Evaluation of a Novel Hospice-Specific Patient Decision Aid

Abstract

Background:

We tested a novel hospice-specific patient decision aid to determine whether the decision aid could improve hospice knowledge, opinions of hospice, and decision self-efficacy in making decisions about hospice.

Methods:

Two patient-level randomized studies were conducted using two different cohorts. Recruitment was completed from March 2019 through May 2020. Cohort #1 was recruited from an academic hospital and a safety-net hospital and Cohort #2 was recruited from community members. Participants were randomized to review a hospice-specific patient decision aid. The primary outcomes were change in hospice knowledge, hospice beliefs and attitudes, and decision self-efficacy Wilcoxon signed rank tests were used to evaluate differences on the primary outcomes between baseline and 1-month.

Participants:

Participants were at least 65 years of age. A total of 266 participants enrolled (131 in Cohort #1 and 135 in Cohort #2). Participants were randomized to the intervention group (n = 156) or control group (n = 109). The sample was 74% (n = 197) female, 58% (n = 156) African American and mean age was 74.9.

Results:

Improvements in hospice knowledge between baseline and 1-month were observed in both the intervention and the control groups with no differences between groups (.43 vs .275 points, P = .823). There were no observed differences between groups on Hospice Beliefs and Attitudes scale (3.29 vs 3.08, P = .076). In contrast, Decision Self-Efficacy improved in both groups and the effect of the intervention was significant (8.04 vs 2.90, P = −.027).

Conclusions:

The intervention demonstrated significant improvements in decision self-efficacy but not in hospice knowledge or hospice beliefs and attitudes.

Introduction

Hospice enrollment has increased over the last several decades yet it remains an underutilized healthcare benefit.¹ Fewer than 50% of hospice eligible Medicare decedents enrolled in hospice prior to death.^2,3 Evidence shows hospice patients experience better health and quality of life outcomes compared to patients not receiving hospice care.^4–6 Additionally, bereaved family and caregivers often report wishing they had engaged hospice services much sooner to provide care for their loved ones.^7,8 Many clinicians and patients are misinformed about hospice care and its services, leading to underuse and late referrals which further complicates this issue.^7,9 Prior research shows there is a strong desire for patients and families to engage in conversations about hospice well before times of crisis,^7,10–12 but many physicians are reluctant to initiate these difficult conversations and prefer for patients and families to initiate conversations.^13,14

Shared decision-making (SDM) is an evidence-based approach shown to improve patient-provider communication, information sharing, improved decision quality, and decreased decisional regret.^15–18 One strategy to support SDM is the use of patient decision aids (PtDAs). PtDAs engage patients in the process of decision-making by providing them with information on treatment options, eliciting patient values, and facilitating deliberation between patients and providers on treatment choices.^19,20 Critical to PtDAs is the presentation of evidence-based and unbiased information.^21,22

Our prior work included a needs assessment to determine what information patients and families needed to make informed hospice decisions.⁷ Next we developed a hospice specific PtDA to support these complex conversations.^7,8,10 Herein, we report pilot data evaluating whether a hospice PtDA can improve hospice knowledge, opinions of hospice care, and decision self-efficacy in two unique cohorts of participants.

Methods

Overview/Study Design

We collected data from two cohorts. Both studies were intentionally designed a priori to use the same data collection tools and time points to create datasets that could be analyzed together. The pilot studies were distinct as one assessed older adults in clinical settings and the other explored hospice decision making in older African American (AA) community members. Participants recruited from clinical settings were randomization using a ratio of 2:1 (intervention: control) to maximize the number of participants in the intervention group who reviewed the PtDA. AA participants recruited from community settings where randomized using a 1:1 ratio (intervention: control). The rationale for synergizing datasets was to leverage resources and increase sample power when evaluating overall trends in the data. While this is an unconventional method, there is precedent for combing datasets in the literature (see limitations below).^23–25 The studies were registered at clinicaltrials.gov and both studies were approved by the Colorado Multiple Institutional Review Board.

Settings and Recruitment

Recruitment for Cohort #1 took place at two different healthcare systems in the Denver Metro area. One was a large academic hospital, and the other was a large safety-net hospital. Participants were recruited from primary care clinics that specialize in providing care to geriatric patients at each institution and from an outpatient palliative care clinic affiliated with the academic hospital.

Cohort #2 focused on recruitment of older AA respondents residing in the Denver Metro area. Advertisements were placed in local newspapers and publications. Additionally, recruitment events were held at local community centers, senior living communities, recreation centers, and faith-based organizations. AA participants self-referred themselves to take part in the study after reviewing recruitment material in publications or attending a community recruitment event. We chose to utilize this, community outreach-based recruitment method as it is difficult to enroll people of color in research using the traditional medical research framework.

Randomization for both Cohorts was done using a REDCap database which was assigned participants to the intervention or control groups. Data for both cohorts was collected from March 2019 through May 2020.

Participants

Participants were at least 65 years old, spoke English (as the PtDA is currently only in English), and were able to provide informed consent. Participants received two $25 gift cards ($50 total) for participation in the study.

Intervention

The intervention consisted of a novel hospice specific PtDA.^8,21,26 The PtDA was available in two formats: a 12-page paper booklet and an accompanying 17-minute video.⁸ Participants received both the booklet and the video.

Control

Participants assigned to the control group received no intervention at the time of randomization. However, all participants, including those randomized to the control, were mailed a copy of the PtDA (both the video and booklet) after completing the 1-month follow-up appointment.

Data Collection

Participants completed 2 or 3 study visits depending on whether they were randomized into the intervention or control groups. Intervention participants completed the baseline study visit, a 1-week follow-up visit to assess acceptability of the intervention (i.e. hospice PtDA), and a 1-month follow-up visit. Control participants completed only the baseline visit and the 1-month follow-up.

In Cohort #1, a trained research assistant (RA) identified eligible participants in the electronic medical records for both the academic and safety-net clinics. The RA obtained permission from providers to approach each participant about the study during a regularly scheduled clinic appointment. The RA met with the participant to obtain informed consent and asked them to complete the baseline surveys (described below). After completing the baseline surveys, participants were randomized to the control or intervention group. Individuals randomized to the intervention group received the hospice PtDA and were asked to review the materials over the next week. The RA scheduled a time to follow-up with participants within the next week to conduct a brief phone interview about the PtDA. Participants randomized to the control completed the baseline assessment and the RA scheduled a 1-month follow-up phone visit. All individuals in the control group received a copy of the PtDA after completing the 1-month follow-up visit.

In Cohort #2, AA participants enrolled by contacting the study PI after reviewing the advertisements placed in local publications or attending a recruitment event hosted by the study PI. The RA met with the participant to obtain informed consent and asked them to complete the baseline surveys (described below). After completing the baseline surveys, participants were then randomized to the control or intervention group. Participants randomized to the intervention received the PtDA at the baseline visit and were asked to review the materials over the following week. The RA scheduled a time to follow-up with them to conduct a brief phone interview about the PtDA one week after baseline visit. Participants randomized to the control completed the baseline assessment and the RA scheduled a 1-month follow-up call with participants randomized to the control group. All control participants received a copy of the PtDA after completion of the 1-month follow-up.

Measures

The primary outcomes were measured through three surveys. The Hospice Knowledge Test is a 23-item true/false assessment of participant knowledge.⁹ Participants’ knowledge score was then calculated by summing the number of hospice knowledge questions the respondent answered correctly with the highest possible score of 23. Higher scores indicated more hospice knowledge.

The Hospice Beliefs and Attitudes Survey (HBAS) is an 8-item assessment of one’s beliefs and attitudes, e.g., their opinions, about hospice care.²⁷ The survey uses a 5-point Likert scale ranging from Strongly Disagree to Strongly Agree. The 8 items were then summed for the highest possible score of 40, with higher scores indicating more favorable opinions of hospice.

The Decision Self-Efficacy Scale is an 11-item assessment of participants’ confidence when making medical decisions. Questions are answered on a 5-point Likert scale ranging from Not at All Confident to Very Confident. Self-Efficacy Scores range from 0 (extremely low self-efficacy) to 100 (extremely high self-efficacy).²⁸

Additionally, we collected basic demographics (age, race, education, income, comorbidities) at baseline. Finally, RAs called all intervention participants approximately 1-week after baseline to determine the acceptability and usability of the PtDA. We measured three key domains of acceptability, using questions adapted from Barry et al: how well the PtDA explained hospice care, how balanced the PtDA is (i.e. is it slanted more towards enrolling in hospice vs not enrolling), and the usefulness of the tool in making end-of-life care decisions.²⁹

Analysis

We compared baseline characteristics between the control and intervention groups using χ2 and t-tests. Change in Knowledge Scores, Hospice Belief Scores, and Self-Efficacy Scores were calculated between baseline and one week (data not shown), baseline and one month, and between one week and one month (data not shown). The Wilcoxon Signed-Rank test was used to evaluate changes in scale scores between time points.

Results

Participants

A total of 266 individuals participated in this study, with 131 participants enrolled in Cohort #1, and 135 participants enrolled into Cohort #2 (Figure 1). All participants from Cohort #2 self-identified as AA. An additional, 21 individuals from Cohort #1 self-identified as AA. Thus, in the total sample analyzed 156 participants were AA and 97 participants were White. There were no statistical differences between groups except with regards to race and ethnicity (Table 1).

Figure 1.

Consort diagram-describes Cohort #1 and Cohort # 2.

Table 1.

Combined Demographics.

Variable	Control (n = 109)	Intervention (n = 156)	P-Value
Age (Years)
Mean	74.8	75.0
Range	65-92	65-98	.797
Sex (N, %)
Male	24 (22.02)	44 (28.21)
Female	85 (77.98)	112 (71.79)	.257
Race/Ethnicity (N. %)
Black/African American	76 (69.72)	80 (50.96)	.002
White	28 (25.6)	69 (43.95)	.002
LatinX	4 (3.67)	6 (3.82)	.949
Native American/Alaskan Native	2 (1.83)	3 (1.91)	.964
Other	3 (2.75)	2 (1.27)	.383
Marital Status (N, %)
Single/Never Married	11 (10.09)	11 (7.10)
Married/Partnered	33 (30.28)	55 (35.48)
Separated	6 (5.50)	5 (3.23)
Divorced	23 (21.10)	39 (25.16)
Widowed	36 (33.03)	45 (29.03)	.594
Education (N, %)
Less than High School	4 (3.67)	10 (6.41)
High School/GED	18 (16.51)	33 (21.15)
Some College/Associates Degree	36 (33.03)	48 (30.77)
College Degree	27 (24.77)	30 (19.23)
Master’s Degree	27 (24.77)	19 (12.18)
Professional/Doctoral Degree	7 (6.42)	16 (10.26)	.509
Income (N, %)
Less than $25,000	38 (36.89)	59 (36.56)
$25,001-$50,000	29 (28.16)	42 (27.45)
$25,001-$50,000	20 (19.42)	19 (12.42)
Greater than $75,000	16 (15.53)	19 (12.42)	.361

Primary Outcomes

Combined Analysis

In the combined analysis both the control and intervention groups had relatively high mean baseline hospice knowledge scores, 17.6 and 17.5 respectively, out of a possible score of 23. Hospice knowledge significantly improved from baseline to 1-month in both groups by a mean of .43 points (P < .001) in the intervention and by .275 point (P = .027) in the control group, with no statistically significant difference observed between the groups (P = .823) (Figure 2(a)).

Figure 2.

Primary Outcomes-Graph A describes the change in hospice knowledge between the control and intervention. Graph B describes the change in hospice beliefs and attitudes between the control and intervention. Graph C describes the change in decision self-efficacy between the control and intervention.

Baseline mean HBAS scores were 27.7 and 28.9, out of a possible score of 40, respectively for the control and intervention. HBAS improved by mean of 3.29 points in the intervention (P < .001) and by 3.08 points in the control (P < .001). The overall change in HBAS between the control and intervention at 1-month was not statistically significant (P = .076) (Figure 2(a)).

Mean Decision Self-Efficacy scores for the control and intervention were 79.9 and 83.1 respectively out of a possible score of 100. Scores increased by a mean of 8.04 (P < .001) in the intervention and by 2.90 points (P = .430) in the control. The overall change in Decision Self-Efficacy between the control and intervention from baseline to 1-month was significant (.027) (Figure 2(a)).

Cohort #1

Both the control and intervention had relatively high mean baseline hospice knowledge scores, 18.14 and 18.01, respectively. Hospice knowledge decreased slightly in both groups (.49 and .093 intervention and control respectively), but neither decrease was statistically significant with no difference between groups (P = .76) (Figure 2(b)).

Mean baseline scores on the HBAS were 29.69 and 29.51 for the intervention and control groups respectively. HBAS scores increased by 3.65 points (P < .001) in the intervention group and 3.12 points (P < .001) in the control groups with no difference between groups (P = .98) (Figure 2(b)).

Baseline Decision Self-Efficacy Scores for the intervention and control were 87.27 and 91.75 respectively. Decision Self-Efficacy increased in the intervention group by 6.33 points (P = .003) but decreased by 3.12 points in the control group (P = .15). The change between groups was statistically significant (P = .003) (Figure 2(b)).

Cohort #2

Baseline scores for the Hospice Knowledge Test were 16.78 and 17.23 for the intervention and control groups respectively. Knowledge scores improved by 1.61 points (P = .02) in the intervention group and by .51 points (P = .29) in the control group with no difference between groups (P = .18).

Baseline mean HBAS scores were 26.55 and 28 for the intervention and controls groups respectively. At one month the HBAS scores increased by 3.73 points (P < .001) in the intervention and by 3.28 points (P < .001) in the control with no difference between groups (P = .65).

Baseline Decision Self-Efficacy scores were 77.62 and 72.11 for the intervention and control groups respectively. At one month Decision Self-Efficacy scores increased by 11.63 points (P < .001) in the intervention and by 6.04 points (P = .013) in the control with no significant difference between groups (P = .17) (Figure 2(c)).

Patient Decision Aid Acceptability

We asked participants to evaluate acceptability of the PtDA in making decisions about hospice care. When asked how well the PtDA explained what hospice care is, 96.4% of respondents found the explanation either excellent or good, while 3.6% found it fair, and 0% stated it was poor. Similarly, 88.3% of respondents found our discussion of hospice payment mechanisms to be either good or excellent, while 10.2% found the presentation fair, and 1.5% found it poor. When asked if the presentation of the PtDA was biased towards hospice or against hospice care 78.7% of respondents found the PtDA to be balanced and not biased. Further, 93.5% of respondents thought the length of the PtDA was appropriate and 92% found the amount of information presented was appropriate (Figure 3). The individual results for Cohorts 1 and 2 were consistent with the combined results (data not shown). Most respondents reported that the explanation of hospice was good or excellent, the that the PtDA balanced and neutral, and that the PtDA would be useful if there were making hospice decisions.

Figure 3.

Patient Decision Aid Acceptability Outcomes-Figure 3 describes how well the key components of the decision aid were defined, how balanced the decision aid was, and how useful participants found the decision aid in making hospice decisions.

Discussion

We conducted two pilot studies with two different participant cohorts to explore the preliminary effects of a hospice PtDA on hospice knowledge, opinions of hospice, and decision self-efficacy. Although hospice knowledge improved from baseline to 1-month in the overall sample, there was no difference between groups. Similarly, we found within-group improvements in opinions of hospice from baseline to 1-month, but the overall change between groups was not significant in any of the analyses. We did find significant within group improvements in decision self-efficacy from baseline to 1-month among participants receiving the PtDA compared to controls.

Our findings are consistent with other studies using the hospice knowledge test to measure hospice knowledge in potential hospice enrollees. Cagle et al found overall knowledge scores were high in a cohort of adults, but when stratified by race, both Hispanics and African Americans reported less hospice knowledge.⁹ In our study, overall knowledge scores were high, but Cohort #2, which consisted of only AA respondents, had a lower baseline knowledge score than our White respondents. Similarly, other studies also show that minoritized communities have lower hospice knowledge study scores.³⁰

Our study is unique in two ways. First, we sample respondents that are over the age of 65, the population most likely to use hospice services. Other studies examining hospice knowledge tend to sample younger participants, including one study of college students who possess relatively low hospice knowledge.³¹ Many other studies are cross-sectional in design, and only report one time point of data collection.^30,32,33 The study presented herein is one of few that endeavors to provide education and support to patients and families making decisions about hospice care. An abundance of evidence shows that individuals who are more informed about treatment options are more likely to make decisions consistent with their values, goals of care, and experience less decisional regret.¹⁸ Our intervention showed modest improvements in hospice knowledge in the intervention group. These results provide evidence that hospice specific decisional support materials can improve hospice knowledge, facilitate quality end of life discussions, and improve shared decision making about hospice care.

Our study found that overall opinions of hospice care were favorable. This finding is consistent with other literature evaluating perception of hospice care. El-Jawahri and colleagues found that overall opinions of hospice were favorable; however, knowledge gaps and misperceptions of hospice care were barriers to enrollment.³⁴ Similarly, Van Dussen and colleagues found people generally had favorable opinions about hospice care but struggled to understand the payment mechanisms for hospice care.³³ Further there is evidence that some minoritized groups developed more favorable views of hospice care after misconceptions of payment, concerns about cultural and religious beliefs being valued, and potential impact of hospice care on familial caregivers were explained.³⁵ Common across all studies is the conclusion that patients need to be better informed about hospice care and that exposure to hospice should occur well before times of imminent death or crisis.

Our most notable finding was the significant improvement in decision self-efficacy amongst patients who received the PtDA. There is evidence that patients with more knowledge and self-efficacy are less likely to experience decisional conflict after making medical decisions.¹⁸ Patient decision aids are designed to help increase knowledge and thus increase self-efficacy. This association has been shown in a variety of diseases and conditions ranging from diabetes management, to cancer screening and treatment decisions.^36–38 Further decision self-efficacy has shown to be critical in supporting older adults making healthcare decisions.³⁹ The increases in self-efficacy are especially significant in underserved and underrepresented decision makers. Given our finding of a significant increase in self-efficacy there is evidence that our PtDA might improve decision making agency around end of life and warrants more investigation in future studies.

Additionally, we acknowledge that our findings were largely negative, however we feel some important lessons were learned from this work. Given the complexities of end-of-life decision-making, assessing hospice knowledge quantitatively (e.g. with a survey) may not be the most important measure. Our own work and the work of others suggest that qualitatively assessing hospice knowledge provides more nuanced and accurate evaluation of hospice knowledge. Though most Americans have heard of hospice care,^40–45 only 34% of Black adults are able to accurately describe hospice care and its services, as compared to 66% White adults.^40,43,44 Our decision aid was intentionally designed to provide a general overview of hospice care with the intent to facilitate more detailed conversations at the point of care. There is ample evidence that decision aids improve discussions and collaboration in decision making especially in complex and emotional medical decisions.¹⁸ Our most significant finding is that self-efficacy improved. This indicates that there is something about review the decision aid that empowers decision makers to make informed and value concordant decision at the end of life.

The complexity and intensity of end-of-life care and decision-making during times of crisis cannot be overstated. Thus, we intentionally enrolled healthy individuals in this study. We chose this approach because research indicates that patients and families want to engage in end-of-life conversations well before times of crisis.^7,33,34,46 Our goal was to provide information and facilitate conversations well before the emotion, distress, and urgency of end-of-life decision-making occurred during times of crisis. Our study found that our decision aid was neutral and not biased toward enrolling or not enrolling in hospice care and that 97% of patients found our PtDA helpful in making decisions about hospice. One of our next steps is to evaluate the decision aid in a population of individuals with serious illness to determine if the decision aid improves conversations about hospice transitions.

Strengths and Limitations

Several limitations should be considered when interpreting our data. We made an a priori decision to combine data from two different study cohorts. Our choice to combine two cohorts with different recruitment strategies is unusual but combining datasets is a common practice throughout research and medical literature. The method we employed is consistent with other studies harmonizing multiple datasets. Though not common, there is a precedent to use only two studies and apply a prospective data collection method.^23–25 Utilizing a combined data framework allowed us to recruit a diverse sample with more than half of the participants self-identifying as AA.

Another limitation of our study is that our sample of participants were highly educated and well-informed about hospice care at baseline. The mean knowledge scores for the control and intervention were both of 17 out of a possible score of 23. There may be numerous explanations for this, but one may be the oversaturation of hospice care facilities in the Denver Metro Area. However, it should be noted that we did see improvements in hospice knowledge despite the highly educated sample. Additionally, our sample of AA respondents were middle class. Most research in underrepresented communities focuses on the extremes of low or high SES, thus we present data from a population that is not often captured in the literature. Lastly, given our community recruitment strategy, it is possible that there may have been a spillover effect or selection bias by recruiting individuals from the same social networks. Although, we did analyze each cohort separately and found similar results which suggested that combining the cohorts did not affect the results. Further our study highlights that community-focused recruitment methods may help increase diversity in research.

Conclusions

There are few, if any, shared decision-making tools to help patients and families make end-of-life decisions. We conducted two pilot studies to evaluate the efficacy of a novel PtDA specific to hospice care. We found improvements in hospice knowledge, opinions of hospice, and improved decision self-efficacy in making hospice-related decisions. These findings are encouraging but a larger multisite study is planned to validate and replicate results in both healthy participants and those with serious illness.