Restricted mean survival time versus conventional effect summary for treatment decision-making: A mixed-methods study

Abstract

Background:

Treatment effect is typically summarized in terms of relative risk reduction or number needed to treat (“conventional effect summary”). Restricted mean survival time (RMST) summarizes treatment effect in terms of a gain or loss in event-free days. Older adults’ preference between the two effect summary measures has not been studied.

Methods:

We conducted a mixed methods study using a quantitative survey and qualitative semi-structured interviews. For the survey, we enrolled 102 residents with hypertension at five senior housing facilities (mean age 81.3 years, 82 female, 95 white race). We randomly assigned respondents to either RMST-based (n = 49) or conventional decision aid (n = 53) about the benefits and harms of intensive versus standard blood pressure-lowering strategies and compared decision conflict scale (DCS) responses (range: 0 [no conflict] to 100 [maximum conflict]; <25 is associated with implementing decisions). We used a purposive sample of 23 survey respondents stratified by both their random assignment and DCS from the survey. Inductive qualitative thematic analysis explored complementary perspectives on preferred ways of summarizing treatment effects.

Results:

The mean (standard deviation) total DCS was 22.0 (14.3) for the conventional decision aid group and 16.7 (14.1) for the RMST-based decision aid group (p = 0.06), but the proportion of participants with a DCS <25 was higher in the RMST-based group (26 [49.1%] vs 34 [69.4%]; p = 0.04). Qualitative interviews suggested that, regardless of effect summary measure, older individuals’ preference depended on their ability to clearly comprehend quantitative information, clarity of presentation in the visual aid, and inclusion of desired information.

Conclusions:

When choosing a blood pressure-lowering strategy, older adults’ perceived uncertainty may be reduced with a time-based effect summary, although our study was underpowered to detect a statistically significant difference. Given highly variable individual preferences, it may be useful to present both conventional and RMST-based information in decision aids.

INTRODUCTION

Decision aids are increasingly used to enhance shared decision-making for older adults, allowing for a more nuanced understanding of treatment effects by balancing information about potential benefits and harms.¹ Most decision aids present treatment effects in terms of relative risk reduction or number needed to treat (“conventional effect summary”). Restricted mean survival time (RMST) is an increasingly used measure that summarizes treatment effect in terms of a gain or loss in event-free days.^2,3 While the time-based RMST may improve evidence communication in treatment decision-making,⁴ whether older adults prefer one summary measure over the other has yet to be demonstrated. An online survey found similar levels of decisional conflict between those who were provided with RMST-based decision aid and those who were provided with conventional decision aid about intensive versus standard blood pressure-lowering strategies.⁵ However, this was conducted in relatively healthy older people (mean age, 72 years) who signed up for an online panel and, therefore, may not generalize to older or multimorbid populations. Therefore, we conducted a telephone survey and semi-structured interviews of older adults who reside at senior care facilities to examine their preference between RMST-based and conventional decision aids using blood pressure-lowering strategies as an example.

METHODS

Study design

This mixed-methods, sequential explanatory study integrated quantitative and qualitative approaches when sampling and when analyzing data for complementary perspectives on preferred ways of learning about treatment benefits and harms.⁶ The two datasets derived, respectively, from: (1) a quantitative survey in which participants were randomly assigned to react to either an RMST-based or a conventional decision aid on the benefits and harms of blood pressure-lowering strategies; and (2) qualitative semi-structured interviews of purposively sampled survey participants to understand complementary perspectives on their preferred way of summarizing treatment benefits and harms. The quantitative survey occurred between September 2020 and March 2021, and the semi-structured interview between September 2021 and December 2021. This study was approved by the Institutional Review Board of Hebrew SeniorLife/Advarra.

Participants

Participants were recruited using flyers and internal television advertisements from 5 independent living sites within Hebrew SeniorLife, a large senior care organization in the New England region. Individuals were eligible if they were ≥65 years old, with hypertension, and self-reported taking one or more blood pressure-lowering drugs. We excluded those who were non-English speaking, unable to read written materials, or with the Brief Interview for Mental Status score <7. For the semi-structured interviews, we invited 30 participants purposefully stratified by the assignment group (RMST-based or conventional decision aid), and Decisional Conflict Scale (DCS) score from the survey (high conflict [25–100 points] vs low conflict [0–24 points]; see detailed description below). All participants provided verbal informed consent.

Quantitative survey

Participants were randomized to receive a decision aid that used either RMST-based summary or conventional effect summary about the benefits and harms of intensive versus standard blood pressure-lowering strategies based on the published reports of the Systolic Blood Pressure Intervention Trial.^7,8 These decision aids have been published previously,⁵ with examples in Figure 1, and detailed aids and questions provided in Appendix S1. The assigned decision aid was sent to a participant by mail or electronically before telephone survey. During the survey, the study coordinator (D.C.) collected demographic information, education level, cardiovascular disease history, limitations in activities of daily living, and number of prescription drugs, as well as comorbidities to calculate the Lee prognostic index score for an estimate of 4-year mortality risk.^9,10 Participants reviewed the decision aid with the study coordinator and then completed the DCS questionnaire. The DCS is a 16-item scale measuring patients’ perceived conflict and uncertainty in choosing options on a scale of 0 (no conflict) to 100 (maximum conflict) in the domains of feeling uninformed, feeling unclear about values, feeling unsupported in decision-making, feeling uncertain, and perceived effectiveness in decision-making.¹¹ A DCS total score <25 points is associated with implementing decisions.¹² Finally, participants were asked about their preference between the intensive and standard blood pressure lowering strategy.

FIGURE 1.

Conventional versus restricted mean survival time-based decision aid graphics. Participants were randomized to receive information as either a conventional decision aid (top) or restricted mean survival time-based (bottom). For the qualitative study, they were shown both decision aids.

Semi-structured interview

We developed and pilot-tested a semi-structured interview guide de novo (Appendix S2). The trained interviewer (N.N.) began the telephone interview with a question about “what matters most” to the participant when making medical decisions for themselves. Next, the interviewer verbally presented the RMST-based summary and conventional effect summary while the participant viewed each depiction in tandem. Subsequent questions asked for the participant’s perception of each presentation in relation to, first, the benefits and then the harms of the intensive versus standard blood pressure-lowering strategies. Questions pertained to (1) how well the RMST-based and conventional decision aids matched what the participant would want to know to make a decision about the treatment; and (2) which method would be more helpful in making the treatment decision. If the preferred method was discrepant across the benefits and harms sub-sections, the interviewer probed for the participant’s rationale underlying the discrepancy. All interviews were digitally recorded and transcribed verbatim. Average duration of interviews was 30.5 min (range: 15–82).

Data analysis

For analysis of the quantitative survey data, we compared the two decision aid groups for the total DCS score and the subdomain scores using a two-sample t-test and the proportion of participants who chose the intensive strategy over the standard treatment using a chi-square test. We also compared the proportion of participants with a total DCS score <25 between the two groups, with a chi-square test. Last, we examined whether the effect of RMST-based versus conventional decision aid differed by pre-specified subgroups by age (<75 or ≥75 years), sex, self-reported race (white or non-white race), education (less than bachelor’s degree or otherwise), cardiovascular disease, and the Lee index scores (0–5, 6–9, or ≥10). A two-sided p-value <0.05 was considered statistically significant.

We conducted inductive qualitative thematic analysis of semi-structured interview data based upon rapid assessment procedures.¹³ Analysts performed three stages of analysis: (1) summarizing transcripts in a template organized by questions from the semi-structured interview guide (N.N., N.S.); (2) transferring the templated summary data to a pilot-tested matrix to conceptualize patterns within and across participants’ narratives (N.N.); and (3) developing thematic structures based upon the inductively emergent patterns (N.N., N.S.).

Several steps were taken to ensure analytic rigor. First, the two analysts developed transcript summary templates for the same 3 transcripts to verify cross-analyst consistency; subsequently, the two analysts evenly divided up the transcripts to summarize them independently. For the remaining stages, the analysts worked independently, cross-checking each other’s work at regular intervals. Second, throughout the rapid assessment procedures, when new findings emerged later in the analytic process, the two analysts re-reviewed earlier analyzed data to ensure all instances of the emergent findings were captured. Third, the analysts monitored the data for positive and negative evidence affirming each theme as they developed thematic structures. Finally, a doctoral-level qualitative methodologist and health services researcher (J.A.P.) trained the two analysts in rapid assessment procedures with continuous supervision. When discrepancies in the analysts’ interpretations existed, consensus was reached by involving J.A.P., who additionally guided the team in identifying the final thematic structure and supporting quotes.

RESULTS

Quantitative survey

We randomized 102 participants to the conventional decision aid (n = 53) and the RMST-based decision aid (n = 49) (Table 1). The sample had a mean age (standard deviation [SD]) of 81.3 (8.0) years, 82 (80.4%) females, 95 (93.1%), white participants, 37 (36.3%) with less than a bachelor’s degree education, 18 (17.7%) with cardiovascular disease, and 33 (33.3%) with a high risk of mortality (Lee index score ≥10 points).

TABLE 1.

Characteristics of survey participants

Characteristics	Conventional decision aid (n = 53)	RMST-based decision aid (n = 49)	p-value
Age, years, mean (SD)	80.9 (8.1)	81.8 (8.0)	0.60
Female, n (%)	43 (81.1)	39 (79.6)	0.85
White race, n (%)	48 (90.6)	47 (95.9)	1.0
Education level, n (%)			0.45
High school	3 (5.7)	7 (14.3)
Some college	10(18.9)	7 (14.3)
Associate	5 (9.4)	5 (10.2)
Bachelor	14 (26.4)	14 (28.6)
Advanced	21 (39.6)	16 (32.7)
Body mass index, kg/m2, mean (SD)	28.1 (5.9)	27.6 (4.3)	0.60
Cancer, n (%)	6 (11.3)	6 (12.2)	1.0
Chronic obstructive pulmonary disease, n (%)	8(15.1)	1 (2.0)	0.03
Congestive heart failure, n (%)	3(5.8)	3(6.1)	1.0
Diabetes mellitus type 2, n (%)	11 (21.2)	10 (20.4)	1.0
Myocardial infarction, n (%)	2(3.8)	3(6.1)	0.67
Stroke, n (%)	7(13.5)	2(4.1)	0.16
Lee prognostic index (range: 0–26 points)a			0.93
0–5 (n = 21)	10(18.9)	11 (22.5)
6–9 (n = 48)	25 (47.2)	23 (46.9)
≥10 (n = 33)	18 (34.0)	15 (30.6)

Abbreviations: RMST, restricted mean survival time; SD, standard deviation.

^aHigher Lee index predicted risk of 4-year mortality for Lee index category is <4% for 0–5 points, 15% for 6–9 points, and 42%–64% for ≥10 points.

There was no statistically significant difference in the mean total DCS between the conventional decision aid group and RMST-based decision aid group (mean [SD]: 22.0 [14.3] vs 16.7 [14.1]; p = 0.06) (Table 2). However, those who received the RMST-based decision aid were significantly more likely to have a score <25, compared to those who received a conventional decision aid (34 [69.4%] vs 26 [49.1%], p = 0.04). In most subgroups, the mean total DCS scores seemed lower in the RMST-based decision aid group. The proportion of participants who preferred the intensive treatment was similar between the groups (conventional vs RMST-based decision aid: 13.2% vs 14.3%; p = 0.87).

TABLE 2.

Decisional conflict scale between conventional versus RMST-based effect summary

Population	Conventional decision aid (n = 53)	RMST-based decision aid (n = 49)	p-value
Total population, mean (SD)
Total scorea	22.0 (14.3)	16.7 (14.1)	0.06
Uncertainty subscale	28.1 (23.5)	19.4 (19.9)	0.046
Informed subscale	18.9 (13.3)	14.1 (14.6)	0.09
Values Clarity subscale	22.2 (15.8)	17.2 (14.4)	0.10
Support subscale	18.9 (15.6)	16.5 (15.4)	0.44
Effective Decision subscale	22.1 (15.9)	16.3 (15.4)	0.07
Total score <25a	26 (49.1)	34 (69.4)	0.04
Subgroups, mean (SD)
Age
<75 years (n = 24)	22.6 (19.9)	7.8 (10.0)	0.04
≥75 years (n = 78)	21.8 (12.2)	19.2 (14.1)	0.39
Sex
Male (n = 20)	26.6 (19.6)	13.1 (17.4)	0.12
Female (n = 82)	21.0 (12.8)	17.6 (13.2)	0.24
Race
White (n = 95)	21.3 (14.1)	16.7 (14.0)	0.12
Non-White (n = 7)	29.1 (15.2)	15.6 (22.1)	0.38
Education
Less than bachelor's degree (n = 37)	21.0 (14.2)	16.4 (12.4)	0.30
Bachelor's degree or higher (n = 65)	22.5 (14.5)	16.9 (15.3)	0.13
Cardiovascular diseaseb
Present (n = 18)	17.8 (15.4)	14.3 (14.0)	0.62
Absent (n = 84)	23.0 (14.0)	17.1 (15.4)	0.06
Lee prognostic index (range: 0–26 points)c
0–5 (n = 21)	16.9 (3.9)	4.8 (7.6)	0.01
6–9 (n = 48)	24.9 (15.9)	21.0 (14.7)	0.38
≥10 (n = 33)	20.8 (12.6)	18.8 (12.4)	0.64

Abbreviations: RMST, restricted mean survival time; SD, standard deviation.

^aThe Decisional Conflict Scale measures the level of uncertainty about a treatment choice, on a range from 0 (no conflict) to 100 (high conflict). A total score <25 points is associated with implementing decisions.

^bCardiovascular disease was defined as self-reported history of myocardial infarction, congestive heart failure, or stroke.

^cHigher Lee index predicted risk of 4-year mortality for Lee index category is <4% for 0–5 points, 15% for 6–9 points, and 42%–64% for ≥10 points.

Semi-structured interview

The 23 participants of the semi-structured interviews (one participant’s interview was not recorded) had a mean (SD) age of 80.7 (7.0) years, 16 (70.0%) female, 22 (96.0%) white participants, 6 (26%) with less than a bachelor’s degree education, 1 (4.0%) with congestive heart failure, 1 (4%) with heart attack, 5 (22.0%) with stroke, and 6 (26.0%) with the Lee index score ≥10 points. Their mean (SD) DCS score was 17.6 (19.0), and 5 (21.7%) had a DCS score ≥25 points.

Three themes emerged from the qualitative data as to the desired characteristics of a decision aid about blood pressure treatment: (1) conceptual clarity of the treatment effect information; (2) clarity of presentation in the visual decision aid; and (3) inclusion of desired information. Each theme is presented below, along with illustrative participant quotes. Table 3 provides additional quotes supporting each theme.

TABLE 3.

Additional quotes supporting each theme

Theme	Quotes
Theme 1: Conceptual clarity of the treatment effect information	“I would choose the one on the right [RMST Decision Aid]. It is just bar [referring to graph] - it shows the days and then you can look and see what you are comparing. So, I think the one on the right [RMST Decision Aid], are a little bit more helpful to the average person … And this is the number of days that this can happen and this is the numbers that can happen, so I definitely like the clearer one on the right [RMST Decision Aid].” [ID 118] “I do not think people rationalize things in terms of days [as done in the RMST Decision Aid]. I think they rationalize it more in terms of numbers [as done in the Conventional Decision Aid].” [ID 121] “…it [RMST Decision Aid] is more meaningful, because I donť know enough about statistics, to tell me 5 people out of whatever or 7 people out of whatever - just doesnť register for me.” [ID 129]
Theme 2: Clarity of presentation in the visual aid	“I would say it [RMST Decision Aid] is a little frustrating to make the effort to figure out what these graphs are about and find out that all they do is to illustrate the initial statement of 14 more days without cardiovascular events … So if you just look at the underlined words, you have 8 or 9 words, which tell you what you want to know without bothering to figure out what is being communicated visually here in the graphics.” [ID 114] “I asked myself ‘why are the same number of figures blacked out and the other 3 red?’ [as done in the Conventional Decision Aid] I am wondering if maybe instead of seeing exactly the same as the left side with standard strategy if it may be blacked out figures were a little bit different…” [ID 211]
Theme 3: Inclusion of desired information	“…what specific side effects and how they enter into these statistics … I would want to know what were the effects … how serious was it and could it be remedied easily or would that lead to a serious downhill slope and ultimately death? The side effects should not be all lumped together.” [ID 105] “I donť know what a kidney injury means … so I want to know more about that … about the harms that I would experience or could experience, what they would be, whether I would want to endure that or not.” [ID 130] “One of the things that is missing is the intensity of the cardiovascular event that occurred. Was it serious? Did it affect the lifestyle? Did it cause death? [RMST Decision Aid] really doesnť give you enough information to make much of a choice.” [ID 134]

Abbreviation: RMST, restricted mean survival time.

Theme 1: conceptual clarity of the treatment effect information

Conceptual clarity was viewed as the extent to which a decision aid facilitated comparisons across treatment strategies, provided abstract information, and presented understandable units of measurement. A participant (ID 145) who preferred the RMST-based decision aid for both the benefits and harms scenarios emphasized this point: compared to the “complexity of the silhouettes” [in the conventional decision aid], the bar graphs [in the RMST-based decision aid] can be “apples to apples” and “easier … for most people to understand.”

Participants’ preferences also pertained to a decision aid’s level of conceptual abstraction; this consideration proved relevant to preferences for both types of decision aids. One participant (ID 169), who preferred the conventional decision aid, described the RMST decision aid as “abstract … the words – they do not mean anything… You have no real understanding. It is talking about 1506 days versus 1342 days. What does that mean?” Meanwhile, another participant (ID 130) perceived the RMST decision aid to be “more tangible” since it “…translates that [information about treatment harms] into something that I can more easily identify with.”

Whether commenting on the RMST-based or conventional decision aid, participants found the presented units of measurement as influencing their ability to interpret information. One participant (ID 208) did not understand the meaning of the percentages in the conventional decision aid: the participant alluded to the latitude in interpretation in that a “25% increase” in cardiovascular events “could be a lot” and that they were unsure in relative terms as to whether it indeed was. A different participant (ID 129) felt that the RMST decision aid’s information was “more meaningful” than that of the conventional decision aid because “…I don’t know enough about statistics to tell me 5 people out of whatever or 7 people out of whatever – [the conventional decision aid] just doesn’t register for me.” However, several participants found it hard to understand the RMST decision aid’s use of days in place of years as the unit of measurement. A participant (ID 121) noted: “…it is very hard to interpret days”; this participant thus felt the need to translate the data back into a calculation of years.

Theme 2: clarity of presentation in the visual aid

Clarity of the visual aids was a consideration when selecting a preferred decision aid; this consideration was consistent regardless of the ultimate decision aid preference. Some participants’ comments in support of this finding were general in content. In contrast, others specifically referred to whether the text was needed along with the images, the specifics of the images’ formatting, and the images’ gestalt.

A couple of participants considered the two decision aids’ clarity of presentation or lack thereof similar, using general terms to explain their reasoning. One participant (ID 154) endorsed the RMST method in both the benefits and harms scenarios as “the … picture translates that into something that I can more easily identify with, is more tangible”, though still cited both RMST and conventional decision aids’ images as “good” and “very graphic.”

The need to supplement images with text also informed participants’ preferences. A participant (ID 154) described needing the text to clarify the conventional decision aid’s image. In contrast, another participant (ID 133) stated that they preferred the conventional decision aid’s image when considering treatment harms since the image was clear “straight away” and did not require reading the text.

Participants referred to specific features (e.g., color and font size) of the images’ formatting. A participant (ID 133) interpreted the red shading in the conventional decision aid’s image as illustrating the “danger” of choosing the more intensive treatment; this participant’s perception of the coloring as symbolic led to their view of the conventional decision aid as “effective.” A different participant (ID 154), who preferred the RMST decision aid in both the benefits and harms scenarios, found the blue color from the conventional decision aid’s image “kind of blah” and the text too small and “blurry”; their recommendation was to display fewer words or to display the words in bigger print.

Participants’ preferences were guided not just by specific formatting but also by the images’ visual gestalt. For example, a participant (ID 145) chose RMST as their preferred option due to the following thoughts: “…from a visual standpoint and a cognitive standpoint, the bar graph [referring to RMST], as long as it is well labeled and not complicated, it is easier to understand than a whole bunch of these silhouettes [referring to conventional decision aid].” Of note, the participant’s mention of RMST as visually easier to understand supports the current theme (i.e., clarity of presentation), while the mention of RMST as conceptually easier to understand supports Theme #1 (i.e., conceptual clarity).

Theme 3: Inclusion of desired information

Participants described how both decision aids did not include all information desired to guide treatment decision-making. First, participants felt the decision aids needed more detail on the characteristics and impact of each treatment option. Several participants wanted to know more about medication side effects, including what they were, whether they were short or long-term, their frequency, whether they would change with time, and the percentage of people who might experience them. Contraindications due to interactions with other medications also surfaced as desired information. Participants further stated that they would want to know about administration forms (i.e., via pills vs injection, size of pills) and the cost and related insurance coverage. In addition, a few participants thought that the decision aids should focus on how treatment impacted not just clinical outcomes but also quality of life. One participant’s (ID 163) statement encapsulated a range of desired information (e.g., long-term effects, quality of life) that touched on both decision aids: “…what would the cardiovascular events be in the next 4 years that might occur with a standard strategy versus the intensive strategy? Are those going to be life-threatening? I guess that is what I would like to know. So, what about those 14 days? What type of cardiovascular events could there be from those in the next 4 years from the 14 days? What would be my quality of life and as far as the effect of being on those medicines for 4 years versus those 3 side effects of stroke, heart attack, or cardiovascular death.”

Second, participants detailed how the personal relevance of decision aids’ information was a critical feature of their design. One participant (ID 116) had a similar reaction to both decision aids; when referring to the conventional decision aid, the participant explained explicitly: “…it is primitive or simple or basic as it is … without, you know, any explanation or even relevance to a person…”. A participant (ID 121) noted how having personalized information about treatment options based on individual demographics would be easier to understand and relate to: “…there are some older people who have difficulty interpreting these graphs, and I think maybe if it was put down on a more personal level - as it is just numbers now. To me, it is just projections and stuff…”. As some participants elaborated, such information should incorporate, for example, age, gender, medical history including comorbidities, and genetics.

DISCUSSION

In recent years, RMST, which summarizes treatment effect in terms of a gain or loss in the event-free days,^2,3 has been proposed as a methodologically robust and clinically intuitive measure of treatment effect. To our knowledge, this is the first mixed-methods study to understand older adults’ preference between conventional and RMST-based decision aids. We found no clear preference for one type of effect summary over the other among older adults. Individuals’ preference was informed by conceptual clarity of the treatment effect information, clarity of presentation in the visual aid, and inclusion of desired information in the decision aid.

In our quantitative survey, the difference in total DCS between conventional and RMST-based decision aids did not meet a pre-specified level of statistical significance. Notably, DCS scores were low in our study, suggesting that most participants were comfortable with their decision and did not have a high level of uncertainty. Therefore, adequately powering for a reduction in DCS score was more challenging than anticipated. Nonetheless, there was a trend toward lower mean DCS score in the group presented with RMST-based summary (mean [SD]: 22.0 [14.3] vs 16.7 [14.1]; p = 0.06), which was consistent across most DCS subscales and participants’ demographic and health status subgroups. In addition, we found more people in the RMST-based decision aid group had a DCS score <25, which was associated with implementing decisions,¹² compared with the conventional decision aid group. In our previous online survey, the total DCS was similar between the conventional and RMST-based decision aid groups (mean [SD]: 25.6 [14.1] vs 25.2 [15.0]; p = 0.84).⁵ Compared to the respondents of the online survey,⁵ the participants of the current survey were older (mean age: 81 vs 72 years) and at higher risk of mortality (32.4% vs 18.0% with Lee index score ≥10). Taking this into account, these results suggest that an RMST-based decision aid may suggest that time-based presentations can be useful in certain subgroups of older adults.

Our semi-structured interviews suggest that the preference depended on an individual’s ability to comprehend quantitative information on treatment effect presented in the decision aid (theme #1: conceptual clarity of the treatment effect information). Some participants expressed that the number of event-free days or RMST was more intuitive than the conventional probability-based summary, while others felt that the use of “days” was more abstract than the use of “years” in presenting RMST. Several participants had difficulty understanding absolute and relative risk reduction and how two seemingly different numbers were related.

We also found that the format of visual aids influenced the participants’ understanding and preference (theme #2: clarity of presentation in the visual aid). The bar graph comparing the event-free days between the treatment groups (RMST decision aid) was easier to read and facilitated the understanding of RMST differences. The person-icon figure presenting the risk of events in each treatment group (conventional effect decision aid) was difficult to read and seemed to add to the difficulty of understanding absolute and relative risk reduction. Some participants thought that the conventional decision aid was more effective because it used red color to indicate the number of people who would develop an event. Use of larger text, fewer words, and clear color contrast in visual aids was recommended by several participants.

Regardless of presented decision aid, participants felt that the decision aids were not specific enough to their unique situations and lacked personalizing information, as well as more details about what interventions would be (theme #3: inclusion of desired information). Much of the information that participants felt would be informative pertained to side effects and quality of life. This highlights the limitations of decision aids (presented in a paper format as in our study), which may not provide patient-centered information in sufficient detail. The use of video-based decision aids or interactive websites may be helpful in presenting additional information in a more personalized and comprehensible way.¹⁴ Because the desired information by individual patients is often not examined in RCTs, all decision aids should be accompanied by a discussion with knowledgeable health care providers.

Our study has limitations that deserve mention. First, our quantitative survey had insufficient statistical power to detect a small difference in DCS score. However, we found a statistically significant difference in the proportion of participants with a DCS score <25. Second, our quantitative and qualitative samples consisted predominantly of self-reported white race and older adults with at least a Bachelor’s degree or more. This sample homogeneity limits the generalizability of our quantitative results and the transferability of our qualitative results to non-white races or less educated populations. An additional example relates to our stratified purposive sampling approach when drawing the qualitative subset of participants from the quantitative sample. That is, we were unable to draw an interview sample that was evenly distributed across the strata given the proportionally lower number of survey respondents with high DCS scores. Third, given the COVID-19 pandemic, we needed to conduct the surveys and interviews by telephone rather than in-person. Participants may have found the decision aid materials more confusing to understand than they would have with a more easily interactive in-person approach.

In conclusion, our findings do not suggest a single effect summary measure for decision aids that is universally better understood by older adults. Given that different older adults have different familiarity with quantitative information, it may be useful to present treatment effect in terms of gain or loss in event-free days as well as absolute and relative risk reductions. Visual aids can facilitate understanding of the evidence when they present information consistent with the text description in a simple format (e.g., bar graphs showing an event-free time in years). Whether RMST-based decision aid is more useful than conventional decision aid in certain older people (e.g., those with a limited life expectancy, minorities, or those with less education) warrants further research.

Supplementary Material

supinfo

Appendix S1. Decision aid and semi-structured interview guide.

Key points

• Restricted mean survival time (RMST) summarizes treatment effect in terms of a gain or loss in the event-free days, which may provide a more intuitive interpretation compared to relative risk reduction or number needed to treat (“conventional effect summary”).

• In a randomized trial of decision aids for standard versus intensive blood pressure treatment among well-educated predominantly white older adults, we found no statistically significant difference in overall older adults’ perceived uncertainty.

• In secondary analyses, there was suggestion that time-based RMST decision aids may reduce uncertainty in decision-making and be preferred to conventional decision aids. Qualitative work pointed more toward a balanced view of both aids.

Why does this paper matter?

Given the variable individual preference in receiving information on treatment effect among older adults, it may be useful to present both conventional effect summary and RMST-based effect summary in decision aids to facilitate treatment decision-making.