Patient Education and Counseling 69 (2007) 145–157

www.elsevier.com/locate/pateducou

Patient education about anticoagulant medication: Is narrative

evidence or statistical evidence more effective?
Kathleen M. Mazor a,*, Joann Baril a, Elizabeth Dugan b, Frederick Spencer c,
Pamela Burgwinkle d, Jerry H. Gurwitz a
a
Meyers Primary Care Institute, University of Massachusetts Medical School, Fallon Clinic Foundation,
and Fallon Community Health Plan, United States
b
University of Massachusetts Medical School, United States
c
McMaster University Faculty of Health Sciences, Canada
d
University of Massachusetts Memorial Health Care, United States
Received 14 December 2006; received in revised form 6 August 2007; accepted 20 August 2007

Abstract
Objective: To determine the relative impact of incorporating narrative evidence, statistical evidence or both into patient education about warfarin,
a widely used oral anticoagulant medication.
Methods: 600 patients receiving anticoagulant therapy were randomly assigned to view one of three versions of a video depicting a physician–
patient encounter where anticoagulation treatment was discussed, or usual care (no video). The videos differed in whether the physician used
narrative evidence (patient anecdotes), statistical evidence, or both to highlight key information. 317 patients completed both the baseline and post-
test questionnaires. Questions assessed knowledge, beliefs and adherence to medication and laboratory monitoring regimens.
Results: All three approaches positively effected patients’ warfarin-related knowledge, and beliefs in the importance of lab testing; there was also
some indication that viewing a video strengthened belief in the benefits of warfarin. There was some indication that narrative evidence had a greater
impact than statistical evidence on beliefs about the importance of lab testing and on knowledge. No other evidence of the differential effectiveness
of either approach was found. No statistically significant effect was found on intent to adhere, or documented adherence to lab monitoring.
Conclusion: Videos depicting a physician–patient dialogue about warfarin were effective in educating patients about anticoagulant medication,
and had a positive impact on their beliefs. The use of narrative evidence in the form of patient anecdotes may be more effective than statistical
evidence for some patient outcomes.
Practice implications: Patients on oral anticoagulant therapy may benefit from periodic educational efforts reinforcing key medication safety
information, even after initial education and ongoing monitoring. Incorporating patient anecdotes into physician–patient dialogues or educational
materials may increase the effectiveness of the message.
# 2007 Elsevier Ireland Ltd. All rights reserved.

Keywords: Physician–patient communication; Patient education; Medication adherence; Patient safety

1. Introduction may be complex with frequent changes due to the impact of

Warfarin is a widely used anticoagulant medication, with
significant therapeutic benefits, but also significant risks [1].
Managing warfarin to achieve the maximum benefit can be
challenging for both provider and patient [2]. Dosing schedules
* Corresponding author at: University of Massachusetts Medical School,
Meyers Primary Care Institute, 630 Plantation Street, Worcester, MA 01605,
United States. Tel.: +1 508 791 7392; fax: +1 508 595 2200.
E-mail address: Kathleen.Mazor@umassmed.edu (K.M. Mazor).
0738-3991/$ – see front matter # 2007 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.pec.2007.08.010
diet, illnesses, and other medications on warfarin absorption
and metabolism. These issues complicate adherence and
increase patient errors [3]. Importantly, the risks associated
with warfarin, including life-threatening bleeding, increase
substantially if patients do not follow recommendations
regarding dosing and monitoring. Laboratory testing to monitor
international normalized ratios (INRs) must be conducted
frequently to ensure that INRs are maintained within the
therapeutic range, and to prevent serious bleeding events.
Patients who do not follow dosing and monitoring regimens are
not only decreasing potential therapeutic benefits, but are also
146 K.M. Mazor et al. / Patient Education and Counseling 69 (2007) 145–157
increasing their risk of suffering an adverse event. As many as
36,000 patients are treated at hospital emergency rooms
annually for adverse events related to anticoagulant treatment
[4]. A recent national survey found that 93% of patients on oral
anticoagulants had experienced adverse events [2], and other
researchers have found that many of the adverse events in
patients on anticoagulants are due to preventable patient errors
[5,6]. At least some patient errors are probably attributable to
patients0 inadequate knowledge of anticoagulant therapy and
the associated risks [7]. There is a clear need for more effective
patient education about anticoagulant therapy [2,7,8].
While evidence-based medicine has become the standard for
clinicians [9], it is not necessarily true that patients find
statistical evidence useful, or even comprehensible. In fact,
there is a growing body of literature suggesting that many
patients, even highly educated ones, have considerable
difficulty understanding quantitative information [10,11].
Within warfarin patients in particular, there appears to be a
link between time in therapeutic range and numeracy skill [12].
Difficulties with statistics and numeracy has led to suggestions
for contextualizing risk messages, for instance through
testimonials or other narratives, which describe the antecedents
and/or consequences of a potential health problem [13].
In fact, narrative health information is ubiquitous. People are
exposed to narrative health messages when they talk about
health experiences with friends and family members. Findings
from a recent qualitative study suggest that patients’ treatment
decisions were influenced by ‘‘naturally occurring’’ narratives
(e.g., family members’ outcomes after treatment of similar
problems). People are also exposed to health narratives in the
media. Journalists routinely and deliberately use narratives to
exemplify issues, and to increase vividness, authenticity, and
audience interest [14]. In the context of patient education,
patient narratives have been collected, edited and made
available to patients and caregivers in book, pamphlet and
video form with the intent of educating and supporting patients
[15,16].
There are several characteristics that differentiate narrative
messages from statistical messages. Narratives are typically
considered more familiar, concrete, and vivid than statistical
messages. Another important distinguishing characteristic of
narratives is that events either happen or not. Fagerlin et al. [17]
refers to this as presenting a dichotomy; in fact, narratives
typically present one side of a dichotomy, i.e., a particular event
is specified. In contrast, statistical representations almost
always contain uncertainty, e.g., a 10% risk of an outcome.
These qualities (familiar, concrete, vivid, and event specificity)
may help to make narrative messages more engaging,
memorable, realistic, comprehensible, and believable than
statistical messages. However, to say that a narrative message is
‘‘engaging’’ or ‘‘realistic’’ does not explain the mechanism by
which such messages have an impact. Narratives have been
called the ‘‘basic mode of human interaction [18].’’ Given this,
it is possible that people are either ‘‘hard wired’’ to receive and
process narrative information, or that extensive experience
results in adeptness. There is evidence that stories and technical
information are processed differently by the brain [19], but
again, this work is descriptive rather than explanatory. Hinyard
and Kreuter provide a detailed overview of the issues around
using narrative messages in health communication, including a
review of potentially relevant models of persuasion and health
behavior change [18]. Ultimately, Kreuter and co-workers
suggest that narratives may function by facilitating information
processing, reducing counter-arguing, facilitating observa-
tional learning, and influencing perceptions susceptibility
and social norms [18,20].
Despite the apparent intuitive appeal of narrative messages,
and these plausible descriptions of how such messages might
influence health behaviors, the evidence on whether narrative
messages influence attitudes, beliefs, behaviors or knowledge
more strongly than statistical messages is mixed, with some
reviewers concluding that narratives are more influential, and
others concluding that statistics have a greater impact [18,21–
23]. Four recent studies compared the relative impact of
narrative and statistical messages on health-related behaviors
specifically; again results have been mixed [24–27]. Impor-
tantly, two of these studies [25,26] and many of the studies of
the impact of other important variations in message design on
knowledge, beliefs, behaviors or behavioral intent have studied
the responses of undergraduate college students [23,28–33],
significantly limiting the extent to which findings are general-
izable to older adults.
It is important to note that the use of narrative messages in
patient education and counseling may be controversial.
Selective use of narrative evidence has the potential to distort
reality, even if the information in the narrative is technically
accurate [14]. The concern that narratives may induce
inaccurate beliefs or result in non-optimal decision-making
is well-founded. Not only is it well documented that many
people have difficulty understanding and using numerical
information [34], there is also evidence that people may
disregard statistical data in favor of narrative information, such
as testimonials or anecdotes [17,35–37].
The limited research and contradictory findings on whether
narrative or statistical messages are more effective in commu-
nicating health information, and the absence of research on the
effectiveness of these two sorts of evidence combined, provide
the motivation for the present study. Medication safety in
general, and warfarin in particular provides an appropriate
context for studying the impact of variations in oral commu-
nication, because of the potential for serious or life threatening
adverse events if communication is not effective. Therefore, the
purpose of the present study was to compare three approaches to
communicating key information about anticoagulation (inte-
grating narrative evidence, statistical evidence or a combination
of the two) on patients’ knowledge, beliefs and behaviors around
anticoagulant therapy and monitoring.
2. Methods
2.1. Design
The study utilized a randomized clinical trial design, testing
three approaches to communicating with patients about safe
 K.M. Mazor et al. / Patient Education and Counseling 69 (2007) 145–157
warfarin use: incorporating narrative evidence (patient anec-
dotes), statistical evidence, and narrative and statistical
evidence combined, and compared them to a usual-care
control. Each participant completed a baseline questionnaire.
Approximately 3 weeks after the baseline questionnaire was
received at the research office, the second set of study materials
was sent. For participants assigned to one of the video
conditions, the mailing included a video depicting a physician–
patient encounter about oral anticoagulant medication manage-
ment; the physician used narrative evidence, statistical
evidence or both to support his recommendations. The post-
questionnaire was included with the video for patients in the
three intervention conditions; patients assigned to usual care/
control received the post-questionnaire but no video. The study
was conducted between March and July 2005.
2.2. Participants
Adult patients receiving care from an anticoagulation clinic
located in an academic health center (UMassMemorial Medical
Center, Worcester, Massachusetts) were eligible for the study if
they had received services for at least 3 months prior to the
recruitment date and had current active prescriptions for
warfarin according to clinic records. The clinic staff reviewed
the list of potentially eligible patients and excluded those
known to be unable to understand English, to have cognitive
impairment or medical problems that would preclude
participation (e.g., hospitalization), or not residing in the area
at the time of the study (e.g., patients who resided in Florida a
portion of the year). A total of 680 patient names, randomly
selected from the clinic list of current patients, were reviewed
by clinic staff; 80 were excluded, resulting in 600 patients
potentially eligible for the study.
The Institutional Review Board of the University of
Massachusetts Medical School reviewed and approved the
study protocol.
2.3. Materials
2.3.1. Questionnaires
Questionnaires administered at baseline and post-interven-
tion included items intended to assess warfarin-related
knowledge, belief that warfarin has a positive impact on one’s
health, belief that taking warfarin as prescribed is confusing or
difficult, self-reported adherence to warfarin dosing regimen,
belief that lab testing is important, and intent to adhere to
recommendations for lab testing. The questionnaire was
developed by the authors after review of existing instruments
[38–42]. None of the reviewed instruments provided sufficient
coverage of all content areas considered important for the
present study. Eight items were drawn directly from the Beliefs
About Medication Scale [41]. A small number of background
items were included (e.g., age, education, self-rated health). All
other items were developed specifically for this study. All items
were reviewed by two physicians (F.S., J.H.G.) prior to
administration. Preliminary versions of the full questionnaire
were pre-tested using cognitive interviewing with a conve-
147
nience sample of 4 anticoagulation clinic patients in 1:1
sessions. A brief description of each set of items follows.
2.3.1.1. Warfarin-related knowledge. A 22 item test of
warfarin-related knowledge was developed for this study.
Items tested knowledge of foods that may affect INR levels,
medications (over the counter and prescription) that may affect
INR levels, consequences of over- and under-dosing, and
symptoms that should be reported to the anticoagulation clinic
staff. Response options were ‘‘True’’ ‘‘False’’ and ‘‘Don’t
Know’’. All knowledge items and the scoring key were
reviewed by 2 physicians (J.H.G., F.S.). Each item was scored
correct/incorrect (‘‘Don’t Know’’ responses were scored
incorrect); total score was calculated as the percent correct
across all 22 items.
2.3.1.2. Beliefs. A total of 25 new items were developed to
assess beliefs about warfarin, the warfarin regimen, and the
laboratory regimen; in addition, 8 items from the Beliefs About
Medication (BAM) scale [41] were included, for a total of 33
items. Response options were Strongly Agree, Agree, Neutral,
Disagree, and Strongly Disagree, except for the BAM items,
where the middle response option was labeled Uncertain. After
psychometric analyses (described below), a total of 22 items
were retained, grouped into 5 subsets. These items and the
groupings used are provided in Table 1.
2.3.1.3. Adherence. Intent to adhere to recommended labora-
tory testing was assessed by two items developed for this study:
‘‘If the doctor recommended that I get my blood tested every
day, I would do it.’’ and, ‘‘If the doctor recommended that I get
my blood tested every week, I would do it.’’ Response options
were Strongly Agree, Agree, Neutral, Disagree, and Strongly
Disagree. Documentation of actual adherence to laboratory
monitoring was extracted from clinic records related to
attendance at scheduled lab appointments for 3 months prior
to the receipt of the completed pre-questionnaire, and 3 months
after receipt of the post-intervention questionnaire for patients
who provided written consent giving access to this information.
For the analysis, records of adherence were dichotomized so
that any record of non-attendance at a scheduled lab
appointment was coded 1, and the absence of such a record
was coded 0. Self-reported adherence to warfarin dosing
regimen was assessed using two questions developed for this
study: ‘‘I sometimes take a little extra warfarin when I feel it’s
too low’’, and ‘‘I sometimes skip a dose of warfarin because I
feel better without it’’. Self-reported adherence was also scored
dichotomously; patients who responded Strongly Agree, Agree,
or Neutral to either of the warfarin adherence question were
counted as non-adherent; patients who responded Disagree or
Strongly Disagree were counted as adherent.
2.3.2. Intervention videos
Three videos were produced for this study; patients were
randomly assigned to one of the video conditions, or to usual
care (control). Randomization was done prior to the first
mailing announcing the study, but study procedures were
148 K.M. Mazor et al. / Patient Education and Counseling 69 (2007) 145–157

Table 1
Items used to assess warfarin-related beliefs
Warfarin is beneficial (alpha = .83)
Higher score represents greater belief in the benefits of warfarin
My health, at present, depends on warfarina (R)
My life would be impossible without warfarina (R)
Without warfarin I would be very illa (R)
My health in the future will depend on warfarina (R)
Taking Warfarin is Worrisome (alpha = .75)
Lower score represents greater worry
I worry about my health more since I started taking warfarin
I sometimes worry a bit about the side effects of taking warfarin
Having to take warfarin worries me a
I sometimes worry about the long-term effects of warfarin a
Taking Warfarin is Confusing or Difficult (alpha = .70)
Higher score represents less confusion or difficulty
Sometimes I get confused about how much warfarin I should take
Some days I can’t remember whether I’ve already taken my warfarin
Taking my warfarin exactly as prescribed is easy (R)
Taking my warfarin exactly as prescribed is confusing
Taking my warfarin exactly as prescribed is inconvenient
Lab Testing is Important (alpha = .72)
Higher score represents greater belief that testing is important.
Missing a lab appointment once in a while is not a big deal
Often things come up that are more important to me than getting my blood
tested
If I miss lab appointments I am taking a chance with my health (R)
I would never miss an appointment to get my blood tested except in an
emergency (R)
Responses were coded as follows: strongly agree, 1; agree, 2; Uncertain or
neutral, 3; disagree, 4; strongly disagree, 5; except where indicated by (R)
where coding was reverse (e.g., strongly disagree, 1, etc.).
a
Item was drawn from the Beliefs about Medication scale [27].
identical for all groups until the videos were mailed. Each of the
videos depicted a physician talking with a patient about
anticoagulant medication management. In each video the
physician covered key information on anticoagulant medica-
tion management (Table 2). These were the key points that the
Table 2
Information on warfarin management presented via video
 Warfarin dose must be within a specific range measured by blood test of INR
levels
 Underdosing can result in loss of therapeutic effectiveness and can increase
risk of dangerous clotting
 Overdosing can result harms such as serious or even fatal bleeding
 To stay within the optimal therapeutic range patients must take medications
exactly as prescribed and have blood work performed as often as
recommended
 Taking certain other medications can cause INR levels to go out of range
 Dietary changes can cause INR levels to go out of range
 Patients may not be aware whether INR levels are within range; INR testing
is therefore critical
 Symptoms such as bruising could indicate serious problems and should be
reported immediately
clinical members of the research team judged most important
for patients to be aware of. A draft script was created that
covered these main points, without referring to specific patient
anecdotes or statistical evidence. The script was then reviewed
to identify points where the patient anecdotes or statistical
evidence might be inserted and not interrupt the flow of the
dialogue. Our goal was to keep the encounter as realistic
sounding as possible. The two versions (narrative evidence and
statistical evidence) were developed simultaneously to ensure
that they were as comparable as possible. In the narrative
evidence video, the physician highlighted key points by citing
anecdotes about other patients’ experiences. In the statistical
evidence version, the physician covered the same key
information on warfarin management, but referred to scientific
evidence drawn from the medical literature, using lay language.
Wherever evidence was inserted in one version, evidence was
also inserted in the other version, in an effort to keep the
dialogues as comparable as possible in terms of what was
emphasized. In the combined version, the physician cited both
narrative and statistical evidence. With the exception of the
patient anecdotes (narrative evidence) or statistical evidence,

Table 3
Sample dialogue extracts illustrating differences between videos
Video version Sample dialogue
Narrative Evidence . . . watch for any unexplained bruising. Bleeding or bruising can indicate that your INR is too high. Just a few days ago a patient
called because he noticed a bruise and didn’t know how he got it. He said he was embarrassed because he was pretty sure it
was nothing. We had him come in and the bruise was the size of a baseball. It turned out that his INR was very high, over 6,
but luckily we were able to deal with that before he developed more serious bleeding problems. He really dodged a bullet.
So, don’t ever feel embarrassed or like you’re bothering us if you have a question about a bruise or signs of bleeding.
We’d rather be safe than sorry
Statistical Evidence . . . watch for any unexplained bruising. Bleeding or bruising can indicate that your INR is too high. Studies have shown that 9 out
of every 100 patients on warfarin who have an INR over 6 need to seek medical care for bleeding. Half of these patients suffer
very serious bleeding- more serious than just bruising. So, don’t ever feel embarrassed or like you’re bothering us if you have a
question about a bruise or signs of bleeding. We’d rather be safe than sorry
Combined . . . watch for any unexplained bruising. Bleeding or bruising can indicate that your INR is too high. Just a few days ago a patient
called because he noticed a bruise and didn’t know how he got it. He said he was embarrassed because he was pretty sure it
was nothing. We had him come in and the bruise was the size of a baseball. It turned out that his INR was very high, over 6,
but luckily we were able to deal with that before he developed more serious bleeding problems. He really dodged a bullet.
Studies have shown that 9 out of every 100 patients on warfarin who have an INR over 6 need to seek medical care for
bleeding. Half of these patients suffer very serious bleeding- more serious than just bruising. So, don’t ever feel embarrassed
or like you’re bothering us if you have a question about a bruise or signs of bleeding. We’d rather be safe than sorry
Note: Sample dialogue is from the physician portion of the script. Bold text in dialogue highlights anecdote or evidence. Complete scripts are available from the first
author.
 K.M. Mazor et al. / Patient Education and Counseling 69 (2007) 145–157 149

the dialogue was exactly the same across videos. The same
actors were used in all three versions, and where possible, the
same footage was used. Draft versions of videos were pilot
tested with four anticoagulation clinic patients. Final versions
of scripts were reviewed by the two physician authors for
clinical accuracy. Sample extracts illustrating the differences
between the three versions are presented in Table 3. The full
script used in the combined evidence condition, with the
narrative evidence and statistical evidence portions marked, is
provided in Appendix A.
2.4. Study procedures
Recruitment and flow are summarized in Fig. 1. As noted
above, 600 anticoagulation clinic patients were identified after
screening 680 randomly selected patient names. These 600
patients who met inclusion criteria and who were not excluded
by clinic staff were randomly assigned to one of the 3 video
conditions, or the control condition prior to the first mailing.
The first mailing described the study, and stated that a
questionnaire would be mailed if the patient did not request
otherwise. The second mailing contained the baseline ques-
tionnaire, and a US$ 5 cash incentive. A fact sheet and a form
for the patient to provide written permission for access to
protected patient information were also included. Completion
of the questionnaire was sufficient to participate in all other
aspects of the study. Approximately 3 weeks after a completed
questionnaire was returned to the study office, a video and
second questionnaire was mailed to the patient if he or she had
been assigned to a video intervention condition, with a US$ 10

Fig. 1. Study participant flow.

150 K.M. Mazor et al. / Patient Education and Counseling 69 (2007) 145–157
incentive; patients in the control condition received the second
questionnaire only, with a US$ 5 incentive. The higher
incentive was included with the video because of the extra effort
and time required to watch the video. This paper reports results
for those patients who participated in the intervention phase of
the study, and who completed both the baseline and post-
intervention questionnaires (N = 317). Of these patients, 270
(85%) also provided consent for research staff to access clinic
records to review documentation related to adherence to
laboratory monitoring recommendations.
2.5. Analysis
For the full set of 22 knowledge items, Cronbach’s alpha was
computed to assess score reliability. The knowledge score
reflects the total percent correct responses across the 22 items.
For the items intended to assess warfarin-related beliefs, factor
analysis was used to sort beliefs items into subsets. Items with
factor loadings of .40 or less were discarded; items with
loadings of .40 to .50 were retained if content review and
reliability analysis suggested the items contributed to better
measurement of the underlying construct. The reliability of
each belief subscore was assessed using Cronbach’s alpha. The
beliefs scores reported here and the intent to adhere score are
calculated as the mean of the items included in the scale, and
thus can range from 1 to 5. Within each group of items, items
were recoded as needed so that item coding within a set was
consistent (e.g., reflecting either negative or positive beliefs;
see Table 1).
Chi-square tests were used to compare study groups on age,
gender, education, race, adherence to lab appointments, and
adherence to warfarin dosing (all dichotomously coded).
ANOVA was used to test whether groups differed on baseline
measures of knowledge, beliefs and intent to adhere to lab
testing.
Frequencies for each questionnaire item at baseline and
post-intervention were examined; summary scores for each set
of knowledge or beliefs items were calculated. Regression
analyses were conducted to examine differences in perfor-
mance as a function of intervention. Three comparisons were
made: (1) subjects viewing any video were compared to control
subjects; (2) subjects viewing either of the two videos
containing narrative evidence (narrative evidence only or
narrative evidence plus statistical evidence) were compared to
subjects viewing the statistical evidence only video; (3)
subjects viewing the narrative evidence only video were
compared to subjects viewing the narrative evidence plus
statistical evidence video. For each comparison, regression
analyses were used, with change scores (i.e., the difference in
scores from baseline to post-intervention) as the dependent
variable. Analyses were conducted both with and without the
corresponding baseline score included in the model as a
covariate (e.g., the knowledge score at baseline was included
when evaluating knowledge score differences at post-test; the
analysis was then repeated without this covariate). All analyses
included adjustments for age (age less than 65/age 65 or older),
education (high school education or less/at least some college)
and gender. Tests of significance were two-tailed. A criterion of
p < .05 was used to determine statistical significance.
3. Results
Characteristics of the 317 patients participating in this study
are described in Table 4. Patients randomized to video
intervention versus control did not differ significantly with
respect to the aforementioned variables. Comparisons of scores
on the baseline knowledge, beliefs and intent to adhere to lab
testing also revealed no differences between groups.
3.1. Score reliability
Cronbach’s alpha for the 22 item knowledge test was .79; for
the 4 item measure of beliefs about medication as beneficial
alpha was .83, for the 4 item measure of beliefs that warfarin is
worrisome alpha was .75; for the 5 item measure of the belief
that taking warfarin is confusing/difficult alpha was .70, and for
the belief that lab monitoring is important alpha was .72.
3.2. Baseline status
Baseline scores for each subset of questionnaire items and
the clinic data are summarized in Table 5. Frequencies for
critical knowledge items were examined: 5% of respondents
did not know they should call the anticoagulation clinic if they
had a cut that would not stop bleeding; 13% did not know to call
if they noticed any unexplained bruising; 13% did not know to
call if they noticed blood in their urine or stool. Considering
these three items together, 23% did not know to call under at
least one of the three conditions. Most respondents knew that
too little warfarin in their blood could cause a stroke (83%), but
only 21% knew that too much warfarin could cause a stroke.
More than 1 in 10 respondents (13%) did not know that too
much warfarin could cause bruising, and more than 1 in 3 (36%)
did not know that too much warfarin could cause stomach
bleeding. With regard to the items intended to assess beliefs
about anticoagulation therapy, 13% of respondents believed
‘‘missing a lab appointment once in a while is not a big deal’’
and 12% believed they were able to tell if their warfarin was at
the right level by how they felt.
3.3. Impact of the narrative evidence, statistical evidence
and combined evidence patient education messages
Unadjusted means and proportions by intervention condition
are presented in Table 5. Comparing the results for those who
watched any video to the control group (with the appropriate
baseline measure included in the analysis) revealed that those
who watched any video showed greater gains on the knowledge
test than those in the control condition ( p < .001), showed
greater positive shifts in their beliefs in the importance of
laboratory testing ( p = .010), and in their beliefs that taking
warfarin is beneficial ( p = .012). There was also some
indication that patients in the control condition were less
likely to exhibit improved lab attendance during the second
 K.M. Mazor et al. / Patient Education and Counseling 69 (2007) 145–157 151

Table 4
Characteristics of participants
Characteristic Intervention arm
Narrative evidence Statistical Combined Control (N = 90)
(N = 77) evidence (N = 82) (N = 77)
N % N % N % N %
Gender
Men 55 67.1 42 61.8 44 57.1 55 61.1
Women 27 32.9 25 36.8 33 42.9 34 37.8
Missing 0 0.0 1 1.5 0 0.0 1 1.1
Race/ethnicity
American Indian or Alaska Native 0 0.0 0 0.0 0 0.0 0 0.0
Asian 0 0.0 0 0.0 1 1.3 0 0.0
Black or African American 2 2.4 1 1.5 0 0.0 1 1.1
Hispanic or Latino 0 0.0 0 0.0 2 2.6 2 2.2
White 78 95.1 65 95.6 71 92.2 80 88.9
Other 1 1.2 0 0.0 2 2.6 3 3.3
Multiple 1 1.2 1 1.5 1 1.3 2 2.2
Missing 0 0.0 1 1.5 0 0.0 2 2.2
Education
8th grade or less 4 4.9 4 5.9 4 5.2 1 1.1
Some high school 8 9.8 5 7.4 5 6.5 11 12.2
High school graduate or GED 18 22.0 13 19.1 20 26.0 24 26.7
Some college or 2 year degree 22 26.8 10 14.7 16 20.8 28 31.1
4 year college graduate 9 11.0 17 25.0 12 15.6 8 8.9
More than 4 year college degree 20 24.4 18 26.5 19 24.7 17 18.9
Missing 1 1.2 1 1.5 1 1.3 1 1.1
Age
44 years 0 0.0 8 11.8 3 3.9 6 6.7
45–54 years 14 17.1 9 13.2 5 6.5 11 12.2
55–64 years 19 23.2 14 20.6 21 27.3 21 23.3
65–74 years 13 15.9 11 16.2 18 23.4 23 25.6
75 years or older 36 43.9 25 36.8 30 39.0 27 30.0
Missing 0 0.0 1 1.5 0 0.0 2 2.2
Self-rating of overall health
Excellent 4 4.9 5 7.4 5 6.5 4 4.4
Very good 17 20.7 18 26.5 11 14.3 17 18.9
Good 37 45.1 24 35.3 42 54.5 38 42.2
Fair 20 24.4 14 20.6 16 20.8 26 28.9
Poor 3 3.7 5 7.4 3 3.9 3 3.3
Missing 1 1.2 2 2.9 0 0.0 2 2.2
Usually take warfarin. . .
Exactly as prescribed. 79 96.3 62 91.2 71 92.2 84 93.3
Mostly as prescribed, but sometimes 1 1.2 2 2.9 4 5.2 4 4.4
get confused or make mistakes.
Mostly as prescribed, but sometimes 2 2.4 3 4.4 2 2.6 1 1.1
adjust how much or how often.
Missing 0 0.0 1 1.5 0 0.0 1 1.1
GED, General Educational Development Test.

measurement period (according to clinic records) than patients
who watched a video, but the difference did not achieve
statistical significance ( p = .07). These results were consistent
whether or not the corresponding baseline measure covariate
was included as a covariate ( p values shifted by .01 or less).
Comparing results for patients who viewed either video
containing narrative evidence (either the narrative only version
or narrative plus statistical evidence version) revealed that
patients viewing a version with narrative evidence showed
stronger belief that lab testing is important compared to patients
viewing the statistical evidence only version ( p = .05). When
the baseline knowledge score was included as a covariate, the
results suggested that patients viewing a version with narrative
evidence showed greater knowledge gains ( p = .006), but this
difference did not achieve statistical significance if the baseline
score was not included as a covariate ( p = .15). Comparing
results for patients who viewed the narrative evidence only
video to the narrative plus statistical evidence video, no
statistically significant differences were found ( p values ranged
from .17 to .86).
152 K.M. Mazor et al. / Patient Education and Counseling 69 (2007) 145–157

Table 5
Baseline and post-intervention scores and rates by intervention group
Measure Time, mean (S.D.)
Baseline Post-test
Narrative Statistical Combined Control Narrative Statistical Combined Control
evidence evidence evidence evidence
Warfarin-related Knowledge; 60 (17) 57 (14) 56 (17) 54 (18) 69 (17) 70 (15) 68 (16) 57 (17)
Higher scores represent
greater knowledge
Belief that Labs are Important; 4.15 (.74) 4.24 (.65) 4.10 (.81) 4.06 (.71) 4.25 (.69) 4.29 (.64) 4.25 (.71) 4.03 (.70)
Higher scores represent stronger
belief in importance of labs
Belief that Warfarin is Beneficial; 3.39 (.75) 3.51 (.81) 3.48 (.80) 3.47 (.75) 3.51 (.83) 3.75 (.86) 3.60 (.73) 3.44 (.74)
Higher scores represent stronger
belief in the benefits of warfarin
Warfarin is Worrisome; 3.10 (.89) 3.08 (1.03) 3.16 (.93) 3.21 (.89) 2.99 (.96) 2.93 (1.06) 2.97 (.94) 3.13 (.85)
Higher scores represent less worry
Warfarin Regimen as Confusing 4.30 (.57) 4.37 (.56) 4.17 (.53) 4.21 (.63) 4.26 (.62) 4.30 (.59) 4.25 (.53) 4.19 (.65)
or Difficult; Higher scores represent
less confusion or difficulty
Intent to Adhere to Recommended 3.63 (1.14) 3.61 (1.2) 3.85 (1.00) 3.70 (1.01) 3.95 (1.00) 3.88 (1.02) 3.96 (.96) 3.81 (.96)
Lab Testinga; Higher scores
represent stronger intent to adhere
Non-adherence to Warfarin Dosing; 6.1 (5) 6.4 (4) 8.0 (6) 5.8 (5) 3.7 (3) 3.2 (2) 4.0 (3) 3.5 (3)
% self-reporting non-adherence
(% (N))
Non-attendance at lab appointments; 26.6 (17) 19.2 (10) 13.6 (8) 17.5 (10) 26.6 (17) 19.2 (10) 18.6 (11) 33.3 (19)
% with 1 missed appointment
documenteda (% (N))
Sample sizes varied by analysis due to missing responses on some questionnaire items and omission of self-testers from some analyses. In addition, not all respondents
granted permission for the research team to review clinical records. Sample sizes for the Narrative Evidence, Statistical Evidence, Combined and Control conditions
respectively, for each analysis, are as follows: knowledge Ns = 81, 65, 74, 87; labs important Ns = 82, 64, 75, 89; warfarin beneficial Ns = 81, 65, 75, 89; warfarin is
worrisome Ns = 82, 62, 75, 89; warfarin as confusing Ns = 78, 64, 73, 85; intent to adhere to lab testing Ns = 76, 64, 73, 85; non-adherence to dosing Ns = 82, 63, 75,
86; non-attendance at lab appointments Ns = 64, 52, 59, 57.
a
Indicates analyses where patients who reported self-testing at home (N = 8) were omitted.

4. Discussion and conclusion
4.1. Discussion
In this study, we found some indication that the use of
narrative evidence (i.e., patient anecdotes) may be more
effective in communicating with patients about warfarin at least
in some areas. While we had predicted that the narrative
evidence presentation would be more effective in all areas, the
absence of a consistent effect may not be surprising given the
mixed results reported in prior studies [21–27].
These findings are promising, but clearly far from definitive.
Clearly, there is need for future empirical work that system-
atically investigates the factors that influence whether or under
what conditions narrative evidence has an impact, what that
impact is, and what the relevant interactions are. We concur with
Hinyard and Kreuter, who close their excellent overview of
narrative communication as a behavior change tool by
recommending organizing future research according to the
basic components of communication: source, message, channel
and receiver [18]. In fact, an expanded version of this
communication model, the Communication-Human Information
Processing (C-HIP) model, has been proposed and productively
applied in an extensive body of research around communicating
warning information [43–45]. This model expands the simple
communication model described above to include substages of
information processing that occur within the receiver. These
substages include attention capture and maintenance, compre-
hension, memory, attitudes and beliefs, and motivation; the final
outcome of the process is behavior. Processing can be impeded or
enhanced at any stage or substage, and multiple feedback loops
exist. Applying this model to the narrative messages, one might
postulate how manipulations of the various characteristics of
narrative messages (e.g., vividness, event specificity, etc.) would
be expected to influence how the message is processed at each of
these stages, and ultimately how these manipulations might
influence behavior. For instance, this model suggests that
attention capture and maintenance are critical early stages in
processing; it would be possible to design an experiment that
specifically investigated whether narrative or statistical evidence
differentially influenced attention, which might be measured by
gaze, eye-contact, self-report, or other measures. The specific
narratives tested could also be systematically varied to determine
whether the effects (if found) were due to the characteristics that
have been suggested in the literature, such as vividness, realism,
or event specificity. Comparable experiments could be carried
out to investigate differences in processing at each substage, as
well as the ultimate stage, behavior change.
 K.M. Mazor et al. / Patient Education and Counseling 69 (2007) 145–157
An important secondary finding of this study is that videos
portraying a physician–patient encounter about oral antic-
oagulant medication management and the importance of
monitoring did have a positive impact on patient knowledge.
The impact of the videos on patients’ beliefs was mixed;
viewing a video was associated with a stronger belief in the
importance of adhering to recommendations for lab monitor-
ing, and may be associated with stronger belief in the benefits of
warfarin. There was no evidence of an effect of the video on
patients’ tendency to worry about warfarin, or on perceptions of
taking warfarin as confusing or difficult.
While not a primary focus of the present study, several
findings from the baseline questionnaire are worth noting
because of their clinical importance. First, many patients
demonstrated potentially dangerous beliefs and gaps in their
warfarin-related knowledge. Second, non-adherence is a
significant concern. Many patients missed at least one lab
appointment over the 6 months period study here. Fewer
reported being non-adherent to dosing regimens, but ideally no
patients would self-adjust dosing of anticoagulants. Non-
adherence to either dosing or lab monitoring recommendations
is a risky behavior that has the potential to lead to life-
threatening adverse events. Further, it is important to note that
the patients in this study were receiving services from an
anticoagulation clinic, where they received education about
anticoagulant medication upon enrollment, and during regular
interactions with clinic staff. Thus, these patients might be
expected to be more knowledgeable than patients receiving care
in primary care practices.
The videos used in this study did not eliminate patients’
knowledge gaps, incorrect beliefs or non-adherent behaviors.
Nonetheless, the finding that the videos had some positive impact
is encouraging and suggests that further research in this area is
warranted. Videos provide a relatively efficient way of conveying
a large amount of complex information to patients, and may be a
cost-effective way to supplement clinicians’ efforts to educate
patients about anticoagulant medications. In addition, videos
also have the advantage of being able to repeat the patient
education message; some participants in this study commented
that they watched the video more than once, or kept it longer to
share with a family member. The value of providing written take
home instructions for patients is widely acknowledged; for
patients who have difficulty with reading, a video to take home
may be even more effective. For patients who do not have access
to a VHS player, alternatives such as DVDs, web-delivery, or
opportunities to view the video in the clinic might be considered.
Our findings suggest that all patients may benefit from
routine ‘‘refresher’’ education about warfarin periodically, and
that clinicians should not assume that patients who have been
on anticoagulants for months or years have adequate knowledge
of how to manage their medication safely.
This study has limitations. First, not all patients invited to
participate completed the study. Approximately 53% of those
receiving an invitation completed the pre-questionnaire, viewed
a video (if so assigned) and completed a post-questionnaire. A
total of 193 patients (32% of the invited sample) either explicitly
opted out or did not respond to at least one of the mailings. Other
153
reasons for loss were determination of ineligibility after the first
mailing (N = 24); lack of access to a VCR (N = 44); and mail
returned as undeliverable (N = 22). Because the questionnaires
and videos were distributed via mail, we did not supervise
patients as they completed the questionnaires or watched the
videos. This allowed us to recruit a larger sample, and to include
people who would have difficulty coming to a study office (e.g.,
people with mobility restrictions, or limited access to
transportation). However, it also means that patients could have
discussed the questions with others at home. Also important to
note are potential limitations on the generalizability of the
findings. The results reported here are likely to be generalizable
to comparable populations, specifically, to warfarin patients
managed by an anticoagulation clinic, but it is not known whether
these finding are generalizable to patients who receive care in
other settings. Similarly, most participants were majority group
members, drawn from a single geographic area; again, general-
izability of these findings to minority group members, or to other
geographic regions is not known. Finally, this study did not test
the long-term impact of the communication strategies evaluated
here, since the post-questionnaire was completed immediately
after viewing.
4.2. Conclusion
This study provides some support for the hypothesis that
narratives are more effective in communicating information
about medication safety and usage than statistical evidence
alone. These findings also demonstrate that videos offer an
efficient method of providing patient education, and patients on
long-term medications may benefit from periodic educational
‘‘refreshers’’ about appropriate medication use.
4.3. Practice implications
Clinicians should be aware that many patients receiving oral
anticoagulants do have significant gaps in their warfarin-related
knowledge, and may hold potentially dangerous beliefs even
after receiving initial education about warfarin and frequent
monitoring through an anticoagulation clinic. Therefore it is
advisable to assess patients’ knowledge, beliefs and practices
with respect to anticoagulation management and monitoring on
an ongoing basis, to provide frequent reinforcement of key
information, and to offer ‘‘refresher’’ educational materials
periodically. Incorporating patient anecdotes into physician–
patient dialogues or educational materials may increase the
effectiveness of the message.
Acknowledgements
This study was supported by the Meyers Primary Care
Institute, a joint endeavor of the University of Massachusetts
Medical School, Fallon Clinic Foundation, and Fallon
Community Health Plan, and the Rosalie Wolf Interdisciplinary
Geriatric Health Care Research Center at the University of
Massachusetts, which is funded through a joint program of
RAND Health and the John A. Hartford Foundation.
154 K.M. Mazor et al. / Patient Education and Counseling 69 (2007) 145–157

Appendix A
K.M. Mazor et al. / Patient Education and Counseling 69 (2007) 145–157 155
156 K.M. Mazor et al. / Patient Education and Counseling 69 (2007) 145–157
References
[1] Horton JD, Bushwick BM. Warfarin therapy: evolving strategies in
anticoagulation. Am Fam Physician 1999;59:635–46.
[2] Harris Interactive. The Use of Anticoagulant Medications: Key Findings
From Surveys of Patients, Caregivers and Physicians; 2005.
[3] Wong W, Wilson Norton J, Wittkowsky AK. Influence of warfarin regimen
type on clinical and monitoring outcomes in stable patients in an antic-
oagulation management services. Pharmacotherapy 1999;19:1385–91.
[4] Budnitz DS, Pollock DA, Mendelsohn AB, Weidenbach KN, McDonald
AK, Annest JL. Emergency department visits for outpatient adverse drug
events: demonstration for a national surveillance system. Ann Emerg Med
2005;45:197–206.
[5] Gurwitz JH, Field TS, Harrold LR, Rothschild J, Devellis K, Seger AC,
Cadoret C, Fish LS, Garber L, Kelleher M, Bates DW. Incidence and
preventability of adverse drug events among older persons in the ambu-
latory setting. J Amer Med Assoc 2003;289:1107–16.
[6] Field TS, Mazor KM. Briesacher B, DeBellis KR. Gurwitz JH Adverse
drug events resulting from patient errors among older adults J Am
Geriatrics Soc 2007;55:271–6.
[7] Fang MC, Machtinger EL, Wang F, Schillinger D. Health literacy and
anticoagulation-related outcomes among patients taking warfarin. J Gen
Int Med 2006;21:841–6.
[8] Newall F, Monagle P, Johnston L. Patient understanding of warfarin
therapy: a review of education strategies. Hematology 2005;10:437–42.
[9] Sackett DL, Straus SE, Richardson WS, Rosenberg W, Haynes RB. Evi-
dence-based medicine, 2nd ed., Edinburgh: Churchill Livingstone; 2000.
[10] Schwartz LM, Woloshin S, Black WC, Welch HG. The role of numeracy
in understanding the benefit of screening mammography. Ann Intern Med
1997;127:966–72.
[11] Lipkus IM, Samsa G, Rimer BK. General performance on a numeracy
scale among highly educated samples. Med Decis Making 2001;21:37–44.
[12] Estrada CA, Martin-Hryniewicz M, Peek BT, Collins C, Byrd JC. Literacy
and numeracy skills and anticoagulation control. Am J Med Sci
2004;328:88–93.
[13] Rothman AJ, Kiviniemi MT. Treating people with information: an ana-
lysis and review of approaches to communicating health risk information.
J Natl Cancer Inst Monogr 1999;25:44–51.
[14] Brosius H-B, Bathelt A. The utility of exemplars in persuasive commu-
nications. Commun Res 1994;21:48–78.
[15] Wachters-Kaufmann CS. Personal accounts of stroke experiences. Patient
Educ Couns 2000;41:295–303.
[16] Swift TL, Dieppe PA. Using expert patients’ narratives as an educational
resource. Patient Educ Couns 2005;57:115–21.
[17] Fagerlin A, Wang C, Ubel PA. Reducing the influence of anecdotal
reasoning on people’s health care decisions: is a picture worth a thousand
statistics? Med Decis Making 2005;25:398–405.
[18] Hinyard LJ, Kreuter MW. Using narrative communication as a tool for
health behavior change: a conceptual, theoretical, and empirical overview.
Health Educ Behav 2007;34:777–92.
[19] Ornstein R, Herron J, Johnstone J, Swencionis C. Differential right
hemisphere involvement in two reading tasks. Psychophysiology
1979;16:398–401.
[20] Kreuter MW, Green MC, Cappella JN, Slater MD, Wise ME, Storey D,
Clark EM, O’Keefe DJ, Erwin DO, Holmes K, Hinyard LJ, Houston T,
Woolley S. Narrative communication in cancer prevention and control: a
framework to guide research and application. Ann Behav Med
2007;33:221–35.
[21] Reinard JC. The empirical study of the persuasive effects of evidence: The
status after fifty years of research. Hum Commun Res 1988;15:3–59.
 K.M. Mazor et al. / Patient Education and Counseling 69 (2007) 145–157
[22] Allen M, Preiss R. Comparing the persuasiveness of narrative and
statistical evidence using meta-analysis. Commun Res Rep 1997;14:
125–31.
[23] Baesler E, Burgoon J. The temporal effects of story and statistical
evidence on belief change. Commun Res 1994;21:582–602.
[24] Golding D, Krimsky S, Plough A. Evaluating risk communication:
narrative vs. technical presentations of information about radon. Risk
Anal 1992;12:27–35.
[25] Greene K, Brinn LS. Messages influencing college women’s tanning
bed use: statistical versus narrative evidence format and a self-assessment
to increase perceived susceptibility. J Health Commun 2003;8:
443–61.
[26] Kopfman J, Smith S, Ah Yun J, Hodges A. Affective and cognitive
reactions to narrative versus statistical evidence organ donation messages.
J Appl Commun Res 1998;26:279–300.
[27] Cox D, Cox AD. Communicating the consequences of early detection: the
role of evidence and framing. J Market 2001;65:91–103.
[28] Pfau M, Holbert RL, Zubric SJ, Pasha NH, Lin W-K. Role and influence of
communication modality in the process of resistance to persuasion. Media
Psychol 2000;2:1–33.
[29] Ottati V, Rhoads S, Graesser AC. The effect of metaphor in processing
style in a persuasion task: A motivational resonance model. J Personality
Soc Psychol 1999;77:688–97.
[30] Holtgraves T, Lasky B. Linguistic power and persuasion. J Lang Social
Psychol 1999;18:196–205.
[31] Hosman LA, Huebner TM, Siltanen SA. The impact of power-of-speech
style, arguments, and need for cognition on impression formation, cog-
nitive responses and persuasion. J Lang Social Psychol 2002;21:361–79.
[32] Ratneshwar S, Chaiken S. Comprehension’s role in persuasion: the case of
its moderating effect on the persuasive impact of source cues. J Consumer
Res 1991;18:52–62.
[33] Tal-Or N, Boninger DS, Poran A, Cleicher F. Counterfactual thinking as a
mechanism in narrative persuasion. Hum Commun Res 2004;30:301–28.
157
[34] Peters E, Hibbard J, Slovic P, Dieckmann N. Numeracy skill and the
communication, comprehension, and use of risk-benefit information.
Health Aff (Millwood) 2007;26:741–8.
[35] Ubel PA, Jepson C, Baron J. The inclusion of patient testimonials in
decision aids: effects on treatment choices. Med Decis Making
2001;21:60–8.
[36] Denberg TD, Melhado TV, Steiner JF. Patient treatment preferences in
localized prostate carcinoma: The influence of emotion, misconception,
and anecdote. Cancer 2006;107:620–30.
[37] Ajzen I. Intuitive theories of events and the effects of base-rate informa-
tion on prediction. J Personal Soc Psychol 1977;35:303–14.
[38] Tang EO, Lai CS, Lee KK, Wong RS, Cheng G, Chan TY. Relationship
between patients’ warfarin knowledge and anticoagulation control. Ann
Pharmacother 2003;37:34–9.
[39] Svarstad BL, Chewning BA, Sleath BL, Claesson C. The Brief Medication
Questionnaire: a tool for screening patient adherence and barriers to
adherence. Patient Educ Couns 1999;37:113–24.
[40] Taylor FC, Ramsay ME, Tan G, Gabbay J, Cohen H. Evaluation of
patients’ knowledge about anticoagulant treatment. Qual Health Care
1994;3:79–85.
[41] Horne R, Weinman J, Hankins M. The Beliefs about Medicines Ques-
tionnaire (BMQ). The development and evaluation of a new method for
assessing the cognitive representation of medication. Psychol Health
1999;14:1–24.
[42] Arnsten JH, Gelfand JM, Singer DE. Determinants of compliance with
anticoagulation: A case-control study. Am J Med 1997;103:11–7.
[43] Wogalter MS, DeJoy DM, Laugherty KR. Warnings and risk commu-
nication. London, UK: Taylor & Francis; 1999.
[44] Conzola VC, Wogalter MS. A communication-human information pro-
cessing (C-HIP) approach to warning effectiveness in the workplace. J
Risk Res 2001;4:309–22.
[45] Wogalter MS, editor. Handbook of warnings. New Jersey: Lawrence
Erlbaum Associates, Inc.; 2006.