delivery, patients or SDM’s may not have been able to rate the nephrology team in isolation of their entire hospital or ICU experience. Therefore, although an effort was made to approach study participants within a short window of dialysis initiation, their responses to questionnaire items is likely influenced by external factors. The high score on the DAS reported in this study could be related to the fact that study participants who were very unsatisfied with care or were extremely ill declined to participate when they were approached for enrollment. It is difficult for us to comment on the prevalence of chronic illness in the study cohort because the protocol for medical record review did not include an assessment for underlying cardiovascular disease or cancer. Lastly, there may have been a Hawthorne effect prompting positive behavior change by nephrology teams resulting in higher than expected DAS scores.

Conclusions

This study adds important insight into the largely unexplored questions surrounding decisional capacity and decisional satisfaction for dialysis initiation in hospitalized patients. Although additional studies are needed, our data suggest that a substantial percentage of patients who initiate dialysis in the hospital may be unable to participate in that decision. In addition, despite the high mortality associated with severe AKI requiring dialysis initiation, patients and their SDM’s are either not being told about alternatives to dialysis or cannot recall a discussion about alternatives. Further qualitative work is needed to clarify how patients who survive critical illness with ongoing dialysis needs feel about various aspects of their care. This kind of inquiry can help improve the delivery of patient-centered care, especially among those patients who emerge from acute illness with ESRD.