Total hip arthroplasty, an effective intervention for patients
with moderate-to-severe arthritis, results in dramatic relief of
pain and improvement in function7.
Because the primary purpose of total hip arthroplasty is to relieve
pain and increase functional capacity, the individual patient's
main goals need to be clearly determined to evaluate the result
of treatment appropriately. Prior research by one of us (J. G. W.)
and colleagues showed that patients interviewed prior to total hip
arthroplasty differed both in the spectrum of their concerns and
in the importance that they attached to the relief of those concerns14. For example, discussions with patients
during the development of the Patient-Specific Index revealed concerns
such as fear of falling and difficulties with sexual function that
are not found in many standard hip-rating scales. Furthermore, although
only twenty-three (32 percent) of the seventy-two patients interviewed
prior to total hip arthroplasty in that study indicated that the
hip interfered with sexual activity, relief of this difficulty was
extremely important for these patients14.
Thus, methods for evaluating the outcome of treatment that incorporate
the preferences of individual patients are important.
The Patient-Specific Index is a standardized method that allows
patients who are scheduled for a total hip arthroplasty to indicate
the type, severity, and importance of their concerns, which can
be aggregated in a single summary score. This index has been shown
to have excellent test-retest reliability and to be a valid method
for the evaluation of patients before and after total hip arthroplasty15. Furthermore, the Patient-Specific
Index has been shown to be more responsive than other hip-rating scales15. Thus, an index that focuses on
the concerns of individual patients, such as the Patient-Specific
Index, indicates even greater improvement than that detected by
other types of hip-rating scales. This increased responsiveness
is important because it decreases the sample size required in randomized clinical
trials5,12. An additional advantage
of the Patient-Specific Index is that it allows a surgeon to document
a patient's concerns and expectations of surgery and thereby provides
an opportunity for a surgeon to address any unrealistic expectations
of a patient prior to surgery15.
The main limitation of the Patient-Specific Index is the need
for an interviewer to administer the questionnaire. A self-administered
questionnaire would improve the feasibility of use of the questionnaire
by eliminating the need to train an interviewer and by decreasing
the time and cost of administration. In current protocols for same-day admission
to a hospital, a self-administered version would substantially simplify
the logistics of data collection by permitting the patient to complete
the questionnaire at home or in the physician's office prior to
surgery.
The purposes of this study were to create a self-reported version
of the Patient-Specific Index, to test the reliability of this new
self-reported version, and to determine the validity of this version
by assessing the relationship between its scores and those on the
interviewer-administered version of the Patient-Specific Index.
Formulation of the Questionnaire
Previous use of the interviewer-administered questionnaire revealed
two areas that consistently required an explanation to patients13,14. This information was used to
revise the questionnaire into a format that was easier for patients
to follow. First, the previous version of the Patient-Specific Index
contained intermixed questions about activities and symptoms. In
the self-reported version, these questions were segregated to improve
comprehension. Second, in the prior format, patients had been asked
to rate the importance of their concerns with consideration given
to how important it was to them that the symptom or disability be
"relieved after your new hip joint."14 That
wording of the question would not make sense for patients who complete
the questionnaire after total hip arthroplasty or for those treated
with interventions for hip arthritis other than total hip arthroplasty.
Thus, the wording of the question was changed to "How important
is it to be free from . . .?" A pilot study of the first draft of
the self-administered version of the questionnaire was performed
on a sample of ten patients. We arbitrarily chose ten patients for
the pilot study, with plans to expand the number of patients if
we found that the questionnaire was not readily understood. The
patients were debriefed after completing the questionnaire. They
reported that they had easily completed the questionnaire and had
understood all of the individual questions. Thus, only minor changes
in the wording of a few of the questions were required.
Scoring of the Questionnaire
The Patient-Specific Index is intended to evaluate the concerns
of individual patients before and after total hip arthroplasty.
Patients rate twenty-four concerns (and any additional ones) with
regard to severity and importance (see the Appendix). Severity was
defined as the degree of the concern, such as extreme difficulty
with stair-climbing, and importance was defined as the level of
concern. For example, although a patient may have extreme difficulty
with stair-climbing, this activity may be of minimal importance
to the patient if there are no stairs in his or her residence. The
severity of each concern was rated according to seven response categories,
which depended on the question. The importance with regard to the
relief of the concern was rated according to seven response categories, ranging
from "not important at all" to "extremely important." In previous
research, two of us (J. G. W. and N. L.Y.) compared four different
methods of combining the patients' rating of the severity and the
importance of the concern to create a summary score15. Although all of the methods provided
similar results, the simplest to use and the most responsive was
a method whereby the ratings for severity and importance were simply
added together. In the present study, we focused on this simpler
version of the Patient-Specific Index. The severity (Si) and importance
(Ii) assigned by the patient to each concern were simply added together
(Si + Ii), and these sums were totaled for all concerns to yield
a summary score (S [Si + Ii]). Thus, patients with more severe or
important concerns are rated as more disabled and symptomatic.
Three alternative methods of creating summary Patient-Specific
Index scores were also used. First, the patients' ratings of the
severity (Si) and the importance (Ii) of each concern were multiplied
together (Si × Ii), and these products were
summed across all concerns to yield the multiplicative cumulative Patient-Specific
Index score (S [Si × Ii]). Second, the total
possible maximum score was calculated as the sum of the importance
ratings (Ii) multiplied by the maximum severity score (SM) for each
concern (S [SM × Ii]). The ratio of the sum
of the patients' severity-importance products divided by the maximum score
(100 × S [Si × Ii] /
S [SM × Ii]) forms a multiplicative percentile
Patient-Specific Index score, ranging from 0 to 100, with 0 indicating
the worst rating and 100, the best. Third, a similar percentile
method was used to create the additive percentile Patient-Specific
Index. (The denominator for the additive percentile method is S
[SM + Ii].) Thus, both the multiplicative and the additive methods
have a cumulative and a percentile aggregate, yielding four possible
scores.
Reliability
In order to determine the test-retest reliability, the patients
completed the self-reported Patient-Specific Index at home twice,
two weeks apart. The interval of two weeks was chosen because the
clinical status of the patients was not expected to change substantially
during this time and the chance that patients would recall their
previous responses was minimized. Both the patients who participated
preoperatively and those who participated postoperatively were included
in the reliability testing so that the reliability coefficient would not
exclusively reflect the reliability of the preoperative assessments.
The sample-size calculation for test-retest reliability was based
on the random-effects intraclass correlation coefficient2. The intraclass correlation coefficient
(ranging between 0 and 1) is an index of concordance for continuous
data. An intraclass correlation coefficient of less than 0.4 is
considered poor, one between 0.4 and 0.75 is considered fair, and
one greater than 0.75 is considered excellent10.
On the basis of a projected reliability coefficient of 0.9 or greater
and an intent to detect an intraclass correlation coefficient of
at least 0.8 with an alpha of 0.05, a beta of 0.20, and two measurements
per subject, approximately fifty patients were required2. Although the measurements obtained
with use of hip-rating scales are more likely to be ordinal than continuous,
the intraclass correlation coefficient is easier to use with large
data sets and yields the same results as the quadratically weighted
kappa used for ordinal data3.
Validity Testing
Validity was assessed by a comparison of the scores on the self-reported
Patient-Specific Index and those on the interviewer-administered
Patient-Specific Index. Construct validity for health-measurement
scales typically is assessed on the basis of correlation9,11. However, given that the same
scale, with different modes of administration, was being compared with
itself, criterion validity or concordance was a more appropriate
and rigorous test of validity. The sample-size estimate was based
on detection of an intraclass correlation coefficient of 0.8, given
a true intraclass correlation coefficient of 0.9; it was estimated
that fifty subjects were required2.
Thus, all subjects who completed the reliability testing were included
in the validity testing.
Upon return of the second self-reported Patient-Specific Index,
subjects were contacted in order to schedule a time for interviewer
administration of the Patient-Specific Index. When possible, subjects
were interviewed in person in their home (if it was in the greater
metropolitan region), at St. Michael's Hospital (if the patient
was scheduled for a clinical visit), or by telephone. We chose several
modes of administration to reflect the various settings in which
interviewer-administered questionnaires are completed in research
and clinical situations. The decision to conduct the interview last
rather than in random order was based on a concern about a potential
teaching effect of the interview. If the interview was conducted
last, the teaching would not falsely elevate the result of the self-reported
Patient-Specific Index reliability testing. Concordance was assessed
with use of the intraclass correlation coefficient random-effects model.
Rosner stated that an intraclass correlation coefficient of more
than 0.75 would be considered excellent10.
Patients
Seventy-nine patients were considered for enrollment in the study;
five of them were ineligible. Of the seventy-four eligible patients,
fifty-five (74 percent) were enrolled and completed the study, twelve
(16 percent) could not be contacted by telephone, six (8 percent)
consented to be enrolled but did not complete the questionnaires,
and one patient (1 percent) refused to participate.
We interviewed twenty-five of the patients preoperatively and
thirty patients six months after the total hip arthroplasty. Consecutive
patients who were scheduled to have an elective primary or revision
total hip arthroplasty were considered for enrollment in the study.
Patients who could not speak English or who were to have a total
hip arthroplasty as part of a reconstruction after excision of a
tumor, after an acute fracture, or because of an infection at the
site of a previous total hip arthroplasty were excluded. All patients
had the total hip arthroplasty performed at St. Michael's Hospital by
the same surgeon, with use of the same operative approach and a
consistent perioperative protocol. This approach was used to minimize
the effect of multiple variables on the outcome of total hip arthroplasty.
Prostheses that were inserted with cement were of the Contemporary
design, while those inserted without cement were of the St. Michael's
design (both manufactured by Howmedica, Pfizer Hospital Products,
Guelph, Ontario, Canada).
All patients were questioned with regard to demographic information,
diagnosis, previous hip disease, and previous operative and nonoperative treatment.
The subjects were contacted by mail and were provided with information
on the study and with consent forms. The initial mailing was followed
by a telephone call to obtain informed consent. Those who consented
to participate in the study were mailed a package containing the
questionnaire, a formal consent form, and a postage-paid return
envelope. Upon return of the questionnaire and the written consent,
a second package was mailed with directions to open and complete the
questionnaire two weeks after completion of the first questionnaire.
The second package included the same material as well as a card
on which the patient was asked to indicate whether his or her physical
function had changed since completion of the first questionnaire.
The patients were instructed to provide their own answers. The study
received approval by the Institutional Review Board of the University
of Toronto and St. Michael's Hospital, and all patients provided
informed consent.
We interviewed twenty-five patients preoperatively and thirty
patients six months after the total hip arthroplasty. The mean age
of the fifty-five enrolled patients was 65.6 years (range, twenty-eight to
eighty-four years). The mean age of the unenrolled patients was
67.5 years (range, thirty-nine to eighty-five years). Twenty-three
(42 percent) of the enrolled patients were male. Forty-seven patients
(85 percent) had had a primary total hip arthroplasty. Thirty-four
patients (62 percent) were married, nine (16 percent) were never
married, five (9 percent) were widowed, four (7 percent) were divorced
or separated, and three (5 percent) responded by indicating "other"
(for example, a nun) or did not provide a response. Forty-two patients
(76 percent) lived with someone, and thirteen (24 percent) lived
alone. Seventeen patients (31 percent) indicated that their financial
resources were limited; thirty (55 percent), that their lifestyle was
moderately comfortable; and eight (15 percent), that it was very
comfortable. These results demonstrated the heterogeneity of the
study population, which supports the generalizability of the results.
Because comparisons between groups were not performed, we have not
provided stratified descriptive statistics.
The twenty-five patients in the preoperative group were interviewed
at a mean of 2.6 months (range, 0.4 to 5.7 months) prior to surgery.
The thirty patients in the postoperative group were interviewed at
a mean of 9.5 months (range, 0.3 to 12.9 months) following the procedure.
Interviewer-administered questionnaires were conducted by telephone
for thirty patients (55 percent) and in person for twenty-five (45
percent). The questionnaires were mailed to the patients in advance.
Of the twenty-five in-person interviews, twenty-two (88 percent)
were performed in the patient's home and three (12 percent) were
performed in a clinical setting. It took a mean of eighteen minutes
(range, four to sixty-eight minutes) in order to complete the self-reported
Patient-Specific Index questionnaire and a mean of eighteen minutes
(range, two to fifty-five minutes) in order to complete the interviewer-administered
Patient-Specific Index questionnaire.
The patients were asked about any change in their clinical status
between the administration of the questionnaires to determine test-retest
reliability. Thirty-four patients (62 percent) indicated that there
had been no change; ten (18 percent), that their condition was slightly
better; and eleven (20 percent), that their condition was slightly
worse. The mean difference (and standard deviation) in the Patient-Specific
Index scores was 0.9 ± 3.8 for those who
were slightly better, -3.3 ± 7.9 for those
who had no change, and -6.6 ± 12.9 for those
who were slightly worse. The test-retest reliability of the self-reported
Patient-Specific Index was 0.79, suggesting excellent reliability
as defined by Rosner10 (Fig. 1).
The mean differences between the scores on the self-reported
Patient-Specific Index and the scores on the interviewer-administered
Patient-Specific Index were small (Table I) and were not found to be significant
with the Student t test (additive cumulative scores, p = 0.45; multiplicative
cumulative scores, p = 0.64; multiplicative percentile scores, p
= 0.45; and additive percentile scores, p = 0.59). The concordance
of the two methods of administration was 0.78, suggesting excellent
agreement (Table II and Fig. 2).
The Patient-Specific Index is a method of outcome assessment
that addresses the concerns of individuals. Attention to the concerns
of individual patients is particularly relevant to orthopaedics because
many adult orthopaedic procedures are elective and are directed
toward the amelioration of the patient's symptoms and disability6. The first prerequisite of a scale
is reliability - that is, the same result is obtained when the measure
is repeated at different times and the patient's clinical status
remains stable13. The test-retest
interval of two weeks was chosen because we believed that this would
minimize the chances that the patients would recall their previous
responses and that not many patients would be expected to have a
change in clinical status during this interval. However, even with
an interval of only two weeks, 38 percent of the patients indicated
that their condition had either slightly improved or worsened. If
the clinical status of the patients had truly changed, inclusion
of these patients in the reliability testing would have reduced
the reliability coefficient. Despite the inclusion of these patients,
the reliability coefficient was excellent10,
indicating that the self-reported version was reliable.
After reliability is established, a measure must be shown to
be valid and responsive13. In
a prior study, we established the validity of an interviewer-administered
questionnaire15. Responsiveness,
the ability to accurately measure change over time, was not evaluated
in the present study because it had been evaluated in a previous investigation15. The interviewer-administered questionnaire
can be administered in person or by telephone. Administration by
telephone is useful when patients are unable to return for an in-person
assessment because of geographic distance or ill health. We initially
chose the interviewer mode of administration because we were concerned
about the ability of patients to understand and distinguish the
concepts of the importance and the severity of their concerns. However,
we hoped ultimately to develop a self-reported version. The purpose
of the present study was to improve the feasibility of the use of
the questionnaire. We compared the scores on the self-reported version with
the scores on the interviewer-administered version to assess criterion
validity. The concordance of the two modes of administration was
excellent, indicating that the self-reported Patient-Specific Index
was valid.
The advantages of interviewer-administered questionnaires are
that the interviewer is available to ensure that questions are interpreted
appropriately and that the patient is the actual respondent to the questionnaire.
The two main disadvantages of interviewer-administered questionnaires
are a decreased feasibility of utilization and a concern about how
the interviewer may affect the patient's responses to the questions.
In general, interviewer-administered questionnaires are much more
expensive to administer because it is often necessary to train the
interviewer and to pay him or her to administer the questionnaire.
Many clinicians, outside of sponsored research studies, are unable
to afford or to identify an appropriate person to administer the
questionnaire. Furthermore, most clinical practices either do not
have the space that needs to be set aside to interview patients
or do not have the personnel to conduct interviews in the patients'
homes. Patients who move away, either permanently or temporarily
(for example, elderly patients who travel south in the winter),
may also be inaccessible for interviewer-administered questionnaires.
Thus, for many practical reasons, self-reported questionnaires are
preferred over interviewer-administered questionnaires.
As mentioned, a second potential concern with regard to any interviewer-administered
questionnaire is how an interviewer affects the patient's responses
to the questionnaire. Social desirability bias occurs when a respondent
answers questions with what he or she thinks are acceptable answers1. This phenomenon is of particular
concern in orthopaedics because many patients may want to please
their surgeons and thus underrate their concerns. Previous research
has shown that patients portray themselves as much better when they
talk with their surgeons8. In
the present study, we were reassured by our finding of little difference
between the scores for interviewer-administered questionnaires and
those for self-reported questionnaires. The fact that the interviewer
was a therapist in a research setting with no direct ties to the
surgeon may explain why patients provided similar responses with
the two modes of administration.
The main limitation of our study is the generalizability of the
results obtained from the patients of a single surgeon. We chose
a single surgeon to minimize extraneous factors that might affect
the patients' scores. Furthermore, the patients in this study were
typical of patients undergoing total hip arthroplasty, and we have
no reason to believe that the results would not apply to patients
in other settings. However, this study should be repeated in other
contexts to ensure the generalizability of the results. Also, the
questionnaire is appropriate only for patients who are fluent in
written English. The questionnaire would need to be translated before being
used for patients who speak other languages4.
In conclusion, a self-reported version of the Patient-Specific
Index is a reliable and valid method for the assessment of the individual
concerns of patients who have had or are scheduled to have a total
hip arthroplasty.
Self-Reported Patient-Specific Index
The questions in the tables on the following pages ask about
symptoms you may be experiencing, aspects of disease that may be
bothersome (or annoying) to you, activities that you may have difficulty
with, and the relative importance of these activities.Put an "X"
in one box beside each question that best represents your hip function
during the past month.For example, we might ask:
Question: How severe has your dizziness been
on average this month?
Answer: If during the past month you were moderately
dizzy twice daily, but only for brief periods, you might consider
your dizziness to be "somewhat severe."Sample table: