Vol. 41 No.

3 March 2011 Journal of Pain and Symptom Management 535

Original Article

Development and Validation

of an Instrument for Rapidly Assessing
Symptoms: The General Symptom Distress
Scale
Terry A. Badger, PhD, PMHCNS-BC, RN, FAAN, Chris Segrin, PhD,
and Paula Meek, PhD, RN, FAAN
College of Nursing (T.A.B.) and Department of Communication (C.S.), The University of Arizona,
Tucson, Arizona; and College of Nursing (P.M.), University of Colorado Anschutz Health Science
Center, Aurora, Colorado, USA

Abstract
Context. Symptom assessment has increasingly focused on the evaluation of
total symptom distress or burden rather than assessing only individual symptoms.
The challenge for clinicians and researchers alike is to assess symptoms, and to
determine the symptom distress associated with the symptoms and the patient’s
ability for symptom management without a lengthy and burdensome assessment
process.
Objectives. The objective of this article was to discuss the psychometric
evaluation of a brief general symptom distress scale (GSDS) developed to assess
specific symptoms and how they rank in relation to each other, the overall
symptom distress associated with the symptom schema, and provide an assessment
of how well or poorly that symptom schema is managed.
Methods. Results from a pilot study about the initial development of the GSDS
with 76 hospitalized patients are presented, followed by a more complete
psychometric evaluation of the GSDS using three samples of cancer patients
(n ¼ 190) and their social network members, called partners in these studies
(n ¼ 94). Descriptive statistics were used to describe the GSDS symptoms,
symptom distress, and symptom management. Point biserial correlations indexed
the associations between dichotomous symptoms and continuous measures, and
conditional probabilities were used to illustrate the substantial comorbidities of
this sample. Internal consistency was examined using the KR-20 coefficient, and
test-retest reliability was examined. Construct validity and predictive validity also
were examined.
Results. The GSDS demonstrated satisfactory internal consistency and test-
retest reliability, and good construct validity and predictive validity. The total score
on the GSDS, symptom distress, and symptom management correlated
significantly with related constructs of depression, positive and negative affect,

Address correspondence to: Terry A. Badger, PhD, RN, Accepted for publication: June 1, 2010.
College of Nursing, The University of Arizona,
1305 N. Martin, Tucson, AZ 85721-0203, USA.
E-mail: [email protected]

Ó 2011 U.S. Cancer Pain Relief Committee 0885-3924/$ - see front matter
Published by Elsevier Inc. All rights reserved. doi:10.1016/j.jpainsymman.2010.06.011
536 Badger et al. Vol. 41 No. 3 March 2011

and general health. The GSDS was able to demonstrate its ability to distinguish
between those with or without chronic illness, and was able to significantly predict
scores on criterion measures such as depression.
Conclusion. Collectively, these results suggest that the GSDS is a straightforward
and useful instrument for rapidly assessing symptoms that can disrupt health-
related quality of life. J Pain Symptom Manage 2011;41:535e548. Ó 2011 U.S.
Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved.

Key Words
Symptoms, symptom distress, instrumentation, psychometric evaluation

Introduction treatment.5 These symptoms can be pervasive, se-

vere, and debilitating. Cancer-related fatigue is
The assessment of symptoms has increasingly
the most frequently reported side effect across
focused on the evaluation of total symptom
cancer diagnosis, disease stage, and treatment
distress or burden rather than assessing only
regimen, ranging from 40% to 100%.6,7 In a re-
individual symptoms.1 Patients rarely present
cently published study, 534 patients with various
with just one symptom, and often it is the pa-
forms of advanced cancer were asked what were
tient’s distress associated with the symptoms
the ‘‘most important symptoms or concerns to
experienced coupled with their abilities to man-
monitor.’’8 Fatigue emerged as the top-rated
age these distressing symptoms that prompt
symptom and was ranked most important for
patients to seek care.2,3 The challenge for clini-
10 of the 11 cancer diagnoses. Psychological
cians and researchers alike is to assess the symp-
symptoms are also frequently reported. Between
toms relevant to their particular population,
10% and 25% of women with breast cancer have
and to determine the symptom distress associ-
a concurrent episode of major depressive disor-
ated with the symptoms and patients’ ability to
der,9 and among cancer patients more generally,
manage their symptoms without a lengthy and
16%e26% experience significant depressive
burdensome assessment process.4
symptoms.10 Clinically significant symptoms of
In certain settings, an in-depth and multidi-
anxiety in cancer survivors range from 20% to
mensional assessment of key symptoms such as
50%.11,12 These anxiety symptoms are remark-
depression and fatigue is indicated. However,
ably persistent, remaining elevated for periods
there are other clinical and research contexts
of one to five years after diagnosis.13
where it is desirable to secure rapid and straight-
Findings such as these demonstrate the need
forward assessments of diverse symptoms with
for rapid assessment and monitoring of a wide
minimal response burden. Although such brief
range of symptoms experienced by patients in
self-report assessments are not a substitute for
both clinical and research settings, especially
more detailed evaluation of patients’ symptoms,
when there is a focus on quality of life. The as-
they can play an important role in clinical set-
sessment of symptoms among cancer patients
tings to efficiently identify patients with com-
has been shown to improve physician-patient
promised quality of life who are in need of
communication about chronic symptoms.14
further assessment and intervention. In longitu-
When feedback about these symptoms was pro-
dinal research, brief symptom checklists can
vided to patients, many in the Velikova et al.’s14
play a vital role in gathering symptom informa-
investigation evidenced improved quality of life
tion with minimal response burden. This is an
and emotional functioning.
especially important feature of any instrument
Although there have been several instruments
designed for use with populations known to
developed to assess multiple symptoms,15,16
suffer from fatigue or difficulty concentrating.
these instruments tend to remain focused on
The treatment of cancer is frequently associ-
an individual’s symptoms and not the relation-
ated with a multitude of symptoms such as fa-
ship of each symptom to each other or the pa-
tigue and difficulty concentrating that adversely
tient’s distress about the symptoms. No brief
affects quality of life. The abundance of symp-
instrument to date asks individuals to rank their
toms stems from both the disease itself and its
Vol. 41 No. 3 March 2011 Development and Validation of the GSDS 537

symptoms neither when compared with other half were male, and the majority was non-
symptoms nor to rate the overall distress and Hispanic white (80%). About half were married,
management ability of their particular symptom and the majority reported at least some college.
schema. The purpose of this article was to dis- Most participants worked full time or part time.
cuss the psychometric evaluation of a brief gen-
eral symptom distress scale (GSDS) developed Methods in Pilot Study. Following admission
to assess specific symptoms and how they rank and consent, participants were given a question-
in relation to each other, the overall symptom naire packet that included the GSDS, Symptom
distress associated with their symptom schema, Distress Scale (SDS),17 Medical Outcomes
and provide an assessment of how well or poorly Study Short-Form-36 (MOS SF-36),18 and Cen-
that symptom schema is managed. The initial ter for Epidemiological Studies-Depression
development of the GSDS and results from a pi- Scale19 (CES-D). A subset of participants
lot study are presented, followed by a more com- (n ¼ 31) was also re-tested on the GSDS 24
plete psychometric evaluation of the GSDS hours postadmission and 24 hours postdi-
using three samples of cancer patients. scharge to determine test-retest reliability using
the intraclass correlation coefficient (ICC).
Pilot Study of the GSDS Construct validity was also examined using
The GSDS was developed to assess symptoms these questionnaires. Research assistants
in chronically ill populations, rank symptoms picked up the questionnaire packets after the
compared with each other, and to assess general participants completed them. The average
symptom distress and general ability to manage time between admission baseline and the 24
symptoms as a brief symptom assessment instru- hours postadmission assessment was three days
ment. The GSDS has three sections. In the first and between admission and postdischarge five
section, the respondent is presented with a list days.
of 12 symptoms (fatigue, sleep difficulties, anxi-
ety, pain, difficulty concentrating, depression, Results from Pilot Study. Participants averaged
shortness of breath, nausea, vomiting, bowel 5.90 (standard deviation [SD] ¼ 2.75) symp-
problems, loss of appetite, cough) that are com- toms, with fatigue (M ¼ 1.79, SD ¼ 2.1) ranked
mon among people diagnosed with illnesses. as the most distressing symptom followed by
Respondents simply indicate whether they have pain (M ¼ 1.87, SD ¼ 2.0). Test-retest reliability
the symptom. Next, respondents return to the was ICC 0.76 (P < 0.001) for symptom distress
list of checked symptoms and are asked to rank and ICC 0.62 (P < 0.01) for symptom manage-
them from most distressing (rank of 1) to least ment. The ICC for the individual symptoms
distressing. Lower scores indicate greater dis- ranged from 0.45 to 0.87, with loss of appetite
tress associated with the symptom. In the second being the least and cough being the most stable.
section of the GSDS, participants are asked to These values were considered reasonable given
provide a global rating of how distressing their the rapidly changing symptoms in these acute
symptoms are on a 10-point scale, with higher care patients. Validity was partially supported
scores indicating more distress. In the third using a similar measure, the 13-item SDS,15,17
part, participants are asked to provide a global with the correlation between the GSDS and
rating of how well they are able to manage SDS of r ¼ 0.75 (P < 0.01), indicating a strong
their symptoms on a 10-point scale, with higher relationship with the GSDS. Construct validity
scores indicating better ability to manage symp- was further supported by correlating the dis-
toms (Appendix). tress measure with related constructs such as
the global health item from the MOS SF-3618
Participants in Pilot Study. The GSDS was pilot (r ¼ 0.41, P < 0.05), and depression (CES-
tested with a sample of 76 hospitalized inpatients D19) (r ¼ 0.58, P < 0.01), indicating that with in-
with a variety of chronic conditions (e.g., cardio- creased symptom distress, there was decreased
vascular disease, pulmonary disease, and can- perceived general health and increased depres-
cer) who had been admitted to the hospital for sion. The ability to manage symptoms was nega-
treatment. Average length of stay on these tively correlated with the CES-D (r ¼ 0.53,
medical-surgical units is about three days. Partic- P < 0.05), indicating that as the ability to man-
ipants ranged in age from 18 to 80þ years, about age symptoms increased, depression decreased,
538 Badger et al. Vol. 41 No. 3 March 2011

further supporting construct validity. Our pre- Spanish-speaking breast cancer and English-
liminary pilot study findings provided basic speaking prostate cancer samples completed
support for the reliability and validity of the the GSDS. Further information on the rela-
self-report GSDS as a brief GSDS. tionship of these 94 partners to the cancer sur-
vivors appears on the bottom of Table 1.

Methods Procedure
Participants On enrollment in any of the three cancer
The analyses presented in this report are studies, all the participants completed a base-
based on data collected from three different line assessment (T1) in which all of the mea-
samples of breast or prostate cancer survivors sures described below were administered over
(total n ¼ 190) who resided in the community. the telephone by a trained data collector. All
None of these participants were hospitalized at of the results reported herein are based on
the time of the study. The first sample con- the baseline data of the cancer survivors and
sisted of 96 English-speaking women with their partners, taken before any intervention
breast cancer. The second sample consisted of started, unless otherwise noted. Immediately
38 Spanish-speaking women with breast cancer. after the baseline assessment, participants
The third sample included 56 men with prostate were randomly assigned to one of several inter-
cancer. All the participants were recruited to ventions that lasted from six to eight weeks.
participate in a clinical trial testing the effective- Immediately after completion of the interven-
ness of an educational, interpersonal counsel- tion, the same set of measures was completed
ing, or exercise intervention developed to over the telephone (T2). Finally, there was
enhance and maintain quality of life for people a follow-up assessment three to four months af-
recently or currently in treatment for cancer. ter the baseline (T3) in which the measures
Details of these interventions have been pre- were again completed over the telephone. In
sented elsewhere20,21 and only those proce- the sample of English-speaking women with
dures and measures relevant to the current breast cancer (n ¼ 96), the T1eT2 attrition
analyses are presented. Participants were re- rate was 3% and the T1eT3 attrition rate was
cruited from regional cancer centers, oncolo- 5%. Final attrition rates for the Spanish-
gists’ offices, regional Veterans Administration speaking sample of women with breast cancer
medical centers, cancer support groups, and cannot yet be assessed as the investigation is
research study websites. Enrollment was open still ongoing, but for men with prostate cancer
to any person diagnosed with breast or prostate the attrition rate has been less than 10%. In
cancer who was currently undergoing treat- these analyses, T1eT3 data from 38, 22, and
ment or who had completed treatment within 18 Spanish-speaking women with breast cancer
the prior six months. The sample of English- cases, respectively, are included. Similarly,
speaking women with breast cancer was re- T1eT3 data from 56, 46, and 42 prostate can-
stricted to those with Stages IeIII disease, and cer cases, respectively, are included. All partic-
the sample of Spanish-speaking women with ipants received a gift card to one of several
breast cancer required that the participant national retail stores after each of the three as-
spoke Spanish as a primary language. Additional sessments to thank them for their time.
information on the demographics, disease state,
and treatment regimens of the participants ap- Measures
pears in Table 1. All of the measures described below were ad-
All the participants were asked to nominate ministered over the telephone. The measures
a partner, defined as any member of their so- were translated into Spanish for the Spanish-
cial network whether or not related by blood speaking breast cancer sample. All other par-
or marriage who played an important role in ticipants completed the measures in English.
their recovery, to participate along with them We used an adaptation of Brislin’s22 transla-
in the investigation. Partners also were ran- tion/back translation process23,24 to translate
domly assigned to one of the interventions, the GSDS into Spanish. This approach blends
yoked to that of the patient, with identical pro- symmetrical and asymmetrical approaches to
cedures and measures. Only partners in the provide for flexibility while maintaining the
Vol. 41 No. 3 March 2011 Development and Validation of the GSDS 539

Table 1
Participant Demographics, Disease State, and Treatments
English-Speaking Breast Spanish-Speaking Breast English-Speaking Prostate
Participant Characteristics Cancer Survivors (n ¼ 96) Cancer Survivors (n ¼ 38) Cancer Survivors (n ¼ 56)

Mean age (SD) 54.11 (10.60) 48.95 (10.56) 67.54 (9.95)

Race/ethnicity (%)
Black 0 0 7
Latina/o 14 100 7
White 85 0 86
Other/unknown 1 0 0
Currently married (%) 73 66 80
Education (%)
<High school 2 29 2
High school 19 29 16
Vocational/technical 34 18 29
College 23 16 32
Postgraduate 22 8 21
Cancer stage (%)
I 33 21 13
II 53 43 7
III 14 29 9
IV 0 7 9
Unknowna 0 0 62
Treatmentb (%)
Chemotherapy 75 82 16
Radiation 54 42 64
Hormone 64 11 57
Surgery 51 84 91
Other chronic illnessc (%) 28 47 77
Partner relationship (%)
Spouse/significant other 41 84
Son/daughter 24 3
Other family 19 2
Friend 5 4
Other 11 7
Partner mean age (SD) 45.24 (14.04) 61.58 (11.73)
a
63% of the prostate cancer survivors did not know their disease stage, but 61% knew their Gleason score, which was on average 6.28 (SD ¼ 1.50).
b
Currently receiving or completed within the past six months. Percentages do not add up to 100 because of multiple treatments.
c
For example, diabetes, arthritis, heart disease, respiratory illness.

goal of functional and cultural equivalence be- cancer were 0.86 or greater for T1eT3. Respec-
tween the original written English versions and tive reliabilities for Spanish-speaking women
the Spanish-language versions. The GSDS, for with breast cancer were a $ 0.91. T1eT3 reli-
example, was translated into Spanish by three abilities for men with prostate cancer were
cultural experts and then back translated into a $ 0.89.
English. Next, a focus group of six Spanish-
speaking Latinas discussed the translation in
terms of understandability (language level and Negative and Positive Affects. Negative and pos-
complexity), use of idioms, and consistency of itive affects were measured by the 20-item Pos-
meaning across measures. Finally, the Spanish itive and Negative Affect Schedule (PANAS).26
version was field tested with 10 Latinas. Scores range from 10 to 50, with higher scores
reflecting greater positive or negative affect.
The PANAS has been used extensively with
Depression. Symptoms of depression were mea- general populations and cancer patients with
sured using the 20-item CES-D.19 The CES-D has satisfactory reliability and validity.27,28 In this
been used in numerous studies with general study, PANAS (þ) reliabilities at T1eT3 were
and cancer populations with satisfactory reli- a $ 0.84 for English-speaking women with
ability and validity.25 Cronbach’s alphas in this breast cancer, a $ 0.82 for Spanish-speaking
study for English-speaking women with breast women with breast cancer, and a $ 0.86 for
540 Badger et al. Vol. 41 No. 3 March 2011

men with prostate cancer. Reliabilities for symptom distress, and symptom management
PANAS () at T1eT3 were a $ 0.88 among items to measures of related constructs that in-
English-speaking women with breast cancer, cluded depression (CES-D), positive affect,
a $ 0.90 for Spanish-speaking women with negative affect, and general health. The gen-
breast cancer, and a $ 0.90 for men with pros- eral health measure was only available in the
tate cancer. English-speaking breast cancer subsample.
To evaluate the predictive validity of the
General Health. General health was assessed GSDS, a series of hierarchical regression analy-
with the global health item from the MOS SF- ses were conducted. For each regression model,
3629 in the English-speaking women with breast the measure (e.g., depression) was treated
cancer only. This instrument, developed for the as the dependent variable. On the first step of
Medical Outcomes Study, has been used exten- the analysis, the corresponding T1 criterion
sively as a measure of health-related quality of variable (e.g., depression) was entered. On
life throughout the health sciences. For the the second step, the GSDS measure was en-
global health rating, participants responded to tered. This procedure was repeated for each
the statement ‘‘In general would you say your criterion variable, and for each of the three
health is .’’ followed by five response options GSDS measures (the total number of symp-
that ranged from ‘‘excellent’’ to ‘‘poor.’’ Higher toms, symptom distress, and symptom manage-
scores indicate better self-rated health. This ment). These analyses test if the GSDS can
item is commonly used as an overall indicator significantly predict scores on a measure
of perceived general health. assessed at a subsequent point in time.31 The
final method for establishing construct validity
Symptom Distress and Symptom Management. used the known-group technique, which in-
Symptom distress and symptom management volves administration of the measure to groups
were assessed with the GSDS. Complete ver- that are known to differ in ways that should be
sions of the scale in the English and Spanish captured by the instrument.32 More advanced
appear in the Appendix. disease stage is generally associated with poorer
quality of life and higher symptom prevalence
Analytic Strategies and burden in breast and prostate cancer pa-
Descriptive statistics were used to describe the tients.33,34 We had data on cancer stage for
GSDS symptoms, symptom distress, and symp- 145 participants and compared these different
tom management. Point biserial correlations staged groups on the GSDS from the baseline
indexed the associations between dichotomous assessment.
symptoms and continuous measures such as
global ratings of symptom distress and manage-
ment. Conditional probabilities were used to
illustrate the substantial comorbidities of this Results
sample. Descriptive Statistics
Descriptive statistics for the GSDS symptoms
Reliability Analysis. Internal consistency was (fatigue, sleep difficulties, anxiety, pain, diffi-
examined using the Kuder-Richardson 20 coeffi- culty concentrating, depression, shortness of
cient,30 which is analogous to Cronbach’s alpha, breath, nausea, vomiting, bowel problems, loss
because the GSDS is composed of dichotomous of appetite, cough) appear in Table 2. Symp-
variables. Test-retest reliability was examined. toms experienced most frequently by the cancer
The interval for the test-retest for the cancer survivors were fatigue, sleep difficulties, pain,
survivors and partners was considered reason- anxiety, and difficulty concentrating. The most
able based on the anticipated length of time of distressing symptoms by mean rank were pain,
the intervention.31 Interclass correlation coeffi- vomiting, nausea, anxiety, and depression
cients were examined using a random one-way (Table 2). These ranks indicate that although
model. nausea and vomiting were relatively rare among
these respondents, when they did occur they
Validity Analyses. Construct validity was exam- tended to be ranked among the most distressing
ined by correlating the GSDS total score, the symptoms, along with pain.
Vol. 41 No. 3 March 2011 Development and Validation of the GSDS 541

Table 2
Descriptive Statistics for General Symptom Distress Scale Items
Correlated with Correlated with
Symptom Frequency (%) Mean Ranka Symptom Distressb Symptom Managementc

Depression 35.4 3.16 0.33d 0.24e

Anxiety 46.7 3.12 0.35d 0.32d
Fatigue 68.8 4.40 0.30d 0.22e
Shortness of breath 27.9 5.05 0.18f 0.11
Nausea 21.2 2.91 0.19f 0.15f
Vomiting 5.3 2.83 0.18f 0.15f
Pain 48.7 2.64 0.21e 0.22e
Sleep difficulties 56.6 4.31 0.26d 0.24e
Bowel problems 32.8 4.15 0.14c 0.05
Difficulty concentrating 41.3 3.81 0.36d 0.20e
Loss of appetite 17.5 4.68 0.12 0.18f
Cough 15.5 4.67 0.10 0.07
a
Lower ranks correspond to most distressing.
b
Values are point biserial r.
c
P ¼ 0.06.
d
P < 0.001.
e
P < 0.01.
f
P < 0.05.

Participants reported an overall average various symptoms of the GSDS appears in

symptom distress of 5.11 (SD ¼ 2.55) on a 10- Table 4. The values in Table 4 are phi coeffi-
point scale, and an average of 7.24 (SD ¼ 2.13) cients, which are a specialized version of corre-
on how well they were able to manage their lation when both variables are dichotomous.
symptoms on a 10-point scale. These results in Phi is interpreted the same as Pearson’s r. As
the two right-most columns of Table 2 show noted previously, 94 partners of the English-
that difficulty concentrating, anxiety, depres- speaking prostate cancer or Spanish-speaking
sion, and fatigue were the symptoms that breast cancer patients also completed the
are most strongly associated with the overall GSDS. Among these partners, the GSDS inter-
symptom distress item. These are also the same nal consistency reliability was a ¼ 0.79.
symptoms that are generally most negatively
correlated with participants’ ratings of their
ability to manage their symptoms (Table 2). Test-Retest Reliability
Conditional probabilities among symptoms Participants completed the GSDS scale at
in the GSDS are presented in Table 3. These baseline before any intervention, six to eight
conditional probabilities are arrayed row weeks later (T2) and then six to eight weeks
wise. So for example, looking at Row 1, partic- after that (T3). The GSDS total score demon-
ipants who indicated that they had depression strated satisfactory test-retest reliability for
had a 0.78 probability of also having anxiety, T1eT2 (r ¼ 0.72, P < 0.001) and T2eT3
a 0.91 probability of having fatigue, and (r ¼ 0.67, P < 0.001). It is noteworthy that the
a 0.69 probability of having sleep difficulties. test-retest reliability did not decrease signifi-
Numerous high probabilities for any given cantly when comparing either six- or eight-
symptom are indicative of substantial comor- week intervals. The ICC for individual symptoms
bidity. Extreme caution must be exercised for T2eT3, for example, ranged from 0.45 to
when interpreting the conditional probabili- 0.87, with pain being the least stable and dif-
ties associated with the vomiting symptom as ficulty concentrating being the most stable
that had a baseline occurrence of only 5% among women with breast cancer.
in our cancer participants. A subset of 94 partners also completed the
GSDS, although T2 data were available from
only 66 partners and T3 data were available
Internal Consistency Reliability from 60 partners. Among these partners, the
The internal consistency reliability of the GSDS also exhibited satisfactory test-retest reli-
GSDS symptoms was a ¼ 0.75. A correlation ability for the T1eT2 (r ¼ 0.60, P < 0.001) and
matrix that illustrates associations among the T2eT3 (r ¼ 0.64, P < 0.001) intervals.
542 Badger et al. Vol. 41 No. 3 March 2011

Table 3
Conditional Probabilities Among Symptoms in the General Symptom Distress Scale
Index Symptom 1 2 3 4 5 6 7 8 9 10 11 12

1. Depression d 0.78 0.91 0.40 0.28 0.08 0.57 0.69 0.45 0.63 0.21 0.15
2. Anxiety 0.59 d 0.84 0.31 0.28 0.07 0.59 0.73 0.41 0.57 0.25 0.19
3. Fatigue 0.47 0.57 d 0.37 0.28 0.08 0.60 0.69 0.41 0.55 0.22 0.16
4. Shortness of breath 0.52 0.52 0.92 d 0.37 0.14 0.64 0.75 0.50 0.58 0.21 0.31
5. Nausea 0.48 0.63 0.90 0.48 d 0.25 0.65 0.75 0.50 0.65 0.35 0.30
6. Vomiting 0.50 0.60 1.00 0.70 1.00 d 0.90 0.60 0.80 0.60 0.60 0.30
7. Pain 0.41 0.57 0.85 0.36 0.28 0.10 d 0.70 0.42 0.54 0.22 0.22
8. Sleep difficulties 0.43 0.60 0.84 0.36 0.28 0.06 0.60 d 0.41 0.53 0.21 0.20
9. Bowel problems 0.48 0.58 0.86 0.42 0.32 0.13 0.63 0.71 d 0.52 0.27 0.21
10. Difficulty concentrating 0.54 0.64 0.92 0.37 0.33 0.08 0.64 0.73 0.41 d 0.24 0.18
11. Loss of appetite 0.42 0.68 0.85 0.33 0.42 0.18 0.61 0.67 0.52 0.58 d 0.18
12. Cough 0.35 0.57 0.72 0.55 0.41 0.10 0.69 0.72 0.45 0.48 0.21 d
Note. Table values represent the conditional probability of experiencing the column-wise symptom, given the row-wise index symptom.

Construct Validity management scores exhibited evidence of con-

The correlations between the GSDS total struct validity, correlating significantly, and in
score, symptom distress, and symptom man- the predicted direction with depression, nega-
agement items with depression, positive affect, tive affect, and positive affect. However, the
negative affect, and general health appear in GSDS symptom distress score was not signifi-
Table 5. All GSDS measures were significantly cantly associated with any of the three criterion
correlated with the related constructs in the variables among partners.
predicted direction. More symptoms were asso-
ciated with higher depression, higher negative Predictive Validity
affect, lower positive affect, and lower ratings Table 7 documents that the GSDS total sum
of general health. The global symptom distress was a statistically significant predictor in three
rating was similarly correlated with these con- of the six tests of predictive validity, and was
structs, and the global symptom management marginally significant in a fourth test. The
scores were all correlated in the opposite direc- symptom distress and symptom management
tion (e.g., greater ability to manage symptoms scores were similarity predictive of changes in
was associated with lower depression). the measures, with statistically significant or
The partners who completed the GSDS also marginally significant trends, in six of the 12
completed measures of depression, negative tests. The high construct stability of the de-
affect, and positive affect. Correlations among pression, positive affect, and negative affect
these variables for partners appear in Table 6. measures as indicated by their large bs make
The GSDS total symptom score and symptom these rather stringent tests of predictive

Table 4
Inter-item Symptom Phi Correlations in the General Symptom Distress Scale
Symptoms 1 2 3 4 5 6 7 8 9 10 11 12

1. Depression d
2. Anxiety 0.46a d
3. Fatigue 0.37a 0.31b d
4. Shortness of breath 0.21b 0.07 0.31a d
5. Nausea 0.13 0.17c 0.24b 0.23b d
6. Vomiting 0.07 0.06 0.16c 0.23b 0.46a d
7. Pain 0.11 0.19b 0.34a 0.18c 0.17c 0.20b d
8. Sleep difficulties 0.18c 0.30a 0.38a 0.23b 0.19b 0.02 0.26a d
9. Bowel problems 0.19b 0.16c 0.25b 0.23b 0.19b 0.24b 0.20b 0.20b d
10. Difficulty concentrating 0.32a 0.30a 0.43a 0.21b 0.25b 0.09 0.26a 0.28a 0.15c d
11. Loss of appetite 0.07 0.19c 0.16c 0.06 0.24b 0.27a 0.11 0.09 0.18c 0.15c
12. Cough 0.01 0.10 0.03 0.26a 0.17c 0.10 0.17c 0.14 0.11 0.06 0.04 d
Note. Table values are Phi coefficients.
a
P < 0.001.
b
P < 0.01.
c
P < 0.05.
Vol. 41 No. 3 March 2011 Development and Validation of the GSDS 543

Table 5
Correlations Between GSDS Scales and Related Constructs in Cancer Patients
Variables CES-D Positive Affect Negative Affect General Health (SF-12)a

GSDS 0.62b 0.39b 0.52b 0.42b

Symptom distress 0.54b 0.36b 0.49b 0.32a
Symptom management 0.43b 0.45b 0.36b 0.37b
Note. n ¼ 180e189, except for general health, where data were from the English-speaking breast cancer subsample only (n ¼ 96).
a
P < 0.01.
b
P < 0001.

validity. After controlling for T1 variance, (M ¼ 3.27, SD ¼ 2.80) on the GSDS than those
there is little residual variance left to be ex- with no chronic health condition (M ¼ 1.57,
plained by any other predictor. It is, therefore, SD ¼ 1.93), t(87) ¼ 3.32, P < 0.001, once again
reasonable to consider marginally significant providing evidence of known-groups validity for
trends (P ¼ 0.06e0.09) for the GSDS as favor- the GSDS.
able evidence of the instrument’s predictive
validity. The strongest and most consistent pre-
diction occurred when depression was used as
the measure.
Discussion
This investigation was designed to evaluate
the utility and psychometric properties of
Known-Groups Validity a newly developed instrument for rapidly as-
The GSDS exhibited a linear association sessing symptoms experienced by patients: the
with disease stage, with means (SD) for those GSDS. Conditional probabilities associated
at Stages IeIV being 3.51 (2.65), 4.39 (2.57), with each GSDS symptom suggest that the expe-
4.81 (2.77), and 6.57 (3.55), respectively. rience of any given symptom on the scale is asso-
This statistically significant difference across ciated with a substantial likelihood of having
groups, F(3,141) ¼ 3.10, P < 0.05, shows that any of the other symptoms. Intercorrelations
the GSDS can discriminate between cancer among the symptoms indicate associations of
patients of different stages, as predicted. weak to moderate magnitude. The GSDS
Among 89 partners of the cancer patients with symptom total score and the global ratings of
available data, 45 indicated that they had at least symptom distress and symptom management
one chronic health condition from a list that demonstrated construct validity, predictive val-
included heart disease, diabetes, chronic ob- idity, and know-groups validity. Collectively,
structive pulmonary disease/respiratory illness, these results suggest that the GSDS is a straight-
arthritis, stroke, high blood pressure/hyperten- forward and useful instrument for rapidly as-
sion, and ‘‘other,’’ and 62% of these 45 partners sessing symptoms in cancer patients that can
indicated that they had more than one chronic disrupt health-related quality of life.
health condition. The remaining 44 partners The data on GSDS symptom prevalence
indicated no chronic health condition. An inde- (Table 2) corroborate other published findings
pendent samples t-test was conducted to com- showing that fatigue, sleep difficulties, pain,
pare those partners with and without a chronic and anxiety are among the most common symp-
health condition on the GSDS. Those partners toms experienced by cancer patients.35e38
with a chronic health condition scored higher These were also among the symptoms that

Table 6
Correlations Between GSDS Scales and Related Constructs in Cancer Patients’ Partners
Variables CES-D Positive Affect Negative Affect

GSDS 0.40 a
0.48 a
0.29b
Symptom distress 0.18 0.01 0.17
Symptom management 0.37b 0.44a 0.31b
Note. n ¼ 70e89. Data are from partners of English-speaking prostate cancer and Spanish-speaking breast cancer samples only.
a
P < 0.001.
b
P < 0.01.
544 Badger et al. Vol. 41 No. 3 March 2011

Table 7
Multiple Regression Results for Tests of the Predictive Validity of the GSDS Scales
Psychological State Criterion Variables

T2 T2 T2 T3 T3 T3

Predictor Variable CES-D þAffect Affect CES-D þAffect Affect

a a a a a
T1 criterion variable 0.58 0.58 0.64 0.41 0.50 0.54a
GSDS 0.17b 0.13b 0.01 0.21b 0.10 0.13c
T1 criterion variable 0.61a 0.58a 0.63a 0.46a 0.47a 0.59a
Symptom distress 0.12c 0.13c 0.03 0.14c 0.15b 0.03
T1 criterion variable 0.59a 0.61a 0.63a 0.44a 0.49a 0.56a
Symptom management 0.20d 0.05 0.06 0.22d 0.09 0.10
Note. Table values are standardized regression coefficients (b). T1eT2 interval: six to eight weeks. T1eT3 interval: three to four months.
CES-D ¼ Center for Epidemiological Studies-Depression scale.
a
P < 0.001.
b
P < 0.05.
c
P ¼ 0.06e0.09.
d
P < 0.01.

correlated most strongly with participants’ rat- in this heterogeneous sample, it is important
ings of symptom distress and symptom manage- to recognize that symptom profiles could easily
ment. It is understandable that these particular be influenced by a number of these variables.
symptoms are so frequent, as they have obvious For example, people currently receiving che-
causal relationships with each other (e.g., pain motherapy might be considerably more in-
or anxiety / sleep difficulties / fatigue). Un- clined to experience nausea and vomiting
fortunately, for many cancer patients these are than people who have either recently com-
not only frequently experienced, but also associ- pleted or never been exposed to chemotherapy.
ated with high levels of distress and symptom Also, the clustering of symptoms may become
management difficulties. weaker as more time elapses from completion
Survivors’ self-rated self-efficacy and subse- of treatment.41 This is particularly important
quent ability to manage their symptoms were when interpreting the associations among vari-
both concurrently and prospectively associated ous symptoms that appear in Table 4. Certain
with quality-of-life outcomes. We found cancer symptoms that might be associated (e.g., bowel
patients’ perceptions of the ability to manage problems and fatigue) could reflect a common
symptoms significantly associated with self- causal factor (e.g., radiotherapy for prostate
efficacy scores (r ¼ 0.28, P ¼ 0.04) to (r ¼ 0.48, cancer patients). Presumably the association
P < 0.001), with this relatively well-educated between these two symptoms could dissipate
sample demonstrating high self-efficacy scores. after completion of treatment.
This is consistent with related research showing In addition to cancer patients, the GSDS was
that improvements in symptom management given to a subset of the cancer patients’ part-
skills and abilities can substantially reduce suf- ners. About half of these partners were ex-
fering in cancer patients.39 Montazeri40 argued periencing a chronic health condition of their
that recognition and management of these own. The GSDS clearly distinguished those part-
symptoms is a critical issue in cancer care be- ners with a chronic health condition from those
cause of their ability to impair health-related who did not have one. Also, the GSDS total sum
quality of life. and symptom management scores demon-
The utility of the GSDS as a relatively ge- strated construct validity among the partners,
neric measure of symptoms was demonstrated correlating as predicted with measures of de-
by administering the GSDS to a diverse group pression, negative affect, and positive affect.
of survivors diagnosed with either breast or These findings indicate that the GSDS may
prostate cancer, who spoke English or Spanish, have utility for populations other than cancer
and who were either in treatment or had re- patients. Many major illnesses that have a multi-
cently completed treatment. Although the in- variate symptom profile such as respiratory ill-
strument had good psychometrical qualities ness and heart disease could have symptom
Vol. 41 No. 3 March 2011 Development and Validation of the GSDS 545

expressions that are adequately indexed by VA Health Care System, Tucson; and Joanne
the GSDS. Harrington, PhD, RN, Phoenix VA Health Care
System. We thank the Arizona Cancer Center,
Recommendations for Use of the GSDS and Phoenix and Southern Arizona VA Health
Participants experienced no difficulties in Care Systems that served as recruitment sites for
completing the scale either as a self-report in these studies.
the pilot study or over the telephone in any of The authors declare no conflicts of interest.
the three cancer studies. Participants experi-
enced no difficulties having items read to
them and responding with the appropriate References
scales. Given the straightforward nature of the
1. National Institutes of Health State of the Science
GSDS, there is reason to believe that most adults
Panel. Symptom management in cancer: pain, de-
with at least modest reading skills could easily pression, and fatigue. J Natl Cancer Inst 2003;95:
complete the instrument however adminis- 1110e1117.
tered. Self-report vs. researcher-administered 2. Miaskowski C, Aquizaret BE, Dodd M,
versions of the instrument only require minor Cooper B. Conceptual issues in symptoms clusters
rewording of some of the instructions. research and their implication for quality of life as-
sessment in patients with cancer. J Natl Cancer Inst
Monogr 2007;37:39e46.
Conclusion 3. Miaskowski C, Dodd M, Lee K. Symptom clus-
The GSDS is a brief measure that clinicians ters: the new frontier in symptom management re-
search. J Natl Cancer Inst Monogr 2004;32:17e21.
and researchers alike can use to assess symp-
toms relevant to their particular population, 4. Walke LM, Byers AL, McCorkle R, Fried TR.
Symptom assessment in community-dwelling older
to determine symptom distress associated
adults with advanced chronic disease. J Pain Symp-
with the symptoms, and to assess the patients’ tom Manage 2006;31:31e37.
ability to manage their symptoms without
5. Fan G, Filipczak L, Chow E. Symptom clusters
a lengthy and burdensome assessment process. in cancer patients: a review of the literature. Curr
This evaluation demonstrates that the GSDS is Oncol 2007;14:173e179.
a reliable and valid measure for rapid assess-
6. Mock V. Evidenced-based treatment for cancer-
ment of symptoms, how symptoms rank in rela- related fatigue. J Natl Cancer Inst Monogr 2004;
tion to each other, and of symptom distress 32:112e118.
and the ability to manage symptoms in both 7. Mock V, Pickett M, Ropka ME, et al. Fatigue and
English-speaking and Spanish-speaking cancer quality of life outcomes of exercise during cancer
and chronically ill populations. treatment. Cancer Pract 2001;9:119e127.
8. Butt Z, Rosenbloom SK, Abernethy AP, et al.
Fatigue is the most important symptom for ad-
Disclosures and Acknowledgments vanced cancer patients who have had chemother-
Funding for the pilot study was provided by apy. J Natl Compr Canc Netw 2008;6:448e455.
a Vice President’s Research Award, University of 9. Fann JR, Thomas-Rich AM, Katon WJ, et al.
Arizona, Tucson, AZ. Collaborators in the pilot Major depression after breast cancer: a review of ep-
study were Paula Meek, PhD, RN, FAAN, Joyce idemiology and treatment. Gen Hosp Psychiatry
Verran, PhD, RN, FAAN, and Judith Effken, 2008;30:112e126.
PhD, RN, FAAN, University of Arizona, Tucson, 10. Stommel M, Kurtz ME, Kurtz JC, Given CW,
The clinical site was the University Medical Cen- Given BA. A longitudinal analysis of the course of
ter, Tucson. Funding for the cancer studies using depressive symptomatology in geriatric patients
with cancer of the breast, colon, lung or prostate.
the GSDS was provided by the Oncology Nursing Health Psychol 2004;23:564e573.
Foundation, National Institute of Nursing Re-
11. Osborne RH, Elsworth GR, Hopper JL. Age-
search, Lance Armstrong Foundation, and specific norms and determinants of anxiety and
National Cancer Institute to Terry Badger. Collab- depression in 731 women with breast cancer re-
orators on the cancer studies were Ana Maria cruited through a population-based cancer registry.
Lopez, MD, Karen Weihs, MD, and Richard Eur J Cancer 2003;39:755e762.
Ahmann, MD, all from the Arizona Cancer Cen- 12. Stark D, Kiely M, Smith A, et al. Anxiety disor-
ter, Tucson; Maria Bishop, MD, Southern Arizona ders in cancer patients: their nature, associations,
546 Badger et al. Vol. 41 No. 3 March 2011

and relation to quality of life. J Clin Oncol 2002;14: 28. Melvin GA, Molloy GN. Some psychometric
3137e3148. properties of the positive and negative affect sched-
13. Drageset S, Lindstrom TC. The mental health ule among Australian youth. Psychol Rep 2000;86:
of women with suspected breast cancer: the relation- 1209e1212. Available from http://jcc.sagepub.com/
ship between social support, anxiety, coping and de- cgi/external_ref?access_num¼000088334500023&
fense in maintaining mental health. J Psychiatr link_type¼ISI. Accessed November 5, 2010.
Ment Health Nurs 2003;10:401e409. 29. Ware JE Jr, Sherbourne CD. The MOS 36-item
Short-Form health survey (SF-36): I. Conceptual
14. Velikova G, Booth L, Smith AB, et al. Measuring
quality of life in routine oncology practice improves framework and item selection. Med Care 1992;30:
473e483.
communication and patient well-being: a random-
ized controlled trial. J Clin Oncol 2004;22:714e724. 30. Nunnally J, Bernstein I. Psychometric theory,
3rd ed. New York: McGraw-Hill, 1994.
15. McCorkle R. The measurement of symptom dis-
tress. Semin Oncol Nurs 1987;3:48e56. 31. Knapp TR, Brown JK. Focus on psychometrics:
ten measurement commandments that often should
16. Watanbe S, Nekolaichuk C, Beaumont C,
be broken. Res Nurs Health 1995;18:465e469.
Mawani A. The Edmonton symptom assessment
systemdwhat do patients think? Support Care Can- 32. Polit DF, Beck CT. Nursing research: Principles
cer 2009;17:675e683. and methods, 7th ed. Philadelphia, PA: Lippincott,
Williams, & Wilkins, 2003.
17. McCorkle R, Young K. Development of a symp-
tom distress scale. Cancer Nurs 1983;1:373e378. 33. Gwede CK, Small BJ, Munster PN,
Andrykowski MA, Jacobson PB. Exploring the differ-
18. Johnson PA, Goldman L, Orav EJ, et al. Com-
ential experience of breast cancer treatment-related
parison of the Medical Outcomes Study Short-
symptoms: a cluster analytic approach. Support
Form 36-Item health survey in black patients and
Care Cancer 2008;16:925e933.
white patients with acute chest pain. Med Care
1995;33:145e160. 34. Rosenfeld B, Roth AJ, Gandhi S, Penson D. Dif-
ferences in health-related quality of life of prostate
19. Radloff LS. The CES-D scale: a self-report de-
cancer patients based on stage of cancer. Psychoon-
pression scale for research in the general popula-
cology 2004;13:800e807.
tion. Appl Psychol Meas 1977;1:385e401.
35. Chang VT, Hwang SS, Feuerman M, Kasimis BS,
20. Badger T, Segrin C, Dorros SM, Meek P,
Thaler HT. The memorial symptom assessment
Lopez AM. Depression and anxiety in women with
scale short form (MSAS-SF). Cancer 2000;89:
breast cancer and their partners. Nurs Res 2007;
1162e1171.
56:44e53.
36. Chen ML, Lin CC. Cancer symptom clusters:
21. Badger T, Segrin C, Meek P, Lopez AM,
a validation study. J Pain Symptom Manage 2007;
Bonham E. Profiles of women with breast cancer:
34:590e599.
who responds to a telephone interpersonal counseling
intervention? J Psychosoc Oncol 2005;23:79e100. 37. Portenoy RK, Thaler HT, Kornblith AB, et al.
The Memorial Symptom Assessment Scale: an in-
22. Brislin RW. Back-translation for cross-cultural
strument for the evaluation of symptom prevalence,
research. J Cross Cult Psychol 1970;1:185e216.
characteristics and distress. Eur J Cancer 1994;30:
23. Jones E, Kay M. Instrumentation in cross-cultural 1326e1336.
research. Nurs Res 1992;41:186e188.
38. Stein KD, Denniston M, Baker F, et al. Valida-
24. Jones PS, Lee JW, Phillips LR, Zhang XW, tion of a modified Rotterdam Symptom Checklist
Jaciedo KB. An adaptation of Brislin’s translation for use with cancer patients in the United States.
model for cross-cultural research. Nurs Res 2001; J Pain Symptom Manage 2003;26:975e989.
50:300e304.
39. Benor DE, Delbar V, Krulik T. Measuring im-
25. Visser MR, Smets EMA. Fatigue, depression and pact of nursing intervention on cancer patients’
quality of life in cancer patients: how are they re- ability to control symptoms. Cancer Nurs 1998;21:
lated? Support Care Cancer 1998;6:101e108. 320e334.
26. Watson D, Clark L, Tellegan A. Development 40. Montazeri A. Health-related quality of life in
and validation of brief measures of positive and neg- breast cancer patients: a bibliographic review of
ative affect: the PANAS scale. J Pers Soc Psychol the literature from 1974 to 2007. J Exp Clin Cancer
1988;54:1063e1070. Res 2008;27:32.
27. Manne S, Schnoll R. Measuring cancer patients’ 41. Kim HJ, Barsevick AM, Tulman L, McDermott PA.
psychological distress and well-being: a factor analytic Treatment-related symptom clusters in breast cancer:
assessment of the Mental Health Inventory. Psychol a secondary analysis. J Pain Symptom Manage 2008;
Assess 2001;13:99e109. 36:468e479.
Vol. 41 No. 3 March 2011 Development and Validation of the GSDS 547

Appendix

General Symptom Distress Scale

548 Badger et al. Vol. 41 No. 3 March 2011

Escala del Sı ntoma de la Angustia General (GSDS)