External Support for Collaborative Problem Solving in a Simulated Provider/Patient Medication Scheduling Task

Journal of Experimental Psychology: Applied
Copyright 2008 by the American Psychological Association
2008, Vol. 14, No. 3, 288 –297
1076-898X/08/$12.00
DOI: 10.1037/a0012809
External Support for Collaborative Problem Solving in a Simulated
Provider/Patient Medication Scheduling Task
Daniel Morrow, Liza Raquel, Angela Schriver, Seth Redenbo, David Rozovski, and Gillian Weiss
University of Illinois at Urbana-Champaign
Taking medication requires developing plans to accomplish the activity. This planning challenges older
adults because of age-related cognitive limits and inadequate collaboration with health providers. The
authors investigated whether an external aid (medtable) supports collaborative planning in the context of
a simulated patient/provider task in which pairs of older adults worked together to create medication
schedules. Experiment 1 compared pairs who used the medtable, blank paper (unstructured aid), or no aid
to create schedules varying in complexity of medication constraints (number of medications and
medication co-occurrence restrictions) and patient constraints (available times during the day to take
medication). Both aids increased problem-solving accuracy and efficiency (time per unit accuracy)
compared to the no-aid condition, primarily for more complex schedules. However, benefits were similar
for the two aids. In Experiment 2, a redesigned medtable increased problem-solving accuracy and
efficiency compared to blank paper. Both aids presumably supported problem solving by providing a
jointly visible workspace for developing schedules. The medtable may be more effective because it
externalizes constraints (relationships between medication and patient information), so that participants
can more easily organize information.
Keywords: distributed cognition, external aids, health communication, problem solving, medication
adherence
Older adults’ self-care is a crucial health care issue because our
adverse drug events, often reflecting medication misuse (Budnitz,
society is aging at a time of increasing patient responsibility for
Pollock, Weidenbach, Mendelsohn, Schroeder, & Annest, 2006).
self-care. Self-care activities such as taking medication challenge
Among other factors, nonadherence has been linked to age-related
older adults, leading to errors that threaten safety and undermine
differences in cognitive function and health literacy combined with
treatment efficacy. A recent study found that almost 25% of older
inadequate collaboration between patients and providers, leading
adults took 5 or more and 12% took 10 or more prescribed
to calls to improve self-care by supporting patient/provider collab-
medications (Kaufman, Kelly, Rosenberg, Anderson, & Mitchell,
oration and planning (for review see Aspden, Wolcott, Bootman,
2002). On average, 50% of older adults do not take their medica-
& Croenwett, 2007).
tions as prescribed (Haynes, McKibbon, & Kanani, 1996). More-
over, over one third of older adults’ hospital admissions are due to
Cognitive and Collaborative Determinants
of Medication Planning
Daniel Morrow, Institute of Aviation and Beckman Institute of Ad-
Creating and implementing self-care plans can tax cognitive
vanced Science and Technology, University of Illinois at Urbana-
abilities, such as working memory and processing speed. Devel-
Champaign; Liza Raquel, Beckman Institute of Advanced Technology,
oping medication plans requires understanding and integrating
University of Illinois at Urbana-Champaign; Angela Schriver, Institute of
information (e.g., when to take medication; warnings to keep in
Aviation and Beckman Institute of Advanced Technology, University of
mind), and implementing these plans requires prospective memory
Illinois at Urbana-Champaign; Seth Redenbo, Institute of Aviation, Uni-
(Park & Jones, 1997). Indeed, performance on measures of cog-
versity of Illinois at Urbana-Champaign; David Rozovski, Institute of
nitive ability predict errors in taking medication (Insel, Morrow,
Aviation, University of Illinois at Urbana-Champaign; and Gillian Weiss,
Beckman Institute of Advanced Technology, University of Illinois at
Brewer, & Figuerosa, 2006) and loading pill organizers (Carlson,
Urbana-Champaign.
Xue, Fried, Tekwe, & Brandt, 2005) among older adults.
This material is based upon work supported by the National Institute of
Ideally, health care providers mitigate these cognitive demands
Aging and the UIUC Center for Healthy Minds. Any opinions, findings,
by collaborating with patients, as suggested by patient-centered
and conclusions or recommendations expressed in this publication are
approaches to communication (Bodenheimer, Lorig, Holman, &
those of the authors and do not necessarily reflect the views of the NIH.
Graumbach, 2002) and conceptions of adherence that focus on
Partial results (the problem-solving accuracy and efficiency analyses in
patient/provider concordance about adherence goals (Vermeire,
Experiment 2) were presented at the 51st Annual Meeting of the Human
Hearnshaw, Van Royen, & Denekens, 2001). Unfortunately, pa-
Factors Society (Morrow, Raquel, Schriver, Redenbo, & Rozovski, 2007).
tient/provider collaboration can be inadequate because of system
Correspondence concerning this article should be addressed to Daniel
barriers, such as limited patient contact time and inadequate com-
Morrow, Beckman Institute of Advanced Science and Technology and
Institute of Aviation University of Illinois at Urbana-Champaign, 405
munication training (Aspden et al., 2007). Providers omit infor-
North Mathews Avenue, Urbana, IL 61801. E-mail: dgm@uiuc.edu
mation, present dense, disorganized information, and rarely check
288

---

EXTERNAL SUPPORT FOR COLLABORATION
289
patients’ comprehension (Ley, 1997; Schillinger et al., 2003). As a
medtable (see Figure 1) on collaboration involved in creating
result, patients leave consultations without a clear plan and either
multimedication schedules typical of chronically ill older adults.
end up calling back to clarify confusion or taking medication
This aid may support collaborative processes, such as sharing
incorrectly.
information about medications (e.g., how much and often to take,
drug interactions) and patients (daily routine, constraints on when
Improving Patient/Provider Collaborative Planning
patients can take medication) and integrating this information into
a schedule.
We investigated the role of external aids as environmental
The medtable supports these processes in several ways. First, it
support for collaborative problem solving relevant to self-care.
reduces the cognitive demands of understanding and integrating
According to distributed cognition theories, complex task perfor-
problem-relevant information by externalizing this information
mance depends on the interaction of internal and external resources
(Larkin & Simon, 1987). As shown in Figure 1, the table’s rows
(Hutchins, 1995). Older adults may reduce reliance on retrieval,
allowed participants to list each medication, and the columns
computation, or other mental operations that become less reliable
corresponded to times to take these medications during the day.
with age by using environmental support (e.g., memory retrieval
Familiar icons and verbal labels indicated daily events, such as
cues), which decreases need for self-initiated mental processing
meals, and the row below the icons allowed participants to write
(Craik & Jennings, 1992; Morrow & Rogers, 2008). While age
specific times corresponding to these events that were consistent
differences occur for instrumental problem solving domains, such
as medication management, as they do for other complex domains,
with their routine (similar to the timeline previously shown to
the role of environmental support for this important everyday
improve memory for medication schedules; Morrow, Hier, Men-
ability is rarely investigated (for review see Thornton & Dumke,
ard, & Leirer, 1998). Columns also indicated intervals before and
2005). Collaboration can sometimes provide environmental sup-
after meals that were consistent with restrictions about taking
port for older adults’ problem-solving because partners provide
medications with or without food. Thus, the medtable should help
retrieval cues or suggest problem-solving strategies (e.g., Gould,
participants develop schedules by explicitly representing relation-
Dixon, & Kurzman, 1994), but collaboration may itself impair
ships between medications and daily event times (Day, 1988). It
problem-solving because of the cognitive demands of joint perfor-
may especially benefit older adults’ problem solving by reducing
mance (Schwartz, 1995). We investigated whether external aids
the need for age-vulnerable working memory processes such as
support collaborative problem solving relevant to medication use.
storing and accessing multiple sources of information (Bopp &
External aids are ubiquitous in health care settings, with pro-
Verhaeghen, 2007). Thus, the aid may provide environmental
viders relying on white boards and other “cognitive artifacts” to
support for specific processes known to decline with age without
support routine tasks such as ordering tests or scheduling patients
itself taxing elders’ cognitive processes (Morrow & Rogers, 2008).
(Nemeth, Cook, O’Connor, & Klock, 2004). Yet their potential for
Second, by providing a shared workspace, the medtable supports
supporting patient/provider collaboration has been under investi-
joint attention to critical information, reducing the need to describe
gated (Aspden et al., 2007). We investigated the impact of the
information (e.g., proposed medication times), acknowledge
Figure 1.
Original medtable aid (Experiment 1).

---

290
MORROW ET AL.
partners’ contributions, or other processes involved in sharing
problem-solving (Hart & Staveland, 1988). Both aids provide
information (Gergle, Kraut, & Fussell, 2004). Finally, it reduces
environmental support in the form of a jointly visible workspace
the “process loss” sometimes associated with collaboration (e.g.,
for sharing information and developing schedules, and therefore
one partner’s contribution interferes with another’s retrieval;
should increase problem-solving accuracy and reduce time, im-
Schwartz, 1995). In short, the medtable may improve problem-
proving efficiency compared to the no-aid condition. The medtable
solving by reducing the cognitive demands of collaboration, which
should be more effective than paper because it externalizes con-
ideally translates into better patient self-care. Similar aids have
straints, allowing participants to easily organize information. Com-
been included in education interventions designed to improve
plexity of medication (number of medications and co-occurrence
patient adherence (e.g., Murray et al., 2007). Kripalani et al. (2007)
restrictions) and patient (flexible/inflexible schedule) constraints
evaluated benefits of a medication schedule card similar to the
was also varied. We expected larger aid-related benefits for more
medtable in a randomized trial and found many patients thought it
complex problems, which would impose greater cognitive de-
was helpful. However, these studies have not directly investigated
mands in the absence of the aid.
benefits of visual aids for patient/provider collaboration.
Experiment 1
Overview of Experiments
Method
The overall goal of our research program is to test whether the
medtable improves the ability of patients and providers to develop
Participants.
Ninety-six community-dwelling older adults
medication schedules that support adherence. As a first step, the
participated (mean age
69, 56 – 84 years; 52% female). They
present study investigated medtable benefits for problem solving
were screened to ensure they were native speakers of English with
in a simulated patient/provider collaborative planning task. Pairs of
no physical or cognitive impairments that could limit participation
older adults, randomly assigned to be provider or patient, worked
(e.g., stroke in the last 3 years). They were randomly assigned to
together to create schedules that satisfied medication and patient
serve as patient or provider, with pairs randomly assigned to the
constraints. The task was intended to simulate some aspects of
three aid conditions. Speed of mental processing was measured by
planning between patients and those providers responsible for
the Letter Comparison and Pattern Comparison tasks (Salthouse,
patient education (e.g., pharmacists and nurses), such as sharing
1991). In these paper-and-pencil tasks, participants decide as rap-
information. However, the simulation likely has limited general-
idly as possible whether pairs of letter sets or line patterns are the
izability to actual clinical situations. For example, “providers”
same or different.
were not clinicians. In addition, older adults served as both patient
Participants in the three groups differed in mean number of
and provider even though providers are often younger than chron-
prescribed medications but not in terms of the other measured
ically ill patients. However, we note that many nurses and phar-
variables (see Table 1). Participants assigned as patient and pro-
macists are middle-aged or older (Buerhaus, Donelan, Ulrich,
vider did not differ on any of the variables. Because partner
Norman, & Dittus, 2006). Older adults served as patients in our
familiarity can influence collaborative performance (Andersson &
study because this age group is most often prescribed complex
Ronnberg, 1995), we controlled for this variable by selecting
medication regimens. Providers were of similar age to the patients
partners who did not know each other before the study.
to avoid potential intergenerational communication problems, such
Materials and design.
Complexity of the medication and the
as younger adults using “elderspeak” (Ryan, Merideth, MacLean,
patient information given to provider and patient participants (re-
& Orange, 1995).
spectively) was varied. In the complex medication condition, in-
Experiment 1 compared problem solving time and accuracy
formation about four medications commonly used by older adults
for pairs who used the medtable, blank paper (unstructured aid
was presented, including purpose, number of pills and times per
to control for general effects of external support), or no aid (talk
day to take them, dose spacing, and special instructions or
only). Because the aids might reduce problem-solving effort
warnings (an example is presented in Appendix 1, adapted from
without increasing accuracy, we also included the NASA-TLX
www.drugs.com). In the simple condition, information about two
instrument to measure the subjective workload associated with
medications was presented, with fewer co-occurrence constraints
Table 1
Mean (and Standard Deviation) Values for Continuous Demographic and Individual Difference
Variables, by Condition (N
32 per Group) in Experiment 1
Variable
Medtable
Paper
No aid
p value
Age (years)
70.1 (5.4)
69.5 (5.8)
68.8 (5.6)
.10
Education (years)
14.3 (2.7)
15.3 (3.9)
15.0 (2.6)
.10
Comparison task score
43.4 (7.7)
44.3 (11.2)
44.0 (9.0)
.10
Self-rated health score
5.0 (1.4)
4.8 (1.5)
5.1 (1.4)
.10
Mean number of medications
4.0 (3.3)
2.4 (1.6)
3.5 (2.2)
.05
Note.
Comparison task is a measure of processing speed (Salthouse, 1991). The mean of the Letter (Maximum
score
42) and Pattern Comparison (Maximum score
30) tasks was used because performance on the two
tasks were correlated (r
.79). Self-rated health is a 7-point scale from 1 (very poor health) to 7 (very good
health).

---

EXTERNAL SUPPORT FOR COLLABORATION
291
on when these medications could be taken. In the complex patient
was created by dividing solution time by accuracy, indicating time
information condition, patients had a strict daily work routine (e.g.,
needed to achieve the same level of accuracy across participants.
afternoon/night shift) and they could only take medications at
Mean solution time (sec), accuracy (percent correct), and effi-
lunch or after work. In the simple condition, there were no daily
ciency (time per unit accuracy) were analyzed by an Aid (med-
restrictions and patients could adjust wake-up times and meal-
table, paper, none)
Medication Complexity
Patient Complex-
times. Complexity of daily routine was investigated because older
ity mixed design ANOVA with the latter two variables repeated
adults’ adherence is influenced by this factor (Park et al., 1999).
measures. A composite mental workload measure was created
We also measured how many participants took the medications
based on a factor analysis (Principal Components extraction meth-
used in the study because patients may better remember informa-
od; Varimax rotation; factors with Eignenvalues
1.0 were con-
tion about medications they actually take (Morrow et al., 2005).
sidered) of the five NASA-TLX scales for each of the four med-
The three groups did not differ in the percent who took one or
ication x patient complexity conditions (total of 20 variables). Four
more of these medications (44% of participants in no-aid group,
factors accounted for 72% of the variance. Factor loadings showed
31% in the paper group, and 28% in the medtable group, 2
1.9,
that 4 of the 5 scales (mental demand, time pressure, mental effort
p
.10). Problems were presented blocked by medication com-
required, and frustration) loaded on the same factor within each
plexity with order of simple and complex medication problems,
complexity condition (loadings
.35-.91). Composite variables
and the order of simple versus complex patient schedules within
created by averaging scores for these four variables within each
the simple and complex medication condition, counterbalanced
complexity condition were analyzed by an Aid
Role (Provider
across participant pairs in each group.
vs. Patient)
Medication Complexity
Patient Complexity
Participants in one group used the medtable (see Figure 1) to
ANOVA, with the latter two variables repeated measures.
help create their schedules. Participants in the paper group re-
ceived a blank 8”
11” sheet of paper (same size as the med-
Results
table). Finally, participants in the no-aid group did not receive an
external aid.
Problem solving accuracy.
Participants created more accurate
Procedure.
Pairs of participants (one randomly assigned as
schedules for problems with simpler medication information, F(1,
provider and one as patient) completed four problems, one from
45)
77.7, p
.001, eta2
.63, and with simpler patient
each of the four conditions created by combining medication and
information, F(1, 45)
6.5, p
.05, eta2
.13, (see Table 2).
patient complexity. They were given 1 min for each simple med-
They were also more accurate when using the aids, F(2, 45)
ication problem and 2 min for each complex medication problem
14.8, p
.001, eta2
.40. Planned comparisons showed that
to become familiar with their assigned information. They were told
participants were more accurate with either aid compared to no aid,
to share the information verbally, but not to look at each others’
with no difference between aid conditions. Aid-related benefits
information sheets. Next, they worked together to create schedules
depended on medication complexity, F(2, 45)
16.9, p
.001,
consistent with the medication and patient constraints. After they
eta2
.43. This interaction was analyzed by comparing the aid
agreed on the schedule, the patient described it to the experimenter
conditions for problems with simple and complex medication
(reported schedules were taped for later scoring). A maximum
information. An aid effect in the simple medication condition, F(2,
limit of 4 min for simple medication problems and 10 min for
45)
4.2, p
.05, eta2
.16, showed that participants were more
complex problems was imposed to be consistent with limited
accurate when using paper rather than no aid ( p
.01), with no
patient contact time in routine primary care visits (Braddock &
difference between the paper and medtable or no-aid and medtable
Snyder, 2005).
conditions ( ps
.10). An aid effect in the complex medication
After each schedule was reported, provider and patient sepa-
condition, F(2, 45)
15.6, p
.001, eta2
.41, showed that
rately completed the NASA-TLX measure (Hart & Staveland,
participants were more accurate with either aid than no aid ( p
1988). This instrument, composed of 5-point Likert scales that
.001), with no difference between aids.
measure mental demand, time pressure, mental effort required,
Problem solving time.
Solution time results paralleled accu-
assessed performance, and frustration, has been used to measure
racy (see Table 2). Participants were faster for problems with
subjective workload associated with a wide range of tasks (includ-
ing problem solving) and people (Tsang & Wilson, 1997). Partic-
ipants practiced the task with sample medication and patient in-
Table 2
formation of moderate complexity (3 medications).
Mean (and Standard Deviation) Values for Medication Schedule
Dependent variables and plan of analysis.
Problem-solving
Solution Measures in Experiment 1
accuracy, completion time, and efficiency were measured. Accu-
Performance measure
racy was measured by the total points (out of 24 or 29, depending
on specific medications used) awarded for meeting medication
Accuracy %
Efficiency
requirements, which included medication name, number of pills,
Medication
correct
Time sec
sec/acc
times per day, dose spacing, whether scheduled times met food and
complexity
Aid
M (SD)
M (SD)
M (SD)
water restrictions, medication co-occurrence restrictions, and pa-
Simple
None
93 (8.7)
132 (57.7)
1.6 (1.0)
tient schedule restrictions. Problem-solving time was measured
Paper
99 (1.8)
117 (45.7)
1.2 (0.5)
from when provider and patient began sharing medication and
Medtable
96 (5.1)
159 (42.2)
1.7 (0.5)
patient information until the patient indicated that he or she was
Complex
None
52 (25.3)
563 (60.4)
17.8 (21.1)
ready to describe the schedule or until the time limit was reached
Paper
84 (17.1)
510 (80.3)
6.8 (3.9)
Medtable
88 (10.9)
527 (70.5)
6.3 (2.6)
(measured during the session by stopwatch). An efficiency score

---

292
MORROW ET AL.
simpler medication, F(1, 45)
1851.0, p
.001, eta2
.98, and
paper in similar ways, usually creating verbal lists by organizing
simpler patient information, F(1, 45)
11.0, p
.001, eta2
.20.
information in terms of medication name rather than times to take
While there was no overall effect of aid, F(2, 45)
1.9, p
.10,
the medications (66% of pairs, p
.06, binomial test, N
32).
eta2
.08, the Aid
Medication complexity interaction was
Then, for each medication they wrote down times to take it. Those
significant, F(1, 45)
4.1, p
.05, eta2
.15. In the simple
few who created time-based schedules were as likely to embed
medication condition, an aid effect, F(2, 45)
6.1, p
.01,
notes (e.g., about warnings) in the schedule as to segregate them
showed that participants created schedules more slowly when
(64% embedded notes, p
.10, N
11).
using the medtable ( p
.05), with no difference between paper
Participants also used the medtable in similar ways. Most wrote
and no aid ( p
.10). In the complex medication condition, an aid
down medication names for each row of the table under the
effect, F(2, 45)
4.7, p
.05, showed that participants created
medication column (81%, p
.01), and special instructions for
schedules more slowly when they did not have an aid ( p
.05),
each medication next to the name (69%, p
.05). They tended to
with no difference between the aid conditions.
use the timeline that organized the table columns by writing daily
Problem solving efficiency.
Problem solving was more effi-
event times below corresponding icons (81%, p
.01). The
cient for problems with simpler medication information, F(1,
majority wrote medication names or times in the table body (72%,
45)
98.0, p
.001, eta2
.70, but not for those with simpler
p
.05), suggesting difficulty with using the medtable. For
patient information ( p
.10). Problem solving was more efficient
example, participants who correctly filled in the time row often
when participants used either aid rather than no aid, F(2, 45)
wrote additional times in the body. Other strategies also suggested
4.9, p
.05, eta2
.19, with no difference between the aids
difficulty. For example, they did not always fill in the time row of
(planned comparisons p
.10). An Aid
Medication Complexity
the table exactly as the column headers indicated they should: For
interaction, F(2, 45)
5.4, p
.01, eta2
.20, showed that the
the column “1 hour before meal” next to the column “dinner,” they
aids increased efficiency for problems with complex medication
often filled in a time 30 min before dinner because one medication
information, F(2, 45)
4.2, p
.05 (no difference between aid
had to be taken 30 min before a meal.
conditions), but not for those with simple medication information,
F(2, 45)
2.1, p
.10.
Discussion
Subjective mental workload.
Problem-solving was associated
with lower workload for problems with simple versus complex
Both aids increased collaborative problem-solving accuracy and
medication information, F(1, 90)
288.9, p
.001, eta2
.76,
efficiency while reducing subjective workload compared to the
and in the no-aid than in either aid condition, F(2, 90)
11.6, p
no-aid condition, primarily for the complex medication problems.
.001, eta2
.21 (see Figure 2). The cost of not having an aid for
The time measure may have been hampered by the imposed time
perceived workload was greater for complex than for simple
limit. This limit was reached for 72% of the complex medication
medication problems, F(2, 90)
8.1, p
.001, eta2
.15.
problems in the no-aid condition and for 34% in each aid condi-
Workload ratings were greater for patients than for providers, F(1,
tion, so that aid-related effects would more likely occur for prob-
90)
7.2, p
.01, eta2
.07.
lem solving accuracy than for time. However, we note that
Analyses of external aid strategies.
We analyzed how partic-
problem-solving was slower as well as less accurate when not
ipants used the aids to support problem solving. They used the
supported by an aid in the complex medication problem condition.
Figure 2.
Mean subjective workload ratings (NASA-TLX, Hart & Staveland, 1988), by participant role, aid,
and medication complexity (Experiment 1).

---

EXTERNAL SUPPORT FOR COLLABORATION
293
Creating the complex schedules required integrating informa-
solving is less limited by working memory. The medtable, revised
tion about four medications in terms of the patient’s daily routine,
to better reflect constraints relevant to creating medication sched-
which likely imposed heavy demands on cognitive resources.
ules, was compared to the unstructured aid in Experiment 2.
Participants using either aid could offload these demands to an
external representation. Yet, the medtable was no more effective
Experiment 2
than the unstructured aid in supporting problem-solving. Analysis
of how participants used the aids suggested difficulty using the
Method
medtable’s timeline to map medication times onto daily events,
perhaps because before/after meal time columns were too specific
Participants.
Sixty-four community-dwelling older adults par-
to support flexible use.
ticipated (mean age
69, 60 – 86 years; 55% female). As before,
Therefore, the aid was revised to more clearly represent task
they were randomly assigned to be provider or patient, with pairs
constraints (see Figure 3). First, the primary function of mapping
randomly assigned to the two aid conditions. Participants in these
medication onto patient constraints was highlighted by using color
conditions did not differ on any variable except processing speed
to distinguish and relate these two types of information: the med-
scores (see Table 3), so this variable was included as a covariate
ication column was outlined in red while the time row was outlined
(mean processing speed score per pair) for the analyses.
in blue. In addition, the “Medications” and “Times” headings were
Procedure.
The procedure was the same as before except for
separated to clarify that the top row was for time information only.
the following. First, the no-aid condition was eliminated because
Second, the time row was separated from the table body to clarify
Experiment 1 clearly documented the cost of no aid for this task.
that it represented daily event times rather than times for only the
Second, participants were given more practice using the two aids.
first medication. Third, colons were added to the time row to
In the paper condition, they were given sample medication and
clarify where medication-taking times should be written. Fourth,
patient information and asked to consider how they would use the
the titles “2 hours after meal” and “1 hour before meal” were
aid to develop a schedule. In the medtable condition, they were
eliminated because they were too specific. Finally, more space was
given a completed table and asked to describe the schedule. Next,
provided next to the medication names on the rows along with the
the simple medication problems from Experiment 1 were used to
labels “Name” and “Instructions” so patients could write special
practice the specific task. Thus, they were presented before the
instructions. If the revised medtable is easier to use as a joint
complex medication problems rather than counterbalancing the
workspace, it may encourage partners to access information from
order of simple and complex problems as in Experiment 1, and
the aid rather than memory (Fu & Gray, 2000), so that problem
data collection was restricted to the latter. This was done because
Figure 3.
Revised medtable aid (Experiment 2).

---

294
MORROW ET AL.
Table 3
vestigated in the context of complex medication schedules. Thus,
Mean (and Standard Deviation) Values for Continuous
we used a mixed design with Aid as between-groups variable and
Demographic and Individual Difference Variables, by Condition
Complexity of patient information as repeated measure to analyze
(N
32 per Group) in Experiment 2
problem solving performance, measured by accuracy (percent of
points out of 24 or 29), time, and efficiency.
Variable
Medtable
Paper
p value
Age (years)
68.6 (5.4)
68.6 (6.7)
.10
Results
Education (years)
15.4 (2.9)
15.0 (2.7)
.10
Comparison task score
45.8 (4.6)
41.6 (3.6)
.05
Problem solving accuracy.
Participants created more accurate
Self-rated health score
5.1 (1.4)
5.0 (1.4)
.10
schedules when using the medtable rather than blank paper, F(1,
Number of prescribed medications
3.3 (2.8)
3.2 (2.9)
.10
29)
7.3, p
.05, eta2
.20 (see Table 4). Complexity of patient
Note.
Comparison task is a measure of processing speed (Salthouse,
information did not influence accuracy, nor interact with Aid,
1991). The mean of the Letter (Maximum score
42) and Pattern
ps
.10. Participants presumably created more accurate schedules
Comparison (Maximum score
30) tasks was used because performance
because they effectively used the medtable as an external work-
on the two tasks were correlated (r
.79). Self-rated health is a 7-point
space. To explore this possibility, we scored each pair’s medtable
scale from 1 (very poor health) to 7 (very good health).
or paper for accuracy of information written on the aid, using the
same scoring scheme as for the verbally reported schedules. Find-
accuracy for the simple conditions had been near ceiling in Ex-
ings were similar to those from the reported schedules: Accuracy
periment 1 so the problems were unlikely to discriminate effects of
was higher for medtable versus paper, F(1, 29)
5.8, p
.05,
the two aids.
eta2
.16, and for simpler patient information, F(1, 29)
4.6,
Third, maximum time limit was increased from 10 to 15 min in
p
.05, eta2
.13.
order to address the problem in Experiment 1 that many partici-
Problem solving time and efficiency.
Few pairs reached the
pants timed out before completing their schedule. Fourth, fictional
time limit before finishing their schedules (M
9%, P
18%,
medication names were used to avoid potential problems of some
2
1.4, p
.10). A nonsignificant trend suggested faster
participants having taken medications used in the study. Fifth, a
solutions when participants used the medtable, F(1, 29)
3.1, p
delayed cued recall task was included to explore whether the
.10, eta2
.10 (see Table 4). Complexity of patient information
medtable would improve memory for schedules once they were
did not influence time nor interact with Aid ( ps
.10). Problem-
created. This finding would have implications for improving pa-
solving was more efficient when supported by the medtable, F(1,
tients’ ability to remember medication information after consulting
29)
5.1, p
.05, eta2
.15.
with providers. However, the measure had limited sensitivity be-
Subjective workload.
As in Experiment 1, a composite work-
cause it occurred after participants had created two complex sched-
load measure was created from a factor analysis of the five
ules (with 8 medications), so the potential for memory interference
NASA-TLX scales for the simple and complex patient schedule
was high. The delayed test was given after participants completed
conditions (total of 10 variables). Two factors explained 69% of
the medication schedule problems and the individual difference
the variance. Factor loadings (.62
.90) showed that mental de-
measures (approximate delay 10 min). They were asked to write
mand, time pressure, mental effort expended, and frustration
down any information they could remember about the medications
loaded on the same factor within each complexity condition.
and schedules they had created from the two complex problems.
Composite variables were analyzed by an Aid
Role
Patient
They were cued with the medication names and the patient sched-
Complexity ANOVA, with the latter variable a repeated measure.
ule information from each problem.
While the pattern across conditions was similar to the two aid
Dependent variables and design.
We investigated the impact
conditions in Experiment 1, only the complexity variable was
of external aid and patient schedule complexity on problem solv-
significant (Simple
2.3, Complex Patient Information
2.9;
ing. Participants either used the medtable or blank paper as they
F(1, 60)
17.3, p
.001, eta2
.22). Ratings did not differ for
worked together to create medication schedules. Complexity of
patients and providers (2.6 vs. 2.5), nor for medtable and paper
patient information was varied as in Experiment 1, but only in-
groups (2.5 vs. 2.6). Absence of significant effects for the role and
Table 4
Mean (and Standard Deviation) Values for Medication Schedule Solution Measures in
Experiment 2
Performance measure
Reported
Aid
accuracy %
accuracy %
Efficiency
Patient
correct
correct
Time sec
sec/acc
complexity
Aid
M (SD)
M (SD)
M (SD)
M (SD)
Simple
Paper
90 (7.0)
89 (7.0)
702 (192.1)
7.8 (2.2)
Medtable
96 (4.7)
95 (4.7)
553 (160.5)
5.8 (1.8)
Complex
Paper
84 (16.8)
85 (16.8)
734 (154.5)
10.0 (6.5)
Medtable
92 (7.0)
91 (7.0)
694 (187.4)
7.6 (2.2)

---

EXTERNAL SUPPORT FOR COLLABORATION
295
aid variables may reflect the fact that the more discriminating
enced while developing the schedules, so that problem solving was
conditions (no-aid vs. aid; simple vs. complex medication infor-
limited more by perceptual access (from the aid) than by memory
mation) were excluded in Experiment 2. In addition, power to
retrieval. Similar benefits of external aids for problem-solving
detect effects was limited by the small sample size (Costello &
have been found for younger and older adults (Morrow & Rogers,
Osborne, 2005).
2008).
Delayed cued recall task.
Accuracy of delayed recall, scored
Our findings can be explained by distributed cognition theories
the same as accuracy in the problem-solving task, was low, re-
that view performance as emerging from interacting internal and
flecting the difficulty of this task. A nonsignificant trend suggested
external task components, with people managing demands on
that patients recalled more information than providers (Patient
limited cognitive resources by using external representations
28%, Providers
22%, F(1, 60)
3.6, p
.10, eta2
.06).
(Hutchins, 1995; Zhang & Norman, 1994). A key insight of this
There was no difference between the two aid conditions (Med-
approach is that successful performance hinges on how easily
table
24%, Paper
26%, F(1, 60)
1.0). This measure may
internal and external components are coordinated to accomplish
have been insensitive because of the potential for interference
task goals. External aids should be more effective to the extent
among different medications. It is also possible that using the
they clearly represent task-relevant constraints. The medtable’s
external aid reduced elaborate encoding of, and thus later memory
organization (explicitly mapping medications in rows onto daily
for, information because cognitive effort was offloaded to the
event times in columns) externalized relationships between multi-
environment. Future studies should explore cognitive costs as well
ple information sources, reducing the need to store, manipulate,
as benefits of external aids by including the no-aid condition and
and access this information from working memory. Because these
eliminating possible interference from multiple schedules.
processes are age-sensitive (Bopp & Verhaeghen, 2007), the med-
Analyses of external aid strategies.
Participants again consis-
table may have provided environmental support for older adults’
tently used paper to support problem solving. They organized
problem solving.
information into lists by medication name rather than by times
The medtable may have supported collaborative as well as
(69% of pairs, p
.05, N
32), and for each medication they
individual problem-solving processes. External aid benefits have
wrote times to take it. Those who created time-based schedules
been found for collaborative as well as individual problem solving
were as likely to embed notes in the schedule as to segregate them
(Heiser, Tversky, & Siverman, 2004) and learning (Fischer &
(64% embedded notes, p
.10, N
11).
Mandl, 2005), although these studies did not investigate the health
Participants also consistently used the medtable and were more
care domain. By providing a shared workspace, the aids in our
likely to take advantage of its organization than in Experiment 1.
study may have supported processes such as jointly attending to
All patients wrote down medication names for each row of the
information as well as proposing and critiquing schedules. The
table, under the Medication column, and included special instruc-
medtable may have been most effective because it represented
tions in the notes section next to the name (this strategy was more
task-relevant constraints that support management of joint atten-
frequent than in Experiment 1:
2
4.6, p
.05). All patients
tion critical to collaboration (Schwartz, 1995).
used the timeline that organized the table columns by writing times
of daily events such as meals in the available space (again, this
Implications for Health Care Practice: Study Limitations
strategy was more frequent than in Experiment 1: 2
4.6, p
and Future Research
.05). Finally, most patients wrote the number of pills (dose) in the
table cells (84%, p
.01), rather than writing medication names or
There are some constraints on our ability to generalize these
times as they did in Experiment 1, suggesting they better under-
findings to clinical practice. First, participants were not actual
stood how to use the medtable.
patients and providers. Nurses or pharmacists might make better
use of the aids because of medication knowledge. However, while
General Discussion
practitioners often rely on external aids such as white boards to
support many tasks (Nemeth et al., 2004), there is little evidence
Older adults who used external aids created more accurate
that they systematically use such aids to foster the patient/provider
medication schedules in a simulated provider/patient collaborative
partnership
essential
for
patient
self-care
and
safety
problem-solving task. While problem solving in Experiment 1 was
(Schwartzberg, Cowett, VanGeest, & Wolf, 2007). Because the
more accurate and efficient when partners used an aid compared to
medtable is designed to integrate information given by both pro-
when they only shared information verbally, there was no evidence
viders (e.g., about medications) and by patients (e.g., about daily
that the medtable was more effective than an unstructured aid
routine), it may help bridge the gulf between divergent conceptions
(blank paper). In Experiment 2, a redesigned medtable that more
providers and patients bring to clinical encounters (Patel, Arocha,
clearly represented task-relevant constraints (relationships be-
& Kushniruk, 2002).
tween medication and patients’ daily routine information) sup-
Second, communication barriers in the current health system,
ported more accurate and efficient collaborative problem-solving
such as limited time with patients, may also preclude routine use
compared to the unstructured aid.
of the medtable. Such tools may be most suitable for pharmacists
or nurses who help chronically ill older adults manage complex
External Aids and Collaborative Problem Solving
medication regimens. This suggestion is consistent with our find-
ing that aid-related benefits primarily occurred for the complex
The external aids may have reduced problem-solving demands
schedule problems. The medtable may also be appropriate for
on working memory. They were used as a notepad to write down
other situations where time is less of a barrier, such as nurses or
information such as medication names and times and then refer-
caretakers working with older patients at home.

---

296
MORROW ET AL.
Finally, the medtable may be most effective if implemented
Gould, O. N., Kurzman, D., & Dixon, R. A. (1994). Communication during
electronically, which would be more flexible (e.g., tailoring to
prose recall conversations by young and old dyads. Discourse Processes,
diverse patients) and expedite processes such as updating compre-
17, 149 –165.
hensive medication lists. Electronic collaborative aids might also
Hart, S. G., & Staveland, L. E. (1988). Development of a multi-
help mitigate problems associated with using electronic medical
dimensional workload rating scale: Results of empirical and theoretical
research. In P. A. Hancock & N. Meshkati (Eds.), Human mental
records during office visits, such as one-sided communication
workload (pp. 139 –183). Amsterdam: Elsevier.
(Patel et al., 2002). However, it would be important for patients to
Haynes, B., McKibbon, A., & Kanani, R. (1996). Systematic review of
take home a copy of their medtable schedule to guide adherence.
randomised trials of interventions to assist patients to follow prescrip-
We are now investigating the use of the medtable by pharmacy and
tions for medications. Lancet, 348, 383–386.
nurse providers working with patients with low health literacy
Heiser, J., Tversky, B., & Silverman, M. (2004). Sketches for and from
in clinical settings. Eventually, we will test whether collabora-
collaboration. In J. S. Gero, B. Tversky, & T. Knight (Eds.).Visual and
tive use of the medtable improves older adults’ health behaviors
spatial reasoning in design III (pp. 69 –78). Sydney, Australia: Key
and outcomes.
Centre for Design Research.
Hutchins, E. (1995). How a cockpit remembers its speed. Cognitive Sci-
ence, 19, 265–288.
References
Insel, K. C., Morrow, D. G., Brewer, B. B., & Figuerosa, A. J. (2006).
Cognitive function and medication adherence. Journal of Gerontology:
Andersson, J., & Ronnberg, J. (1995). Recall suffers from collaboration:
Psychological Sciences 60B, P102–P107.
Joint recall effects of friendship and task complexity. Applied Cognitive
Kaufman, D. W., Kelly, J. P., Rosenberg, L., Anderson, T. E., & Mitchell,
Psychology, 9, 199 –211.
A. A. (2002). Recent patterns of medication use in the ambulatory adult
Aspden, P., Wolcott. J. A., Bootman, J. L., & Croenwett. L. R. (2007).
population of the United States: The Slone survey. Journal of the
Preventing medication errors. Washington, DC: The National Acade-
American Medical Association, 287, 337–344.
mies Press.
Kripalani, S., Robertson, R., Love-Ghahhari, M. H., Henderson, L. E.,
Bodenheimer, T., Lorig, K., Holman, H., & Grumbach, K. (2002). Patient
Praska, Strawder, A., et al. (2007). Development of an illustrated med-
self-management of chronic illness in primary care. Journal of the
ication schedule as a low-literacy patient education tool. Patient Edu-
American Medical Association, 288, 2469 –2475.
cation and Counseling, 66, 368 –377.
Bopp, K. L., & Verhaeghen, P. (2007). Age-related differences in control
Larkin, J. H., & Simon, H. A. (1987). Why a diagram is (sometimes) worth
processes in verbal and visuospatial working memory: Storage, trans-
ten thousand words. Cognitive Science, 11, 65–99.
formation, supervision, and coordination. Journal of Gerontology: Psy-
Ley, P. (1997). Communicating with patients: Improving communica-
chological Sciences and Social Sciences. 62, P239 –P246.
tion, satisfaction and compliance (1st ed.). London, England: Stanley
Braddock, C. H., & Snyder, L. (2005). The doctor will see you shortly: The
Thorne, Ltd.
ethical significance of time for the patient-physician relationship. Jour-
Morrow, D. G., Hier, C. M., Menard, W. E., & Leirer, V. O. (1998). Icons
nal of General Internal Medicine, 20, 1057–1062.
improve older and younger adults’ comprehension of medication informa-
Budnitz, D. S., Pollock, D. A., Weidenbach, K. A., Mendelsohn, A. B.,
tion. Journal of Gerontology: Psychological Sciences, 53B, P240 –P254.
Schroeder, T. J., & Annest, J. L. (2006). National Surveillance of
Morrow, D. G., Raquel, L. M., Schriver, A. T., Redenbo, S., & Rozovski,
Emergency Department visits for outpatient adverse drug events. Jour-
D. (2007). External support for collaborative medication planning by
nal of the American Medical Association, 296, 1858 –1866.
patients and providers. Proceedings of the Human Factors and Ergo-
Buerhaus, P. I., Donelan, K., Ulrich, B. T., Norman, L., & Dittus, R.
nomics Society 51st Annual Meeting. Santa Monica, CA: Human Factors
(2006). State of the registered nurse workforce in the United States.
& Ergonomics Society.
Nursing Economics, 24, 6 –12.
Morrow, D. G., & Rogers, W. (2008). Environmental support: An integra-
Carlson, M. C., Xue, Q., Fried, L. P., Tekwe, C., & Brandt, J. (2005).
tive framework. In press.
Validation of the Hopkins Medication Schedule to objectively measure
Morrow, D. G., Weiner, M., Young, J., Steinley, D., Deer, M., & Murray,
compliance. The Women’s Health and Aging Study II. Journal of
M. (2005). Improving medication knowledge among older adults with
Gerontology: Medical Sciences. 60A, 217–223.
Costello, A. B., & Osborne, J. W. (2005). Best practices in exploratory
heart failure: A patient- centered approach to instruction design. The
factor analysis: Four recommendations for getting the most out of your
Gerontologist, 45, 545–552.
analysis. Practical Assessment, Research & Evaluation, 10, 1–9. Avail-
Murray, M. D., Young, J., Hoke, S., Tu, W., Weiner, M., Morrow, D. G.,
able online: http://pareonline.net/getvn.asp?v 10&n 7
et al. (2007). Pharmacist intervention to improve medication adherence
Craik, F. I. M., & Jennings, J. M. (1992). Human memory. In F. I. M. Craik
in heart failure: A randomized trial. Annals of Internal Medicine, 146,
& T. A. Salthouse (Eds.), The handbook of aging and cognition (pp.
714 –725.
51–110). Hillsdale, NJ: Erlbaum.
Nemeth, C. P., Cook, R. I., O’Connor, M. F., & Klock, P. A. (2004). Using
Day, R. (1988). Alternative representations. In G. Bower (Ed.), Psychology
cognitive artifacts to understand distributed cognition. IEEE Transac-
of learning and motivation. New York, NY: Academic Press.
tions on Systems, Man and Cybernetics-Part A: Systems and Humans,
Fischer, F., & Mandl, H. (2005). Knowledge convergence in computer-
34, 726 –735.
supported collaborative learning: The role of external representations.
Park, D. C., Hertzog, C., Leventhal, H., Morrell, R. W., Leventhal, E.,
The Journal of the Learning Sciences, 14, 405– 441.
Birchmore, D., et al. (1999). Medication adherence in rheumatoid ar-
Fu, W., & Gray, W. D. (2000). Memory versus perceptual-motor tradeoffs
thritis patients: Older is wiser. Journal of the American Geriatrics
in a blocks world task. Proceedings of the Twenty-second Annual Con-
Society, 47, 172–183.
ference of the Cognitive Science Society (pp. 154 –159). Hillsdale, NJ:
Park, D. C., & Jones, T. R. (1997). Medication adherence and aging. In
Erlbaum.
A. D. Fisk & W. A. Rogers (Eds.), Handbook of human factors and the
Gergle, D., Kraut, R. E., & Fussell, S. R. (2004). Action as language in a
older adult (pp. 257–287). San Diego, CA: Academic Press.
shared visual space. Proceedings of the ACM Conference on Computer
Patel, V. L., Arocha, J. F., & Kushniruk, A. W. (2002). Patients’ and
Supported Cooperative Work, Chicago, IL.
physicians’ understanding of health and biomedical concepts: Relationship

---

EXTERNAL SUPPORT FOR COLLABORATION
297
to the design of EMR systems. Journal of Biomedical Informatics, 35,
Communication techniques for patients with low health literacy: A
8 –16.
survey of physicians, nurses, and pharmacists. American Journal of
Ryan, E. B., Merideth, S. D., MacLean, M. J., & Orange, J. B. (1995).
Health Behavior, 31, S96 –S104.
Changing the way we talk with elders: Promotion of health using the
Thornton, W. J., & Dumke, H. A. (2005). Age differences in everyday
communication enhancement model. International Journal of Aging and
problem-solving and decision-making effectiveness: A meta-analytic
Human Development, 41, 89 –107.
review. Psychology and Aging, 20, 85–99.
Salthouse, T. A. (1991). Mediation of adult age differences in cognition by
Tsang, P. S., & Wilson, G. F. (1997). Mental workload. In G. Salvendy
reductions in working memory and speed of processing. Psychological
(Ed.), The Handbook of Human Factors and Ergonomics (pp. 417– 449).
Science, 2, 179 –183.
New York: Wiley.
Schillinger, D., Piette, J., Grumbach, K., Wang, F., Wilson, C., Daher, C., et
Vermeire, E., Hearnshaw, H., Van Royen, P., & Denekens, J. (2001).
al. (2003). Closing the loop: Physician communication with diabetic pa-
Patient adherence to treatment: Three decades of research. A com-
tients who have low health literacy. Archives of Internal Medicine, 163, 83–90.
prehensive review. Journal of Clinical Pharmacy and Therapeutics,
Schwartz, D. (1995). The emergence of abstract representations in dyad
26, 331.
problem solving. The Journal of the Learning Sciences, 4, 321–354.
Zhang, J., & Norman, D. A. (1994). Representations in distributed cogni-
Schwartzberg, J. G., Cowett, A., VanGeest, J., & Wolf, M. S. (2007).
tive tasks. Cognitive Science, 18, 87–122.
Appendix
Example of Medication Information Given to Provider Participants for Complex Medication Problems
1. Quinoxin (lowers cholesterol):
Special instructions:
Take medication on an empty stomach (at least 1 hour before or 2 hours after a meal).
Do NOT take any other medication within 1 hour.
Dose:
Take 1 pill twice a day. Space doses by at least 8 hours.
2. Fosavin (prevents osteoporosis):
Special instructions:
Do NOT eat for 30 minutes afterwards.
Do NOT lie down up to 1 hour afterwards.
Dose:
Take 1 pill twice a day. Space doses by at least 8 hours.
3. Spirotar (reduces water retention):
Special instructions:
Take with food.
Do not take medication within 4 hours of bedtime.
Dose:
Take 1 pill twice a day. Space doses by at least 8 hours.
4. Zanoxin (antibiotic):
Special instructions:
Take medication with a full (8 oz.) glass of water.
Dose:
Take 2 pill twice a day. Space doses by at least 8 hours.
Received August 3, 2007
Revision received March 6, 2008
Accepted March 21, 2008

---