ORIGINAL ARTICLE
0, MONTH 2017
Identification, Associated Factors, and
Prognosis of Symptom Clusters in Taiwanese
Patients With Heart Failure
Tsuey-Yuan Huang1* & Debra K. Moser2 & Shiow-Li Hwang3
1
RN, PhD, Associate Professor, Department of Nursing, Chang Gung University of Science and Technology, Taoyuan,
Taiwan, ROC & 2RN, PhD, FAAN, Professor, College of Nursing, University of Kentucky, Lexington, USA &
3
RN, DNSc, Chair Professor, Department of Nursing, Asia University, Taichung, Taiwan, ROC.
ABSTRACT
Background: Patients with heart failure (HF) have multiple
distressing symptoms that are associated with poor outcomes.
These symptoms do not occur in isolation from each other but
likely occur as discrete clusters that may prove helpful to
clinicians trying to counsel patients about symptom monitoring
and management. Defining common symptom clusters and
determining the associations between symptom clusters and
outcomes may improve patient management.
Purpose: The aim of this study was to define symptom clusters
and their association with event-free survival in terms of cardiac
hospitalization and all-cause mortality in patients with HF.
Methods: Patients were recruited from outpatient HF clinics.
Physical symptoms (dyspnea, fatigue, edema, sleeplessness,
anorexia, and poor memory) were measured using the modified
Pulmonary Function Status and Dyspnea Questionnaire and the
Minnesota Living with Heart Failure Questionnaire. A two-stage
cluster analysis was conducted to identify subgroups of patients
based on the self-perceived severity of the six symptoms. The
KaplanYMeier survival curve and log-rank test were used to
assess whether symptom clusters were associated with event-
free survival through a 12-month follow-up.
Results: Two hundred fifty-eight patients (mean age = 61.2
T 12.3 years, 75% male, 41% New York Heart Association
class III/IV) participated. Three symptom clusters were identi-
fied based on the severity of symptoms. These clusters were
called the nonsevere symptom cluster (all symptoms were
rated with low severity), the typical severity symptom cluster
(high level of severity for dyspnea and fatigue, low level of
severity for edema, and moderate level of severity for all other
symptoms), and the atypical severity symptom cluster (low level
of severity for dyspnea and fatigue, high level of severity for
edema, and moderate level of severity for all other symptoms).
Symptom clusters were associated with event-free survival
(p G .001). A post hoc comparison showed that the prognosis
was better in the nonsevere symptom cluster than both the
typical symptom (p G .001) and nontypical symptom (p G .001)
clusters and that the prognoses for the latter two clusters did
not differ significantly.
Copyright © 2017 Taiwan Nurses Association. Unauthorized reproduction of this article is prohibited.
The Journal of Nursing Research h VOL. 00, NO.
Conclusions/Implications for Practice: Symptom clusters
play an important role in the prognoses of patients with HF.
Both patients and healthcare providers may use the informa-
tion that is provided by this study to improve the surveillance
and management of related symptoms.
KEY WORDS:
heart failure, symptom cluster, hospitalization, mortality.
Introduction
Heart failure (HF) is a global health problem, with a growing
prevalence that impacts the individual, family, and society
(Mozaffarian et al., 2015). Chronic HF is a syndrome rather
than a specific disease. Decline in cardiac output causes a
failure to meet the basic needs for metabolic function, re-
sulting in numerous related symptoms. Therefore, patients
with HF have multiple distressing symptoms that are asso-
ciated with poor outcomes (Moser et al., 2014).
According to the Middle Range Theory of Unpleasant
Symptoms, symptoms are influenced by physiological, psy-
chological, and situational factors. Moreover, symptom expe-
rience is a subjective perception that varies based on different
ascendants, situational issues, and outcomes (Lenz, Pugh,
Milligan, Gift, & Suppe, 1997). The related psychosocial
Accepted for publication: August 8, 2016
*Address correspondence to: Tsuey-Yuan Huang, No. 261, Wenhwa
1st Road, Kweishan 333, Taoyuan, Taiwan, ROC.
Tel: 886-3-2118999 ext. 3326; Fax: 886-3-2118610;
E-mail: d92426002@ntu.edu.tw
The authors declare no conflicts of interest.
Cite this article as:
Huang, T. Y., Moser, D., & Hwang, S. L. (2017). Identification,
associated factors, and prognosis of symptom clusters in Taiwanese
patients with heart failure. The Journal of Nursing Research, 00(0),
00Y00. doi:10.1097/jnr.0000000000000199
The Journal of Nursing Research
associates include depression, anxiety, perceived control,
and perceived social support (Huang et al., 2013). As
different symptoms may be caused by diverse mechanisms
and pathways, symptoms from the same disease may share
certain pathophysiological mechanisms (Lenz, Suppe, Gift,
Pugh, & Milligan, 1995). Patients with HF usually complain
of various symptoms. These symptoms do not occur in iso-
lation from each other and likely occur as discrete clusters
that may prove helpful to clinicians trying to counsel patients
about symptom monitoring and management (Moser et al.,
2014). Defining patterns in common symptom clusters and
determining the association of these symptom clusters
with outcomes in patients with HF may improve patient
management.
Examining symptom clusters in patients with HF is par-
ticularly important, as prior research has found that some
symptoms tend to cluster together and impact significantly
on HF patient outcomes (Moser et al., 2014). Symptom
clusters may be defined as multiple symptoms that are re-
lated to each other and that are experienced simultaneously
(Aktas, Walsh, & Rybicki, 2010; Lin et al., 2016). Generally,
symptoms belonging to a specific cluster pattern may have
a common biological mechanism for which the treatment
of one symptom may result in the alleviation of the other
symptoms of that cluster (Kirkova, Aktas, Walsh, & Davis,
2011). Symptom cluster studies are important for manifesting
the total symptom burden in appropriate treatment plans
(Aktas et al., 2010). The two approaches to conducting
symptom cluster analyses are clinical and statistical (Kirkova
& Walsh, 2007). The clinical symptom cluster analysis ap-
proach primarily employs correlation methods and is used
in studies that focus on a small number of preselected symp-
toms (Kirkova et al., 2011). The statistical symptom cluster
analysis approach primarily employs multisymptom assess-
ment tools or symptom checklists and is used in studies that
focus on a large number of symptoms (Kirkova et al., 2011).
The starting point for the latter approach is no assumption
of associations between symptoms, with factor or cluster
analysis employed to discover symptom clusters from the
statistically significant interrelationships among the dispa-
rate symptoms. Statistical analyses of multiple symptoms
may identify clusters that were not identified in the clinical
assessment (Aktas et al., 2010). For this reason, statistically
derived symptom clusters appear to be more objective than
clinical clusters and are useful to develop a new perspective
on categories of symptom clusters. However, few studies
have adopted this perspective to examine the phenomenon
of symptom clustering in clinical patients, especially in pa-
tients with HF (Aktas et al., 2010; Baggott, Cooper, Marina,
Matthay, & Miaskowski, 2012; Namisango et al., 2015).
Patterns of symptom clustering may correlate with patient
characteristics. Co-occurring symptoms appear to have more
complicated and synergistically detrimental associations with
the treatment outcomes, prognosis, functional status, and
quality of life of patients than the individual symptoms.
Therefore, understanding how individual symptoms clusters
Copyright © 2017 Taiwan Nurses Association. Unauthorized reproduction of this article is prohibited.
Tsuey-Yuan Huang et al.
are affected by demographic and clinical characteristics is
likely to facilitate early detection and diagnosis. Symptoms
cluster similarly among patients of the same cultural group.
Thus, identification of symptom clusters is a clinically
relevant approach to detecting specific demographic groups
that are most likely to experience similar symptoms with
readily identifiable characteristics (Hwang, Ahn, & Jeong,
2012; Moser et al., 2014). Therefore, the purpose of this
study was to define symptom clusters and their association
with event-free survival (cardiac hospitalization and all-
cause mortality) in patients with HF using a two-stage cluster
analysis that was designed to identify subgroups of patients
based on the self-perceived severity of six different physical
symptoms.
Methods
Study Design
This study used a prospective, 12-month longitudinal
follow-up approach, with data collected at baseline and
at 12 months posttest for the following physical and
psychosocial symptoms: dyspnea, fatigue, leg edema,
sleeplessness, reduced appetite, poor memory, depression,
anxiety, perceived control, perceived social support, and
cardiac event-free survival.
Sample and Data Collection Procedure
Inclusion criteria were (a) age of older than 18 years and
(b) a confirmed, cardiologist-verified diagnosis of HF. Pa-
tients with a history of cerebrovascular accident, recent
myocardial infarction (within the previous 6 months), or a
major comorbid terminal illness or those who were receiving
treatment such as hemodialysis or concurrent cancer treat-
ment were excluded from participation. Patients with cog-
nitive impairment or residing in institutional settings were
also excluded. Patients were recruited at outpatient HF clinics
from four medical centers located in Taiwan (two in northern
Taiwan and two in southern Taiwan). Approval to conduct
this study was obtained from the appropriate hospital insti-
tutional review boards, and informed consent was obtained
from each patient before participation. Nurse research
assistants and cardiologists identified eligible adults with HF
and invited them to participate in the study. Participants
completed the questionnaires in the presence of study nurses,
who were available to answer any questions and, when nec-
essary, read the questionnaires to the patients. Most patients
completed the questionnaires within 30 minutes.
Measurement via Validated
Questionnaires
Physical symptoms (dyspnea, fatigue, edema, sleeplessness,
anorexia, and poor memory) were measured using the
modified Pulmonary Function Status and Dyspnea Ques-
tionnaire (PFSDQ-M) and the Minnesota Living with Heart
Symptom Clusters in Patients With HF
Failure Questionnaire (MLHFQ). Psychosocial status (de-
pression, anxiety, perceived control, and perceived social
support) was measured using the Brief Symptom Inventory
(BSI), Control Attitudes Scale-Revised (CAS-R), and
Multidimensional Scale of Social Support Scale (MSPSS).
The modified pulmonary function status and
dyspnea questionnaire
The PFSDQ-M was used to measure the dyspnea of patients
with HF. The PFSDQ-M was developed in 1994 for use with
patients with chronic obstructive pulmonary disease and
modified in 1998 (Lareau, Carrieri-Kohlman, Janson-Bjerklie,
& Roos, 1994; Lareau, Meek, & Roos, 1998). The psy-
chometric properties of both the original and Chinese ver-
sions of the PFSDQ-M are satisfactory (Huang et al., 2008).
The Chinese-version PFDSQ-M is a 15-item instrument that
measures a respondent’s self-perceived assessment of the main
components of the target symptom: experience, frequency,
and intensity. The PFSDQ-M is composed of five general
questions regarding the patient’s experience of dyspnea, the
frequency of symptom occurrence over the past month, and
the overall intensity of shortness of breath (i.e., most days,
today, and during periods of normal activity). The other
10 questions are related to the intensity of dyspnea when
performing different activities, rated on a scale from 0 to 10,
with higher scores indicating greater severity. In this study,
the Cronbach’s " of the PFSDQ-M was .92.
The Minnesota living with heart failure questionnaire
The MLHFQ is a 21-item, disease-specific measure of health-
related quality of life for patients with HF. This questionnaire
includes two subscales: physical (eight items) and emotional
(five items). Scores for the two subscales and the questionnaire
were calculated. Each questionnaire item score ranges from
0 to 5, with the total possible scores for the questionnaire
ranging from 0 to 105 and higher scores indicating a more
negative perception of quality of life. The MLHFQ has ex-
cellent psychometric properties, and the Chinese version
has been validated (Ho, Clochesy, Madigan, & Liu, 2007).
Individual items from the MLHFQ have been used to define
a respondent’s subjective perception of the effects of a clin-
ical symptom (leg edema, sleeplessness, reduced appetite,
and poor memory) on his or her daily life. The Cronbach’s "
of the MLHFQ in this study was .88.
The brief symptom inventory
The BSI depression and anxiety subscales (Derogatis, 2006),
each containing six items, have been used to measure the
levels of depression and anxiety in patients with HF (Derogatis,
2006). Patients rate their level of distress related to a given
item on a scale of 0Y4 (0 = not at all and 4 = extreme). The
mean score (ranging from 0 to 4) for the six items reveals
the depression/anxiety status of patients with HF, with higher
scores indicating higher levels of depression and anxiety. The
Cronbach’s " of the BSI depression subscale was .82 for
Copyright © 2017 Taiwan Nurses Association. Unauthorized reproduction of this article is prohibited.
VOL. 00, NO. 0, MONTH 2017
Taiwanese patients with HF and .89 for American patients
with HF. This study calculated a Cronbach’s " of .78 for the
BSI anxiety subscale.
The control attitudes scale-revised
CAS-R (Moser et al., 2009) is an eight-item instrument
that has been modified from the original Cardiac Attitudes
Index and CAS. The CAS-R evaluates an individual’s per-
spective of his or her ability to control his or her heart
disease. Moser et al. (2009) developed this instrument to
measure the degree to which patients feel they have control
(and, conversely, a sense of helplessness) with regard to their
cardiac disease (Moser et al., 2009). Subjects rate their level
of agreement with the statements on a Likert scale (1Y5 or
1Y7), with the sum of all scores used as the overall score.
Higher scores indicate higher levels of perceived control.
The Cronbach’s " of the CAS-R in this study was .80.
The multidimensional scale of social support scale
MSPSS (Bruwer, Emsley, Kidd, Lochner, & Seedat, 2008)
is a 12-item, self-reported measurement that is used to mea-
sure social support. The total score and subscale scores for
family, friends, and significant others indicate the level of
perceived support. The MSPSS uses a 7-point Likert scale
with responses ranging from 1 to 7, with 1 = very strongly
disagree and 7 = very strongly agree and higher total scores
indicating more perceived support. The Cronbach’s " of the
MSPSS questionnaire in this study was .89.
End point of cardiac event-free survival
Cardiac event-free survival is defined in this study as living
with HF without an emergency department visit for HF ex-
acerbation or without a cardiac-event-related hospital ad-
mission or death. Patient or family interviews and medical
records were used to collect these data.
Data Analysis
IBM SPSS Statistics Version 15 (IBM, Inc., Armonk, NY,
USA) was used to conduct all data analyses. A two-stage
cluster analysis was conducted to identify subgroups of pa-
tients based on their perceived severity in terms of the six
physical symptoms (Hair, Black, Babin, Anderson, & Tatham,
2006). In the first stage, an agglomerative hierarchical cluster
analysis was performed, with squared Euclidean distances
used in the proximities matrix and between-group linkage
used as the clustering method. Moreover, between sum of
squares (BSS) was used to determine the number of clusters.
An increase in the BSS from k to k j 1 clusters was used to
indicate that the heterogeneity between clusters increased in
k j 1 clusters and should be stopped in the kth cluster
solution. After determining the number of clusters, a non-
hierarchical k-means cluster analysis was performed to
assign patients to their corresponding cluster. The clinical
characteristics among the clusters of patients were compared
The Journal of Nursing Research
using either a one-way analysis of variance or a chi-square
test. A Bonferroni post hoc multiple comparison was con-
ducted if the difference between clusters was significant.
Furthermore, a multivariable multinomial logistic regression
was applied to explore the factors that were associated with
the above clusters.
The KaplanYMeier method was used to estimate the 1-year
survival rate, and the log-rank test was used to examine the
differences in this survival rate among symptom clusters. The
prognostic value of the symptom clusters in predicting 1-year
outcomes was investigated using a multivariable Cox propor-
tional hazard model with the adjustment of other covariates.
There is no formal statistical power analysis of sample
size in cluster analysis (Hair et al., 2006). However, one
suggested guideline is that the number of observations should
be at least 2m, where m is the number of clustering variables
(Sarstedt & Mooi, 2014). This study included six symptoms,
resulting in a minimum required sample size of 26 = 64
participants to guarantee substantial segments.
Results
Results of the Cluster Analysis
Data from the 258 participants who provided complete data
on all of the six symptoms were entered into the cluster
analysis. In the first stage, hierarchical cluster analysis
revealed a BSS of 34.1, 36.2, 36.8, 41.5, and 46.8 for six-,
five-, four-, three-, and two-cluster solutions, respectively.
The change rate was relatively large between four and three
clusters (12.8% [41.5, 36.8]/36.8) and between three and
two clusters (12.9% [46.8, 41.5]/41.5), indicating the ap-
propriate number of clusters as four or three. Because of
considerations of limited sample size and the principle of
parsimony, we decided to use the three-cluster solution and
further assigned cluster numbers in the nonhierarchical
k-means cluster analysis (second stage).
The symptom scores for the three patient subgroups are
shown in Figure 1. The first cluster, categorized as the non-
symptom group, included 191 participants who reported
low levels for each symptom (symptom scores = 0.7Y1.5).
Figure 1. The mean score for each symptom, by symptom
cluster.
Copyright © 2017 Taiwan Nurses Association. Unauthorized reproduction of this article is prohibited.
Tsuey-Yuan Huang et al.
The second cluster, categorized as the typical symptom group,
included 28 participants who reported moderate levels of all
six symptoms. The third cluster, categorized as the atypical
symptom group, included 39 participants who reported low
levels of dyspnea and fatigue (1.6 and 1.1) and moderate
levels of the remaining four symptoms (from 1.9 to 3.7). The
naming of the clusters was performed post hoc based on the
analysis of the findings.
Clinical Characteristics of the Three
Symptom Clusters
Table 1 lists the clinical characteristics according to symp-
tom clusters. The results show that the clusters differed
significantly in terms of the following variables: living ar-
rangements, left ventricular dysfunction as measured by the
ejection fraction (LVEF G 30%), New York Heart Associa-
tion (NYHA) class, perceived depression, perceived anxiety,
and perceived control (p G .05). The proportion of living
alone was substantially higher in the typical group (36%)
than in either the nonsymptom group (12%) or the atypical
group (13%). The prevalence of left ventricular dysfunction
was most common in the atypical group (49%), followed
by the typical group (37%) and the nonsymptom group
(21%). The proportion of NYHA class III or IV was much
lower in the atypical group (28%) than in the two other
groups (970%). Bonferroni multiple comparison indicated
that the perceived level of depression and anxiety was higher
in the typical group than in the two other groups. In ad-
dition, the perceived level of anxiety was higher in the
atypical group than in the nonsymptom group. In contrast,
the level of perceived control in the typical group was lower
than in the two other groups.
Factors Associated With Symptom Clusters
Table 2 lists the association between clinical characteristics
and symptom clusters. Using the nonsymptom group as the
reference group, participants with NYHA class III or IV were
more likely to be in the typical symptom (OR = 6.53, p G .01)
or atypical symptom (OR = 7.01, p G .001) group than their
NYHA class I or IV peers. Participants who were living with
someone were less likely to be in the typical symptom group
(OR = 0.14, p G .05) than their peers who were living
alone. Furthermore, a higher level of perceived anxiety was
associated with an increased likelihood of being in the
typical symptom group (OR = 1.23, p G .05). In contrast,
a higher level of perceived control was associated with a
decreased likelihood of being in the typical symptom group
(OR = 0.93, p G .05).
Symptom Clusters and Chronic Heart
Failure Prognosis
Figure 2 illustrates the cumulative event-free survival curves
for 1-year cardiac outcomes for the three clusters. During the
1-year follow-up period, 75 (29.1%) participants experienced
Symptom Clusters in Patients With HF VOL. 00, NO. 0, MONTH 2017

TABLE 1.
The Characteristics of Participants by Symptom Cluster
Nonsymptom Typical Total Atypical Total
Total (N = 258) (n = 191) (N = 28) (N = 39)
Characteristic n % n % n % n % p

Gender .396
Male 194 75 146 76 22 79 26 67
Female 64 25 45 24 6 21 13 33
Age (years) .082
G65 143 55 100 52 15 54 28 72
Q65 115 45 91 48 13 46 11 28
Marital status .055
Without spouse 79 31 51 27 13 46 15 38
With spouse 179 69 140 73 15 54 24 62
Living arrangement .003
Live alone 37 14 22 12 10 36 5 13
Live with someone 221 86 169 88 18 64 34 87
LVEF (n = 239) .002
Q30% 174 73 139 79 17 63 18 51
G30% 65 27 38 21 10 37 17 49
NYHA class G.001
I/II 153 59 137 72 6 21 10 26
III/IV 105 41 54 28 22 79 29 74
Depression (M, SD) 3.4 4.4 2.8 3.9 6.7 5.6 3.7 4.4 G.001
Anxiety (M, SD) 3.2 3.6 2.5 3.1 6.5 5.2 4.1 3.2 G.001
Perceived social support (M, SD) 53.5 16.3 53.6 16.7 51.6 15.4 53.8 14.7 .821
Perceived control (M, SD) 67.9 9.6 68.6 9.6 61.6 10.2 68.5 7.6 G.001

Note. LVEF = left ventricular ejection fraction; NYHA = New York Heart Association.

a cardiac event, and 183 (70.9%) were censored. The mean three symptom clusters significantly differed in terms of
follow-up time was 10.1 months, with a standard deviation of survival rate (p G .001). Post hoc comparisons suggest that
3.4 months. The result of the log-rank test revealed that the the typical symptom and atypical symptom groups had

TABLE 2.
Factors Associated With Cluster of Physical Symptom (Multivariable Multinomial
Logistic Regression)
Typical Symptom Atypical Symptom
Variable OR 95% CI p OR 95% CI p

Female gender 0.55 [0.15, 1.95] .352 1.03 [0.40, 2.67] .944
Age Q 65 years 1.15 [0.38, 3.44] .803 0.50 [0.20, 1.28] .149
With spouse 1.06 [0.25, 4.40] .939 0.74 [0.25, 2.18] .583
Living with someone 0.14 [0.03, 0.65] .012 0.87 [0.21, 3.56] .845
LVEF G 30% 1.10 [0.36, 3.40] .862 1.91 [0.80, 4.56] .143
NYHA class III/IV 6.53 [2.14, 19.9] .001 7.01 [2.71, 18.2] G.001
Depression 1.01 [0.88, 1.17] .856 1.01 [0.88, 1.15] .942
Anxiety 1.23 [1.04, 1.45] .017 1.07 [0.91, 1.25] .402
Perceived social support 1.01 [0.97, 1.05] .573 1.00 [0.97, 1.03] .913
Perceived control 0.93 [0.87, 0.99] .022 1.02 [0.96, 1.08] .611

Note. LVEF = left ventricular ejection fraction; NYHA = New York Heart Association.

Copyright © 2017 Taiwan Nurses Association. Unauthorized reproduction of this article is prohibited.
The Journal of Nursing Research Tsuey-Yuan Huang et al.

anorexia, and poor memory. As described, participants were

Figure 2. The KaplanYMeier survival curves of 1-year cardiac
outcome, by symptom cluster.
significantly higher event rates than the nonsymptom group
(ps = .016 and .001, respectively). In contrast, this study
found no difference between the survival rates of the typical
symptom and atypical symptom groups (p = .704).
We further compared the prognostic effect of symptom
clusters and 1-year cardiac outcome using the multivariable
Cox proportional hazard model, which was adjusted for
gender, age, and LVEF. The atypical symptom group had
a statistically higher level of outcome risk than the non-
symptom group (hazard ratio = 2.11, 95% CI [1.15, 3.88]).
Although the typical symptom group had a higher outcome
risk than the nonsymptom group, the difference was not
statistically significant (p = .061; see Table 3).
Discussion
Symptom Cluster of Patients With
Heart Failure
On the basis of the respective levels of distress that were
caused by the six symptoms, patients with HF complained
frequently of dyspnea, fatigue, leg edema, sleeplessness,
classified into three symptom cluster patterns: nonsymptom
group, typical symptom group, and nontypical symptom
group. Participants with relatively low levels of each symp-
tom were assigned to the nonsymptom group, participants
who reported moderate levels of all six symptoms were as-
signed to the typical symptom group, and participants who
reported low levels of dyspnea and fatigue and moderate
levels of the remaining four symptoms were assigned to the
nontypical symptom group. Patients with HF have multiple
symptoms because of decompensated, reduced cardiac
output. Some typical symptoms such as dyspnea and fatigue
are commonly observed and are similar to the trajectory of
the disease in patients with HF (Tsai et al., 2013). Patients
with HF and physicians usually are alert for these symptoms
and call for help soon after the symptoms become obvious.
Comparatively, leg edema, sleeplessness, anorexia, and poor
memory are not typical symptoms for HF for less stressful,
less specific, and less attentive sufferers. In this study,
symptom clusters reflect the intensity of symptoms as
experienced by the participants (Moser et al., 2014). The
nature of the three symptom cluster patterns is consistent
with the experiences of other patients in clinical settings.
Factors Associated With Symptom Clusters
Factors that this study found to differ significantly between
symptom clusters were living arrangements, left ventricular
dysfunction (LVEF G 30%), NYHA class, perceived depres-
sion, perceived anxiety, and perceived control. A significantly
larger number of participants who lived alone were allocated
to the typical group than to the other two groups; participants
with left ventricular dysfunction were predominately assigned
to the nontypical group, followed by the typical group and the
nonsymptom group; and fewer participants in the nontypical
group were NYHA class III or IV than their peers in the two
other groups.
Participants with NYHA class III or IV were most likely
to be in the typical symptom or nontypical symptom group,
with participants with NYHA class I or IV most likely to be

TABLE 3.
Prognostic Effect of Symptom Clusters and 1-Year Cardiac Outcome
(Multivariable Cox Proportional Hazard Regression)
Variable HR 95% CI p

Female gender 1.35 [0.80, 2.29] .263

Age Q 65 years 0.96 [0.59, 1.58] .873
LVEF G 30% 1.44 [0.85, 2.44] .178
Symptom clustera
Typical symptom 1.91 [0.97, 3.77] .061
Atypical symptom 2.11 [1.15, 3.88] .016

Note. LVEF = left ventricular ejection fraction; NYHA = New York Heart Association; HR = hazard ratio.
a
Reference category was the nonsymptom group.

Copyright © 2017 Taiwan Nurses Association. Unauthorized reproduction of this article is prohibited.
 Symptom Clusters in Patients With HF
in the nonsymptom group. This suggests that patients with
HF in the worst functional class tend to perceive a variety of
symptoms, including typical and nontypical ones. The NYHA
functional classification is a valid measure of functional
statusVa concept that is distinct from functional capacity
AQ5 and functional performance (Alhusseiny, Al-Nimer, Latif, &
Ibrahim, 2013). Patients with NYHA functional class III or
IV experience more distress from symptoms while perform-
ing daily life activities. In addition, symptoms due to de-
compensated cardiac function usually share the same pathway
of mechanism (Kirkova et al., 2011). Therefore, participants
with worse functional class levels experienced both types of
symptoms.
Participants who lived with someone were less likely to be
in the typical symptom group than those who lived alone. In
Taiwan, most patients with HF live with their family, with
only 14% living alone. Dyspnea is a life-threatening symptom
and a sign for patients with HF of disease aggravation,
whereas fatigue causes intolerance and disability in daily life.
Both of these symptoms should be closely monitored in
patients with HF who live alone.
The psychological status of participants in this study re-
lated significantly to symptom cluster assignment. The per-
ceived level of depression and anxiety was higher in the typical
group than the two other groups. In addition, the perceived
level of anxiety was higher in the nontypical group than the
nonsymptom group. Moreover, the perceived level of control
was lower in the typical group than the two other groups.
A higher level of perceived anxiety was associated with an
increased likelihood of being assigned to the typical symp-
tom group. In contrast, a higher level of perceived control
was associated with a decreased likelihood of being in the
typical symptom group. Psychosocial factors may affect
physical symptoms of patients with HF directly or indirectly
(Huang et al., 2013; Riegel et al., 2010). Anxiety has been
shown to enhance the association between clinical charac-
teristics and dyspnea in American patients (Huang et al.,
2013). Thus, higher depression and anxiety levels have less
LVEF effect on dyspnea. In other words, regardless of LVEF
status, patients with HF in the aforementioned study who
were relatively more depressed or more anxious experienced
more severe dyspnea (Huang et al., 2013). Patients with
higher perceived control tend to better recognize their typical
HF symptoms and pay more attention to the possibility of
condition change (Riegel et al., 2011; Yu et al., 2012). The
relationship between psychosocial factors and typical symp-
toms that were identified in this study confirms the results
that were obtained by previous investigators.
Symptom Clusters and Chronic Heart
Failure Prognosis
The atypical symptom group showed a higher risk than the
nonsymptom group. However, a borderline trend (p = .061)
was found in this study of a slightly higher risk in the typical
symptom group than the nonsymptom group, although the
sample size for symptom cluster was sufficient (Hair et al.,
Copyright © 2017 Taiwan Nurses Association. Unauthorized reproduction of this article is prohibited.
VOL. 00, NO. 0, MONTH 2017
2006). The predicting effect of atypical symptoms on the
rehospitalization and mortality of patients with HF was
thus confirmed in this study. Comparatively, leg edema,
sleeplessness, anorexia, and poor memory are not typical
symptoms for HF due to their being relatively less stressful,
less specific, and less attended to by healthcare providers
than dyspnea and fatigue. Although the effects of specific
symptoms on the outcomes of patients with HF have been
shown (Tsai et al., 2013), to our knowledge, this study is
the first to examine the relationship of symptom clusters in
terms of long-term outcomes in patients with HF. Thus, the
conclusions of this study are unique and provide a pre-
cursor for future research.
Limitations
This study was a cross-sectional survey. Therefore, the causal
relationship between related factors and symptom clusters
cannot be inferred. The physical symptoms were measured
using two different instruments. For more stringent and con-
sistent consideration, only one symptom measurement should
be used in future studies.
Conclusion
This was the first study to address symptom clusters based
on the intensity of symptom distress in patients rather than
on simply the types of symptoms and the predictive effect
of symptom cluster on the prognosis of patients with HF.
Furthermore, this study identified the critical symptom clusters
patterns and factors that are related to these symptoms in
patients with HF. On the basis of the findings of this study, it
is important to evaluate patients with HF in both typical and
atypical symptom clusters. In particular, atypical physical
symptoms such as leg edema, sleeplessness, anorexia, and
poor memory significantly influence the risk of 1-year cardiac
events in patients with HF. Both patients and healthcare
providers may use the information that is provided in this
study to improve their surveillance and management of
related symptoms.
References
Aktas, A., Walsh, D., & Rybicki, L. (2010). Symptom clusters:
Myth or reality? Palliative Medicine, 24(4), 373Y385. doi:10
.1177/0269216310367842
Alhusseiny, A. H., Al-Nimer, M. S. M., Latif, I. I., & Ibrahim, A. K.
(2012). Heart Failure: Discrepancy Between NYHA Function-
al Classification, Serum NT-pro Brain Natriuretic Peptide
and Ejection Fraction. European Journal of General Medi-
cine, 10(1), 26Y31.
Baggott, C., Cooper, B. A., Marina, N., Matthay, K. K., &
Miaskowski, C. (2012). Symptom cluster analyses based on
symptom occurrence and severity ratings among pediatric on-
cology patients during myelosuppressive chemotherapy. Cancer
Nursing, 35(1), 19Y28. doi:10.1097/NCC.0b013e31822909fd
Bruwer, B., Emsley, R., Kidd, M., Lochner, C., & Seedat, S.
(2008). Psychometric properties of the Multidimensional Scale
The Journal of Nursing Research
of Perceived Social Support in youth. Comprehensive Psychi-
atry, 49(2), 195Y201. doi:10.1016/j.comppsych.2007.09.002
Derogatis, S. (2006). Pearson assessmentVBrief symptom inven-
tory (BSI). Retrieved from http://www.pearsonclinical.com/
psychology/products/100000450/brief-symptom-inventory-bsi.
html?origsearchtext=100000450
Hair, J. F., Black, B., Babin, B., Anderson, R. E., & Tatham, R. L.
(2006). Multivariable data analysis. (6th ed.). Upper Saddle
River, NJ: Prentice-Hall.
Ho, C. C., Clochesy, J. M., Madigan, E., & Liu, C. C. (2007).
Psychometric evaluation of the Chinese version of the Minnesota
Living with Heart Failure Questionnaire. Nursing Research,
56(6), 441Y448. doi:10.1097/01.NNR.0000299849.21935.c4
Huang, T. Y., Moser, D. K., Hsieh, Y. S., Gau, B. S., Chiang, F. T.,
& Hwang, S. L. (2013). Moderating effect of psychosocial
factors for dyspnea in Taiwanese and American heart failure
patients. The Journal of Nursing Research, 21(1), 49Y58. doi:10
.1097/jnr.0b013e3182828d77
Huang, T. Y., Moser, D. K., Hsieh, Y. S., Lareau, S. C., Durkin, A. C.,
& Hwang, S. L. (2008). Validation of Chinese version of the
Modified Pulmonary Functional Status and Dyspnea Ques-
tionnaire with heart failure patients in Taiwan. American Journal
of Critical Care, 17(5), 436Y442.
Hwang, S. Y., Ahn, Y. G., & Jeong, M. H. (2012). Atypical
symptom cluster predicts a higher mortality in patients with
first-time acute myocardial infarction. Korean Circulation Journal,
42(1), 16Y22. doi:10.4070/kcj.2012.42.1.16
Kirkova, J., Aktas, A., Walsh, D., & Davis, M. P. (2011). Cancer
symptom clusters: Clinical and research methodology. Journal of
Palliative Medicine, 14(10), 1149Y1166. doi:10.1089/jpm.2010.0507
Kirkova, J., & Walsh, D. (2007). Cancer symptom clustersVA
dynamic construct. Supportive Care in Cancer, 15(9), 1011Y1013.
doi:10.1007/s00520-007-0259-2
Lareau, S. C., Carrieri-Kohlman, V., Janson-Bjerklie, S., & Roos, P. J.
(1994). Development and testing of the Pulmonary Functional
Status and Dyspnea Questionnaire (PFSDQ). Heart & Lung,
23(3), 242Y250.
Lareau, S. C., Meek, P. M., & Roos, P. J. (1998). Development and
testing of the modified version of the pulmonary functional
status and dyspnea questionnaire (PFSDQ-M). Heart & Lung,
27(3), 159Y168.
Lenz, E. R., Pugh, L. C., Milligan, R. A., Gift, A., & Suppe, F.
(1997). The middle-range theory of unpleasant symptoms: An
update. Advances in Nursing Science, 19(3), 14Y27.
Lenz, E. R., Suppe, F., Gift, A. G., Pugh, L. C., & Milligan, R. A.
(1995). Collaborative development of middle-range nursing
Copyright © 2017 Taiwan Nurses Association. Unauthorized reproduction of this article is prohibited.
Tsuey-Yuan Huang et al.
theories: Toward a theory of unpleasant symptoms. Advances
in Nursing Science, 17(3), 1Y13.
Lin, W. C., Huang, T. Y., Liu, C. Y., Yeh, M. L., Yu, C. H., & Hwang,
S. L. (2016). Validation of the Clinical COPD Questionnaire in
Taiwan, COPD. Journal of Chronic Obstructive Pulmonary
Disease, 13(3), 360Y366. doi:10.3109/15412555.2015.1094456
Moser, D. K., Lee, K. S., Wu, J. R., Mudd-Martin, G., Jaarsma, T.,
Huang, T. Y., I Riegel, B. (2014). Identification of symptom
clusters among patients with heart failure: An international
observational study. International Journal of Nursing Studies,
51(10), 1366Y1372 doi:10.1016/j.ijnurstu.2014.02.004
Moser, D. K., Riegel, B., McKinley, S., Doering, L. V., Meischke, H.,
Heo, S., I Dracup, K. (2009). The Control Attitudes Scale-
Revised: Psychometric evaluation in three groups of patients
with cardiac illness. Nursing Research, 58(1), 42Y51.
Mozaffarian, D., Benjamin, E. J., Go, A. S., Arnett, D. K., Blaha, M. J.,
Cushman, M., I American Heart Association Statistics
Committee and Stroke Statistics Subcommittee. (2015). Heart
disease and stroke statisticsV2015 update: A report from the
American Heart Association. Circulation, 131(4), e29Ye322.
doi:10.1161/cir.0000000000000152
Namisango, E., Harding, R., Katabira, E. T., Siegert, R. J.,
Powell, R. A., Atuhaire, L., I Taylor, S. (2015). A novel
symptom cluster analysis among ambulatory HIV/AIDS
patients in Uganda. AIDS Care, 27(8), 954Y963. doi:10.1080/
09540121.2015.1020749
Riegel, B., Dickson, V. V., Cameron, J., Johnson, J. C., Bunker, S.,
Page, K., & Worrall-Carter, L. (2010). Symptom recognition in
elders with heart failure. Journal of Nursing Scholarship,
42(1), 92Y100. doi:10.1111/j.1547-5069.2010.01333.x
Riegel, B., Lee, C. S., Albert, N., Lennie, T., Chung, M., Song, E. K.,
I Moser, D. K. (2011). From novice to expert: Confidence and
activity status determine heart failure self-care performance.
Nursing Research, 60(2), 132Y138. doi:10.1097/NNR
.0b013e31820978ec
Sarstedt, M., & Mooi, E. (2014). A concise guide to market
research. New York, NY: Springer.
Tsai, M. F., Hwang, S. L., Tsay, S. L., Wang, C. L., Tasi, F. C.,
Chen, C. C., & Huang, T. Y. (2013). Predicting trends in dyspnea
and fatigue in heart failure patients’ outcomes. Acta Cardiologica
Sinica, 29(6), 488Y495.
Yu, H. C., Yeh, M. L., Tsai, M. F., Hwang, S. L., Lennie, T. A.,
Moser, D. K., & Huang, T. Y. (2012). Ethnicity, functional class,
depression, anxiety, and perceived control are predictors of
health-related quality of life in patients with heart failure. Journal
of Cardiac Failure, 18(8), S84YS85. doi:10.1016/j.cardfail
.2012.06.513