Journal ofCemnlology: MEDICAL SCIENCES In the Public Domain

1998, Vol. 53A, No. 5, M35I-M360

Symptom Severity of Osteoarthritis of the Knee:

A Patient-Based Measure Developed
in the Veterans Health Study
Jack A. Clark,12 Avron Spiro III,23 Graeme Fincke,45 Donald R. Miller,12 and Lewis E. Kazis12

'Center for Health Quality, Outcomes, and Economic Research,

Edith Nourse Rogers Memorial Veterans Hospital, Bedford, Massachusetts.
2
Boston University School of Public Health, Boston.
•"Veterans Affairs Normative Aging Study, Boston Veterans Affairs Outpatient Clinic, Boston.
"Boston Veterans Affairs Medical Center, Boston.
'Boston University School of Medicine, Boston.

Background. Our objective was to develop a patient-based measure of the severity of osteoarthritis of the knee,
focusing on symptomatology, that may be used in conjunction with measures of health-related quality of life in moni-
toring the health status of outpatients.

Methods. We surveyed a random sample of male outpatients at Boston-area Veterans Affairs medical centers who
were identified as having osteoarthritis of the knee according to a three-question screen. Structured interviews included
12 items covering five domains of symptoms (global severity, 4 items; pain with activity, 3 items; pain at rest, 2 items;
impaired mobility, 2 items; and sensations of crepitus, 1 item), which were derived from clinical texts, consensus state-
ments, and previously developed severity indices. Interviews also included a detailed medical history. Health-related
quality of life was measured by the Medical Outcomes Study Health Status Survey (SF-36). Factor analysis and evalu-
ation of multiattribute scales were used to evaluate the structural relationships between and within the five domains of
symptoms.

Results. We identified 415 of the 1770 screened outpatients as having osteoarthritis of the knee. Internal consisten-
cies of the five domains ranged from .50 to .72, with substantial convergence between domains. The 12 items comprise
a summary index with high internal consistency (a = .88). Overall severity, defined as the mean of the 12 items after
standardization, was moderately correlated with the SF-36 component summaries: r = -.48 for physical; r = -.30 for
mental.

Conclusions. Our measure provides a reliable index that represents symptomatic severity of osteoarthritis of the
knee, which may be useful in comparing patient groups and assessing health outcomes; subscales may help character-
ize temporal changes, including responses to treatment.

R ESEARCH on the processes and outcomes of health

care relies heavily on patient-based measures of health
status. Patients are often regarded as the primary observers of
both their health service needs and the outcomes of their care.
Patients' perceptions of their health status are particularly rel-
evant when evaluating health services for patients with
chronic illness, where the therapeutic goals include preserv-
ing and optimizing patients' functional status and well-being;
that is, their health-related quality of life (1). Accordingly,
investigators in this field have devoted considerable effort to
improving standardized measures of patients' perceptions of
their health status. These efforts have yielded a variety of dis-
ease-specific and generic instruments that assess the manifest,
psychosocial consequences of disease that patients can reli-
ably observe and report, such as physical, emotional, and
social function (2). Data derived from these instruments have
been used in monitoring patients' health status, in modeling
the need for care, and in estimating the effects of processes of
care on patients' health-related quality of life.
A comprehensive assessment of health status requires
additional information regarding the severity of underlying
health problems. Objective parameters of disease severity
are most reliably obtained from physicians or patient
records. However, patients can provide reliable information
regarding the presence of chronic conditions (3-7) and the
severity of their symptoms (8,9). In this article we describe
the development of a patient-based measure of symptom
severity for osteoarthritis of the knee that will complement
more broadly focused measures of health-related quality of
life that assess behavioral and emotional disabilities.
Osteoarthritis of the knee provides a useful index condi-
tion for observing processes and outcomes of ambulatory
care, especially in settings that serve predominantly geri-
atric populations. The disease is chronic, with high preva-
lence that increases with age (10,11). It is associated with
substantial disability, including major activity limitations
and lost work (12,13), as well as much demand for health
services (13). Moreover, the principal goals of therapy are
symptom relief and reduction of disability (14), making it
an ideal focus for patient-based studies of health outcomes.

M351
M352 CLARK ETAL.
The severity of osteoarthritis of the knee may be indi-
cated by radiographic measures or by specific symptoms,
including pain, stiffness, and diminished knee mobility,
which reflect knee joint deterioration and remodeling
(15,16). Although there are instruments that have been
applied in assessing the functional consequences of osteo-
arthritis of the knee, including the Sickness Impact Profile
(17) and the Medical Outcomes Study Health Status Survey
(SF-36) (18), these instruments do not directly address the
specific symptomatology of the disease. For example, the
SF-36 contains a two-item bodily pain scale that assesses
pain intensity and pain-related interference in daily activi-
ties without regard to physical site, which confounds the
assessment of knee arthritis in patients with coexisting dis-
eases. The Arthritis Impact Measurement Scales (AIMS)
has been applied to studies of osteoarthritis (19). However,
AIMS was designed for rheumatoid arthritis and thus con-
tains a pain scale that addresses pain in multiple, unspeci-
fied sites and morning stiffness that persists more than 1 h
after waking up.
Previous efforts, similar to ours, have yielded the Index
of Severity of Osteoarthritis (20) and the Western Ontario
and McMaster Universities Osteoarthritis Index (WOMAC)
(21). Both were developed as measures of the outcomes of
clinical trials with patients with osteoarthritis of either the
knee or the hip. Both discriminate between treatment and
control groups in medication trials. However, neither was
developed for or applied to problems of monitoring health
status in general samples of ambulatory patients. For exam-
ple, questions in the WOMAC address the intensity of pain
in "study joints" in patients participating in a study
designed to focus on a specific joint. Their nonspecific for-
mulation may limit the reliability of these questions in
observational studies, where patients may have pain in mul-
tiple sites and have multiple comorbidities. In addition, nei-
ther has been evaluated with respect to measures of health-
related quality of life. The WOMAC also combines
disease-specific functional limitations with symptomatol-
ogy in a single instrument. In contrast, we have focused on
symptom severity alone to complement other measures that
focus on quality of life. Our objective was to develop a
measure of severity attributable to osteoarthritis of the knee
in a general ambulatory population characterized by sub-
stantial comorbidity.
Hence, we sought to construct a multi-item measure con-
sisting of questions addressing the symptomatology of
osteoarthritis of the knee and reflecting its severity. To
complement measures of disease-related functional status
and well-being, we focused exclusively on symptoms of
osteoarthritis of the knee that represent the extent of pro-
gressive knee joint remodeling caused by osteoarthritis,
which is manifested by specific experiences that are per-
ceived by patients. It would be correlated with relevant
clinical characteristics of patients receiving treatment for
osteoarthritis of the knee, including time since diagnosis,
physical examination findings, use of medications, and the
use of therapeutic and evaluative procedures that are likely
to be administered in response to increasing severity (e.g.,
arthrocentesis, arthroscopy) (22-25). It would also be cor-
related with measures of health-related quality of life.
Hence, it would be consistent with clinical markers of
severity as well as measures of the functional consequences
of disease, rendering it useful as a component of compre-
hensive health status assessment.
METHODS
Data on patients' perceptions of osteoarthritis of the knee
were collected as part of structured, baseline interviews of
patients in the Veterans Health Study (VHS), an ongoing,
prospective, observational study of health status and pro-
cesses of care in Veterans Affairs (VA) outpatient settings
(26,27). The study sites include four Boston-area VA outpa-
tient centers that comprise a small but heterogenous mix of
facilities. Within each site, eligible patients were identified
at the time of a clinic visit during specified time periods
distributed throughout the year in order to account for sea-
sonal variations in clinic use. Patients were eligible if they
had a medical visit to a VA ambulatory clinic during the
preceding 12 months and provided us with a telephone
number and address for subsequent contact. The sample
excluded women because their representation in the outpa-
tient population of these VA medical centers is very low.
Potential participants were sampled at random from the
pool of eligible male patients and recruited for the study.
The findings reported here are based on data from 415
patients with osteoarthritis of the knee who represent a sub-
set of 1,770 VHS participants recruited between August
1993 and December 1995.
We developed a three-item screen to identify patients
with osteoarthritis (shown in Table 1). The first item asked
whether a doctor had ever disclosed a diagnosis of the
disorder to them. The second item, based on American
College of Rheumatology clinical criteria (28), assessed
whether they had pain, aching, or stiffness in one or both
knees on most days. The third item assessed the presence of
pain, aching, or stiffness after rest. Patients were considered
to have osteoarthritis of the knee if they gave affirmative
responses to the first question and to either the second or
the third question. Patients who reported either a diagnosis
of rheumatoid arthritis (a physician had told them they had
rheumatoid arthritis and they were currently being treated
with gold, penicillamine, plaquenil, or methotrexate; 6
patients) or a history of total knee replacement (27 patients)
were excluded from the analysis.
Design of the Severity Measure
Pain is often regarded as the cardinal symptom of
osteoarthritis of the knee. Instruments developed by
Lequesne (20) and Bellamy et al. (21) addressed pain in the
Table 1. Screening Questions for Osteoarthritis of the Knee
No. Question
1 Has a doctor ever told you that you have arthritis in your knees?
2 Do you have pain, aching, or stiffness in one or both knees on most
days?
3 Do you have knee pain, aching, or stiffness after sitting for a long
time?
 SEVERITY OF OSTEOARTHRITIS OF THE KNEE
context of either activity or rest. For example, the WOMAC
included five questions that elicited ratings of the amount of
pain that is felt when walking, using stairs, and standing
upright (that is, when stressing an arthritic joint), as well as
when sitting or lying and at night while in bed. These ques-
tions may address pain that is directly attributable to dis-
ease-related changes in a weight-bearing joint. However,
the expression of osteoarthritis pain may also be more
diffuse. Questions referring to pain with specific activities
or events may underrepresent patients' experiences of os-
teoarthritis pain by missing knee pain that patients cannot
easily attribute to specific activities or occasions. Patients
may also express qualities of knee pain other than intensity,
such as persistence, frequency of recurrence, or day-to-day
variation. Hence, we developed questions to address (i) the
persistence and intensity of pain in each knee without attri-
bution to specific activities or occasions, (ii) the diurnal
duration of pain (i.e., how long pain usually lasted during
the day), (iii) the relative frequency of severe pain in the
past 4 weeks, and (iv) pain associated with knee joint stress.
We also decided to combine pain and stiffness rather
than assess them separately. Pain and stiffness may have
distinct clinical implications, with the former suggesting
the need for anti-inflammatory medication and the latter a
mechanical intervention. Thus, clinicians may carefully for-
mulate their queries to distinguish between these two symp-
toms. However, their observation via brief, patient-com-
pleted questionnaires is prone to error. Bellamy et al. (21)
reported that questions focused on stiffness had relatively
low test-retest reliability. Other recent studies also have
suggested that patients may inconsistently discriminate
between pain and stiffness and that assessments of stiffness
alone may be unreliable (29,30).
Therefore we developed focused questions to elicit 12
items of information in five domains of osteoarthritis symp-
toms, as shown in Table 2. The complete set of questions is
presented in the Appendix. Some items were represented by
multiple questions. Pain and stiffness were represented by
Table 2. Symptoms Represented by Osteoarthritis
Severity Scale Items
Pain and Stiffness With Activity
Walking
Using stairs
Standing
Pain and Stiffness With Rest
Resting
Sleeping
Pain and Stiffness Without Regard to Situation
Persistence of pain in worst knee
Intensity of pain in worst knee
Diurnal duration of pain and stiffness in the worst knee
Frequency of severe pain in the worst knee
Sensations of Crepitus
Clicking or sandpaper sensation
Impaired Function
Limping
Knees give way with activity
M353
four sets of questions. First, we developed questions per-
taining to the presence, persistence, and overall intensity of
pain in each knee. That is, for both the right and left knees,
patients could report the presence of any pain, whether or
not it is experienced "most days," and rate it as mild, mod-
erate, or severe. Second, we formulated questions that
addressed the extent of pain and stiffness associated with
different activities. Similar to questions developed by Bel-
lamy et al. (21) in the WOMAC, our questions addressed
the intensity of pain associated with rising, standing, and
walking on a 5-point Likert scale, ranging from "none at
all" to "a great deal." Third, two additional questions also
echoed WOMAC questions by addressing pain and stiff-
ness, while resting and sleeping. Fourth, two questions per-
tained to temporal qualities of pain in the worst knee in the
past 4 weeks. One would elicit a 4-point rating of the diur-
nal duration of pain in the worst knee, when it occurs, rang-
ing from "not at all" to "all day long." The other addressed
the relative frequency of "severe" pain in the worst knee,
ranging from "never" to "always." Thus, we attempted to
describe pain and stiffness both globally and with respect to
particular stresses.
Strictly speaking, crepitus is a physical examination find-
ing. However, in the opinion of the clinicians we consulted,
patients with osteoarthritis of the knee may describe a
clicking or sandpaper feeling in their joints. Thus, we for-
mulated two questions to represent patients' experiences of
feelings analogous to crepitus. One question addressed the
presence of a "clicking or sandpaper sensation" when walk-
ing. The other question requested a 4-point rating of the rel-
ative frequency of this sensation, ranging from "a little of
the time" to "all of the time."
Three questions represented impaired knee function. Two
addressed the presence and the severity of limping because
of knee arthritis. One question addressed the experience of
patella and internal derangement problems, which may be
reported by patients with advanced osteoarthritis of the
knee: "Do your knees ever 'give way' when you walk or
perform other related activities?"
Additional Measures
Baseline interviews included questions pertaining to the
use of medications for osteoarthritis of the knee and history
of arthrocentesis and arthroscopy. Physical examination by
trained research technicians determined the presence of
crepitus, bony enlargement, and joint tenderness on palpa-
tion. Measurements of height and weight were also ob-
tained in order to calculate body mass. Felson et al. (31)
and Verbrugge et al. (32) have shown obesity to be not only
a risk factor for the disease but also a risk factor for
increased severity. Coexistent disease was determined by
tabulating specific conditions and groups of symptoms
reported in a comprehensive medical history interview. The
resulting Disease Burden Index represents a count of dis-
crete morbidities afflicting VA outpatients. Cognitive status
was assessed by administering the memory test of the Neu-
robehavioral Cognitive Status Examination (NCSE) (33)
during the medical history interview. Demographic data
were collected through questionnaires completed at the
time of the interview.
M354 CLARK ETAL.

Health-related quality of life was assessed by the SF-36
(34). The eight scales of the SF-36 represent dimensions of
physical function, role performance, emotional well-being,
and general health perceptions related to disease status and
are valid across a variety of diseases (35,36). Ware et al.
(37) identified two summary scales, the Physical Compo-
nent Summary and the Mental Component Summary, that
represent weighted summaries of these eight scales, normed
with respect to a probability sample of the adult U.S. popu-
lation. The eight scales range from zero to 100, with the lat-
ter representing optimal health, whereas the two summaries
are defined as T scores with a mean of 50 and a standard
deviation (SD) of 10.
nent Summary of the SF-36, tended to be quite low. This is
consistent with their high level of comorbidity; they re-
ported an average of seven diseases in addition to osteo-
arthritis of the knee. Most of these men (72%) reported tak-
ing one or more medications for their osteoarthritis. Eight
percent reported having undergone arthrocentesis in the
past year. Physical findings, as determined by our techni-
cians, indicated that crepitus was very common (70%),
joint tenderness on palpation was present for 38%, and 28%
had bony enlargement.
Responses for the 12 severity items are summarized in
Table 4. Persistent pain was reported by most of the
patients with osteoarthritis of the knee; 85% of them said

Analytic Procedures
Data analysis focused on the construction and psychome- Table 3. Demographic and Health Status Characteristics
of Patients (n = 415)
tric evaluation of a multi-item index of severity. The struc-
tural relationships within and between the five domains Characteristics Values
encompassing the 12 candidate items were evaluated in two Age: median (range, years) 66 (22-90)
ways. First, the domain structure was evaluated by examin- Marital status: married (%) 60
ing the multitrait/multiattribute matrix in which the correla- Education: median (years) 12
tions between each item and its hypothesized scale (item- Race: nonwhite (%) 9
convergence corrected for item-overlap) and the other Income: median ($) 17,863
domains (item-divergence) were calculated. The internal Employment: employed (%) 29
consistencies (Cronbach's alpha) for items within each Taking medications for osteoarthritis of the knee (%) 72
hypothesized domain also were calculated. Second, the 12 History of arthrocentesis in past 12 months (%) 8
candidate items were included in a factor analysis. History of arthroscopy in past 12 months (%) 4
The construct validity of the severity index was exam- Crepitus: worst knee (%) 70
ined by evaluating associations between the index and clin- Bony enlargement: worst knee (%) 28
ical variables as well as health-related quality of life scales. Joint tenderness: worst knee (%) 38
Associations between the index and dichotomous clinical
Physical component summary (mean ± SD) 31.2 ± 10.3
variables (e.g., presence of joint tenderness or recent his- Mental component summary (mean ± SD) 47.7 ±13.4
tory of arthroscopy) were evaluated with t tests, whereas Disease burden index (mean ± SD) 7.2 ±4.1
associations with continuous variables (e.g., SF-36 scores)
were evaluated by calculating Pearson correlation coeffi-
cients. To examine the independent relationships between
Table 4. Severity Items
the severity index and health-related quality-of-life scores,
we estimated a series of multiple regression models using Value
ordinary least squares regression. The severity index was Item (mean ± SD)
first regressed on the three physical examination parame- Persistence of pain/stiffness in worst knee
ters: crepitus, bony enlargement, and joint tenderness. The (pain on most days; 0 = no, 1 = yes) .85
severity index was then included as an independent vari-
Intensity of pain/stiffness in worst knee
able, along with the physical examination findings and the
(0 = no pain to 3 = severe) 1.97 ± .76
demographic and comorbidity covariates, in models to pre-
dict four pertinent health-related quality-of-life scores: Diurnal duration of pain/stiffness in worst knee
physical function, role function with physical limitations, (1 = not at all to 4 = all day long) 2.73 ± .88
bodily pain, and general health perceptions. Frequency of severe pain in worst knee in past 4 weeks
(1 = never to 6 = always) 3.71 ±1.48

RESULTS Pain/stiffness with activity (1 = none at all to 5 = a great deal)

Demographic and health status characteristics of the 415 During or after walking 3.14 ± 1.17
patients with osteoarthritis of the knee in the VHS are Going up or down stairs 3.38 ± 1.19
After standing for 30 min 3.17 ± 1.26
shown in Table 3. Their ages ranged from 22 to 90 years,
although 95% were between 40 and 80 years old. Almost Pain/stiffness with rest (1 = none at all to 5 = a great deal)
all of these patients were white. Cognitive function, as indi- At rest 2.16±1.05
cated by memory test of the NCSE, was comparable to While trying to sleep at night 2.17±1.12
geriatric patients described by Kiernan et al. (33), with only Frequency of clicking or sandpaper sensation
6% scoring less than 1 SD below the mean they report for (0 = never to 4 = all of the time) 1.34 ±1.46
geriatric patients. Severity of limp (0 = no limp to 3 = severe limp) .94 ± .94
The physical aspects of these patients' overall health-
Knees ever give way with activity (0 = no, 1 = yes) .65
related quality of life, as indicated by the Physical Compo-
 SEVERITY OF OSTEOARTHRITIS OF THE KNEE M355

they had pain and stiffness in their worst knee on most defined here by three indicators of pain and stiffness with
days. Many of them also rated the intensity of this pain as walking, using stairs, and standing, may reflect a more gen-
moderate, as indicated by the mean of 1.97 on this 4-point eral tendency to report difficulties with using the knee.
scale. Other items pertaining to pain and stiffness suggested In addition, several other item-total correlations suggest
moderate to high levels of severity, although the intensity that these five domains are not widely separated. For exam-
of pain at rest or while sleeping was notably lower than ple, diurnal duration is most highly correlated not with
pain and stiffness with activity. The severity of limping was global severity (r = .49) but with pain with activity (r =
moderate to low, but almost two thirds said that their knees .52). A similar pattern appears for frequency of severe pain
occasionally gave way with activity. in the worst knee: r = .64 for global severity, and .64 for
The results of both the factor analysis and the evaluation of pain with activity. The correlations between the domain
the multitrait/multiattribute matrix provide only modest sup- scores are consistent with this finding of limited discrimina-
port for the hypothesized domain structure of the 12 symp- tion. The correlations among global severity, pain with
tom items. The multitrait/multiattribute matrix is presented in activity, and impaired mobility range between .64 and .69.
Table 5. Three of the four sets of items (i.e., global severity, These findings suggest that the five domains of symptoms
pain with activity, and pain with rest) had high internal con- converge in reflecting severity of osteoarthritis of the knee,
sistency, with respective a values of .72, .84, and .77. How- and that severity is best represented by a single index
ever, the two items representing impaired mobility had an a derived from all 12 items. The resulting index has high
of .50. Item-total correlations (adjusted for overlap within internal consistency, a = .88 (Table 5).
each set) indicated that most of the items tended to be more The index was scored as the unweighted sum of the 12
highly correlated with their hypothesized domain than with items. Because the item* had varying response scales (e.g.,
other domains. For example, the intensity of pain in the worst dichotomous scales with values 0 or 1 and Likert questions
knee had a higher correlation with global severity (r = .55) with 5-point response sets), each item was standardized (x
than with the other three multi-item domains (r = .42-.53). = 0, SD = 1) prior to calculating the total score. The result-
However, there were several indications of overlap between ing scale score was transformed into a T score (x = 50, SD
domains, which suggest that the five domains are not dis- = 10) that uses the population sampled in the VHS as a nor-
crete. The presence of pain in the worst knee on "most days" mative standard for a heterogeneous population of outpa-
was more highly correlated with pain with activity than with tients with osteoarthritis of the knee. The distribution is
global severity. Likewise, two items ascribed to impaired approximately normal, with less than 1% of the patients
mobility (e.g., knees "giving way," and limping) were more scoring at either the floor or the ceiling of the index.
highly correlated with pain with activity. Pain with activity, The associations between the severity index and clinical
characteristics are summarized in Table 6. The severity
index was correlated with time since diagnosis in the ex-
Table 5. Multitrait/Multiattribute Matrix of Domains
pected direction given the natural progression of the dis-
of Osteoarthritis of the Knee Symptoms
ease; severity was somewhat greater among patients who
Pain Pain reported longer histories of osteoarthritis of the knee.
Global With With Impaired Severity Severity index scores tended to be higher among patients
Symptom Severity Activity Rest Mobility Index
Internal consistency: a .72 .84 .77 .50 .88
Intensity of pain, Table 6. Association Between Severity
worst knee .55 .51 .42 .43 .60 and Clinical Characteristics
Pain most days,
worst knee .35 .42 .24 .24 .40 Absent Present Value
Diurnal duration of pain, Findings (n absent/n present) (mean ± SD) (mean ± SD) (t test)
worst knee .49 .52 .42 .39 .55 Physical Examination Findings
Frequency severe pain, Crepitus in worst knee (112/265) 48.3 ±9.5 50.7 ± 9.9 .03
worst knee .64 .64 .47 .54 .74 Bony enlargement in worst knee
(268/106) 49.4 ±9.9 51.3 ±9.8 .09
Pain/stiffness, walking .63 .75 .48 .50 .72 Joint tenderness in worst knee
Pain/stiffness, using stairs .63 .70 .46 .45 .69 (232/143) 47.9 ±9.6 53.4 ±9.6 <.001
Pain/stiffness, standing .56 .65 .44 .48 .64 Historical Findings
Taking medications for
Pain/stiffness, resting .47 .48 .63 .33 .54
osteoarthritis of the knee
Pain/stiffness, sleeping .49 .48 .63 .37 .57 (112/281) 46.0 ±8.8 51.7 ±9.8 <.001
Knees ever give way .37 .37 .29 .33 .42 History of arthrocentesis in past
12 months (380/35) 49.8 ±10.0 51.9 ±10.4 .23
Extent of limping .51 .53 .35 .33 .57 History of arthroscopy in past
Clicking/sandpaper 12 months (398/17) 49.6 ± 9.9 58.2 ± 8.0 <.00l
sensation .38 .36 .23 .38 .42 Obesity (body mass index > 30)
(245/143) 49.2 ±9.9 50.9 ±10.0 .10
Note: Boxes in table identify correlations between items and assigned
scales. Note: For age, r = -.15; for years since diagnosis, r = .20.
M356 CLARK ETAL.

with findings of crepitus and joint tenderness in the worst
knee, with the biggest difference associated with joint ten-
derness. There was no significant difference in severity
scores with respect to the presence or absence of bony
enlargement. Multivariate adjustment for the simultaneous
effects of all three physical findings, using linear regres-
sion, indicated that joint tenderness and crepitus were sig-
nificant correlates of symptom severity score, accounting
for 7% and 1%, respectively, of the variance in the index.
Although bony enlargement has diagnostic value, it is not
an indicator of severity. Joint tenderness is a clinical sign
indicating severity. Hence, an index that largely reflects
pain and stiffness is correlated with the most pertinent
physical examination finding. Scores on the severity index
were correlated with use of medications and history of
arthroscopy in the previous 12 months. However, severe
obesity was not associated with the severity index.
Correlations between the severity index and the scales
of the SF-36 were all substantial and were all in the
expected negative direction. ThaJ, is, increased severity
was associated with lower functional status. The correla-
tion with the bodily pain scale was the highest: r = -.58
(Table 7). A higher correlation between a generic measure
of pain and an index largely defined by items pertaining to
pain and stiffness might be expected. However, the corre-
Table 7. Correlations Between Severity Index
and Health-Related Quality of Life Scales
Quality of Life Scale
Physical function index
Role performance with physical limitations
Bodily pain
General health perceptions
Vitality
Social function
Role performance with emotional limitations
Mental health index
Physical component summary
Mental component summary
lation is suggestive of construct validity given that the pain
scale of the SF-36 reflects bodily pain from all sources, of
which there are several in this population, whereas the
severity index is focused on pain arising from osteoarthri-
tis of the knee. In addition, the correlation with the physi-
cal function index (r = -.54), which assesses health-related
limitations in physical mobility, is also supportive of con-
struct validity.
The severity index was included in regression models to
estimate the independent effects of the severity index on
SF-36 scales, controlling for demographics, comorbidity,
and physical findings. In these stepwise analyses, the block
of covariates was entered first, followed by the severity
index. Results are shown in Table 8. The combined effects
of age, education, employment, comorbidity, and physical
findings accounted for 19-34% of the variance in physical
function, role function with physical limitations, general
health perceptions, and bodily pain. The independent effects
of the severity index are substantial for physical function
and bodily pain, accounting for 20% and 21% of the vari-
ance in these scales, respectively.
As a final evaluation of the information contained in the
index, we estimated the probabilities of specific responses to
each of the 12 items as functions of scores on the index.
Index scores should accurately predict responses to the 12
items to the extent that the index is an accurate representa-
tion of the severity of osteoarthritis of the knee. Thus, for
each of the 12 items we estimated an ordinal logistic regres-
sion model, using the index score to predict question
responses. The findings in Table 9 illustrate the results of
r Value these analyses. For illustration, we chose three index scores:
-.54
50, which is the mean for all patients with osteoarthritis; 40,
-.40 which is 1 SD below the mean (that is, substantially lower
-.58 severity); and 60, which is 1 SD above the mean (that is, sub-
-.35 stantially greater severity). As shown in Table 9, a patient
-.37 with a score of 40 on the index has an 82% probability of
-.47 endorsing "almost never" or higher on item 4, frequency of
-.33 severe knee pain. This patient has only a 10% probability of
-.31 endorsing "fairly often" or higher to this question. A patient
-.48 with a score of 60 on the index has a 99% probability of
-.30 responding "almost never" or higher to this question. As the

Table 8. Regression of Selected SF-36 Scales on Osteoarthritis of the Knee Severity Index
Physical Function Role Function With General Health
Index Physical Limitations Perceptions Bodily Pain
b P b P b P b P
Age -.14 .22 .14 .42 0.27 .004 -0.09 .30
Education 1.15 .02 .44 .56 0.68 .10 0.33 .41
Employed 7.20 .01 10.55 .02 8.16 .001 3.88 .10
Comorbidity -1.33 <.001 -2.38 <.001 -2.38 <.001 -1.21 <.00l
Crepitus -3.03 .28 -3.53 .41 2.85 .22 -1.81 .43
Bony enlargement 2.00 .48 4.95 .25 2.12 .369 1.96 .40
Joint tenderness 1.07 .69 -0.84 .83 2.21 .31 -1.88 .38
/?2 (covariates) .21 .19 .34 .21
OA severity -1.36 <.001 -1.19 <.001 -0.53 <.001 -1.14 <.001
/?2 (OA severity alone) .20 .08 .05 .21
F 24.51 13.24 22.58 25.57
P <.OO1 <.001 <.001 <.001

Note: b, regression coefficient;/?, significance of regression coefficient; OA, Osteoarthritis of the Knee Severity Index.
 SEVERITY OF OSTEOARTHRITIS OF THE KNEE
Table 9. Probability of Responding at Specific Ordinal Levels or Higher
of Items Comprising the Severity Measure, Given an Index Score of 40, 50, or 60
Item Severity Measure
4 Frequency of Severe Pain > Almost Never
40 82
50 98
60 99
5 Pain/Stiffness Walking > A Little Bit
40 85
50 99
60 99
score on the index increases from 40 to 60, the probability of
endorsing "fairly often" or higher to the question pertaining
to frequency of knee pain increases from 10% to 94%.
DISCUSSION
We have developed a patient-based index of severity for
osteoarthritis of the knee derived from patients' responses to
focused questions that address symptomatic manifestations
of degenerative changes in the knee joint associated with this
disorder. The index discriminates between patients across a
broad range of symptomatic severity of osteoarthritis of the
knee, with virtually no floor or ceiling effects. It yields scores
that are correlated with pertinent clinical variables. Consis-
tent with the natural progression of the disease, the index is
positively correlated with the number of years since patients'
knee problems were diagnosed as osteoarthritis. It is posi-
tively correlated with joint tenderness on examination as well
as patient-reported use of medications for osteoarthritis and
recent history of either arthrocentesis or arthroscopy. Hence,
scores on the index may be interpreted as representing the
severity of clinical osteoarthritis of the knee. The index is
also correlated with pertinent, physical function measures of
health-related quality of life included in the SF-36. These
correlations remain meaningful, as well as significant, when
evaluated beside comorbidity, demographic characteristics,
and the effects of physical examination findings. Hence, the
measure has substantial construct validity as a component of
disease-related health status.
Our objective was to develop a measure that would
closely reflect the patient-perceived severity of osteoarthri-
tis of the knee and account for variation in the health-
related quality of life of patients with this condition.
Dimensions of the latter are measured by instruments such
as the SF-36. The present measure of symptom severity
should complement quality-of-life measures in providing a
more complete patient-based assessment of health status
that takes account of the underlying chronic condition. In
addition, it is potentially useful in providing data for risk-
adjusted estimates of the relationships between health-
related quality of life and processes of ambulatory care.
That is, it may be included as an important covariate in
models of the effects of health-related quality of life on the
use of health services as well as in models of the effects of
M357
Index Score
> Sometimes > Fairly Often > Very Often > Always
58 10 2 0
94 57 21 3
99 94 77 27
> Some > Quite a Bit > A Great Deal
27 4 0
82 34 3
98 87 27
processes of care on health outcomes. However, these
potential uses of our index of severity need to be examined
in longitudinal analyses.
The present study is limited to cross-sectional analyses.
Longitudinal analyses, which are planned for the next
phase of the VHS, will examine the responsiveness of the
measure to significant changes in either the management of
the disease or other clinical indicators of severity, as well as
evaluate the value of the measure in predicting the utiliza-
tion of services. This study is also limited by its exclusively
male sample. Although the index may prove useful in the
major portion of the VA population, it will need further
evaluation before it is applied in the larger population of
patients with osteoarthritis.
We have drawn upon the results of previous research by
Lequesne (20) and Bellamy et al. (21). However, the index
we developed differs from their severity measures in some
important respects. Lequesne's index consisted of 11 items
of information to be elicited and rated by physicians during
clinical interviews, whereas like the WOMAC (21), ours
consists of standardized questions that elicit information,
including ratings of specific symptoms, directly from pa-
tients. In addition to subscales that addressed pain and stiff-
ness, the WOMAC included a 17-item subscale that ad-
dressed physical function with respect to activities, such as
ascending and descending stairs, dressing oneself, shop-
ping, and accomplishing light and heavy domestic tasks.
These functional capabilities are related to knee arthritis.
However, they are also within the domain typically ad-
dressed as health-related functional limitations in measures
such as the SF-36. In contrast to the WOMAC, we have
maintained a careful distinction between these two domains
of health status. Our index is focused exclusively on symp-
toms of the underlying disease, and covers this domain
somewhat more broadly.
The index we developed does not directly represent
underlying anatomical or functional changes, such as those
represented by the best-known, perhaps classic, index of
severity for osteoarthritis of the knee developed by Kell-
gren and Lawrence (38). The disease process, which still
remains poorly understood (15,16), results in osteophyte
formation and joint space narrowing, which may be
observed by way of radiographs and rated according to a
M358 CLARK ETAL.
scale ranging from 0 (the absence of evidence of joint
degeneration) to 4 ("severe joint space narrowing with
cysts, osteophytes, and sclerosis") (38). Measures such as
this and other assessments derived from laboratory tests,
arthroscopic examinations, or other techniques for visualiz-
ing the knee joint (22,23,25) are objective, but they pose
several problems when applied to health services research.
The collection of radiographic data can be a very costly
undertaking in large-scale health outcomes research if it
entails chart abstraction or ordering X rays. Relying on
clinical records for these data may not be fruitful because
outpatient medical records are often incomplete and timely
radiographs, in particular, may not be available.
In the primary care setting, osteoarthritis of the knee is
often diagnosed and treated on the basis of patient history
and presenting symptoms without obtaining X rays of the
knee (39). Furthermore, disease severity based on radio-
graphic evidence is often poorly correlated with sympto-
matic manifestations of the disease (18,40,41). Thus, we
did not seek to evaluate the validity of this index in terms
of its correlation with radiographs or other objective dis-
ease parameters. We relied entirely on the information that
study participants provided in response to our screen (i.e.,
that a doctor had told them they had the disorder and they
reported either knee pain and stiffness on most days or pain
and stiffness after sitting for a long time) in order to iden-
tify patients with this condition.
Our method for ascertaining osteoarthritis of the knee
represents a potential limitation of our findings. We sought
to identify VA outpatients who perceive themselves as hav-
ing osteoarthritis and might seek care as a result. It is pos-
sible that some of the patients may be misinformed or may
have misrepresented their medical condition. In the pilot
phase of the VHS, we compared our ascertainment of
osteoarthritis of the knee, as well as five other chronic con-
ditions that are prevalent in the VA population, with infor-
mation in the VA medical charts (i.e., record of diagnosis of
osteoarthritis or degenerative joint disease, history of total
knee replacement, or positive radiograph) for 402 of
patients. We found moderate concordance (K = .54; predic-
tive value positive = 59.7%) in the case of osteoarthritis of
the knee, but it is likely that their charts were incomplete
for a variety of physician and patient-related reasons.
Our strategy for assessing the severity of osteoarthritis of
the knee in outpatients has been to construct a patient-based
measure. Rather than directly addressing quality of life
issues, it consists of questions that focus upon the clinically
significant, symptomatic manifestations of the underlying
knee joint changes caused by the disease. In the case of
osteoarthritis of the knee, as with many other chronic dis-
eases treated in primary care settings, patients' reports of
their health, including the presence and severity of medical
conditions and their effects on functional status and well-
being, may represent the most important factors that deter-
mine providers' and patients' decisions regarding the uti-
lization of health services, as well as their evaluations of
the effectiveness of those services. Hence, we have devel-
oped a measure of severity based on information that may
be provided by patients directly. Although limited to
patients' perceptions, it may represent the severity of
osteoarthritis in terms that are most relevant to clinical
management and quality of life.
ACKNOWLEDGMENTS
This study was supported by VA Health Services Research and Devel-
opment Service Grant SDR 91-006.S.
A preliminary report of these findings was presented at the Annual
Meeting of the VA Health Services Research and Development Service,
Washington, DC, February 1996.
Address correspondence to Dr. Jack A. Clark, Center for Health Quality,
Outcomes, and Economic Research (152), Edith Nourse Rogers Memorial
Veterans Hospital, 200 Springs Road, Bedford, MA 01730. E-mail:
jaclark@bu.edu
REFERENCES
1. Tarlov A, Ware J, Greenfield S, Nelson E, Perrin E, Zubkoff M. The
medical outcomes study: an application of methods for monitoring
the results of medical care. JAMA. 1989;262:925-930.
2. Bowling A. Measuring Disease: A Review of Disease-Specific Quality
of Life Measurement Scales. Buckingham, U.K.: Open University
Press; 1995.
3. Harlow S, Linet M. Agreement between questionnaire data and medi-
cal records: the evidence for accuracy of recall. Am J Epidemiol.
1989;129:233-248.
4. Kehoe R, Wu S, Leske C, Chylack L. Comparing self-reported and
physician-reported medical history. Am J Epidemiol. 1994; 139:813-818.
5. Katz J, Chang L, Sangha O, Fossel A, Bates D. Can comorbidity be
measured by questionnaire rather than medical record review? Med
Care. 1996;34:73-84.
6. Kriegsman D, Penninx B, van Eijk J, Boeke J, Deeg D. Self-repoits
and general practitioner information on the presence of chronic disease
in community dwelling elderly. J Clin Epidemiol. 1996;49:1407—1417.
7. Haapanen N, Miilunpalo S, Pasanen M, Oja P, Vuori I. Agreement
between questionnaire data and medical records of chronic diseases in
middle-aged and elderly Finnish men and women. Am J Epidemiol.
1997; 145:762-769.
8. O'Leary M, Barry M, Fowler F. Hard measures of subjective out-
comes: validating symptom indexes in urology. J Urol. 1992; 148:
1546-1548.
9. Eakin E, Kaplan R, Ries A. Measurement of dyspnoea in chronic
obstructive pulmonary disease. Qual Life Res. 1993;2:181—191.
10. Verbrugge L, Patrick D. Seven chronic conditions: their impact on US
adults' activity levels and use of medical services. Am J Public
Health. 1995;85:173-182.
11. Felson D. Epidemiology of hip and knee osteoarthritis. Epidemiol
Rev. 1988; 10:1-28.
12. Liang M, Larson M, Thompson M, Eaton H, McNamara E, Katz R,
Taylor J. Costs and outcomes in rheumatoid arthritis and osteoarthri-
tis. Arthritis Rheum. 1984;27:522-529.
13. Felts W, Yelin E. The economic impact of the rheumatic diseases in
the United States. J Rheumatol. 1989; 16:867-884.
14. Schnitzer T. Management of osteoarthritis. In: McCarthy D, Koop-
man W, eds. Arthritis and Allied Conditions. Philadelphia: Lea &
Febiger; 1993:1761-1769.
15. Altman R, Asch E, Bloch D, Bole G, Borenstein D, et al. Develop-
ment of criteria for the classification and reporting of osteoarthritis:
classification of osteoarthritis of the knee. Arthritis Rheum. 1986;29:
1039-1049.
16. Moskowitz R. Clinical and laboratory findings in osteoarthritis. In:
McCarthy D, Koopman W, eds. Arthritis and Allied Conditions.
Philadelphia: Lea & Febiger, 1993:1735-1760.
17. Summers M, Haley W, Reveille J, Alarcon G. Radiographic assess-
ment and psychologic variables as predictors of pain and functional
impairment in osteoarthritis of the knee or hip. Arthritis Rheum.
1988;31:204-209.
18. Bombardier C, Melfi C, Paul J, Green R, Hawker G, Wright J, Coyte
P. Comparison of a generic and a disease-specific measure of pain and
physical function after knee replacement surgery. Med Care.
1995;33:AS131-AS144.
 SEVERITY OF OSTEOARTHRITIS OF THE KNEE M359

19. Mason J, Anderson J, Meenan R. Applicability of a health status
model to osteoarthritis. Arthritis Care Res. 1989;2:89-93.
20. Lequesne M. Indices of severity and disease activity for osteoarthritis.
Semin Arthritis Rheum. 1991 ;20 (Suppl. 2):48-54.
21. Bellamy N, Buchanan W, Goldsmith C, Campbell J, Stitt L. Valida-
tion Study of WOMAC: a health status instrument for measuring clin-
ically important patient relevant outcomes to antirheumatic drug ther-
apy in patients with osteoarthritis of the knee. J Rheumatol. 1988;
15:1833-1840.
22. Ike R. The role of arthroscopy in the differential diagnosis of
osteoarthritis of the knee. Rheum Dis Clin North Am. 1993; 19:673-696.
23. Chan W, Lang P, Stevens M, Sack K, Majumdar S, Stoller D, Basch
C, Genant H. Osteoarthritis of the knee: comparison of radiography,
CT, and MR imaging to assess extent and severity. Am J Roentgenol.
1991; 157:799-806.
24. Brandt K, Fife R, Braunstein E, Katz B. Radiographic grading of the
severity of knee osteoarthritis: relation of the Kellgren and Lawrence
grade to a grade based on joint space narrowing, and correlation with
arthroscopic evidence of articular cartilage degeneration. Arthritis
Rheum. 1991 ;34:1381-1386.
25. McKeag D, Smith B, Edminster R, Laird T, Clark J, Herron S. Esti-
mating the severity of osteoarthritis with magnetic resonance spec-
troscopy. Sem Arthritis Rheum. 1992;21:227-238.
26. Kazis L, Miller D, Skinner K, Clark J, Spiro A, Rogers W, Lee A,
Payne S, Fincke G, Selim A. Health Related Quality of Life in Veter-
ans: The Veterans Health Study. Washington, DC: Department of Vet-
erans Affairs, 1995.
27. Kazis L, Skinner K, Miller D, Clark J, Lee A. Quality of life for the
chronically ill elderly. In: Romeis J, Coe R, Morley J, eds. Applying
Health Services Research for Lang-Term Care. New York: Springer;
1996:43-75.
28. Altman R. Classification of disease: osteoarthritis. Sem Arthritis
Rheum. 1991;20:40-47.
29. Hazes J, Hayton R, Silman A. A reevaluation of the symptom of
morning stiffness. J Rheumatol. 1993;20:1138-1142.
30. Hazes J, Hayton R, Burt J, Silman A. Consistency of morning stiff-
ness: an analysis of diary data. Br J Rheumatol. 1994;33:562-565.
31. Felson D, Zhang Y, Anthony J, Naimark A, Anderson J. Weight loss
reduces the risk for symptomatic knee osteoarthritis in women. Ann
Intern Med. 1992; 116:535-539.
32. Verbrugge L, Gates D, Ike R. Risk factors for disability among U.S.
adults with arthritis. J Clin Epidemiol. 1991 ;44:167-182.
33. Kiernan R, Mueller J, Langston W, van Dyke C. The neurobehavioral
cognitive status examination: a brief but differentiated approach to
cognitive assessment. Ann Intern Med. 1987; 107:481-485.
34. Ware J, Sherbourne C. The MOS 36-item short-form health survey
(SF-36). Med Care. 1992;30:473^83.
35. McHorney C, Ware J, Raczek A. The MOS 36-item short-form health
survey (SF-36); II, psychometric and clinical tests of validity in mea-
suring physical and mental health constructs. Med Care. 1993;31:
247-263.
36. McHorney C, Ware J, Lu R, Sherbourne C. The MOS 36-item short-
form health survey (SF-36); III: tests of data quality, scaling assump-
tions, and reliability across diverse patient groups. Med Care.
1994;32:40-66.
37. Ware J, Kosinski M, Bayliss M, McHorney C, Rogers W, Raczek A.
Comparison of methods for the scoring and statistical analysis of SF-
36 health profile and summary measures: summary of results from the
medical outcomes study. Med Care. 1995;33:AS264-AS279.
38. Kellgren J, Lawrence J. Rheumatism in miners, II: x-ray study. Br J
IndMed. 1952;9:197-207.
39. Bolumar F, Ruiz M, Hernandez I, Pascual E. Reliability of the diag-
nosis of rheumatic conditions at the primary health care level. J
Rheumatol. 1994;21:2344-2348.
40. Felson D, Naimark A, Anderson J, Kazis L, Castelli W, Meenan R.
The prevalence of knee osteoarthritis in the elderly. Arthritis Rheum.
1987;30:914-918.
41. Dekker J, Boot B, van der Woude L, Bijlsma J. Pain and disability in
osteoarthritis: a review of biobehavioral mechanisms. J Behav Med.
1992;15:189.
Received January 29, 1997
Accepted December 9, 1997

Appendix

Items Representing Symptomatic Severity of Osteoarthritis of the Knee

Item Response Scale

Qualities of Pain and Stiffness

Persistence and intensity of pain and stiffness in right and left knees
Do you have any pain, aching, or stiffness in your left knee/right knee? yes/no
If yes: On most days do you have pain, aching, or stiffness in your left knee/right knee? yes/no
Is the pain, aching, or stiffness in the left knee/right knee mild, 0 = no pain
moderate, or severe? 1 = mild
2 = moderate
3 = severe
Diurnal duration of knee pain and stiffness in worst knee
During the past 4 weeks, when you have severe knee pain, aching, or stiffness, 1 = all day long
how long has it usually lasted in the worst knee? 2 = most of the day
3 = part of the day
4 = not at all
Frequency of severe knee pain and stiffness in worst knee
During the past 4 weeks, how often have you had severe knee pain, aching, 1 = always
or stiffness in the worst knee? 2 = very often
3 = fairly often
4 = sometimes
5 = almost never
6 = never
(continued on next page)
M360 CLARK ETAL.

Appendix (continued)

Item Response Scale

Pain and Stiffness With Activity or Rest

Do you have pain, aching, or stiffness in your knee(s) with the following activities?
If so, how much?
During or after walking, going up or down stairs, after standing for 30 min, 1 = none at all
at rest, while trying to sleep at night 2 = a little bit
3 = some
4 = quite a bit
5 = a great deal

Nonpain Sensations
Do you have clicking or a sandpaper sensation in your knee(s) when you walk? no/yes
If yes: How often do you have clicking or a sandpaper sensation in your knees? 1 = all of the time
2 = most of the time
3 = some of the time
4 = a little of the time

Impaired Mobility
Do you walk with a limp because of your knee arthritis? no/yes
If yes: Is the limp mild, moderate or severe? 1 = mild
2 = moderate
3 = severe
Do your knees ever "give way" when you walk or perform other related activities? no/yes