Man-Son-Hing M, Laupacis A, O'Connor AM, Biggs J, Drake E, Yetisir E, Hart RG, for the Stroke Prevention in Atrial Fibrillation Investigators . A Patient Decision Aid Regarding Antithrombotic Therapy for Stroke Prevention in Atrial FibrillationA Randomized Controlled Trial. JAMA. 1999;282(8):737-743. doi:10.1001/jama.282.8.737
Author Affiliations: Clinical Epidemiology Unit, Loeb Health Research Institute (Drs Man-Son-Hing, Laupacis, and O'Connor and Mss Biggs, Drake, and Yetisir), Geriatric Assessment Unit (Dr Man-Son-Hing), Ottawa Hospital, Ottawa, Ontario; Department of Medicine (Drs Man-Son-Hing and Laupacis), School of Nursing (Dr O'Connor), University of Ottawa; and Department of Medicine (Neurology), University of Texas, San Antonio (Dr Hart).
Context Decision aids are tools designed to help patients participate in the
clinical decision-making process.
Objective To determine whether use of an audiobooklet (AB) decision aid explaining
the results of a clinical trial affected the decision-making process of study
Design Randomized controlled trial conducted from May 1997 to April 1998.
Setting Fourteen centers that participated in the Stroke Prevention in Atrial
Fibrillation (SPAF) III trial.
Participants A total of 287 patients from the SPAF III aspirin cohort study, in which
patients with atrial fibrillation and a relatively low risk of stroke received
325 mg/d of aspirin and were followed up for a mean of 2 years.
Intervention At the end of SPAF III, participants were randomized to be informed
of the study results with usual care plus use of an AB (AB group) vs usual
care alone (control group). The AB included pertinent information to help
patients decide whether to continue taking aspirin or switch to warfarin.
Main Outcome Measures Patients' ability to make choices regarding antithrombotic therapy,
and 6-month adherence to these decisions. Their knowledge, expectations, decisional
conflict (the amount of uncertainty about the course of action to take), and
satisfaction with the decision-making process were also measured.
Results More patients in the AB group made a choice about antithrombotic therapy
than in the control group (99% vs 94%; P=.02). Patients
in the AB group were more knowledgeable and had more realistic expectations
about the risk of stroke and hemorrhage (in the AB group, 53%-80% correctly
estimated different risks; in the control group, 16%-28% gave correct estimates).
Decisional conflict and satisfaction were similar for the 2 groups. After
6 months, a similar percentage of patients were still taking their initial
choice of antithrombotic therapy (95% vs 93%; P=.44).
Conclusions For patients with atrial fibrillation who had participated in a major
clinical trial, the use of an AB decision aid improved their understanding
of the benefits and risks associated with different treatment options and
helped them make definitive choices about which therapy to take. Further studies
are necessary to evaluate the acceptability and impact of decision aids in
other clinical settings.
Decision aids are tools designed to help patients participate in the
clinical decision-making process and make informed choices consistent with
their personal values. Compared with general educational materials (such as
informational pamphlets), decision aids provide detailed descriptions of clinically
important outcomes and their consequences, provide quantitative information
about the likelihood of these outcomes (often tailored to the patient's own
clinical risk profile), are more explicit about the therapeutic choices, and
encourage patients to indicate which therapy they currently favor.1 Decision aids are usually developed for clinical situations
in which the relative values of the benefits vs risks are unclear. They are
designed to be adjuncts to the patient-physician interaction rather than substitutes.
A number of formats for decision aids are available, including interactive
videodiscs,2 audiobooklets (ABs),3
and decision boards.4
Patients with atrial fibrillation have an increased risk of stroke,
and antithrombotic therapy is widely recommended for stroke prevention.5 Long-term therapy with warfarin decreases the risk
of stroke by about 68%,6 and aspirin decreases
the risk by 21%.7 However, use of warfarin
is associated with a greater chance of major bleeding8
and is more complicated to use than aspirin.9
Since individual patients with atrial fibrillation are likely to perceive
differently the trade-off between the efficacy and adverse effects of warfarin
vs aspirin therapy, there is no right or wrong choice of antithrombotic therapy
for many of these patients. Thus, a decision aid may be beneficial.
To test the validity of a previously developed risk stratification scheme,
the Stroke Prevention in Atrial Fibrillation (SPAF) investigators recently
completed a trial (SPAF III) in which patients with atrial fibrillation who
were deemed to be at relatively low risk of stroke were treated with 325 mg/d
of aspirin and followed up for a mean of 2 years.10,11
At the end, participants were informed of the study results and, in conjunction
with their physicians, used this information to decide whether they wished
to continue taking aspirin or switch to warfarin therapy. To assess the effect
of a decision aid on the decision-making process, they were randomized to
receive or not receive an AB in addition to usual counseling.
All 20 SPAF centers were invited to participate in the randomized trial.
All patients in participating centers who were in the SPAF III aspirin cohort
study10 were eligible for this study, except
those who had high-risk criteria or had a major hemorrhage during the study.
The ethics review boards of each participating center approved the study,
and informed consent was obtained from all participants. Pertinent patient
characteristics (eg, age, sex, educational level, and previous use of warfarin
and aspirin) were collected at entry into SPAF III.
to a computer-generated scheme, administered from a central location to block
the sequence from previewing, patients were randomized to receive the AB or
usual care. Randomization was also stratified by center and the presence of
a history of hypertension.
The AB decision aid consisted of a 29-page booklet, a personal worksheet,
and a 20-minute audiotape that guided the participants through the booklet
and worksheet. The booklet highlighted key points (similar to a slide presentation),
and the audiotape connected the points in a narrative format, providing more
detail than the booklet. The AB contained descriptions of the consequences
of a minor stroke, a major stroke, and a major hemorrhage; the blood monitoring
required for warfarin therapy; and the 2-year probabilities of stroke and
major hemorrhage for patients taking aspirin or warfarin. Probabilities were
presented in the booklet using 100 icons (Figure 1), whereas the text and AB also presented the chance of
experiencing and not experiencing a stroke in percentages. The probability
of a stroke while taking aspirin was derived directly from the results of
the SPAF III cohort study10 in which the patients
had recently participated (3% over 2 years in patients without a history of
hypertension and 8% in patients with a history of hypertension). The probability
of stroke while taking warfarin was calculated by assuming that warfarin was
approximately 50% more efficacious than aspirin (2% over 2 years for patients
without a history of hypertension and 4% for patients with a history of hypertension).6 The risk of major hemorrhage was presented as 1% over
2 years while taking aspirin and 3% while taking warfarin.6
An accompanying 12-page physician's manual was also developed, which summarized
the material in the AB and provided references.
After reviewing the booklet, patients completed a 1-page worksheet.
It included sections that clarified their values for possible outcomes (eg,
stroke and major bleeding), asked them to list any questions they had about
the decision, and elicited which therapy (aspirin, warfarin, or unsure) they
were inclined to take. Participants also indicated their preferred role in
the decision-making process (ie, their physician should make the decision,
the patient should make the decision, or the decision should be shared).
The AB was provided as an adjunct to each center's usual decision-making
process at the end of the trial. Patients randomized to receive the decision
aid were sent the AB a few days before they met with their physicians. They
reviewed the AB and completed the personal worksheet before their visits with
their physicians. The physicians received copies of the physician's manual
before they met the patients.
Control group subjects received usual care, ie, no change was made to
the usual manner in which each center communicated the results of the SPAF
III study to patients or to the way in which the decision regarding type of
antithrombotic therapy was made. The methods used by centers to inform participants
about the results of SPAF III varied. Of the participating centers, 7 asked
participants to return to the SPAF clinic to discuss the results of the study,
5 sent summaries of the results of the study to the patients' personal physicians
and asked patients to arrange appointments with these physicians, and 2 held
an end-of-study gathering at which participants were given the study results
and asked to follow up with their personal physicians.
One to 4 days after meeting with their physicians, all patients completed
a questionnaire eliciting information about the following outcome measures.
Patients were asked to indicate whether a decision regarding the choice
of antithrombotic therapy had been made in conjunction with their physician
and, if so, what this choice was. On a 5-point Likert scale, they also judged
the relative strength of their personal input into the choice vs their physicians'.
Knowledge was tested using 23 questions about atrial fibrillation, stroke,
and the advantages and disadvantages of taking warfarin or aspirin. These
questions (eg, "Taking aspirin daily means that you have to go for regular
blood testing") had the potential responses "true," "false," and "unsure."
Patients' expectations about the probability of stroke and major hemorrhage
with aspirin or warfarin therapy were quantitatively assessed with 4 questions.
Each question contained 14 response options on a probability scale (eg, "If
you continue to take aspirin, your risk of stroke during the next two years
is. . . ." Response options ranged from "0% to 0.5%" to "80% to 100%").
The decisional conflict scale12 measured
patients' uncertainty about which therapy to choose, modifiable factors contributing
to uncertainty (believing themselves to be uninformed, unclear about values,
and unsupported in decision making), and perceived effective decision making.
The scale is reliable,12,13 discriminates
between those who make or delay decisions,12,13
is responsive to change,3,14,15
and discriminates between different decision-supporting interventions.3,16,17 Two items were added
to elicit patients' perceptions that they were informed about the benefits
and risks of warfarin and, separately, about benefits and risks of aspirin.
This did not affect the scale's reliability in this study (Cronbach α=.92).
Satisfaction with various aspects of the decision-making process was
assessed with 6 questions using a 5-point Likert scale (1, strongly agree;
2, agree; 3, neither agree nor disagree; 4, disagree; and 5, strongly disagree).
Adherence to their decisions regarding antithrombotic therapy was assessed
6 months later using a brief questionnaire administered by telephone. Participants
were asked which therapy they were currently taking and the reasons for any
change from their original decision.
Copies of the study materials are available on the Internet (http://www.lri.ca).
A sample size calculation was not performed because we attempted to
enroll as many patients from the SPAF III aspirin study as possible. Differences
in outcomes between the patients who received the AB and those in the control
group were compared with χ2 and t
tests as appropriate. A forward stepwise logistic regression procedure was
performed to adjust raw outcome proportions using significant covariates to
predict outcomes. Covariates were submitted to the logistic model at P<.10. A priori, the baseline factors thought most likely
to affect the impact of the decision aid, and, thus, included in the model,
were age, sex, education, and whether patients had ever taken warfarin prior
to participation in SPAF III. An α level of .05 was used to indicate
The trial was conducted from May 1997 to April 1998. Two hundred eighty-seven
patients at 14 SPAF centers were randomized to either receive (n=139) or not
receive (n=148) the AB. Figure 2
shows participant flow through the trial. Table 1 compares characteristics of SPAF III patients who did and
did not participate in the AB trial. The mean (SD) length of clinic visits
did not differ significantly between the AB and control groups (AB group,
27  minutes; control group, 25  minutes; P=.51).
A few days after visits with their clinicians, more patients in the
AB group (n=138) were able to make definite choices regarding antithrombotic
therapy compared with those in the control group (n=139; 99% vs 94%; P=.02). Overall, the proportion of patients who decided
to take warfarin was higher in the control group (n=12 [8%] in the AB group,
n=17 [11%] in control group; P=.02). The 119 patients
with hypertension (n=20 [17%] overall; 7 [12%] of 58 in AB group, 13 [21%]
of 61 in control group) were more likely to choose warfarin compared with
the 168 patients without hypertension (9 [5%] overall; 5 [6%] of 81 in AB
group, 4 [5%] of 87 in control group) (P=.003). Patients
taking long-term warfarin therapy prior to enrollment were more likely to
switch to warfarin than those who were not (10 [13%] of 75 vs 15 [7%] of 212,
Eighty-seven (63%) worksheets from the 139 patients in the AB group
were returned. After review of the AB, 79 (91%) of the 87 patients were inclined
to take aspirin, 1 (1%) to take warfarin, and 7 (8%) were unsure. After meeting
with their practitioners, all patients who indicated a preference except 2
decided to take the medication they were favorably disposed toward (those
2 decided to take warfarin). Of the patients who were unsure, 6 decided to
take aspirin and 1 decided to take warfarin.
Similar percentages of participants in the AB and control groups reported
that they, rather than their physicians, made the decision regarding antithrombotic
therapy (n=85 [61%], AB group; n=83 [56%], control; P=.43).
Participants' choice of antithrombotic therapy (aspirin or warfarin) was not
affected by the method with which centers informed participants of the results
of the SPAF III aspirin study (ie, clinic visit, letter to personal physician,
end-of-study gathering) (P=.62).
Patients who reviewed the AB were generally more knowledgeable about
the pertinent clinical issues regarding stroke, atrial fibrillation, and their
treatment and consequences compared with patients who received usual care
(Table 2). Patients receiving
the AB were also more willing than those in the control group to make quantitative
estimates of their chance of stroke and major bleeding when taking aspirin
or warfarin (134 [96%] of 139 vs 113 [76%] of 148, respectively; P<.001). Compared with patients receiving usual care, a higher percentage
of patients reviewing the AB gave correct quantitative estimates of their
stroke and bleeding risks when taking aspirin or warfarin (Table 2). As an example, individual responses of participants with
hypertension regarding their chance of stroke if taking warfarin are shown
in Figure 3.
There was no statistically significant difference in overall decisional
conflict between patients who received the AB and those who did not (P=.14) (Table 3).
When examining subscales of the decisional conflict scale, those receiving
the AB believed they were more informed compared with those who did not (−0.21
units; 95% confidence interval [CI], −0.34, to −0.08).
Use of the AB did not significantly affect patients' satisfaction with
various aspects of their interaction with their practitioner (Table 4), although there was a trend toward patients using the AB
to become more satisfied with the decision-making process (P=.10).
Six-month follow-up data regarding adherence to patients' initial choice
of therapy were available for 92% (263/287) of the participants. A similar
percentage of patients in both groups continued to take the therapy that was
initially chosen (AB group, 123 [95%] of 129; control group, 125 [93%] of
134; P=.44). For patients who initially chose warfarin,
2 (17%) of 12 in the AB group switched to aspirin 6 months later, while 1
(6%) of 17 in the control group switched. For those who initially chose aspirin,
110 (96%) of 114 in the AB group continued taking aspirin, with 4 (3%) switching
to warfarin; for the control group, 104 (93%) of 112 continued taking aspirin
with 8 (7%) switching to warfarin. All patients who were undecided about their
choice of antithrombotic therapy immediately after their clinic visit (n=8)
were taking aspirin 6 months later.
While controlling for use of the AB, stepwise logistic regression revealed
that certain baseline factors were independent predictors of the various study
outcomes. Previous warfarin use (odds ratio [OR], 2.18; 95% CI, 1.01-4.74; P=.04) was an independent predictor of choosing warfarin
as the initial antithrombotic therapy. Age categories were defined as younger
than 60 years, 60 to 75, and older than 75 years. Younger age (OR, 1.49; 95%
CI, 1.03-2.15; P=.04) and male sex (OR, 1.89; 95%
CI, 1.07-3.33; P=.04) were independent predictors
of lower overall decisional conflict score. Higher educational level, defined
as did not complete high school, did complete high school, or greater (OR,
1.96; 95% CI, 1.33-2.89; P=.01) and younger age (OR,
1.67; 95% CI, 1.14-2.45; P=.04) were significantly
associated with higher knowledge scores. No baseline factor was independently
associated with improved satisfaction scores.
Compared with patients in the usual care group, those who used the decision
aid were more likely to make a decision regarding antithrombotic therapy;
were more knowledgeable about treatment options, benefits, and risks; and
had much more realistic expectations about their chance of stroke and major
bleeding. However, they did not demonstrate significant differences in overall
decisional conflict or in satisfaction with the decision-making process.
The results of this trial are compatible with other randomized controlled
trials involving patients faced with treatment decisions about hormone replacement
therapy,3 benign prostatic hyperplasia,2 and coronary artery disease.16
In these trials, compared with patients receiving usual care, patients using
decision aids were consistently more knowledgeable, believed they were more
informed about the pertinent clinical issues, and had more realistic expectations
about the probability of outcomes. Also, another study did find a positive
effect on satisfaction for patients contemplating surgery for benign prostatic
the AB did exert a small influence on the eventual choice of antithrombotic
therapy. With greater knowledge and awareness of pertinent clinical issues,
slightly more patients who used the AB compared with control patients preferred
to continue taking aspirin rather than to switch to warfarin. This result
is supported by results of other trials that evaluated the impact of decision
aids. In those trials,2,16 patients
made more conservative decisions regarding therapy after being informed of
the benefits and risks of therapeutic options, resulting in trends toward
lower rates of surgery for benign prostatic hyperplasia2
and coronary artery disease,16 respectively.
The conservative selection may be due to more realistic expectations of potential
benefits and harms and to correction of exaggerated notions of the baseline
risks and benefits of treatment after exposure to the decision aids.
By participating in the SPAF III aspirin trial, all patients were familiar
with taking aspirin daily, whereas only 38% had taken warfarin previously.
Therefore, given the explicit description of the risks and inconveniences
of aspirin and warfarin therapy in the AB, it was not surprising that those
reviewing the AB demonstrated greater knowledge and awareness of issues regarding
warfarin therapy. More surprisingly, patients receiving the AB, compared with
patients receiving usual care, also demonstrated significantly greater knowledge
and awareness of issues regarding aspirin therapy.
A similar percentage
of patients in the AB and control groups believed that their physicians played
an important role in the decision-making process, showing that decision aids
are supplements, rather than alternatives, to the patient-physician interaction.
Our study has several possible limitations. Contamination may
have caused the limited effect of the decision aid for the outcomes of decisional
conflict and 6-month adherence. During clinic visits with patients receiving
usual care, physicians may have provided patients with information similar
to that contained in the AB, making the benefit of the AB harder to detect.
Randomizing participating centers might have reduced the possibility of contamination
but would have increased the possibility of imbalance for pertinent participant
characteristics between the AB and usual care groups.
in the usual care group reported low overall decisional conflict and high
satisfaction with the patient-practitioner interaction. Thus, there may have
been a ceiling effect, with little chance of the AB significantly improving
overall patient satisfaction with the decision-making process. Patients in
this study may have been more likely to receive greater personalized care
and been better informed about their conditions and treatment compared with
average patients with atrial fibrillation, making the effects of the decision
aid harder to detect. Therefore, the benefits of decision aids may be greater
in usual care settings. Further evaluation of our AB in a general clinic setting
Other possible limitations of our study include the
relatively low percentage of eligible patients who participated. Also, some
baseline characteristics (eg, socioeconomic status) that may have affected
the influence of the AB were not included in our logistic regression analysis,
because they were not recorded at SPAF III entry.
This study is
the largest randomized trial of a decision aid, the patients were evaluated
when they were making clinical decisions, and the outcome measures used were
comprehensive. The decision aid was well accepted by patients, and those who
received the AB were better informed about atrial fibrillation and the benefits
and risks of the treatment alternatives than patients who received usual care.
Previous studies have shown that many patients with atrial fibrillation are
still not being prescribed warfarin therapy.18,19
In some circumstances, physicians may have inappropriately failed to offer
warfarin therapy to high-risk patients (eg, patients with a previous transient
ischemic attack). On the other hand, fully informed patients may have decided
not to take warfarin because they perceived that the benefits did not outweigh
the risks and inconvenience of therapy. Given the prevalence of atrial fibrillation
in older persons, it would seem appropriate to evaluate the acceptability
and impact of decision aids for patients with atrial fibrillation in clinical
practice. Decision aids should be tailored to the risk profile of individual
patients, updated as new information becomes available, and evaluated for