776 Journal of Pain and Symptom Management Vol. 51 No.

4 April 2016

Brief Methodological Report

An Exploratory Factor Analysis of the Scale Structure of the

Patient Care Monitor Version 2.0
Gregory P. Samsa, PhD, Steven Wolf, MB, Thomas W. LeBlanc, MD, MA, and Amy P. Abernethy, MD, PhD
Center for Learning Health Care (G.P.S., S.W., T.W.L., A.P.A.), Duke Clinical Research Institute; and Departments of Biostatistics and
Bioinformatics (G.P.S., S.W.) and Department of Medicine (T.W.L., A.P.A.), Duke University School of Medicine, Durham, North Carolina,
USA

Abstract
Context. The Patient Care Monitor (PCM), version 2.0, is an electronic patient-reported outcomes instrument designed to
be embedded into oncology practices. One key psychometric component of an instrument is its factor structure.
Objectives. To validate the factor structure of the PCM.
Methods. The PCM was administered within various oncology clinics at our institution from 2006 to 2011 as part of
standard of care, yielding a large (n 5624) and diverse data set. An exploratory factor analysis was performed.
Results. The PCM performed well in terms of missing values and floor and ceiling effects. The three scales postulated by
the PCM developers exhibited high internal consistency (Cronbach alpha 0.94e0.95); the six subscales exhibited good
internal consistency (Cronbach alpha 0.80e0.95). A three-factor model approximated simple structure and was consistent
with the constructs of emotional function, physical function, and physical symptoms suggested by the PCM developers.
However, a six-factor model did not support the division of these three constructs into subscales of despair, distress,
ambulation, impaired performance, treatment side effects, and general physical symptoms. Instead, we observed an emotional
factor, a physical functioning factor, a factor including many of the treatment side effects, and three factors consisting of
various clusters of physical symptoms.
Conclusion. Although six subscales postulated by its developers perform reasonably, allocation of the PCM items to three
constructs is more accurate and likely more consistent with how symptoms and concerns are conceptualized by
patients. J Pain Symptom Manage 2016;51:776e783 2016 American Academy of Hospice and Palliative Medicine. Published by
Elsevier Inc. All rights reserved.

Key Words
Exploratory factor analysis, instrument development, instrument validation, psychometrics, oncology

Introduction symptoms, worsening symptoms, and the like. This in-

formation would then be acted on during the clinic
Patient-reported outcome measures describe infor-
visit.
mation that can best be reported by patientsdfor
The 78-item Patient Care Monitor (PCM), version
example, severity of symptoms and quality of life.
2.0, has special potential for real-time application as
Although traditional paper-based data collection re-
previously mentioned, having been originally de-
mains possible, in real-time applications, patient-
signed with the community oncology care provider
reported outcome data could be collected before the
in mind.1 The first version of the PCM, termed the
clinic visit (e.g., by tablet computer in a waiting area,
Cancer Care Monitor (CCM), was a 38-item instru-
via the web) and then forwarded to the clinician in
ment organized around three metaconstructs, each
the form of a report that highlights unresolved
of which was divided into two subscales: physical

Address correspondence to: Gregory P. Samsa, PhD, 11045 Hock Accepted for publication: November 21, 2015.
Plaza, Durham, NC 27705, USA. E-mail: samsa001@
mc.duke.edu

2016 American Academy of Hospice and Palliative Medicine. 0885-3924/$ - see front matter
Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jpainsymman.2015.11.013
Vol. 51 No. 4 April 2016 PCM Psychometric Validation 777

symptoms (subdivided into impaired ambulation and within this cohort to perform an examination of the
impaired performance); emotional issues (subdivided factor structure of the PCM.
into acute distress and despair); and physical symp-
toms (subdivided into general physical symptoms
and treatment side effects). For the CCM, the assign- Instruments and Data Collection
ment of items to subscales was based on a combination The PCM is based on the CCM but reflects substan-
of factor analysis and clinical reasoning. Although tial changes. Among these changes were the imposi-
most items loaded onto the anticipated subscales, tion of consistency on the answer set, rewording of
some loaded onto multiple subscales, and the subscale various items, and addition of new items. The PCM,
with the strongest loading was not always linked with version 2.0, used in this study comprises 78 items for
the item in question. Patients might or might not men and 86 items for women (i.e., the 78 items for
group these items in the same way as clinicians; for men plus an additional eight items). Most of the
example, regardless of whether they are caused by dis- new items pertained to physical symptoms, the inten-
ease progression or treatment side effects, patients tion being to assist clinicians by creating as inclusive
might potentially just lump all physical symptoms a set of symptoms as possible, whether the symptom
together as sources of misery. in question was common and whether it could be as-
The initial validation of the CCM included various signed to a subscale. Forty-seven items were assigned
reports.1e3 Although the PCM was based on the to subscales; 39 were not.
CCM, it is a substantially different instrument (dis- Patients completed the PCM in clinic waiting areas
cussed in the following) and thus merits validation using touch-screen technology (Fig. 1, available at
on its own. Our validation efforts are primarily jpsmjournal.com, presents an illustrative screen
focused on factor structure. shot). Patients also answered questions regarding de-
mographics, personal habits (e.g., smoking), and dis-
ease characteristics. Appendix I, available at
jpsmjournal.com, includes a column denoting the de-
Methods velopers recommendations for how the 86 items
The PCM was administered within various oncology could be assigned to the six putative subscales of the
clinics at our institution during the period from 2006 CCM. The items about hot flashes and vaginal/men-
to 2011, thus producing a large and diverse cohort of strual symptoms were only administered to female
patients. We used the first record from each patient patients.

Fig. 1. A typical screen shot from a PCM 2.0 item. PCM Patient Care Monitor.
778 Samsa et al. Vol. 51 No. 4 April 2016

Statistical Analysis Table 1

To evaluate the internal consistency of the six scales Demographics by Group
recommended by the PCMs developers, we assigned Included Excluded
Patients Patients
the appropriate items to those scales and then calcu- (n 5624) (n 1331)
lated Cronbach alpha coefficients.
% or % or
To evaluate the factor structure of the 78 items that Variable Mean  SD Mean  SD
were administered to patients regardless of gender, we
performed an exploratory factor analysis using orthog- Age at survey 59.6  12.7 65.2  12.0
Gender: female 58.3 55.2
onal rotation. We first used the pattern of eigenvalues Race/ethnicity: 15.7 19.9
to estimate the number of factors, and then consid- Black/African American
ered the pattern of factor loadings to assign items to Education
High School diploma or less 30.4 44.0
putative factors. In essence, we are repeating the orig- Associates, technical, some college 28.1 26.2
inal factor analysis1 (albeit with a notably different in- Bachelors degree 19.2 14.4
strument) to explore whether a similar factor Some grad school/graduate degree 22.3 15.4
Married/Partnered/Remarried 71.5 66.4
structure was observed. Initial Cancer Site
Noting the heterogeneity of benchmarks for factor Breast 13.5 11.9
loadings, our intention was to be inclusive by report- Lung or bronchus 11.2 15.0
Prostate 8.4 9.8
ing all factor loadings exceeding 0.30 (in absolute Other 25.7 28.8
value). When considering the pattern of factor load- Unknown/missing 41.2 34.5
ings, we interpreted the assignment of an item to a fac-
tor as unambiguous (i.e., producing simple
structure) if the factor loading exceeded 0.40 (in ab-
solute value) with no other loadings exceeding 0.30 Missing Values
(in absolute value). Typically, such loadings were Table 2 presents a frequency distribution of the
notably in excess of 0.40. number of missing items per patient, from the entire
6955 patient cohort; 5624 patients (80.9%) had com-
plete data. The modal number of missing items was
one.
Results One of the columns of Appendix I contains a count
Inclusions and Exclusions of the number of missing items for the 1331 patients
The PCM database contained records from 6965 who were excluded from the factor analysis. The
unique patients. (For patients with multiple visits, we only item with more than 10% missing (i.e., of the
used only the data from their first visit.) Ten records set of 6965 unique patients) was the one that referred
contained missing values for all PCM items and were to a paid job (n 822). The item with the second
deleted. Of the remaining 6955 patients, 5624 had re- highest level of missing values was the one that per-
sponses for all PCM items and thus comprise the pri- tained to reduced sexual enjoyment, interest, or per-
mary data set used in the factor analysis. formance (n 248). We hypothesize that some
patients skipped items that did not pertain to them.
We also hypothesize that the problem is not primarily
Demographic Characteristics the potentially sensitive nature of the items because
Table 1 summarizes demographic information for the employment item should be less sensitive than
the 5624 patients included in the factor analysis and the sexuality item yet contains more missing values.
also the 1331 who were excluded. Most included pa- The level of data completeness was generally
tients were white, married, and had at least a high encouraging.
school education. Various sites of cancer were repre-
sented and, indeed, one of the deficiencies of the de-
mographic database was that primary cancer site was Table 2
inconsistently recorded (e.g., because diagnostic infor- Number of Missing Items Per Patient (n 6955)
mation was patient reported, a patient with metastatic Frequency (Cumulative
breast cancer to bone may have entered breast can- Number of Missing Items Percentage)
cer or bone cancer as her primary diagnosis). 0 5624 (80.9)
The excluded patients were notably more likely to 1 619 (89.8)
be older and less educated. Nevertheless, there is 2 225 (93.0)
3 148 (95.1)
considerable overlap between the included and 4 87 (96.4)
excluded patients on all the demographic variables 5e9 142 (98.4)
that were available for analysis. >10 110 (100.0)
Vol. 51 No. 4 April 2016 PCM Psychometric Validation 779

Item Frequencies and Means model. The emotional items approximate simple
For the 5624 included patients, Appendix I also in- structure, in that the 11 items thought to represent
cludes a frequency distribution of the responses for this construct all have loadings onto factor three
each item. All items used the entire 0e10 range of exceeding 0.50. Moreover, none of these 11 items
their scale, suggesting that the scale is not too fine- have loadings exceeding 0.40 for the other two factors.
grained. Moreover, although the most common In addition, none of the other 67 items have loadings
response category for each of the items was 0, this is on factor three that exceed 0.40.
consistent with clinical observation. Ceiling and floor The physical functioning items also approximate
effects appear to be relatively modest, at most. simple structure. The 14 items thought to represent
In order from largest to smallest, the items with the this construct all have loadings onto factor one of at
highest mean values are heavy work, fatigue, running, least 0.50. Moreover, none of these 14 items have load-
sleeping at night, reduced sexual enjoyment, pain, ings exceeding 0.40 for the other two factors. Among
and housework. This pattern is consistent with clinical the other items, decrease in appetite also appeared to
observation.4,5 load onto factor one. Fatigue, daytime sleepiness,
pain, and weakness of body parts appeared to load
Internal Consistency onto factors one and two. These items describe gen-
Table 3 presents Cronbach alpha coefficients for the eral symptoms that could also affect functional status.
factor structure as hypothesized by the PCM devel- The physical symptom items also tended to cluster
opers. These coefficients for the six subscales ranged together. Loadings were relatively weaker than those
from 0.80 to 0.94, supporting their internal consis- for emotional and physical functioning, but a pattern
tency. Alpha coefficients for the three constructs that roughly approximated simple structure was
ranged from 0.94 to 0.95, strongly supporting the in- observed.
ternal consistency of the conceptual grouping of Taken as a whole, the three-factor model is notably
emotional items, physical functioning items, and phys- consistent with the PCM developers conceptualiza-
ical symptom items. In sum, the PCM subscales tion of the three underlying constructs of emotion,
perform well as designed; however, there might be physical function, and physical symptoms.
other ways to assign items to subscales that are equally
(if not more) in line with the viewpoint of the patient.

Six-Factor Model
Number of Factors
Table 5 presents the rotated factor loadings that
Appendix II, available at jpsmjournal.com, lists the
exceed 0.30 (in absolute value) for the six-factor
15 eigenvalues that exceeded unity. The first three
model. In the six-factor model, the 11 emotional items
and six factors account for 38% and 46% of the varia-
continued to exhibit simple structure, strongly
tion in the responses, respectively. Apart from being
loading onto factor two. However, a division of the
plausible cut points for the number of factors, these
items into distress and despair subscales was not
correspond to the division of the items into three con-
observed. In fact, when the 11 emotional items were
structs and six subscales, as conceptualized by the scale
entered into a separate factor analysis, only a single
developers and thus will be considered further.
eigenvalue exceeded unity, that factor being associ-
ated with 64.8% of variation in the responses and ex-
Three-Factor Model hibiting a similar pattern of factor loadings as in
Table 4 presents the rotated factor loadings that Tables 4 and 5.
exceed 0.30 (in absolute value) for the three-factor The 14 physical function items also continued to
exhibit simple structure, strongly loading onto factor
Table 3 one. As in the three-factor solution, items such as fa-
Cronbach Alpha Coefficients (Standardized) tigue, daytime sleepiness, pain, decrease in appetite,
Number and weakness of body parts behaved more ambigu-
Name Type of Items Alpha ously, being somewhat associated with factor one but
Emotional Construct 11 0.94 also with other factors as well. When the 14 physical
Despair Subscale 7 0.91 function items were entered into a separate factor
Distress Subscale 4 0.91
Physical functioning/ Construct 14 0.95 analysis, the largest eigenvalue was 8.69 (associated
functional status with 62.1% of the variance in responses) and the
Impaired ambulation Subscale 4 0.80 next largest eigenvalue was 1.06, suggesting that these
Impaired performance Subscale 4 0.90
Physical symptoms Construct 54 0.94 items form a single construct rather than being
General physical symptoms Subscale 15 0.84 divided into impaired ambulation and impaired
Physical side effects Subscale 13 0.81 performance.
780 Samsa et al. Vol. 51 No. 4 April 2016

Table 4 Table 4
Rotated Factor Loadings for the Three-Factor Solution Continued
Factor Factor Factor Factor Factor Factor
Category Item Name 1 2 3 Category Item Name 1 2 3

E1 Feel hopeless 0.80 S3 Dizziness/ 0.31 0.51

E1 Lost interest in people 0.73 lightheadedness
E1 Feel helpless 0.82 S3 Burning sensation 0.49
E1 Feel I would be 0.66 in hands or feet
better off dead S3 Swelling 0.43
E1 Feel guilty 0.67 S3 Fever 0.31
E1 Lost interest in 0.33 0.68 S3 Chills 0.45
pleasurable activities S3 Trouble seeing 0.48
E1 Feel worthless 0.74 S3 Dry eyes 0.46
E2 Crying/feel like crying 0.76 S3 Eyes tearing 0.48
E2 Worry 0.72 S3 Difficulty breathing 0.36 0.46
E2 Nervous, tense, anxious 0.71 S3 Heartburn 0.40
E2 Sad (depressed) 0.83 S3 Weakness of body parts 0.57 0.47
F1 Sit up 0.50 S3 Joint pain 0.51
F1 Stay out of bed 0.51 0.33 S3 Muscle aches 0.31 0.52
F1 Walk 0.72 S3 Hives 0.36
F1 Bathe/dress 0.69 S3 Nail changes 0.39
F2 Run 0.71 S3 Memory loss 0.51 0.30
F2 Light work or activity 0.82
F2 Heavy work or activity 0.77
F2 Function normally 0.78
F3 Housework 0.83 The remaining four factors involve the 53 physical
F3 Driving 0.75 symptoms. The items did not obviously divide them-
F3 Run errands 0.86
F3 Cook for self 0.81
selves into a single treatment side effects subscale
F3 Attend social activities 0.79 and a single general physical symptoms subscale.
F3 Attend paid job 0.67 Instead, the factors consisted of small numbers of
S1 Fatigue 0.52 0.44
S1 Daytime sleepiness 0.39 0.47 0.30
somewhat related items. Factor 4 was almost entirely
S1 Pain 0.46 0.43 composed of treatment side effects, although some
S1 Reduced sexual treatment side effects were not included in this factor.
enjoyment, interest
or performance
Factor 3 mostly but not entirely consisted of general
S1 Problems with urination 0.35 physical symptoms. Factor 5 mostly but not entirely
S1 Numbness 0.49 consisted of breathing-related symptoms. Factor 6
S1 Coughing 0.43
S1 Wheezing 0.44
mostly but not entirely consisted of skin-related
S1 Rash 0.35 symptoms.
S1 Dry skin 0.55 When these items were entered into a separate fac-
S1 Itching 0.47
S1 Trouble thinking 0.50 0.37
tor analysis, the largest eigenvalue was 13.75 (associ-
S1 Trouble sleeping 0.47 0.33 ated with 24.5% of the variance in responses) and
at night the second largest eigenvalue was 2.41. There was little
S1 Night sweat 0.39
S1 Day sweat 0.37
evidence that these items bifurcated into treatment
S2 Constipation 0.31 0.35 side effects and physical symptoms.
S2 Hair loss 0.34
S2 Dry mouth 0.33 0.53
S2 Change in taste of food 0.37 0.43
S2 Weight loss 0.39
S2 Weight gain 0.33 Discussion
S2 Sore throat 0.46
S2 Trouble swallowing 0.44 We used a large and diverse database to reconsider
S2 Nausea 0.33 0.42 the factor structure of the PCM. We found that alloca-
S2 Vomiting 0.30 tion of the items into the three constructs postulated
S2 Diarrhea 0.30
S2 Increase in appetite 0.36 by its developers exhibits excellent internal consis-
S2 Decrease in appetite 0.48 tency (alpha coefficients ranging from 0.94 to 0.95),
S3 New lump/mass is simple to analyze, and also makes good sense.
S3 Difficulty hearing 0.38
S3 Headache 0.50 We also found that the six subscales postulated by its
S3 Sinus problems 0.50 developers evidenced good to excellent internal con-
S3 Chest pain 0.41 sistency (alpha coefficients ranging from 0.80 to
S3 Rapid heart beat 0.43
S3 Bruising 0.39 0.94) and likely serve the purpose for which they are
S3 Easy bleeding 0.34 intended. However, the subscales from the six-factor
S3 Mouth sores/ulcers 0.39 model were not as postulated. Instead, we observed a
(Continued) unitary emotional factor, a unitary physical
Vol. 51 No. 4 April 2016 PCM Psychometric Validation 781

Table 5
Rotated Factor Loadings for the Six-Factor Solution
Category Item Name Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6

E1 Feel hopeless 0.79

E1 Lost interest in people 0.72
E1 Feel helpless 0.82
E1 Feel I would be better off dead 0.67
E1 Feel guilty 0.66
E1 Lost interest in pleasurable activities 0.31 0.66
E1 Feel worthless 0.74
E2 Crying/feel like crying 0.75
E2 Worry 0.71
E2 Nervous, tense, anxious 0.70
E2 Sad (depressed) 0.83
F1 Sit up 0.48
F1 Stay out of bed 0.48 0.30
F1 Walk 0.71
F1 Bathe/dress 0.68
F2 Run 0.71
F2 Light work or activity 0.80
F2 Heavy work or activity 0.76
F2 Function normally 0.76
F3 Housework 0.82
F3 Driving 0.74
F3 Run errands 0.85
F3 Cook for self 0.80
F3 Attend social activities 0.77
F3 Attend paid job 0.65
S1 Fatigue 0.48 0.35 0.30
S1 Daytime sleepiness 0.36 0.37
S1 Pain 0.44 0.44
S1 Reduced sexual enjoyment, interest or 0.33
performance
S1 Problems with urination
S1 Numbness 0.47
S1 Coughing 0.68
S1 Wheezing 0.69
S1 Rash 0.66
S1 Dry skin 0.57
S1 Itching 0.64
S1 Trouble thinking 0.35 0.42
S1 Trouble sleeping at night 0.47
S1 Night sweat 0.64
S1 Day sweat 0.65
S2 Constipation 0.30
S2 Hair loss
S2 Dry mouth 0.44
S2 Change in taste of food 0.30 0.59
S2 Weight loss 0.32 0.49
S2 Weight gain 0.46
S2 Sore throat 0.55
S2 Trouble swallowing 0.57
S2 Nausea 0.66
S2 Vomiting 0.61
S2 Diarrhea 0.35
S2 Increase in appetite 0.38
S2 Decrease in appetite 0.40 0.59
S3 New lump/mass
S3 Difficulty hearing 0.38
S3 Headache 0.37
S3 Sinus problems 0.42
S3 Chest pain 0.41
S3 Rapid heart beat 0.39
S3 Bruising 0.37 0.34
S3 Easy bleeding 0.42 0.30
S3 Mouth sores/ulcers 0.32
S3 Dizziness/lightheadedness 0.32
S3 Burning sensation in hands or feet 0.41 0.34
S3 Swelling 0.31 0.31
S3 Fever 0.38
S3 Chills 0.45
(Continued)
782 Samsa et al. Vol. 51 No. 4 April 2016

Table 5
Continued
Category Item Name Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6

S3 Trouble seeing 0.31

S3 Dry eyes 0.32
S3 Eyes tearing 0.32
S3 Difficulty breathing 0.69
S3 Heartburn 0.33
S3 Weakness of body parts 0.55 0.38
S3 Joint pain 0.60
S3 Muscle aches 0.31 0.58
S3 Hives 0.57
S3 Nail changes 0.41
S3 Memory loss 0.40 0.34

functioning factor, a factor including most but not all precepts of learning health care systems6 and its clin-
of the treatment side effects items, and three factors ical relevance for day-to-day cancer care and quality
consisting of various small clusters of physical monitoring. Additional validation efforts (e.g., ad-
symptoms. dressing criterion validity, test-retest reliability
To interpret, discrimination between concepts like responsiveness to change) for the PCM are
treatment side effects and general physical symptoms recommended.
may be comprehensible to physicians providing care We also recommend that future analyses that use
for patients but, for the patient, clinical symptoms the PCM as an outcome variable might be triaged ac-
such as painful sores in the mouth are symptoms, cording to application. Patient care focusing on spe-
regardless of whether the cause is chemotherapy, cific symptoms should use individual items. Analyses
the cancer itself, or a viral coinfection. Distress and focusing on clinically derived constructs such as
despair are concepts that may be distinct to psychol- distress versus despair can be performed with confi-
ogists, but, for the cancer patient, the experience dence using the six subscales. Patient-centric analyses,
(and reporting) is similar. And, distinguishing be- such as those pertaining to quality of life and related
tween impaired function and impaired ambulation constructs, should probably use the three constructs
may not be necessary for the person with cancer. In of emotion, physical function, and physical symptoms,
sum, the three-factor model is simpler and does not as this approach is simpler, has clearer face validity,
require the analyst, researcher, clinician, or patient and is more consistent with the experience of pa-
to distinguish between concepts that are difficult to tients. Doing so has the additional advantage of being
separate and moreover might be conceptualized at consistent with how clinicians conceptualize out-
least somewhat differently according to ones comes, when considered at the level of the broad
perspective. construct.
Some limitations of our analysis should be noted.
Patients with missing values on any of the items were
excluded, even if only a single item was missing. To Disclosures and Acknowledgments
avoid methodological problems associated with multi-
This research received no specific funding/grant
ple observations on the same patient, we selected the
from any funding agency in the public, commercial,
first record for each patient, although this could tend
or not-for-profit sectors. The authors declare no con-
to differentially select patients that were either early in
flicts of interest.
their disease course or one-time consults. On the
other hand, our database was larger (on the order of
magnitude of 10 times) than previous PCM validation
efforts. References
Our reexamination of the scale structure of the 1. Fortner B, Okon T, Schwartzberg L, Tauer K,
PCM was primarily motivated by general principles Houts AC. The Cancer Care Monitor: psychometric con-
of instrument developmentdparticularly, that the tent evaluation and pilot testing of a computer adminis-
tered system for symptom screening and quality of life in
changes from the CCM to the PCM were sufficiently
adult cancer patients. J Pain Symptom Manage 2003;26:
extensive as to suggest that additional validation of 1077e1092.
the PCMs putative factor structure would be benefi-
2. Abernethy AP, Zafar SY, Uronis H, et al. Validation of the
cial. In general, our results were supportive of the Patient Care Monitor (version 2.0): a review of system assess-
PCM, an instrument that is noteworthy in that its ment instrument for cancer patients. J Pain Symptom
development was consistent with many of the Manage 2010;40:545e558.
Vol. 51 No. 4 April 2016 PCM Psychometric Validation 783

3. Fortner B, Baldwin S, Schwartzberg L, Houts AC. Valida- 5. Reese JB, Shelby RA, Abernethy AP. Sexual concerns in
tion of the Cancer Care Monitor items for physical symp- lung cancer patients: an examination of predictors and
toms and treatment side effects using expert oncology moderating effects of age and gender. Support Care Cancer
nurse evaluation. J Pain Symptom Manage 2006;31: 2011;19:161e165.
207e214. 6. Smith M, Saunders R, Stuckhardt L, Mcginnis JM, eds.
4. LeBlanc TW, Nickolich M, Rushing CN, et al. What Best care at lower cost: The path to continuously learning
bothers lung cancer patients the most? A prospective, longi- health care in America. Washington, DC: National Acade-
tudinal electronic patient-reported outcomes study in mies Press, 2012. Available at: http://iom.edu/Reports/
advanced non-small cell lung cancer. Support Care Cancer 2012/Best-Care-at-Lower-Cost-The-Path-to-Continuously-Lea
2015;23:3455e3463. rning-Health-Care-in-America.aspx
783.e1 Samsa et al. Vol. 51 No. 4 April 2016

Appendix I
Frequency Distribution of Items (n 5624 for All Columns Except the Number of Missing Items)
Item Missing
Number Category Item Name (n 1331) 0 1 2 3 4 5 6 7 8 9 10 Mean SD

36 E1 Feel hopeless 61 4354 315 206 235 113 126 109 64 27 32 43 0.81 1.90
40 E1 Lost interest in people 59 4798 186 143 167 73 73 71 44 25 19 25 0.53 1.58
42 E1 Feel helpless 57 4233 322 235 238 111 145 117 74 48 53 48 0.93 2.06
43 E1 Feel I would be better off 74 5207 125 63 56 33 38 32 24 7 16 23 0.27 1.22
dead
44 E1 Feel guilty 67 4879 236 127 117 50 76 47 29 15 23 25 0.44 1.45
45 E1 Lost interest in pleasurable 64 4338 321 239 235 100 130 100 51 30 37 43 0.80 1.88
activities
46 E1 Feel worthless 66 4863 194 116 128 58 86 66 35 24 25 29 0.50 1.58
37 E2 Crying/feel like crying 59 4050 472 278 268 117 159 108 69 37 28 38 0.92 1.92
38 E2 Worry 54 2642 731 491 542 275 320 243 123 67 90 100 1.90 2.53
39 E2 Nervous, tense, anxious 50 2801 740 427 493 262 296 238 125 75 85 82 1.79 2.49
41 E2 Sad (depressed) 56 3828 484 344 304 140 170 143 64 44 58 45 1.09 2.10
1 F1 Sit up 97 5043 105 99 106 60 71 48 37 17 24 14 0.40 1.42
2 F1 Stay out of bed 103 4890 136 126 149 73 96 61 29 21 25 18 0.49 1.52
3 F1 Walk 69 3764 284 285 343 201 232 200 123 76 75 41 1.35 2.37
4 F1 Bathe/dress 68 4766 163 142 146 80 99 96 43 34 29 26 0.60 1.71
5 F2 Run 266 2948 283 183 281 203 265 260 234 172 258 537 2.83 3.65
6 F2 Light work or activity 91 3800 256 250 318 175 234 176 129 85 102 99 1.45 2.57
7 F2 Heavy work or activity 179 2500 345 297 357 247 326 356 266 203 322 405 3.06 3.52
8 F2 Function normally 112 3575 287 288 324 193 320 210 121 97 111 98 1.61 2.64
26 F3 Housework 171 3326 286 288 330 201 273 243 154 121 145 257 2.03 3.07
27 F3 Driving 170 4516 157 145 144 69 117 102 59 36 77 202 1.03 2.51
83 F3 Run errands 162 4031 215 201 240 136 184 156 103 61 98 199 1.39 2.72
84 F3 Cook for self 165 4422 176 145 187 94 133 113 78 57 66 153 1.04 2.43
85 F3 Attend social activities 133 4343 168 163 227 103 147 119 84 55 75 140 1.09 2.44
86 F3 Attend paid job 822 4197 85 69 94 69 104 115 119 89 183 500 1.81 3.45
24 S1 Fatigue 24 1663 531 575 548 435 622 560 300 212 110 68 3.02 2.77
28 S1 Daytime sleepiness 46 3011 608 475 502 262 272 228 102 68 64 32 1.60 2.29
29 S1 Pain 55 2866 404 362 401 310 309 334 224 169 128 117 2.20 2.89
33 S1 Reduced sexual enjoyment, 248 3292 309 243 304 157 251 206 208 129 150 375 2.21 3.29
interest or performance
34 S1 Problems with urination 48 4537 268 169 169 111 108 112 55 27 38 30 0.71 1.82
35 S1 Numbness 53 3687 424 315 321 200 204 203 108 68 64 30 1.29 2.26
56 S1 Coughing 36 3771 571 315 285 158 173 144 76 55 41 35 1.09 2.06
57 S1 Wheezing 39 4709 257 158 154 90 73 87 40 23 20 13 0.54 1.54
73 S1 Rash 51 4982 177 98 108 70 61 59 28 21 13 7 0.39 1.33
74 S1 Dry skin 39 3375 570 420 390 199 254 197 94 50 51 24 1.35 2.17
76 S1 Itching 47 4443 317 234 211 102 113 96 40 28 23 17 0.68 1.67
78 S1 Trouble thinking 44 3594 485 412 420 172 192 160 79 50 35 25 1.19 2.05
80 S1 Trouble sleeping at night 43 2367 669 485 491 329 372 340 230 129 128 84 2.30 2.74
81 S1 Night sweat 52 3909 419 306 286 164 146 148 82 57 54 53 1.10 2.16
82 S1 Day sweat 54 4509 269 190 190 113 120 97 43 34 23 36 0.71 1.79
25 S2 Constipation 43 3843 428 354 308 155 164 152 79 49 48 44 1.11 2.11
30 S2 Hair loss 57 4533 289 151 155 66 99 88 51 37 54 101 0.80 2.10
31 S2 Dry mouth 37 3631 492 341 345 189 203 162 100 59 61 41 1.26 2.22
32 S2 Change in taste of food 34 4074 248 244 245 155 191 161 113 70 64 59 1.16 2.31
47 S2 Weight loss 107 4109 288 213 237 149 227 170 88 57 44 42 1.08 2.18
48 S2 Weight gain 182 4082 353 259 245 124 186 177 62 55 39 42 1.02 2.09
54 S2 Sore throat 41 4763 276 149 153 78 62 56 30 18 20 19 0.48 1.45
55 S2 Trouble swallowing 35 4791 207 143 145 74 71 63 41 24 28 37 0.54 1.64
59 S2 Nausea 36 4116 407 241 265 130 146 130 70 44 41 34 0.94 1.99
60 S2 Vomiting 42 5160 127 64 88 36 37 33 26 18 18 17 0.30 1.26
61 S2 Diarrhea 45 4626 270 193 185 87 105 71 38 15 15 19 0.57 1.54
62 S2 Increase in appetite 49 4724 218 168 171 99 89 73 29 14 16 23 0.54 1.53
63 S2 Decrease in appetite 42 4304 230 223 222 125 148 127 79 57 55 54 0.96 2.13
9 S3 New lump/mass 157 5160 93 34 55 39 62 29 31 31 37 53 0.41 1.62
10 S3 Difficulty hearing 40 4233 335 226 224 148 158 128 57 47 39 29 0.89 1.96
11 S3 Headache 65 4024 448 288 273 148 159 124 59 40 38 23 0.94 1.92
12 S3 Sinus problems 64 3595 504 342 408 189 218 179 73 48 40 28 1.22 2.10
13 S3 Chest pain 68 4766 209 147 168 81 83 67 38 27 26 12 0.53 1.54
14 S3 Rapid heart beat 53 4668 274 198 188 90 95 66 19 10 13 3 0.50 1.36
15 S3 Bruising 71 4328 325 257 231 101 147 112 37 34 31 21 0.78 1.79
16 S3 Easy bleeding 75 4694 264 148 186 57 103 74 36 24 20 18 0.55 1.56
17 S3 Mouth sores/ulcers 34 5056 199 92 89 40 55 40 19 12 9 13 0.31 1.19
18 S3 Dizziness/lightheadedness 49 3926 495 326 313 150 157 130 54 26 30 17 0.94 1.84
(Continued)
Vol. 51 No. 4 April 2016 PCM Psychometric Validation 783.e2

Appendix I
Continued
Item Missing
Number Category Item Name (n 1331) 0 1 2 3 4 5 6 7 8 9 10 Mean SD

19 S3 Burning sensation in hands 50 4668 192 150 157 108 88 98 56 49 33 25 0.67 1.80
or feet
20 S3 Swelling 55 4482 217 172 184 113 140 114 78 50 45 29 0.82 1.97
49 S3 Fever 106 5229 143 73 52 31 44 34 7 1 3 7 0.20 .93
50 S3 Chills 74 4729 233 176 166 84 101 75 17 17 14 12 0.51 1.45
51 S3 Trouble seeing 58 4275 393 273 247 114 113 100 40 26 26 17 0.75 1.70
52 S3 Dry eyes 48 4306 399 271 213 132 113 90 47 24 21 8 0.71 1.64
53 S3 Eyes tearing 39 4470 394 231 200 88 91 61 41 24 14 10 0.59 1.50
58 S3 Difficulty breathing 34 3757 456 304 299 191 194 168 93 80 47 35 1.20 2.20
64 S3 Heartburn 37 4004 465 328 242 164 180 116 54 19 28 24 0.91 1.84
68 S3 Weakness of body parts 47 2811 519 445 485 294 311 299 182 122 90 66 1.98 2.62
69 S3 Joint pain 40 2857 514 438 393 310 347 304 191 121 92 57 1.97 2.63
70 S3 Muscle aches 39 3127 481 419 414 247 296 254 153 105 86 42 1.72 2.50
75 S3 Hives 58 5423 62 35 32 13 14 19 11 8 4 3 0.12 .77
77 S3 Nail changes 51 4945 188 133 104 49 68 50 28 22 21 16 0.41 1.40
79 S3 Memory loss 49 3781 536 394 356 144 145 116 59 48 27 18 1.00 1.88
21* S3 Menstrual pain/cramping 139 3034 24 27 24 12 10 12 5 6 3 4 0.15 .88
22* S3 Vaginal pain/itching 115 2901 103 44 51 19 15 17 9 2 3 1 0.23 .94
23* S3 Hot flashes/flushes 129 1924 296 167 201 113 119 123 80 45 39 58 1.54 2.53
65* S3 Vaginal dryness 113 2493 157 93 120 66 64 69 49 21 17 15 0.79 1.90
66* S3 Vaginal bleeding 116 3065 30 19 18 4 13 8 3 3 1 0 0.10 .65
67* S3 Vaginal discharge 117 2963 88 33 27 15 14 14 6 2 2 0 0.17 .81
71* S3 Breast tenderness 110 2424 187 142 146 73 72 51 28 14 17 11 0.77 1.75
72* S3 Nipple discharge 113 3115 17 11 9 3 3 4 2 0 1 0 0.04 .43
E1 emotional construct, despair subscale; E2 emotional construct, distress subscale; F1 physical functioning construct, impaired ambulation subscale;
F2 physical functioning construct, impaired performance subscale; F3 physical functioning construct, not assigned to a subscale; S1 physical symptom
construct, general physical symptoms subscale; S2 physical symptom construct, treatment side effects subscale; S3 physical symptom construct, not assigned
to a subscale.
Items that were administered to female patients only (n 3935) are denoted by an asterisk.

Appendix II
Eigenvalues >1 and Proportion of Variation Explained
Factor Eigenvalue Cumulative Proportion

1 22.29 0.286
2 4.15 0.339
3 3.56 0.385
4 2.38 0.415
5 1.86 0.439
6 1.80 0.462
7 1.51 0.482
8 1.47 0.500
9 1.34 0.517
10 1.27 0.534
11 1.23 0.550
12 1.17 0.565
13 1.15 0.579
14 1.09 0.606
15 1.03 0.619