Prevalence and predictors of decisional conflict among older African Americans with advanced chronic kidney disease

Abstract

Background and Objectives

Decisional conflict (DC), or uncertainty about health-related decisions, is common among individuals with advanced chronic kidney disease (CKD), yet little is known about its prevalence and predictors among older African Americans, a population disproportionately affected by CKD. This study examined the prevalence of DC and how personal characteristics, perceived kidney disease knowledge, and clinical factors are associated with DC in this group.

Research Design and Methods

This was a cross-sectional study involving 125 African American adults aged 50 and older with Stage 4 or 5 CKD who had not yet started dialysis. Participants completed structured telephone interviews to assess the DC Scale, perceived CKD knowledge, depressive symptoms, anxiety, comorbidities, and sociodemographic factors. Hierarchical multiple regression analyses were performed to identify predictors of DC.

Results

Participants reported moderate levels of DC (M = 36.69, SD = 23.2), with approximately 42% scoring above the threshold indicating decisional difficulty. In multivariate analyses, perceived CKD knowledge was the only significant predictor of DC (β = −0.58, p < .001), accounting for a substantial proportion of variance. Personal and clinical characteristics were not significant in the final model.

Discussion and Implications

Findings highlight the crucial role of perceived knowledge in shaping decision-making among older African Americans with advanced CKD. Interventions that improve disease-specific knowledge during the pre-dialysis period may reduce decisional conflict and promote informed, patient-centered care. Future research should explore tailored educational strategies and structural factors affecting decision-making in this population.

Innovation and Translational Significance:

Decisional conflict is common among older African Americans with advanced chronic kidney disease (CKD), with over 40% experiencing uncertainty that may delay treatment decisions. Perceived kidney disease knowledge was the only significant predictor of decisional conflict, underscoring the central role of patient understanding in decision-making. Clinically, these findings support integrating targeted CKD education and decision support into pre-dialysis care, particularly for populations facing structural inequities. Social workers and interdisciplinary teams can play a key role in delivering culturally responsive education. Translationally, developing tailored educational interventions and decision aids may reduce decisional conflict and improve patient-centered outcomes across the CKD trajectory.

Decisional conflict (DC) is central to the management of chronic kidney disease (CKD), particularly as individuals approach end-stage kidney disease (ESKD). Individuals face difficult treatment decisions that can negatively affect psychosocial well-being, delay decision-making, and worsen outcomes (O’Connor, 1995; Song et al., 2015). The process of selecting treatment options includes key decision-making points in the advanced stages of CKD. Decision-making about treatments during these stages is complex and fraught with important, life-altering choices. These decisions include dialysis modalities, kidney transplantation, conservative management, time-limited dialysis trials, or discontinuation of dialysis with end-of-life care. Even though older African Americans with advanced CKD disproportionately bear the burden of morbidity and mortality across the illness continuum (Kucirka et al., 2011), few studies focus on decision-making or DC in this group.

CKD is classified into stages 1–5, with higher stages indicating greater kidney damage. Stages 4 and 5 indicate advanced CKD and ESKD, respectively. “The complexity of interacting diseases and degradation of the quality of life imposed by late-stage CKD makes it a disease entity that is qualitatively and quantitatively different from early-stage CKD, with suboptimal outcomes in advanced-stage CKD” (Fishbane et al., 2015, p. 55). Participants in the present study were at Stage 4 or 5 and are referred to as having advanced CKD. CKD is also more common among individuals aged 65 and older (34%) (Centers for Disease Control and Prevention, 2026). African Americans experience a disproportionate burden of CKD and represent nearly 35% of those with kidney failure (National Kidney Foundation, 2022). This underscores the importance of focusing on older African Americans.

Older African Americans are likely to continue being affected by CKD as the aging population increases. The number of African Americans aged 65 and older, which was nearly 5 million in 2019, is projected to rise to almost 9 million by 2040 (Administration for Community Living, 2020). The prevalence of CKD is over three times higher among African American individuals compared to White individuals, and they constitute nearly 35% of those with kidney failure (National Kidney Foundation, 2022). These disparities are further compounded by the historical use of race-based algorithms in nephrology, particularly race-adjusted estimated glomerular filtration rate (eGFR) equations, which may have delayed diagnosis, referral, and treatment for African Americans with kidney disease. The movement away from race-adjusted eGFR equations was driven in part by growing recognition that race is a social rather than a biological construct and by concerns that race-based algorithms perpetuate inequities in kidney disease diagnosis and treatment (Delgado et al., 2021). Among older African American adults, these structural inequities may increase DC as individuals consider complex treatment decisions about dialysis, transplantation, and conservative kidney management within healthcare systems shaped by longstanding racial disparities.

Older African Americans are a particularly vulnerable group whose needs regarding dialysis decision-making at the start of dialysis have been previously understudied. This group has experienced a unique relationship with racism and discrimination, as many were born before the Civil Rights movement and have lived through overt and legalized race-based discrimination. It is reasonable to assume that older African Americans face a combination of ageism and racism, or both, due to their minority status. Consequently, older racial and ethnic minorities may encounter double injustices in healthcare (Farrell et al., 2022; Stone, 2012). This study aimed to examine the prevalence and predictors of DC among older African Americans with advanced CKD.

Conceptual framework

DC is defined as personal uncertainty about which course of action to take “when the choice among competing actions involves risk, loss, regret, or challenges to personal life values” (LeBlanc et al., 2009, p. 61). The guiding conceptual framework for this investigation is the Ottawa Decision Support Framework, which describes the relationships between participants’ decisional needs, decision quality, and decision support (O’Connor et al., 2002; Legare et al., 2006). O’Connor’s framework builds on earlier work on DC (Janis, 1959; Janis & Mann, 1977). According to O’Connor et al.’s (2002) framework, decisional needs include DC (uncertainty), knowledge and expectations, values, support and resources, decision type, and clinical/personal characteristics.

In the present study, the primary outcome, DC about CKD treatment modalities, was examined within this framework. In the Ottawa Decision Support Framework, knowledge is defined as cognizance of the health problem or situation, options, and outcomes (Legare et al., 2006). Here, knowledge was operationalized as participants’ perceived knowledge about kidney disease. Clinical characteristics were represented by self-rated health, emotional health status (depression and anxiety), and number of comorbidities. Personal characteristics included age, gender, marital status, annual household income, health insurance, and employment status.

Decisional conflict and chronic kidney disease

DC has been examined in a range of chronic conditions such as cancer (Lam et al., 2013; Manne et al., 2016; O’Connor et al., 1996), diabetes (Bailey et al., 2016; Huang et al., 2017), and hypertension (Montgomery et al., 2003). Prior studies on CKD reported that DC was associated with patients’ delayed decision-making, feelings of regret, decision satisfaction, and general uncertainty around treatment choices (Chen et al., 2018; McPherson et al., 2017; Vélez-Bermúdez et al., 2018). While these studies have outlined the relevance of DC in CKD, there is limited evidence regarding the prevalence of DC among patients with CKD.

The few studies focused on the prevalence of DC among pre-dialysis individuals with advanced CKD and ESRD have found moderate to high levels of DC. DePasquale et al. (2022) reported that over 75% of participants (N = 427) with advanced CKD who received nephrology care reported treatment-related DC, while Chen et al. (2018) found that nearly 40% of study participants reported DC levels higher than 37.5, which is the cutoff for difficulty implementing decisions. Goh et al. (2021) investigated treatment-related DC among pre-dialysis CKD patients, finding moderate levels of DC 29.70 ± 14.45, with more than a quarter of participants reporting severe DC (scores ≥37.5). Tan et al. (2019) found that DC 26.00 ± 19.80 was present among older adults (N = 103) who had already initiated dialysis. Additionally, a cross-sectional study of 464 patients with ESKD being evaluated for kidney transplantation found that 49% experienced DC (McPherson et al., 2017). Vélez-Bermúdez et al. (2018) examined the relationships among patient activation (patients’ levels of skill, knowledge, and confidence in managing their own health), treatment satisfaction, and DC among 64 patients approaching ESRD. Participants in this sample reported moderate levels of DC, with an average score of 31.30 ± 20.03, while 65% reported a DC score indicating decisional difficulties (>25), and 43% had a score indicating decisional delays (>37.5) (Vélez-Bermúdez et al., 2018). Little research has focused on identifying the personal and clinical characteristics associated with DC. Furthermore, the personal and clinical characteristics included in prior studies have varied widely and have yielded mixed results. Five previous studies on DC and CKD have included personal characteristics (Chen et al., 2018; DePasquale et al., 2022; Goh et al., 2021; McPherson et al., 2017; Vélez-Bermúdez et al., 2018), with education, work status, and age being the most examined (Chen et al., 2018; Goh et al., 2021; McPherson et al., 2017; Vélez-Bermúdez et al., 2018). No studies have included self-rated health. The number of comorbidities using the Charlson comorbidity index was reported in one study (DePasquale et al., 2022), while few studies included selected health conditions, such as heart disease (Vélez-Bermúdez et al., 2018), cardiovascular disease (McPherson et al., 2017), hypertension, and diabetes (McPherson et al., 2017; Vélez-Bermúdez et al., 2018), yielding mixed results. One study included depressive symptoms (Vélez-Bermúdez et al., 2018), with negative findings, and no study included anxiety diagnosis or symptoms.

Most studies examining DC among ESRD patients have focused on patients who had already initiated dialysis, with limited studies examining pre-dialysis CKD (Chan et al., 2022; DePasquale et al., 2022; Goh et al., 2021; Ho et al., 2020; Vélez-Bermúdez et al., 2018). The present study focused on the pre-dialysis period to contribute to this understudied part of the advanced CKD experience.

The present study is novel in its focus on the pre-dialysis period and in integrating a comprehensive set of predictors of DC in CKD. No studies have examined perceived kidney disease knowledge in African Americans, particularly older African Americans. In prior research, only dialysis knowledge (Chen et al., 2018) and transplant knowledge (McPherson et al., 2017) have been studied. Knowledge is especially critical for individuals living with advanced CKD, who also experience high rates of poor self-rated health (Robinson-Cohen et al., 2014), depression (Shiravanian et al., 2017), anxiety (Murtaugh et al., 2007), and comorbidities (Fraser et al., 2015), yet no studies have examined these clinical predictors simultaneously in relation to DC. Research specifically on older adults with advanced CKD is limited, with most studies including individuals aged 18 and older, and only a few with mean ages above 50 (e.g., Chen et al, 2018; Goh et al., 2021; McPherson et al., 2017). Furthermore, even less is known about the prevalence of DC among African Americans. To date, only two studies have included an exclusively African American sample (Song et al., 2009, 2010); the rest have included few or no African Americans. This is a significant gap in the current empirical research on DC and CKD.

Research questions and hypotheses for the current study

Based on the Ottawa Decision Support Framework (O’Connor et al., 2002) and gaps in the existing empirical evidence, the following questions were addressed. First, how much DC do pre-dialysis older African Americans with advanced CKD report? Next, three research questions (RQ) addressed the lack of evidence about the degree to which personal characteristics, knowledge, and clinical characteristics predict DC among older African Americans living with advanced CKD pre-dialysis.

RQ1: Are personal characteristics (i.e., age, gender, education, income, marital status, insurance status, and employment status) significantly associated with DC? It was hypothesized that older age, being female, lower education, lower income, being unmarried, having government-only insurance, and being unemployed would be associated with higher DC.

RQ2: Does adding perceived kidney disease knowledge explain significant additional variance in DC, controlling for personal characteristics? Higher perceived knowledge was expected to be associated with lower DC.

RQ3: Does adding clinical characteristics (self-rated health, depression, anxiety, and number of comorbidities) significantly increase the variance explained in DC, controlling for personal characteristics and knowledge? It was hypothesized that worse self-rated health, higher depressive symptoms and anxiety, and a greater number of comorbidities would be associated with higher DC.

Methods

Participants and procedure

This cross-sectional, observational study recruited participants aged 50 years or older with CKD stages 4 or 5 from an ambulatory pre-dialysis care clinic staffed with nephrologists, a nurse practitioner, and registered nursing staff at a large not-for-profit hospital system in the Midwest. This research was approved by the hospital IRB (STUDY2021055). Recruitment and data collection took place from July 2021 to March 2022. Data were collected through self-report measures obtained via telephone interviews conducted by the first author and two trained interviewers.

The hospital’s information technology department identified 572 patients who met the study inclusion criteria. A nephrologist or nurse practitioner reviewed medical records to exclude individuals with active psychosis or cognitive impairment. No potential cases were excluded for these reasons. Additionally, individuals who had undergone a kidney transplant or initiated dialysis in the past were excluded because they had already undergone a variant of the decision-making process of interest. After duplicate names were removed, 558 individuals were mailed study invitations. Of these, 300 were screened by phone; the rest (n = 228) were unable to be contacted. After excluding those who declined to participate (n = 96), those who were unable to complete the interview (n = 76), or those who had interviews with extensive missing data (n = 3), the final sample consisted of 125 participants with complete data. Participants received a $20.00 gift card for the study interview.

Measures

Decisional conflict

The outcome was a 16-item, five-response version of the Decisional Conflict Scale (DCS; O’Connor, 1995), with wording modified to fit advanced CKD. The DCS is the most widely used measure of DC or uncertainty across disease conditions (Garvelink et al., 2019). For the present study, the prompt before the DCS questions read: “If your [CKD] gets worse, which one of the following [CKD] decisions do you prefer?” In the original DCS, treatment options are left blank and can be filled out based on the specific illness being studied. For this study, four potential treatment options were provided: “Do you prefer Dialysis, Kidney Transplant, Symptom Management without Dialysis, or Are you Unsure?” After indicating their treatment preference or uncertainty, all participants were read 16 statements (e.g., “I know what options are available to me,” “This decision is easy for me to make”) and asked if they strongly agree (0), agree (1), neither agree nor disagree (2), disagree (3), or strongly disagree (4). Total DCS scores could range from 0 to 100, with higher scores representing higher DC. Scores below 25 suggest that the individual is likely to implement their decision with confidence, while scores equal to or greater than 37.5 are associated with feelings of uncertainty and may indicate a delay in decision-making (O’Connor, 1995). In this sample, the DCS showed acceptable internal consistency (α = .96).

Perceived CKD knowledge

CKD knowledge was measured by the Perceived Kidney Knowledge Survey (PiKS) (Wright-Nunes et al., 2011), a nine-item instrument designed to measure perceived kidney knowledge in patients not yet on dialysis (Wright-Nunes et al., 2011). The prompt was “Please choose the response that most closely reflects how much knowledge you have about each kidney topic” (e.g., “medications that help the kidney,” “medications that hurt the kidney,” and “your goal blood pressure”). Response options were “I don’t know anything” (1), “I know a little amount” (2), “I know a good amount” (3), and “I know a lot” (4). In this sample, the PiKS had acceptable internal consistency (α = .94). Higher scores represent higher levels of perceived knowledge, with a possible range of 9 to 36.

Clinical characteristics

Self-rated health asked, “In general, would you say that your health is (6) Excellent, (5) Very Good, (4) Good, (3) Fair, (2) Poor, or (1) Very Poor?” This was adapted from the Health and Retirement Study (HRS; Fisher et al., 2008) and the Health Information National Trends Survey (HINTS). HRS and HINTS use the same item stem but provide response options of Excellent, Very Good, Good, Fair, and Poor. The option “Very Poor” was added for the current study due to the target population.

Depressive symptomatology was measured by the Patient Health Questionnaire (PHQ-9), a measure of depression severity covering nine DSM-IV criteria symptoms (Kroenke et al., 2001). Participants were asked, “Over the past two weeks, how often have you been bothered by the following problems?,” such as “little interest or pleasure in doing things,” “feeling down, depressed or hopeless,” and “trouble falling or staying asleep, or sleeping too much.” PHQ-9 scores can range from 0 to 27, with each of the nine items scored from 0 (not at all) to 3 (nearly every day). Scores represent minimal depression (0–4), mild depression (5–9), moderate depression (10–14), moderately severe depression (15–19), and severe depression (20–27) (Kroenke et al., 2001). Prior research has indicated scores ≥10 were indicative of having major depressive disorder (Kroenke et al., 2001). In an inquiry of ethnically diverse primary care patients, the PHQ-9 was determined to have good internal consistency across ethnicity (African Americans, α = .80) (Huang et al., 2006). In this sample, the PHQ-9 also had acceptable internal consistency (α = .82).

The seven-item Generalized Anxiety Disorder (GAD-7) measured generalized anxiety (Spitzer et al., 2006). Participants were asked, “Over the last two weeks, how often have you been bothered by the following?” (e.g., “feeling nervous, anxious, or on edge,” “not being able to stop or control worrying,” “worrying too much about too many different things”). Response options were (0) Not at all, (1) Several days, (2) More than half of the days, and (3) Nearly every day. Total scores of 1–4 indicate minimal anxiety, 5–9 mild anxiety, 10–14 moderate anxiety, and 15–21 severe anxiety (Spitzer et al., 2006). When used as a screening tool, further evaluation is recommended when the score is 10 or greater (Spitzer et al., 2006). In this sample, the GAD-7 had acceptable internal consistency (α = .86).

Number of comorbidities (possible range 0–7) was measured by asking participants whether a doctor ever told them (0 = No, 1 = Yes) they have (1) High Blood Pressure or Hypertension, (2) Diabetes or High Blood Sugar, (3) Heart Disease or Heart Attack, Congestive Heart Failure, or Coronary Heart Disease, (4) Cancer, (5) Chronic Lung Disease (COPD, Emphysema, or Chronic Bronchitis), (6) Stroke, or (7) Arthritis. This list was adapted from the Health and Retirement Survey (Fisher et al., 2008).

Personal characteristics

Demographic variables included age, gender, education, income, marital status, insurance type, and employment status, which were assessed using structured survey items.

Analytic plan

All analyses were conducted in SPSS (Version 28). Univariate frequencies, descriptive statistics, and histograms were examined for outliers, adequate variability, skewness, and kurtosis. The frequency, mean, and standard deviation of the DCS were used to address the study’s exploratory question about DC prevalence. Pearson correlations and one-way ANOVA were used to examine bivariate relationships. Cohen’s (1988) guidelines were used to assess effect size. Assumptions for ordinary least squares regression (linearity, homoscedasticity, normality, and independence of residuals) were examined.

Multiple regression analysis

A hierarchical ordinary least-squares (OLS) multiple regression analysis of DC was conducted to address the research questions. Personal characteristics (demographics) were entered first (RQ1), knowledge was added second (RQ2), and clinical characteristics (self-rated health, depressive symptoms, anxiety, and number of comorbidities) were added third (RQ3). R² and its p-value were used to determine how much variance in the outcome was accounted for by each block. Because of the moderate sample size and large number of predictors, the adjusted R² ($R_{\mathrm{adj}}^2$), which adjusts for N and the number of predictors, was also examined as a more conservative estimate of variance explained (Warner, 2012). Change in R² (ΔR²) and its p-value were examined to see how much additional variance was explained as Block 2 and Block 3 variables were added to the model. To test hypotheses, unstandardized coefficients (b), standard errors, p-values, and beta coefficients are reported. Squared semi-partial correlations (sr²) provide information about the proportion of variance uniquely predictable from each independent variable, thus determining the relative importance of each predictor (Warner, 2012, p. 603).

Results

Sample characteristics

Sample characteristics are shown in Table 1. The mean age was close to 70. Almost two-thirds of the sample were women. More than a third of participants were married or partnered, while most reported that their education level was either at a high school level (or equivalent) or some college. Most of the sample reported being retired, although a few were working full or part-time. Over three-fourths (76.4%) reported having either Medicare or some combination of Medicare and another insurance, and nearly one-third of individuals reported earning $20,000 or less. Health insurance was not included in subsequent analyses due to a lack of variability, and income was not included due to extensive missing data. Nearly one-half of the sample reported their health as fair, while almost one-fifth said their health was very poor to poor.

Table 1.

Characteristics of the sample.

Characteristics		n (%)	M (SD)	Observed range
Age			69.96 (9.34)	51–88
Gender	Men	44 (35.2)
	Women	81 (64.8)
Marital status	Married	36 (28.8)
	Significant other	9 (7.2)
	Never married	31 (24.8)
	Widowed	26 (20.8)
	Separated	1 (0.8)
	Divorced	22 (17.6)
Education	6th grade	1 (0.8)
	8th grade	2 (1.6)
	9th grade	2 (1.6)
	10th grade	5 (4.0)
	11th grade	4 (3.2)
	12th grade	41 (32.8)
	1 year college	16 (12.8)
	2 year college	23 (18.4)
	3 year college	5 (4.0)
	4 or more years college	13 (10.4)
	Post college (17 years or more)	13 (10.4)
Employment status	Working full time	8 (6.4)
	Working part time	9 (7.2)
	Retired	77 (61.6)
	Unemployed—looking for work	4 (3.2)
	Unemployed—not looking for work	27 (21.6)
Insurance (n = 122)	Medicare only	23 (18.4)
	Medicaid only	10 (8.0)
	Medicare and Medicaid	32 (25.6)
	CHAMPUS or VA	2 (1.6)
	Private insurance	16 (12.8)
	Medicare and private insurance	37 (29.6)
	Other plan	2 (1.6)
	No insurance	0 (0.0)
Household income (n = 112)	Less than $20,000	41 (32.8)
	$20,000 to $34,999	27 (21.6)
	$35,000 to $49,999	19 (15.2)
	$50,000 to $74,999	13 (10.4)
	$75,000 or more	12 (9.6)
Self-rated health	Very poor	7 (5.60)
	Poor	17 (13.6)
	Fair	61 (48.8)
	Good	28 (22.4)
	Very good	9 (7.2)
	Excellent	3 (2.4)
Perceived CKD knowledge			21.64 (7.6)	9–36
Depressive symptomatology			5.37 (5.3)	0–24
Anxiety			3.04 (4.3)	0–21
Number of comorbidities			3.34 (1.2)	1–6
Decisional conflict			36.69 (23.2)	0–100

Note. CKD = chronic kidney disease. Percentages and n are presented for categorical variables; means and standard deviations are presented for continuous variables. Sample size is N = 125 unless noted otherwise.

On average, participants reported a moderate knowledge score. Slightly more than four out of 10 individuals (44.8%) scored below 19 (the bottom third of the possible range), and approximately three out of 10 participants (28.8%) rated their knowledge of CKD at over 26 (the top third of the possible range). The top three knowledge items that had the most “I don’t know” or “I know a little” responses were medications that help the kidney (n = 91, 73%), medications that hurt the kidney (n = 80, 64%), and understanding treatment options if kidney disease got worse (n = 80, 64%).

The mean depression score falls in the range of mild depression (Kroenke et al., 2001). Around half (52%, n = 66) of the sample met the criterion for no or minimal depression, 28% (n = 36) met the criterion for mild depression, and approximately 18% (n = 23) met the criterion for moderate or greater depression (including three participants who reported depressive symptoms in the severe category, scoring ≥20). The mean score for anxiety falls below the cut-point for mild anxiety (Spitzer et al., 2006). Most participants (n = 100, 80%) reported no or minimal anxiety with scores less than five. Only 20% (n = 25) of the sample had scores of mild or greater anxiety. That 20% included just over 8% (n = 11) who reported scores ≥10, indicative of the need for further evaluation from a clinical standpoint, and 4% (n = 5) who reported severe anxiety (scores ≥15). Respondents reported being diagnosed with three comorbid health conditions on average, in addition to CKD: 96% reported having high blood pressure, nearly two-thirds (62.4%) were diagnosed with diabetes, and almost one-fifth (19.4%) reported a history of stroke.

Prevalence of decisional conflict

Preferences for treatment options were dialysis (n = 31, 24.8%), kidney transplant (n = 53, 42.4%), symptom management with no dialysis or no kidney transplant (n = 28, 22.4%), or unsure (n = 13, 10.4%). Approximately 42% (n = 53) of DC scores were above 37.5, which is the cut-point for scores indicating uncertainty and feeling unsure about implementing decisions (O’Connor, 1995). Nearly 40% of participants reported scores of 25 or less, which is the cut-point for effectively implementing decisions (O’Connor, 2006). A total of 45 individuals scored in the highest quartile, including two participants who scored 100 out of a possible 100.

Factors associated with decisional conflict

Table 2 displays the correlations between all predictors and the outcome. Correlations with DC ranged from .04 (female) to −.61 (knowledge), with knowledge showing statistical significance and a strong effect size. The correlations between predictors ranged from <.001 (knowledge and comorbidities) to .70 (depressive symptomatology and anxiety), the latter representing the only large effect size (Cohen, 1988).

Table 2.

Correlations for study variables (N = 125).

Variable	1	2	3	4	5	6
1. Age	–
2. Female	.01**	–
3. Perceived CKD knowledge	−.21**	.06**	–
4. Depressive symptomatology	−.13**	.12**	−.09**	–
5. Anxiety	.20**	.10**	−.04**	.70**	–
6. Number of comorbidities	.24**	.06**	.00**	.10**	.17**	–
7. Decisional conflict	.16**	.04**	−.61**	.12**	.10**	.05**

Note. CKD = chronic kidney disease.

^* p < .05.

^** p < .01.

Table 3 displays one-way ANOVA results for categorical predictors and DC. Self-rated health was significantly associated with DC, showing a medium effect size. DC was highest in the Very Poor and Poor categories. Post hoc comparisons using the Tukey HSD test indicated mean DC was significantly different only between the Excellent and Very Good category (lower DC) and the Very Poor and Poor category (higher DC). Checks with Levene’s test, prior to the ANOVAs, showed no violation of homogeneity of variances (all p > .05).

Table 3.

Means, standard deviations, and one-way analyses of variance statistics.

Predictor	Decisional conflict		F(df1, df2)	P	Ƞ 2
	M	SD
Marital status			1.91 (3, 121)	.13	0.05
Married or partnered	31.81	21.99
Never married	44.30	25.13
Widowed	34.68	22.87
Separated or divorced	38.25	22.06
Education level			2.32 (3, 121)	.08	0.05
Less than high school	45.87	19.48
High school or equivalent	39.33	23.37
Some college	36.72	24.38
Bachelor’s or higher	27.52	21.02
Employment status			0.05 (2, 122)	.95	0.00
Working full or part-time	35.48	18.20
Retired	36.55	23.95
Unemployed	37.70	24.55
Self-rated health			3.65* (3, 121)	.02	0.08
Very poor or poor	47.59	21.25
Fair	35.68	23.32
Good	35.88	23.42
Very good or excellent	21.88	18.18

^* p < .05.

Multivariate analysis

Table 4 provides the results of the OLS multiple regression analysis. Age and knowledge were both mean-centered to create a meaningful intercept. The first block, addressing RQ1, predicted 13% of the variance in DC, a medium effect size (Cohen, 1988), although this was not statistically significant, F(10, 114) = 1.71, p = .09. The adjusted R², which takes the small sample size into account, indicates the variance explained by the first block may be even smaller. Despite the nonsignificant R², within this initial block, age and one category of marital status had significant effects, a situation attributed by Darlington and Hayes (2017, p. 120) to the differential power of the respective hypothesis tests. For every one-year increase in age, DC increased by .60 points, supporting H1. For those never married compared to those who were married or partnered, DC was 13.12 points higher. This partially supported the marital status hypothesis (H5). No other hypotheses from RQ1 were supported in this first block.

Table 4.

Hierarchical regression results for decisional conflict (N = 125).

Variable	Block 1			Block 2			Block 3
	B	SE B	β	B	SE B	β	B	SE B	β
Intercept	28.10***	7.38		29.57***	6.09		23.78*	9.14
Agea	0.60*	0.27	0.24	0.20	0.23	0.08	0.29	0.24	0.12
Female	0.32	4.64	0.01	3.65	3.86	0.08
Educationb
Less than high school	13.54	8.00	0.18	6.20	6.68	0.08	6.14	7.01	0.08
High school or equivalent	10.13	6.09	0.20	1.12	5.17	0.02	0.35	5.36	0.01
Some college	9.12	5.76	0.19	7.90	4.76	0.16	8.46	4.99	0.17
Employmentc
Retired	−4.40	6.59	−0.09	0.47	5.48	0.01	−0.61v	5.56	−0.01
Unemployed	−2.94	7.13	−0.06	3.30	5.94	0.06	2.04	6.06	0.04
Marital statusd
Never married	13.12*	5.92	0.24	2.50	5.10	0.05	1.45	5.34	0.03
Widowed	−1.78	6.10	−0.03	−2.78	5.04	−0.05	−2.54	5.33	−0.04
Separated/divorced	4.84	6.02	0.08	−1.33	5.04	−0.02	−3.56	5.23	−0.06
Perceived kidney knowledgea				−1.85*	0.25	−0.60	−1.77***	0.26	−0.58
Self-rated healthe
Very poor or poor							13.64	7.89	0.23
Fair							8.42	6.56	0.18
Good							4.05	6.99	0.07
Depressive symptomatology							0.07	0.49	0.02
Anxiety							0.09	0.56	0.02
Number of comorbidities							−0.03	1.66	−0.00
R 2 (R2 adj)	0.13 (0.05)				0.41*** (0.36)		0.44*** (0.35)
ΔR2					0.28***		0.03

^aMean centered.

^bReference category is BA or higher.

^cReference category is working full or part-time.

^dReference category is married or partnered.

^eReference category is very good or excellent.

^* p < .05.

^** p < .01.

^*** p < .001.

The second block (RQ2), adding knowledge, predicted 28% additional variance, a significant increase, and the overall model predicted a significant 41% variance, F(11, 113) = 7.22, p < .001, a large effect size. The $R_{\text{adj}}^2$ indicated a slightly smaller amount of variance explained by the second block (36%). Only knowledge had a significant effect. For each one-point increase in knowledge, DC decreased by 1.85 points. This supported H8.

The third block added clinical characteristics (self-rated health, depression, anxiety, and number of comorbidities) (RQ3). There was no significant change in R² from Block 2 to Block 3 (an additional 3% variance). This final model was significant, however, and predicted 44% of the variance, F(17, 107) = 4.98, p <.001 (35% of the variance according to the more conservative $R_{\text{adj}}^2$). Knowledge was the only factor with a significant effect. For each one-point increase in knowledge, DC decreased by 1.77 points, all else being equal. None of the hypotheses for RQ3 (clinical characteristics) was supported. None of the personal characteristics was significant either. Thus, in the final model, RQ1 (significance of personal characteristics) was answered negatively. The squared semi-partial correlation (sr², not shown in Table 4) indicated that knowledge explained 23% of the unique variance in DC. In the final block, as in every block, many of the standard errors were relatively large.

Sensitivity analyses were performed to ensure that results were not affected by analytic decisions. First, because ANOVA showed a linear, albeit nonsignificant decrease in DC with higher education, the regression was re-run with education as a quasi-interval predictor. Second, in line with many prior studies, the regression was re-run treating self-rated health as quasi-interval. Third, because of the strong correlation (.70) between depression and anxiety, the regression was re-run alternating these two predictors. None of these sensitivity analyses made any significant change in results or conclusions.

Discussion

The purpose of this study was to examine the prevalence and predictors of DC among older African Americans living with advanced CKD. This is the first known study to investigate DC and this set of predictors (personal characteristics, knowledge, and clinical characteristics) in a sample of older African Americans with advanced CKD. Overall, DC scores showed significant variability, spanning the full range. On average, participants reported moderate DC, slightly below the cut point indicative of decision delay, uncertainty or feeling unsure of implementing decisions (O’Connor, 1995). The prevalence of DC in this study is difficult to compare with other DC studies, as only two other studies included an exclusively older African American population (Song et al., 2009, 2010). However, those studies used a different measure of DC. The present study was also conducted in a population of individuals with Stage 4 or Stage 5 CKD who had not initiated dialysis. Many prior studies on DC and CKD consist of patients who have already started dialysis.

Across previous studies on DC and CKD, mean DC scores ranged from 26.00 (Tan et al., 2019) to 67.89 (Ho et al., 2020). The mean score closest to this study’s 36.69 was reported by Vélez-Bermúdez et al. (2018; M = 31.30), who recruited individuals with slightly lower eGFR (eGFR <25 ml/min/1.73m²), aged >18 years, and only 5% were African American.

However, examining the prevalence of DC in terms of the mean score is just one way to understand prevalence. It is also useful to examine what percentage of the sample scored at or above 37.5, the recommended cut point for distinguishing the presence of DC (O’Connor, 1995). In the current study, DC scores for four out of ten participants met or exceeded this criterion. This indicates that a substantial proportion of participants reported DC. This result is greater than several prior inquiries that reported the percentage of DC scores ≥37.5 (Chen et al., 2018, 38%; Goh et al., 2021, 27.5%), but similar to Vélez-Bermúdez et al. (2018, 43%). This is relevant because prior research has linked high levels of DC with delays in decision-making (Chan et al., 2022; McPherson et al., 2017) and decision-making regret and anxiety (Kaplan et al., 2014; Kates, 2018). Comparing DC prevalence in the present study with prior studies is complicated by not only the unique characteristics of the present sample regarding race and age but also by variability across studies in the measurement of DC, including different versions of O’Connor’s (1995) DCS.

Personal characteristics showed nuanced findings regarding the association with DC. When personal characteristics were the only predictors, older age and being never married, compared to those who were married or partnered, were associated with higher DC. Although most prior studies found no relationship between age and DC, the results of the present study are consistent with Vélez-Bermúdez et al. (2018), which included participants with a similar mean age (63.63 years) as the current study. Our finding that those who were never married reported higher DC compared to those who were married or partnered is similar to McPherson et al.’s (2017) finding that individuals who were not married had higher DC compared to those who were married. Once knowledge was added to the models, demographics and other personal characteristics were no longer significant. The disappearance of the significant effect for age may reflect the fact that older age was correlated with lower knowledge.

Contextualizing the present results on personal characteristics and DC is hindered by the limited knowledge base about DC in adults with advanced CKD. The few existing studies have little overlap in the personal characteristics included and have operationalized them in different ways. Additionally, relevant demographics may differ for older African Americans who are primarily retired. Thus, demographic predictors of DC in advanced CKD remain an area in need of research.

Knowledge emerged as a key predictor of DC after multivariate adjustments for clinical and personal characteristics. Two prior studies reported a significant relationship between knowledge and DC in CKD, although the content of their knowledge measure differed from that of the present study. Chen et al. (2018) assessed knowledge about dialysis, and McPherson et al. (2017) measured knowledge about kidney transplantation.

The significant finding for knowledge argues that the type of knowledge assessed in this study (perceived CKD knowledge) is important among individuals living with advanced CKD facing difficult decisions pre-dialysis. This is the first known study to demonstrate the reliability and predictive value of the PiKS in pre-dialysis in a sample of older African Americans with advanced CKD. Although the Ottawa Decision Support Framework states that knowledge consists of critical and essential facts to inform decision-making about a health problem, the theory is vague about which types of knowledge are most relevant to reducing DC (O’Connor, 2006). The importance of knowledge is also evident from prior work, which has shown that CKD education can delay the onset of dialysis and improve patient outcomes after the start of dialysis (Finkelstein et al., 2008) and has emphasized the importance of educational interventions for patients with CKD (Lopez-Vargas et al., 2014). Exploring the types of knowledge that are important to individuals living with CKD or ESRD is a fascinating area of future inquiry that could prove crucial for future education and interventions. In the present study, knowledge about medications that help or hurt the kidney and treatment options as the disease progresses were the areas where participants reported the least knowledge. Last, the timing of educational interventions to increase knowledge is another important factor to consider, as patients with CKD have decision-making needs across the trajectory of their illness (Murray et al., 2009).

There is a lack of theoretical guidance in the Ottawa Decision Framework (O’Connor, 2006) regarding which clinical characteristics to select and minimal empirical evidence in the CKD literature concerning relationships between clinical characteristics and DC. No prior studies of DC in CKD samples examined self-rated health, anxiety, or the number of comorbidities. The limited number of study participants with moderate to severe depression and anxiety may have restricted the ability to detect any relationships with DC in the present study. This is consistent with prior research on African Americans that has shown lower levels of depression and anxiety among this population (Pamplin & Bates, 2021; Williams et al., 2007). Additionally, the relationship between DC and self-rated health, as well as the number of comorbidities, has limited empirical evidence in the literature (McPherson et al., 2017; Vélez-Bermúdez et al., 2018). The current measure of comorbidities did not include symptoms such as fatigue and pain, which may be important for future research. The present study also did not include clinical characteristics related to CKD. Future research on DC may benefit from including measures of CKD-specific symptom burden and an emphasis on common comorbid conditions associated with CKD (i.e., hypertension, diabetes, and cardiovascular conditions).

Strengths

This study makes several unique contributions. The present study added knowledge to the limited body of literature on the prevalence of DC in individuals living with advanced CKD pre-dialysis (Chan et al., 2022; DePasquale et al., 2022; Goh et al., 2021; Ho et al., 2020; Vélez-Bermúdez et al., 2018). This study is the first to examine DC, perceived CKD knowledge, and this range of predictors in an exclusively older African American population. Previous studies have examined some of these predictors in bivariate analyses, but this is the first study to examine this range of both personal and clinical characteristics in multivariate analyses.

Limitations

This study was cross-sectional and restricted to the pre-dialysis period. Longitudinal research is needed because CKD is a progressive illness, and there may be differences in DC as CKD progresses or after someone initiates dialysis. This could help understand how DC changes during disease progression. The Ottawa Decision Support Framework (O’Connor, 2006), the theory used for this study, provided guidance regarding which categories of predictors to include (e.g., clinical characteristics), but it lacked details regarding specific predictors within categories and how to operationalize them. This research was conducted at a single ambulatory pre-dialysis clinic, which was more feasible and allowed for a more in-depth understanding of older African Americans and DC, but it could limit generalizability.

Limited statistical power may explain the nonsignificant results in this study regarding personal and clinical characteristics. The study’s sample size reflected both the limited resources for data collection and the challenges of recruitment during the COVID-19 pandemic. Although a priori power analysis suggested the sample size would provide sufficient power, the effect sizes were smaller than anticipated.

Implications for practice and research

Based on the results of this study, exploring the types of knowledge important to individuals living with advanced CKD is a fascinating line of future investigation that could prove crucial for future education and interventions. The results of this study indicate a need for targeted education about CKD, particularly during the pre-dialysis period. The information gathered from the PiKS scale could help clinicians identify patients with lower scores and target them for additional education. Each of the knowledge items that participants struggled with is an area in which clinicians could have an impact as educators and advocates in the pre-dialysis period and beyond. Additionally, continuing to research the role of DC, personal, and clinical characteristics will remain important as researchers expand the knowledge base on how social factors influence health outcomes, particularly among historically marginalized populations. The present finding that older age was correlated with lower knowledge (and predicted higher DC until knowledge was controlled) suggests the need to target older CKD patients in future research and interventions.

Moving from observational research to intervention research is an obvious next step that could push DC work forward and reduce DC in older African Americans living with advanced CKD. One major challenge in this area is that deciding on an appropriate control group could be difficult, as defining usual care regarding kidney disease education may be challenging. Although there are general standards for kidney disease education, each hospital and clinic may have a slightly different approach to what is included in standard CKD education and how this is packaged and delivered to patients.

Funding

This study was funded by a grant from the Centers for Dialysis Care (Rosenberg Foundation). The funder had no role in study design, data collection, analysis, interpretation, or manuscript preparation.

Conflicts of interest

The authors declare no conflicts of interest. The authors have no financial or personal relationships that could have influenced the work reported in this study.

Data availability

The data that support the findings of this study are not publicly available due to the presence of sensitive health information and institutional review board (IRB) restrictions. This study was not preregistered, and no publicly accessible analysis plan was registered prior to conducting the analyses. De-identified data may be available from the corresponding author upon reasonable request and with appropriate institutional approvals.

Author contributions

Tyrone C. Hamler (Conceptualization, Methodology, Investigation, Data curation, Formal analysis, Writing—original draft, Writing—review & editing, Project administration), Aloen L. Townsend (Conceptualization, Methodology, Supervision, Writing—review & editing), Mirela Dobre (Methodology, Resources, Writing—review & editing), and Anne Huml (Methodology, Resources, Writing—review & editing)