Recruitment of Participants

Inclusion criteria were (1) aged >18 years, and (2) either estimated glomerular filtration rate (eGFR) <25 mL/min/1.73 m² or on dialysis (HD or PD) for at least 3 months. Recruitment and data collection occurred between June and December 2013. We, with the help of some of the advisory panel members, recruited participants both through nationwide social media outreach and locally (Figure 1. Aim 1). The national outreach involved email blasts and postings on Facebook and Twitter in collaboration with the National Kidney Foundation and the American Association of Kidney Patients. We received a high volume of responses primarily through phone messages. Only those we could recontact by phone and who self-identified as CKD-ND, HD, or PD patients were eligible for inclusion in the study. In Michigan, social workers on the study team approached potential participants in person at renal clinics or dialysis units. Participants provided informed consent either verbally before the start of telephone interviews or in person. Participants also received a $25 gift card upon completion of the interview.

Figure 1. Aim 1

Recruitment Flow of Study Participants.

Data Collection

Study investigators conducted a 1-day interviewer training session that offered background information about the study, tips and guidelines on conducting qualitative interviews, coaching, and role playing for various scenarios. Between June and December 2013, 2 trained interviewers conducted digitally recorded and transcribed telephone interviews (30-40 minutes) of study participants.

Data Analysis

Two independent coders entered interview transcripts into NVivo 10 ([computer program]. QSR International Pty Ltd; 2014), coded the qualitative data, and identified common themes using content analysis. Coders discussed and resolved discrepancies. We collected theme categories in a codebook that included both overarching themes identified before coding and subthemes that emerged directly from patient responses. We identified common themes across all patients as well as within each of the 3 patient subgroups. We further classified yes/no responses on the patients' perceived role in selection of dialysis modality into 4 mutually exclusive categories: (1) “strong yes”—yes response was consistent with an informed or deliberate decision; (2) “weak yes”—yes response, but the transcript indicated medical conditions determined modality choice, or that the patient felt pressure to choose 1 type of dialysis; (3) “no”—no response consistent with not having made the decision; and (4) “combined”—response consistent with making the decision collaboratively with a doctor or a family member.

We calculated standard descriptive statistics (ie, means and frequencies) for quantitative data across patient groups using SAS, version 9.2. To test for differences across patient subgroups, we used chi-square tests of homogeneity for categorical variables and analysis of variance with a Bonferroni correction for multiple comparisons for continuous variables.

Study Sample

Among 302 responders to the national outreach, we interviewed 72 patients (Figure 1. Aim 1). In Michigan, we approached 181 patients and interviewed 109 respondents. The most common reason for not completing the interview was inability to reach the participant to conduct the interview after he or she had provided informed consent. We conducted a total of 181 (72 + 109) interviews, and we included 180 in this analysis; we excluded 1 interview because the participant lived outside of the United States, where clinical practices and education on RRT may be different. While participants represented the 4 major geographic regions, most resided in the Midwest (Michigan). Of participants, 65 had CKD-ND, 77 were on HD, and 38 were on PD. Some had prior dialysis experience.

Patient Characteristics

Patient characteristics (demographics and health status) were markedly different across CKD-ND/dialysis modality groups (Tables 1 and 2. Aim 1). Compared with the US Renal Data System data from 2012, the study sample was younger and included a higher percentage of females and African Americans/Black participants.⁴⁰ PD participants ranked their health slightly better than did either HD or CKD-ND patients but were also more likely to report that kidney disease limited their daily activities. A higher percentage of CKD-ND patients reported having diabetes and high blood pressure compared with HD and PD patients.

Table 1. Aim 1

Study Sample's Demographics, Overall and by CKD-ND/Dialysis Modality.

Table 2. Aim 1

Study Sample's Health Characteristics, Overall and by CKD-ND/Dialysis Modality.

Factors Important to Patients When Choosing Dialysis

As Figure 2. Aim 1, Panel A shows, overall, the 3 most important factors identified by patients were keeping as much independence as possible (96%), issues related to quality and quantity of life (94%), and flexibility in daily schedule (92%). The 3 factors less often cited as important were concern about appearance (39%), spending time with other patients at the dialysis center (37% of HD patients; not asked of PD patients), and worrying about how dialysis will affect others (36%).

Figure 2. Aim 1

Factors Important to Patients When Choosing Dialysis.

Differences were observed among patient groups, including CKD-ND/dialysis modality (Figure 2. Aim 1, Panel B), age (Panel C), and marital status (Panel D). Going to work or school was important to a larger percentage of PD patients compared with CKD-ND patients and for participants aged 45 to 49 compared with younger (<45) and older (60 and older) patients. Patients in the youngest age group affirmed concern about physical appearance more often compared with patients in the older age groups.

Further analysis revealed several subthemes. For example, different patient subgroups defined “quality and quantity of life” differently. Compared with PD patients, HD patients were more likely to respond with themes about extending life (quantity of life) and less likely to respond with quality of life themes. A common theme was that quantity of life was most important because patients were on dialysis to stay alive; dialysis had negatively affected the quality of their life, but within their limitations they felt they were choosing the best dialysis option. However, PD patients responded that quality of life was very important and that PD allowed them to take part in hobbies and be engaged in activities and maintain a more normal lifestyle. CKD-ND patients responded that quality of life meant being able to maintain a somewhat normal lifestyle (eg, continue with hobbies or household activities) after starting dialysis, but they worried dialysis would negatively impact their quality of life by making them feel tired or run down.

Choice of Dialysis Modality

Perceived role in the choice

Among 115 participants on dialysis (either HD or PD), more than a third felt that their dialysis modality had not been their choice, 5.2% responded that their modality choice was a combined decision, and 62.6% stated it was their choice (Table 3. Aim 1). However, patients' perceptions of their role were dramatically different between dialysis modalities: 94.7% of PD vs 46.8% of HD patients said the decision was largely their choice. Among patients responding “yes,” a much higher percentage of PD patients were classified as “strong yes” vs “weak yes” (88.9% vs 11.1%) than with HD patients (61.1% vs 38.9%).

Table 3. Aim 1

Perceived Role in Choice of Dialysis Modality: Participants' Responses to the Question, “Do You Feel That the Decision to Go on Hemodialysis/Peritoneal Dialysis Was Largely Your Choice?” (N = 115).

Patients responding “no” identified acute medical need or crisis situation and doctor's decision as the primary factors attributed to not having a choice. The “weak yes” themes suggest that medical conditions strongly governed patients' modality decision. Informed choice, fits lifestyle, switched from HD to PD, and investigated options about dialysis were the main themes within the category of patients grouped as “strong yes.”

Factors contributing to the choice of one dialysis modality over another

Patients were asked to explain why they chose HD or PD (Table 4. Aim 1). Patients who chose HD and classified as “weak yes” identified medical condition as a common theme. For PD patients, a common theme was side effects from HD. Among HD patients classified as “strong yes,” themes included fear of infection from pd and having trained medical personnel administer dialysis therapy. PD patient responses classified as “strong yes” were associated with the themes of better quality of life on PD, convenience of home dialysis, and ability to work. Individual themes emerging from choice of dialysis modality were not mutually exclusive, and based on some patient responses, multiple themes emerged.

Table 4. Aim 1

Factors Contributing to the Choice of One Dialysis Modality Over Another: Participants' Responses to the Question, “What Led You to Choose Hemodialysis/Peritoneal Dialysis”? (N = 115).

Metathemes related to the choice of dialysis modality

We took the responses to “Was the dialysis modality largely your choice?” (Table 3. Aim 1) and the corresponding individual themes for the question “What led you to choose HD/PD?” (Table 4. Aim 1) and mapped them to the 2 metathemes (Table 5. Aim 1). The metathemes represent higher-level conceptual categories containing the individual themes.

Table 5. Aim 1

Metathemes Related to the Choice of Dialysis Modality; Perceived Benefits and Perceived Risks or Constraints.

The 2 metathemes suggest patients primarily considered perceived benefits or perceived risks when making their decision about the type of dialysis. Perceived benefits include maintaining independence, quality of life, continuing daily activities, ability to work, convenience of doing dialysis at home, and making an informed choice about dialysis. Perceived risks or constraints contains themes about medical conditions constraining the decision on modality choice, negative side effects from dialysis, starting dialysis in a crisis situation, fear of infection, and greater comfort having trained medical personnel administering dialysis. In some cases, patient responses resulted in the emergence of multiple themes associated with both metathemes.

Regardless of perceived choice in the dialysis decision, HD patients tended to highlight burden, medical concerns, or limitations of treatment—characteristics associated within the perceived risks or constraints metatheme. In contrast, PD patients highlighted maintaining some semblance of their lifestyle, which we associated with the perceived benefits metatheme.

Survey Design

We developed a 39-question survey to assess patients' experiences with the dialysis modality decision and factors that patients had previously identified as important (patient-centered outcomes).⁵⁴ The advisory panel tested the survey for readability and comprehension and helped review and finalize survey questions.

The survey asked whether the participant was told that he or she had a choice between PD and HD when starting dialysis and to indicate if his or her involvement in this decision was more, less, or just what was desired. The survey proceeded with 3 sets of questions: (1) Patients ranked the degree to which 10 groups of family members, peers, and clinical staff were involved in their dialysis modality decision. (2) Patients rated their level of agreement with 9 statements focused on recollection of their experiences and satisfaction with their dialysis modality decision. Additionally, patients indicated whether the information they had received before starting dialysis was more, less, or just the amount that they had wanted, and whether they and their doctor agreed on the type of dialysis best for them. (3) Patients ranked the degree 16 different factors were affected by dialysis. We designed the survey for both paper and electronic (tablet) formats and provided both English and Spanish versions.

Recruitment of Participants

The DOPPS and PDOPPS are ongoing, international prospective cohort studies of dialysis facility practices and patient outcomes for adult HD and PD patients, respectively.^55-57 The Empowering Patients on Choices for Renal Replacement Therapy (EPOCH-RRT) survey was administered to patients as an additional patient questionnaire in the DOPPS and PDOPPS studies. All DOPPS and PDOPPS consented patients were eligible for the EPOCH-RRT study. Study coordinators targeted eligible patients between February 2015 and August 2015 to participate in the EPOCH-RRT survey based on patient visit schedules and staff availability (Figure 1. Aim 2). Some patients departed the dialysis facility before study coordinators could approach them with the EPOCH-RRT survey or were unable to participate due to other reasons (eg, cognitive, physical, language, or social impediments); others were approached for participation but unwilling to complete the EPOCH-RRT survey. Facilities were randomly assigned to receive the survey on either paper or tablet platforms.

Figure 1. Aim 2

Flow of Study Participants.

Statistical Analysis

For questions related to involvement of families and peers in the dialysis modality decision, we treated the responses as continuous outcomes. For outcomes on experiences and satisfaction with the dialysis modality decision, we dichotomized responses into agreement (agree or strongly agree) vs nonagreement (strongly disagree, disagree, or neither agree nor disagree). For outcomes on factors important to patients, we dichotomized responses into a large impact (very much or extremely) vs not large impact (not at all, somewhat, or moderately). We excluded patients who reported not applicable from analyses of each corresponding question and excluded missing responses for each question. For dichotomized outcomes (experiences and satisfaction with the dialysis modality decision and factors important to patients), we used generalized estimating equation (GEE) logistic regression models to compare outcomes between HD and PD patients. We used an exchangeable working covariance matrix to account for patient clustering within facility in GEE models. For continuous outcomes (involvement of families and peers), we used linear mixed regression models to compare dialysis modality, accounting for clustering by including a random intercept for each facility. In all models the primary predictor was dialysis modality and adjusted for age, sex, Black race, years on dialysis, and diabetes. In sensitivity analyses, we added paper or tablet platform used for collecting survey data as an additional adjustment factor and tested for effect modification by platform by including an interaction term between platform and dialysis modality. We conducted all analyses using SAS, Version 9.4 ([computer program]. Cary, NC: SAS Institute Inc.; 2013).

Study Sample

Out of 807 PD and 1683 HD patients approached for participation in the EPOCH-RRT study, 614 (76.1%) PD patients and 1346 (80.0%) HD patients responded to at least 1 question in the survey (Figure 1. Aim 2). Among patients <65 years old, response rates across platform (ie, paper or tablet) were similar among both PD and HD patients; however, patients >65 years who were offered tablets had lower response rates than those offered paper surveys. Thus, we controlled for platform and explored effect modification by platform in sensitivity analyses. The median (interquartile range) number of questions answered was 36 (33-38) among PD patients and 35 (32-37) among HD patients. The amount of missingness for each question ranged from 3% to 7%, with the exception of question 2 about the amount of patient involvement in the dialysis modality decision compared with what the patient wanted. This question was left unanswered by 11% of PD patients and 35% of HD patients. Table 1. Aim 2 displays patient characteristics. Compared with HD patients, PD patients were, on average, younger, had shorter dialysis vintage, and were less likely to be Black and to have diabetes.

Table 1. Aim 2

Patient Characteristics, by Dialysis Modality.

Experience With Dialysis Modality Choice

PD patients were more frequently (93%) told they had a choice between dialysis modalities than were HD patients (66%). Ten percent of PD patients and 20% of HD patients felt their involvement in the type of dialysis they would start on was either more than or less than they wanted compared with just what they wanted.

Involvement of Family and Peers

Clinical staff members, especially nephrologists, were most frequently involved in the dialysis modality decision overall compared with involvement of family members and friends (Figure 2. Aim 2). Fewer PD patients than HD patients reported at least some involvement of primary care doctors (60% vs 70%) but slightly more involvement of a nephrologist in their dialysis modality decision (94% vs 92%). We observed greater differences in the 2 modalities for lack of involvement of other clinical staff. For example, 40% of HD patients and 22% of PD patients reported no involvement at all of nursing staff in the dialysis decision. More than 35% of all patients reported that they did not know someone on dialysis at the time of their modality decision. Among those who did know a peer on dialysis, more than 50% recalled no peer involvement. More PD patients than HD patients recalled at least some involvement of physician assistants and nursing staff in their dialysis modality decision. Also, more PD patients than HD patients reported at least some involvement of partners/spouses (79% of PD patients, with 55% reporting very much or extremely involved; 70% of HD patients, with 46% reporting very much or extremely involved). For both PD and HD patients, involvement of other family and friends was low to moderate (32% -60%) and mostly similar across dialysis modalities. In adjusted models, PD patients indicated more involvement than HD patients by physician assistants, nursing staff, partner/spouse, and adult child/children.

Figure 2. Aim 2

Involvement of Family and Peers in the Dialysis Modality Decision for PD and HD Patients.

Experiences and Satisfaction With Dialysis Modality Decision

Overall, HD patients felt less informed and less confident than PD patients at the time of the dialysis modality decision and were less satisfied with their dialysis modality choice (Figure 3. Aim 2). PD patients more often felt that the information received was enough and easy to understand, dialysis choices were explained, advantages and disadvantages of PD and HD were understood, and that they were happy with their dialysis decision compared with HD patients. Almost all PD patients felt their dialysis choices were explained easily and understandable, whereas ~20% of HD patients did not. Additionally, 11% of HD patients regretted their dialysis modality choice, compared with 6% of PD patients (P < .001). While 26% of PD patients reported the information, they had before starting dialysis was not the amount that they wanted but, rather, either more or less than they wanted (9% and 17%, respectively), 36% of HD patients reported they had either more or less information (11% and 25%, respectively; P = .178) than they wanted. Finally, 95% of PD patients and 84% of HD patients reported that they and their doctor agreed on the type of dialysis that was best for them.

Figure 3. Aim 2

Proportion of PD and HD Patients Who Agreed With Statements on Experiences and Satisfaction With the Dialysis Modality Decision.

Impact of Dialysis on Patients' Lives

For all factors, many patients reported that dialysis had a large impact (range, 17%-46%; Figure 4. Aim 2). HD patients were more affected than PD patients by 15 of 16 factors, although most differences were small. PD patients more often felt that their dialysis modality largely affected self-reliance compared with HD patients. In contrast, HD patients more often felt their dialysis modality had a large effect compared with PD patients on doing what I want in my free time, doing activities I am interested in (hobbies), drinking as much water as I want, eating what I like, and feeling healthy.

Figure 4. Aim 2

Proportion of PD and HD Patients Indicating a Large Effect of Dialysis on Patient-Centered Outcomes.

For all outcomes, similar results were obtained after adjusting additionally for platform (tablet vs paper). In analyses testing for interactions between modality (PD vs HD) and questionnaire platform (tablet vs paper), we found little effect modification.

Discussion

Through dissemination of our survey to DOPPS and PDOPPS patients, we found that PD patients were more informed and engaged in dialysis modality decision-making compared with HD patients. This may be expected, given that PD patients undergo intense training coordinated by clinical staff and that this dialysis technique impacts household routine, space needs, and organization (eg, space to store PD supplies). Therefore, those who choose PD may already be more involved in their own care and likely more receptive to the education they receive. Nonetheless, the low involvement of several groups in the dialysis modality decision for both PD and HD patients demonstrates an opportunity to increase family and peer engagement to promote shared decision-making. Such engagement may result in a better fit of the dialysis modality with each patient's life as well as improved experience for their families and other caregivers. Furthermore, the large number of dialysis patients who did not know someone else on dialysis highlights a potentially useful but underutilized resource: Peer mentoring programs have proved to be successful in different clinical conditions,^58-61 and anecdotal evidence indicates that existing peer support programs in dialysis are highly valued by patients and their care partners.^62,63 By improving awareness of and access to peers, patients new to dialysis may benefit from increased practical information about dialysis, empathy and understanding, advice on coping strategies, and a greater sense of empowerment and agency.⁶³

We found large differences in understanding and satisfaction with current dialysis modality between PD and HD patients. PD patients were much more likely than HD patients to report that they had enough information during the dialysis modality decision, that the information given was easy to understand, and that they understood differences between dialysis modalities. Previous studies have found that deficiencies in knowledge are a barrier to choosing PD and that educational interventions can increase PD use.^64-66 Thus, those who choose PD are likely to be patients who have sufficient knowledge about dialysis modalities and willingness to participate in self-care. PD patients also more frequently indicated that they were happy with the modality they chose compared with HD patients. This result may reflect a more deliberate and informed decision-making process among PD patients and/or greater involvement in the dialysis modality decision. Still, more than 20% of PD patients did not know the disadvantages of their modality, and more than 10% did not feel they had easy-to-understand written information. Furthermore, some patients from both PD and HD groups reported not receiving enough information and expressed regret in dialysis modality choice. This finding is consistent with previous research that found anecdotal evidence of dialysis patients who were not satisfied with their dialysis modality decision process.^21,22,26 Therefore, opportunities exist to improve CKD education to increase understanding of dialysis modalities and satisfaction with treatment, especially among HD patients.

Patients perceived a moderate to high impact of dialysis on factors previously identified as important to patients in EPOCH-RRT interviews. Particularly, many patients felt their ability to rely on themselves and travel out of town was affected by starting dialysis. Several life-affecting factors were more frequently identified by HD patients than by PD patients, which may be explained by the differences in modalities. For example, clinical characteristics (eg, lack of residual urine output) of HD patients may require more restrictive diets and fluid intake, while technical aspects of HD (eg, intermittent dialysis in a facility setting) often limit the time HD patients have for their own interests. Some HD patients have also reported that dialyzing in a clinical setting and being surrounded by other patients makes them feel less healthy, although this opportunity to interact with other patients in the in-center setting was not always perceived as a negative aspect of HD.¹³

Overall, the proportion of patients who skipped each question was low, providing some evidence that the survey questions were appropriate and easily interpretable by most dialysis patients. This likely reflected the high engagement of the advisory panel in the development of the survey and the reviews of its questions. There was a higher amount of missingness for 1 question about the amount of the patient's involvement in the dialysis modality decision. The reasons for which 11% of PD patients and 35% of HD patients did not answer this question could include not having preconceived desires about involvement in the dialysis modality decision and/or unwillingness to admit low involvement. Both suggest that more effort should be made to give patients adequate choice and involvement in their dialysis modality decision process.

There are a few limitations of the aim 2 survey worth noting. First, survey questions asked the extent to which patients felt affected by dialysis—without options to indicate whether the effects were positive or negative. Therefore, the interpretation of differences between HD and PD patients must be speculated based on what is known about the different modalities. Second, we administered surveys to both incident and prevalent dialysis patients, so the time between dialysis initiation and survey was variable. Particularly for those who had longer dialysis vintage, recall bias may have affected survey responses related to the dialysis modality decision; however, we have no reason to believe that the recall bias would be different across PD and HD patients, indicating that our comparisons of interest may still have little bias. Third, we did not have information on whether patients in the study had contraindications to either dialysis modality, which also may have affected survey responses. For example, some HD patients may not have been eligible for PD, which limited their exposure to PD information. Still, the fact that HD patients sometimes felt that they did not have enough information about their own modality supports the conclusion that increased access to information on dialysis options is warranted.

Despite these limitations, our study has several strengths and important implications for end-stage renal disease patients, their families, and health care providers. By collaborating with an advisory panel and using analyses from qualitative data collected from patient interviews, our survey was specifically designed to focus on patient-centered outcomes. This approach—consistent with PCORI goals for multi-stakeholder engagement in research—was invaluable for informing the survey content and interpretation of results. We were able to compare factors important to patients in choosing a dialysis modality and living with dialysis treatments. We found that dialysis largely affects patients, which emphasizes the need to optimize the dialysis experience. By comparing the experiences of PD and HD patients, we identified significant differences between dialysis modalities. We found several aspects of the dialysis modality decision that require improvement, including patient education, access to peers, and other support. Increased efforts are needed to encourage multidisciplinary care and to provide resources, such as decision aids for patients facing the choice between dialysis modalities.

Decision Aid Development

We collaboratively developed the decision aid content based on literature review, the US Renal Data System data,⁷⁰ and results from aims 1 and 2; further, the advisory panel reviewed and refined the decision aid in an iterative method. Patients and social workers with different dialysis modality experience and who educate CKD patients reviewed the decision aid on several iterations of the refinement process. An additional set of people with no prior exposure to the study nor the decision aid and some with no exposure to kidney disease reviewed the content. We sought additional input from members of a kidney disease patient advocacy organization. We performed formal usability testing using think-aloud methodology by direct observation of 6 patients recruited from the University of Michigan Health System. We recorded reactions to the website and asked participants to describe what they liked or disliked about each page of the decision aid; this feedback further modified its design, structure, and content.

The finalized decision aid contained the following sections: (1) CKD and its progression, information and comparison of PD and HD from the patient perspective, and (3) value clarification exercise to map personal preferences to dialysis modality features. It included information on potential lifestyle changes associated with each option and consequences of changing one's mind after choosing either option. The decision aid integrated quotes from patients collected from interviews in aim 1 and tips for talking with health care professionals collected during the refinement process. Printing options were provided in sections that might be useful when discussing dialysis options with medical staff.

Per IPDAS,^32,71 the decision aid addresses all the qualifying criteria—that is, describes kidney disease, explicitly states the dialysis treatment decision between HD and PD, describes these 2 options, and describes positive and negative features of each option and side effects of both.

The following IPDASi v3.0 certification criteria were addressed: balanced information on both HD and PD; each page on the website references the funding source (Patient-Centered Outcomes Research Institute); offers additional resources and information about research used to develop the decision aid; and the year of website publication and terms of use and privacy policy. The website, which will be managed and updated by Arbor Research Collaborative for Health, can be updated whenever new information is available.

The design focused on intuitive navigation and accessibility of information. For the study, we designed the website to collect questionnaire data and walk users through all the steps without skipping ahead. We provided hover-over definitions for commonly used terms throughout the site and logged progress so that users could resume where they left off when unable to complete in a single session.

Study Design

We recruited advanced CKD adults (eGFR <25 mL/min/1.73 m²) with internet access to test the decision aid from CKD clinics in southeast Michigan; we also conducted national online outreach with the help of the National Kidney Foundation and American Association of Kidney Patients (Figure 1. Aim 3). Each of the 4 recruiters, immediately after obtaining informed consent, provided the participant with a user login ID chosen sequentially from a list generated by the study team. The list of user login IDs provided to each recruiter was ordered to alternate between intervention and control arm user login IDs, but neither the recruiter nor the participant could discern the assignment based on the login ID. This ensured even distribution of intervention and control arm assignments of consented participants.

Figure 1. Aim 3

Recruitment and Flow of Study Participants.

Once the participant granted informed consent, we provided instructions, login information, links to the test website, and contact information for technical or other support. Participants could access the study website from their own computers or portable devices by following the instructions and using the login credentials provided. Study coordinators could track task completion for each participant who logged into the website. They followed up weekly with consented participants to check on any technical issues and to promote study completion.

Participants in the control arm were required only to complete 1 questionnaire and click the submit button before accessing the decision aid. Participation in the control arm was considered complete once the questionnaire was completed and the submit button was clicked. Participants in this arm were included in the analysis if they answered all questions in the control questionnaire. Participants in the intervention arm were required to click on answers for all the pretest questions and click the submit button in order to proceed to the decision aid. Similarly, participants in the intervention arm were asked to click a button to indicate they had completed review of the decision aid and this would enable them to proceed to the posttest. All participants consented to complete each questionnaire in 1 sitting.

We designed the website to force intervention arm participants to click through sections of the decision aid sequentially and at their leisure. Participants could return to a section at any time, as many times as needed. The last page of the decision aid study website required participants to click a button to proceed to the posttest. We considered intervention arm participants to have completed the study if they answered all questions in the posttest and clicked the submit button.

Questionnaire Design

The questionnaires developed to test the efficacy of the decision aid in promoting shared decision-making included several established and/or validated measures of the following parameters: preference for shared decision-making,⁷² decisional conflict,⁷³ decision self-efficacy,⁷⁴ knowledge based on the contents of the decision aid and adapted from Cavanaugh,⁷⁵ literacy,⁷⁶ numeracy,^77,78 and demographics. The posttest also included the Preparation for Decision Making Scale⁷⁹ and questions to help assess usability, satisfaction with the decision aid, adequacy, and relevance and quality of content, as well as open-ended questions for positive and negative feedback.

Recruitment of Participants

Inclusion criteria were (1) aged >18 years, (2) eGFR <25 mL/min/1.73 m², (3) internet access through a computer or tablet, and (4) English language fluency. We recruited participants through both nationwide social media outreach and local efforts (Figure 1. Aim 3). The national outreach involved email blasts and postings on Facebook and Twitter in collaboration with the National Kidney Foundation and American Association of Kidney Patients. We received a high volume of responses, primarily through emails and telephone messages. We tracked only those who could be recontacted by telephone and self-identified as CKD, HD, or PD patients and who met all eligibility criteria. Clinic staff reviewed patient visit schedules to identify potential participants meeting clinical criteria. Social workers on the study team approached these patients at the renal clinic for interest in participation. We obtained informed consent either verbally before the start of telephone interviews or in person. Participants received a $25 gift card upon completion or attempted completion of study questionnaires. Local institutional review boards (Ethical and Independent Review Services E&I #13016, Henry Ford Health Systems IRB #8144, University of Michigan IRBMED HUM00073058) approved all study procedures. We considered enrolled participants who did not complete the study as lost to follow-up after 5 unsuccessful attempts to contact them by phone or email. One month before data collection completion, we attempted to contact every patient who had consented but had not completed the study.

Statistical Analysis

We tested for differences between the intervention and control arms using pretest intervention arm responses and control arm responses. We statistically tested differences in demographic composition regarding age, race, sex, education, ethnicity, and numeracy based on t tests, Pearson's chi-square tests, and Fisher exact tests. We compared outcomes between the pretest and posttest within the intervention arm using paired t tests, Wilcoxon signed rank tests, and tests for marginal homogeneity. We compared the intervention posttest with the control arm using unpaired t tests, Wilcoxon rank sum tests, and Pearson's chi-square tests.

We also assessed whether the effects of the decision aid differed across subgroups regarding self-efficacy, decisional conflict, preparation for decision-making, and knowledge of dialysis options. Thus, we tested whether differences between intervention pretest and posttest responses and differences between intervention posttest and control arms differed across age, sex, education level, or race groups. We used generalized estimating equation logistic or linear regression models for these tests by including an interaction term between subgroups and different arms in models.⁸⁰ Models accounted for the correlations within subjects when comparing pretest and posttest intervention arm responses using an exchangeable correlation structure and sandwich-type estimator for standard errors.

Study Sample

Figure 1. Aim 3 summarized the participant flow. In Michigan, we could not reach all screened patients at the renal clinics; some were ineligible for participation as determined by clinic staff. Several eligible patients could not be contacted either in person or over the phone for consent. Among those who were available and approached, reasons for ineligibility included prior experience of dialysis or patients having started dialysis by the time they were approached for consent. A large number of the patients did not have internet connectivity or access to a computer. Of those patients who declined to participate because they were not interested or were not comfortable providing consent (131), approximately 24% of these patients self-described level of computer literacy (23) or English fluency (8) as not sufficient for study requirements. Other patients were unable to participate because their eyesight was too poor for the study tasks.

We received informed consent from 234 participants. Some consented participants (96) across both control and intervention arms did not start the study, and 78 were not reachable by phone or email (Figure 1. Aim 3). A total of 140 participants logged into the site and started the study. Demographic information was self-reported and collected from participants in the control and pretest questionnaires for the 2 arms, respectively (Table 1. Aim 3). Seven participants in the intervention arm started the study and completed the pretest but did not go on to finish the posttest. The remaining 133 participants started and completed the study. Fifty of the 63 intervention arm participants (79.3%) completed the pretests and posttests within 1 week, with 60% having completed both tests on the same day. Only 5 participants needed more than 1 month to complete both tests.

Table 1. Aim 3

Questionnaire Design, Distribution of Sections in Control and Intervention Arms.

Patient Characteristics

Patient characteristics in the control and intervention arms had similar demographic composition regarding age, race, sex, education, ethnicity, and numeracy (Table 2. Aim 3). Patients were mostly White, and almost all had graduated high school and considered English as their native language. The Subjective Numeracy Scale is a self-report measure of perceived ability to perform various mathematical tasks.^77,78,81 Both arms were similar in terms of overall subjective numeracy as well as ability and preference subscales.

Table 2. Aim 3

Participant Characteristics, by Control and Intervention Arms.

Efficacy of the Decision Aid

Reduction in uncertainty

In the control arm and treatment arm, both before and after using the decision aid, patients reported what type of dialysis they might do when starting treatment. Both the control and pretest results suggest that both arms had similar baseline uncertainty on treatment choice: 40% and 47%, respectively. Those in the treatment arm reported a significant decrease to 16% in uncertainty on choice of dialysis type after using the decision aid (Table 3. Aim 3).

Table 3. Aim 3

Outcome Measures for Decision Aid Efficacy.

Feedback on the relevance, usability, and satisfaction with the decision aid

Greater than 90% of participants in the intervention arm felt that the decision aid helped somewhat to a great deal, both for preparing for dialysis and for follow-up with care providers, with approximately 80% responding “quite a bit” and “a great deal” (Figures 2 and 3. Aim 3).

Figure 2. Aim 3

Summary of Responses Related to the Helpfulness of the Decision Aid.

Figure 3. Aim 3

Opinions on the Decision Aid Website.

Only 1 person did not like the website, and 2 people said they would not recommend the decision aid to others. Most participants (92%) felt the decision aid was balanced and not slanted toward HD or PD, 88% trusted the website content, and 89% agreed/strongly agreed the content was relevant to them, with 49% agreeing the decision aid was extremely helpful in understanding dialysis options.

Participants provided open-ended feedback on what they liked about the decision aid and how it might be improved. Overall, participants most frequently cited that the decision aid was informative (65%) and helpful (40%), with 22% providing critical comments and 3% unsure about what they thought of the decision aid. One participant said, “There was a lot of very good information to assist me to make a very serious decision when and if the time comes. I hope I won't need to make the choice, but if necessary, I can make a knowledge-based decision that is best for me. I do appreciate the unbiased information. Thank you!” Another wrote, “I was knowledgeable already on 80% of the information, but it was helpful. Since I am hopefully still years away from needing dialysis, reviewing this info was a little depressing. I hope all treatments improve.”

When asked what they liked about the decision aid, participants often expressed that they found the website easy to navigate and well organized, and they liked the value clarification exercise and the testimonials from patients. One participant said, “I like the 'feel' perspective… the facts of each treatment can be found everywhere, but not often do you see the feelings of the patient put in consideration. I had mixed feelings about which way to go, but this site helped a great deal.”

A few participants (10%) had suggestions for improvements in usability, such as, “Make this available to patients that do not have a computer” and “Maybe add video clips during the decision aid to make it more interesting and captivating. People typically like to view videos on their computers.” Another suggested, “Speak more about the positives in making this decision. Include more testimonials on improving quality of life.”

We received considerable positive feedback regarding ease of readability and comprehensiveness of content, especially the juxtaposition of PD and HD. The most critical feedback was concerns about missing information on dietary restrictions (29%); how patients deal with side effects; RRT options other than PD, HD, and the slow overnight in-center HD option; effects of PD on family members; complications from treatment; and data on demographics and life expectancy. For example, 1 participant asked, “What about home hemo? And what about info about having to change dialysis types if, for instance, your abdominal wall becomes tough and can no longer filter?” Of respondents, 41% didn't have any suggestions for improvement.

Discussion

Decision aids have demonstrated improved communication between patients and their doctors. A comprehensive review of decision aids suggests that more detailed decision aids are better than simple decision aids in improving knowledge and lowering decisional conflict scores, factors that are related to feeling uninformed and unclear about one's personal values.³³ Decision aids have proved to improve knowledge, increase risk perception, decrease decisional conflict, and enhance participation in shared decision-making among the elderly.⁸² However, few decision aids have been developed specifically for people aged >65 years to characterize important components for this demographic.⁸² We are aware of a small number of decision aids related to dialysis treatment modality choice and that incorporate value clarification tools: 3 in the United States and some developed outside the United States in other health care contexts.^67,68,83,84 The main difference between our choosingdialysis.org decision aid and the others developed in the United States are the use of qualitative interviews and patient-reported data to identify factors important to patients; more detailed information on the 2 main dialysis treatment options; involvement of the advisory panel throughout the study, including the development of the decision aid; the use of quotes and tips from health care professionals reflecting stakeholder perspectives throughout the decision aid; and testing the decision aid with the end-users described in this paper to evaluate the effect of the decision aid on decision-making outcomes.

We developed the content and format of the EPOCH-RRT decision aid for patients deciding on dialysis treatment options in collaboration with an expert advisory panel, and we were guided by decision aid development experts. The advisory panel included patients who had faced this decision and social workers supporting CKD patients in their transition to dialysis. Nephrologists and clinical researchers also reviewed content. Additionally, the EPOCH-RRT decision aid content complements clinical information with the experiences of individuals undergoing dialysis in the form of quotes to share experiences, quotes from their families, and tips from health care professionals to address practical issues. Also included is a value clarification exercise that assists in identifying factors that matter most to the reader or user and how these factors influence which option may best suit him or her. Some participants in our study provided feedback that they found this interactive value clarification tool helpful, but the decision aid literature is uncertain of the benefits of the value clarification exercise to decision-making outcomes.^85,86 Users, patients, and family members are encouraged through different print options to leverage the information gained through the decision aid as a tool to inform discussions and increase communication with their health care teams. Involvement of family members has been suggested as beneficial for patients' health outcomes.⁸⁷

Similar to other decision aids,³³ this dialysis treatment choice decision aid improved knowledge and reduced decisional conflict but did not significantly improve decisional self-efficacy. We carried out subgroup analyses for these outcomes between intervention pretest and posttest responses and differences between intervention posttest and control arms to test for differences across age, sex, education level, or race groups. We observed no differences for any outcomes other than the knowledge test, for which Black participants had lower scores before accessing the decision aid but showed greater improvement in knowledge upon accessing the choosingdialysis.org decision aid than did non-Black participants. Racial disparities in choices for RRT are well documented,^88-90 and improving knowledge through CKD education has been proposed as 1 solution to overcoming identified barriers, such as patients' awareness of choices and disparities in shared decision-making and improved patient-centered care.^87,88,91-93

A recently published randomized controlled study of another web-based decision aid about depression treatment options also showed improvement only in knowledge and decisional conflict outcomes.⁹⁴ It is possible that multiple factors contribute to decisional self-efficacy, which would require a holistic socioecological approach to move the needle on this indicator.

Our work suggests that this decision aid—a new and effective tool developed through a stakeholder-engaged process—informs and supports CKD patients in making the difficult choice of dialysis modality. The broader implementation of this decision aid would complement current CKD education in clinical practice and could support both care providers and patients in shared decision-making, by facilitating communication about treatment options.

A limitation of this study is that self-selected participants were, on average, healthier, more educated, and younger than the US CKD population.⁹⁵ Patients with stage IV kidney disease (eGFR <25 mL/min/1.73 m2), the target population for this study, are often dealing with a heavy medication burden and a multitude of physical and mental symptoms related to kidney failure before the start of RRT.^96,97 These factors could have negatively contributed to willingness to participate in a research study and are reflected in the low consent rate, resulting in a less generalizable participant cohort. We envisioned a web-based format as the ideal way to quickly disseminate the decision aid to the broadest possible audience; however, lack of internet access and computer literacy limitations challenged recruitment efforts, which also contributed to differences in the composition of participants and the broader CKD population. Further studies are needed to address the multiple social and demographic disadvantages that may impact participation in shared decision-making for those facing the dialysis decision.

Seven participants in the intervention arm did not complete the study. Sensitivity analysis to compare the control and intervention groups, with and without the 7 participants in the intervention arm who did not complete the study, suggests that these departures did not affect our study results. Based on the study design, the time between pretest and posttest questionnaire responses might have resulted in drift in the responses in the intervention arm that would affect results, independent of the decision aid. Sensitivity analysis with and without the 13 participants in the intervention arm who had a gap of more than 7 days between completions of the 2 questionnaires did not change any of the measured outcomes nor the reported statistical differences.

Most participants found the decision aid helpful and would recommend it to others. We incorporated specific feedback from study participants and patient advocacy organizations to further modify the layout and readability of the decision aid. We improved the decision aid accessibility by modifying the language to a grade 10 Flesch-Kincaid readability level using the readability statistics package embedded in Microsoft Word. As the content of the website was finalized, we improved the navigation and architecture to ensure that information would be easily discerned. We improved graphics and layout based on feedback from target users. The advisory panel reviewed and approved the final content and design of the decision aid; it was released to the public at http://choosingdialysis.org/.