The Patient-Specific Functional Scale: measurement properties in patients with knee dysfunction.

Health care policy makers and clinical researchers are usually interested in acquiring data on groups, whereas clinicians are usually most interested in obtaining information and making decisions concerning individual patients. Both generic and condition-specific health status measures have been developed to assist in efforts to acquire data on groups.[1-7] More recently, there has been a growing interest in using health status measures at the individual patient level.[8-13] In support of this position, Feinstein and colleagues have suggested that "if the index is intended to demonstrate the patient's improvement, the patient's concept of what should be improved may often be much more cogent than the particular beliefs held by the health care team."[14] In response to this need, a third type of health status measure has emerged, the patient-specific measure. The goal of a patient-specific measure is to aid clinicians in making decisions about the change in health or functional status of individual patients. Several patient-specific measures such as the MacKenzie Questionnaire,[8] the MACTAR Questionnaire,[9] and the Patient-Specific Functional Scale (PSFS)[12] have been described.

The MACTAR Questionnaire was conceived by Tugwell and colleagues[9] with the goal of assessing improvement in physical disability in patients with rheumatoid arthritis. The MACTAR Questionnaire asks patients to (1) identify activities that they are having difficulty with around the house, at work or outside the home, and during social or recreation activities and (2) rank order the activities according to their importance. At subsequent visits, each activity is reviewed and patients are asked: (1) "Have you noticed any change in your ability to [activity identified]?" and (2) "If yes, has your ability to [activity identified] improved or become worse?" Responses on the MACTAR Questionnaire have been shown to be more sensitive to change than those of conventionally structured questionnaires.[9]

MacKenzie et al[8] have devised and tested a patient-specific measure targeting change in maximal function. Their measure addresses physical activity, concentration, and emotional well-being. For example, the physical activity component includes the queries: (1) "Of the things that you usually do, what physical activity requires the most physical exertion or is the most physically strenuous?" and (2) "On your good days, what would be the most [physical activity from question 1] that you could do?" At follow-up, patients are asked to assess change using a transitional approach. Patients are asked (1) whether they are better or worse (five-point scale), (2) whether they are able to perform the activity performed at the previous assessment, (3) whether this activity is the most physically strenuous activity they are able to perform, and (4) to identify the most physically strenuous activity they are now able to perform. This measure was studied on 83 patients with a variety of medical and postsurgical conditions and was shown to be reliable and valid.[8] Neither the MACTAR Questionnaire nor the MacKenzie Questionnaire quantify the amount of disability on the patient-generated activities. Instead, both questionnaires focus on assessing change using a transitional scale in which the patient is asked whether he or she is better or worse that at the previous assessment. In recognition of this limitation, the PSFS was conceived.

The PSFS was developed by Stratford and colleagues[12] to provide a method for eliciting, measuring, and recording descriptions of patients' disabilities. It can be used to guide treatment and assess patient outcome. The PSFS is intended to complement the findings of generic or condition-specific measures. Generic and condition-specific measures were conceived to make decisions regarding function in groups of patients, such as in clinical trials. A shortcoming of such measures is that they are limited in detecting small but important disabilities and changes in disability over time.[2,7,14] For example, consider a runner who experiences knee discomfort only after running several miles over hilly terrain. A condition-specific or generic scale may incorporate many questions that are not sensitive to this problem. A patient-specific measure, however, would best assist the clinician in documenting and evaluating change concerning this specific problem.

The goals that were considered when developing the PSFS were (1) that it be efficient and easy to administer, (2) that it be easy to record in the medical record, (3) that it yield reliable measurements, (4) that it yield valid measurements, including that it assess important change over time, (5) that it provide a comparison of a patient's specified important activity level at any given point in time with respect to the predisability state, and (6) that it be applicable to a large number of clinical presentations (eg, conditions, diseases, problems, ages). In brief, the PSFS is administered at the initial assessment, during the history taking, and prior to the assessment of any impairment measures. The rationale for administration prior to the physical examination is to maximize the patient's focus on function ("I have difficulty walking down stairs") rather than impairment ("I can't flex my knee"). Patients are asked to identify up to five important activities that they are having difficulty with or are unable to perform (Figure). In addition to specifying the activities, patients are asked to rate, on an 11-point scale, the current level of difficulty associated with each activity. The scale anchors are 0 ("unable to perform activity") to 10 ("able to perform activity at same level as before injury or problem"). The clinician's role is to read the script (instructions) to the patient and record the activities, the corresponding numerical difficulty ratings, and the assessment date. At subsequent reassessments, the clinician reads the follow-up script, which reminds the patients of the activities that they identified previously. Once again, the clinician records the ratings specified by each patient and the date. The form provides space for additional activities to be added. Because patients identify between one and five activities and this activity set is unique to each patient, the PSFS is not a comprehensive measure of disability and was not designed to compare disabilities among patients.

The PSFS has been studied on patients with low back pain and was demonstrated to be reliable (intraclass correlation coefficient [ICC]=.97), valid (individual activity-specific correlations with the Roland-Morris Questionnaire, a condition-specific measure, varied from .55 to .74), and sensitive to change over time (Norman's [S.sub.repeat] of .70 to .81).[12] Norman's [S.sub.repeat] coefficient is an ICC that is derived by dividing the group X time variance by the group X time variance plus error variance. A factor that may have influenced the magnitude of these coefficients is that patients were informed of their initial activity scores prior to specifying activity-specific numerical ratings at the follow-up assessment. Guyatt et al[15] have suggested that informed ratings increase the reliability and ultimately the sensitivity to change of a measure.

The intent of our study was to further explore the measurement properties of the PSFS on patients with knee joint dysfunction. Specifically, the purpose was to determine test-retest reliability, construct validity, and sensitivity to change of the PSFS when applied to patients with knee dysfunctions. Sensitivity to change was compared between the PSFS and a generic health status measure, the Medical Outcomes Study 36-Item Short Form Health Survey (SF-36).[2-4]

Method

Sample

The sample consisted of 38 patients with complaints of knee dysfunction who were referred by physicians to one of three physical therapy clinics. All 38 patients participated in the reliability portion of the study, and 32 patients completed the validity and valid change portions of the study. Knee dysfunction was defined as any orthopedic condition of the knee that contributed to a decreased level of function as perceived by the patient. The patients' conditions are listed in Table 1. The sample consisted of 20 women and 18 men whose mean age was 47 years (SD=18, range=19-84) and whose average duration of knee dysfunction was 11.6 weeks (SD=13.3, range=1-52). Non-English-speaking patients and those who were unable to verbally respond to a questionnaire were excluded from the study. Prior to participating in the study, all patients signed a consent form that was approved by the Institutional Review Board of North Georgia College. The facilities were outpatient physical therapy clinics in two north Georgia hospitals and one private practice in Hamilton, Ontario, Canada.

Table 1.
Summary of Patients' Conditions

Condition                                   Frequency

Patellofemoral pain                         12
Osteoarthritis                               7
Total knee arthroplasty                      4
Medial meniscectomy/meniscal repair          4
Collateral ligament sprain                   3
Anterior cruciate ligament reconstruction    3
Arthroscopic loose-body debridement          2
Pes ancerinus bursitis                       1
Patellar/quadriceps tendinitis               2

Measures

PSFS. This measure was described in the introduction, and an example of it appears in the Figure.[12] For the purpose of our study, patients were not asked about additional activities at follow-up.

SF-36. The SF-36 is a multidimensional generic health status instrument that consists of eight health concept scales: (1) physical functioning, (2) role limitation (physical), (3) bodily pain, (4) general health, (5) vitality, (6) social function, (7) role limitation (emotional), and (8) mental health.[2-4] In total, the instrument contains 36 items that represent a broad array of health concepts. All scales are linearly transformed to a 0-to-100 scale, with 100 indicating the most favorable health state. The psychometric properties of the SF-36 have been established on samples from diverse populations, including persons with knee problems.[2-6,16] Of particular interest is a retrospective study performed by Katz et al[16] in which the SF-36 and the Lysholm Knee Scoring Scale were examined on 105 patients (70 male, 35 female) following arthroscopic partial meniscectomy. In that study, the two scales were found to be moderately correlated preoperatively (r =.43) and highly correlated (r=.70) postoperatively. These researchers suggested that there may be several advantages associated with the SF-36 compared with the Lysholm scale. These advantages include the following: (1) Patients have difficulty answering questions on instability and pain, which comprise 50% of the Lysholm scale, (2) patients find the SF-36 easier to understand and complete correctly, and (3) physical function as measured by the SF-36 is more relevant to the patient's perception of surgical results than impairments such as locking or instability, which comprise approximately 40% of the Lysholm scale.[16] The SF-36 was designed for self-, telephone, and interview administration formats.[2-4] The acute version of the SF-36 was used in our study.

Global Rating of Change Scale (GRC). This tool was used to provide an external standard for change. This scale consists of two questions, each of which are scored on a 15-point scale (scores can vary from +7, indicating improvement, to -7, indicating deterioration). One question inquires about the magnitude of change, and the other question asks about the importance of change. Both questions are answered by the patient and the clinician, each being unaware of the other's response. The GRC has been described previously.[12,17,18]

Study Design

The PSFS and the SF-36 were administered at the time of the initial evaluation and 2 to 3 weeks later at a follow-up assessment. The PSFS was also administered within 48 to 72 hours following the initial assessment. The initial and follow-up PSFS ratings were done by the same clinician. To simulate realistic clinical conditions and in light of the ease and simplicity of administration, special training beyond reading the script and clarifying the process to clinicians was not done. The GRC was completed by patients and clinicians at the 2- to 3-week follow-up assessment. It is not possible to administer the PSFS in a manner in which clinicians are blinded; however, patients' SF-36 scores were not known by the clinicians administering the PSFS.

Test-retest reliability of the PSFS was estimated by comparing the initial PSFS score with a second PSFS score obtained 48 to 72 hours following the initial assessment. The assumption was that, on average, the patients' functional status would be stable over this interval. A comparison of the SF-36 and PSFS scores at two points in time (initial assessment and 2- to 3-week follow-up) allowed the assessment of construct convergent validity.

Sensitivity to change is a form of validity. Sensitivity to change is defined as the capacity of a test or measure to distinguish among patients or groups of patients whose health status has improved, deteriorated, and remained stable and to quantify the amount of true change when it has occurred. There is no gold standard against which to evaluate the sensitivity to change of a health status measure. Methodology to examine this attribute has relied heavily, therefore, on the use of various constructs, or hypotheses, for change, against which a health status measure is tested. These constructs for change vary in strength and, therefore, have been used to develop study designs with varying degrees of rigor. For example, a popular but relatively weak construct for change is that a patient's health status will improve with time or treatment. In this design, health status is measured at admission and at a defined follow-up point and amount of change is determined (designs 1a, 1b).[18] A much stronger construct for change is that patients randomly assigned to a group given a treatment that is known to be effective will demonstrate greater change than will patients receiving a placebo treatment (design 2a).18 In the absence of a gold standard, using multiple constructs to evaluate the sensitivity to change of a health status measure would appear to be the most appropriate research model. In our study, sensitivity to change was evaluated using two constructs: (1) that change scores should correlate with patient- and clinician-reported global ratings of change (design 2c) and (2) that the patient's ability to perform easy functional activities should improve more than the ability to perform difficult functional activities (design 2b).[10]

The sensitivity-to-change analysis that involved correlation with the GRC used a 2- to 3-week follow-up time frame. This approach was based on our experience that the functional status of patients changes to varying degrees during this interval. Variation in the extent to which functional status changes is an essential requirement when assessing a measure's capacity to detect change.[18] Although the subjects were actively participating in physical therapy programs during the study, the intent of the study was not to examine the effectiveness of the intervention but to study the capability of the PSFS and the SF-36 to measure change. There was no attempt, therefore, to control interventions.

Procedure

Numbered packets consisting of data collection forms were distributed to the three facilities.. Each packet contained data collection guidelines, a consent form, a demographics form, two copies of the SF-36, a PSFS, and GRC scales for the patient and clinician. All instruments

AB Chatman, PT, is Physical Therapist, Promina Northwest-Kennestone Hospital, Atlanta, Ga.

SP Hyams, PT, is Physical Therapist, Orange Park Physical Therapy Clinic, Jacksonville, Fla.

JM Neel, PT, is Physical Therapist, District Memorial Hospital, Andrews, NC.

JM Binkley, PT, COMP, FAAOMPT, is Physical Therapist, Appalachian Physical Therapy, 109A Tipton Dr, Dahlonega, GA 30533, and Assistant Professor (part-time), School of Rehabilitation Science, McMaster University, Hamilton, Ontario, Canada. Address all correspondence to Ms Binkley at the first address ([e-mail]).

PW Stratford, PT, is Assistant Professor, School of rehabilitation Sciences, and Associate Member, Department of Clinical Epidemiology, McMaster University, Hamilton, Ontario, Canada.

A Schomberg, PhD, PT, is Assistant Professor, Graduate Physical Therapy Program, North Georgia College, Dahlonega, Ga.

M Stabler, PT, SCS, is Senior Physical Therapist, Northeast Georgia Medical Center, Gainesville, Ga.

Mr Chatman, Mr Hyams, and Mr Neel were students when this study was conducted in partial fulfillment of the thesis requirement for their Master of Science (Physical Therapy) degree at North Georgia College.

This study was approved by the Institutional Review Board of North Georgia College. were verbally administered to the patients by one of six clinicians (five physical therapists and one certified athletic trainer). All clinicians had at least 1 year in practice and were working in outpatient orthopedic settings. Patients were not informed of their prior numerical responses on either the PSFS or the SF-36, whereas in the prior study on persons with low back pain, informed ratings were used. Patients were informed of the activities that they had identified on the PSFS. We decided to use uninformed ratings to obtain a more conservative (larger) estimate of measurement error.

Data Analysis

Summary statistics, including means and standard deviations, were calculated for all PSFS activities coded by the order in which the activities were stated by the patient (eg, the first activity mentioned was specified as "activity 1," the second activity mentioned was specified as "activity 2," and so on). An average PSFS score was computed for each patient (ie, sum of activity scores divided by the number of activities the patient specified). Means and standard deviations were also calculated for each of the eight SF-36 dimensions.

Intraclass correlation coefficients (type 2,1) and their corresponding 95% confidence intervals (CIs) were computed for individual activity scores and for the average activity score.[19] Standard errors of measurement (SEMs) and minimum levels of detectable change (significant change index) at the 90% confidence level were calculated.[13] The SEM quantifies measurement error in the same units as the original measurement, in this case, points on the PSFS. Minimum detectable change (MDC) at the 90% confidence level was obtained by multiplying the SEM by [square root]2 and the associated tabled Z value for the 90% confidence level (two-tailed value=1.65).[13] The interpretation is that for a clinician to be confident that a true change has occurred, the observed change in the patient's functional status must be greater than the MDC. The MDC is a variation on the reliability change index described by Ottenbacher and colleagues.[20]

Given that most validity analyses are based on differentiating among patients or groups of patients, assessing the validity of an instrument that was not intended for among-patient comparisons poses a methodological challenge. In our study, we examined the validity of the PSFS at a given point in time and its ability to detect change over time.

The estimate of validity at a single point in time was assessed using two approaches to construct convergent validity.[21] The first approach was the traditional approach of differentiating among patients. Using this approach, we hypothesized that the PSFS activity score assessed at follow-up should demonstrate correlations of about .30 to .50 on the physical function and bodily pain dimensions of the SF-36. We also hypothesized that there would be little to no correlation between the PSFS and the mental health and role limitation (emotional) dimensions of the SF-36. Follow-up scores were targeted because we anticipated a restricted range in initial scores (ie, we hypothesized that most patients would identify activities that were moderately to severely restricted). Pearson correlation coefficients and their respective 95% CIs were calculated to address the association between the PSFS and SF-36 dimension scores at follow-up.[21] The second approach focused on the PSFS as a tool designed to supplement within-patient, rather than between-patient, decisions. The within-patient analysis was based on the premise that when a patient identified activities where an obvious easy-difficult item pairing was provided, the more difficult activity would have a lower PSFS score.

Sensitivity to change was examined using two constructs for change. The first construct specified that when a within-patient activity pairing existed where a sequential recovery can be expected, the change over a relatively short interval, such as that between the initial evaluation and 2- to 3-week follow-up assessment, will be greater for the "easier" activity. Because activity-specific difficulty may vary from patient to patient depending on the presenting condition, our definition of an easy-hard activity pairing was restricted to only those patients who provided paired activities that included walking and running, or walking and a vigorous sports activity. A three-factor (subjects, activity difficulty, and time) repeated-measures analysis of variance was used to test the within-patient analysis pertaining to concurrent and longitudinal validity (ie, change).[21] The validity for a single point in time was tested based on the main effects "activity difficulty" term, and longitudinal validity was evaluated using the "time X activity difficulty" interaction term. Statistical significance was set at the P=.05 level.[21] Sensitivity to change was also assessed by examining the association between the PSFS and SF-36 change scores (ie, follow-up score minus initial score) and the pooled (average of patient's and clinician's scores) GRC score.[18] Using this approach, longitudinal validity was quantified using the Pearson correlation coefficient and 95% CI.[18] The magnitude of the PSFS change correlation coefficient was compared with the change coefficients of the SF-36. Because the measures were dependent, the approach outlined by Williams et al[22] was used.

Results

On average, the PSFS took subjects and therapists approximately 4 minutes (SD=1.9) to administer and the SF-36 took subjects approximately 13 minutes (SD=4.5) to complete. Table 2 provides a summary of the initial PSFS and SF-36 dimension scores. A review of Table 2 indicates that fewer than half of the patients were able to state five activities. For example, at follow-up, only 13 of 32 patients stated five activities.

Table 2. Summary Statistics

                                Initial Test      Retest
Scale/Item                      X      SD    N    X     SD    N
PSFS(a)
  Activity 1                    3.0    2.4   38   3.2   2.6   36
  Activity 2                    2.7    2.6   37   2.9   2.2   35
  Activity 3                    3.1    2.1   36   3.2   2.5   34
  Activity 4                    3.2    2.7   33   3.6   2.6   31
  Activity 5                    3.1    2.2   15   3.3   3.2   14
  Average                       3.1    1.8   38   3.3   1.7   36
SF-36(b) (n=32)
  Physical function             38.1   23.9
  Role limitation (physical)    18.0   31.3
  Bodily pain                   42.4   17.4
  General health                67.8   23.7
  Vitality                      49.7   24.3
  Social function               52.3   28.3
  Role limitation (emotional)   63.5   41.8
  Mental health                 69.8   20.4

                                Follow-up
Scale/Item                      X      SD    N
PSFS(a)
  Activity 1                    6.3    2.4   32
  Activity 2                    5.3    2.7   31
  Activity 3                    5.4    2.9   30
  Activity 4                    5.9    3.1   29
  Activity 5                    6.3    3.1   13
  Average                       5.9    1.8   32
SF-36(b) (n=32)
  Physical function             62.0   20.0
  Role limitation (physical)    44.5   36.3
  Bodily pain                   62.2   17.0
  General health                68.0   21.2
  Vitality                      63.9   18.8
  Social function               79.2   23.9
  Role limitation (emotional)   81.7   35.3
  Mental health                 80.4   14.6

(a) PSFS = Patient-Specific Functional Scale. (b) SF-36 = Medical Outcomes Study 36-Item Short-Form Health Survey.

The PSFS individual activity test-retest reliability coefficient was R=.84 (95% CI=0.78-0.88). The test-retest reliability based on the average score per patient was .87 (95% CI=0.76-0.93). The variance due to time in the type (2,1) reliability analyses was zero, indicating that the small improvement in function between the test and the retest was negligible. The SEM was 1.0 PSFS points (90% confidence level MDC=2.5 PSFS points) for the individual activity analysis and 0.62 PSFS points (90% confidence level MDC=1.5 PSFS points) for the average score analysis.

Table 3 presents the validity coefficients for a single point in time between the PSFS and the SF-36. As hypothesized, the correlations between the physical function and bodily pain dimensions were in the range of .30 to .50 and were greater than those of role limitation (emotional) and mental health. Only nine patients provided activity comparisons that included walking and running/vigorous sports activities. A summary of the mean scores for these patients is provided in Table 4. The results show that "easier activity" scores were greater than the "harder activity" scores at both initial and follow-up assessments (repeated-measures "activity difficulty" P[greater than].001, P[greater than].05 at both points in time). Moreover, it is evident that the amount of change for the "easier activity" was greater than the amount of change for the "harder activity" (repeated-measures "time X activity difficulty" P[greater than].026).

Table 3.
Initial and Follow-up Coefficients Used for Demonstrating
Validity(a)

                              Initial Correlation  Correlation With
SF-36 Dimension               With PSFS (95% CI)   PSFS at Follow-up
                                                   (95% CI)

Physical function             .34(0.05-0.58)       .49(0.17-0.71)
Bodily pain                   .12(-0.18-0.40)      .40(0.06-0.66)
Social function               .32(0.03-0.56)       .36(0.01-0.63)
Vitality                      .24(-0.06-0.50)      .35(0.00-0.62)
Role limitation (physical)    .39(0.11-0.62)       .33(-0.02-0.61)
General health               -.11(-0.39-0.19)      .20(-0.16-0.51)
Role limitation (emotional)   .13(-0.17-0.41)      .07(-0.29-0.41)
Mental health                 .23(-0.07-0.49)      .09(00.22-0.47)

(a) SF-36 = Medical Outcomes Study 36-Item Short-Form Health Survey, PSFS = Patient-Specific Functional Scale, CI = confidence interval.

Table 4.
Within-Patient Concurrent and Longitudinal Validity Scores (n=9)

                 Initial Score   Follow-up Score   Average Score
                 X       SD      X      SD         X       SD

Easier activity  2.1     1.5     5.6    2.0        3.8     2.5
Harder activity  0.2     0.7     0.8    1.6        0.5     1.2
Average score    1.2     1.5     3.2    3.0        2.2     2.6

Coefficients for sensitivity to change, as determined by correlating the PSFS and the SF-36 subscale scores with the GRC, respectively, are reported in Table 5. The PSFS correlation coefficient was greater (P[greater than].002) than the coefficients for the eight SF-36 dimensions.

Table 5.
Sensitivity-to-Change Coefficients for SF-36(a) and PSFS(b)

                                 Correlation With Global
Scale                            Rating (95% CI(c)

Sf-36 dimension
  Physical function              .59 (0.30-0.78)
  Bodily pain                    .46 (0.13-0.70)
  Role limitation (physical)     .33 (-0.02-0.61)
  Vitality                       .26 (-0.10-0.56)
  Role limitation (emotional)    .24 (-0.12-0.54)
  Social function                .12 (-0.24-0.45)
  Mental health                  .12 (-0.24-0.45)
  General health                -.06 (-0.29-0.40)
PSFS                             .77 (0.61-0.89)

(a) SF-36 = Medical Outcomes Study 36-Item Short-Form Health Survey. (b) PSFS = Patient-Specific Functional Scale. (c) CI = confidence interval.

Discussion

The goals in developing the PSFS were (1) that it be efficient and easy to administer, (2) that it be easy to record in the medical record, (3) that it yield reliable measurements, (4) that it yield valid measurements, (5) that it assess important change over time, (6) that it provide a comparison of a patient's specified important activity level at any given point in time with respect to the predisability state, and (7) that it be applicable to a large number of clinical presentations (eg, conditions, diseases, problems, ages). The results of this study appear to support the first two goals in developing the instrument in that the PSFS can be administered and recorded in the medical record in a short period of time (approximately 4 minutes).

Test-retest reliability, as judged by the ICC, for both an individual activity and an average score is excellent. When reliability is expressed in PSFS points, good reliability exists. For example, at the individual activity level, a clinician can be confident (90% confidence level) that a true change in a patient's functional status has occurred when the observed change is equal to or greater than 3 PSFS points. If the clinician elects to make a decision based on the average PSFS score per patient, a change of only 2 PSFS points is required. Although the test-retest reliability at the individual activity level is quite good, it is less than that of .97 obtained when informed ratings (ie, patients were reminded of their initial scores when asked to evaluate disability at follow-up) were provided to patients with low back pain in a previous study.[12]

In the absence of a gold standard against which to judge measures, assessing the validity of health status measures has always posed a methodological challenge. The solution to this problem has relied heavily on construct validity. Typically, scores on the measure under investigation have been correlated with other measures believed to be assessing the same concept, and moderate to good correlations provided support of the new measure's ability. Although this methodology seems reasonable for measures that sum individual item scores to form a single score, it is somewhat problematic for a patient-specific measure such as the PSFS, particularly when the goal is to make decisions concerning the results of individual activities within a patient and not to compare the amount of disability among patients. Thus, although electing to use the traditional approach of correlating the PSFS scores or change scores with the SF-36 scores, we anticipated that the magnitude of the correlation coefficient would be in the range of.30 to.50 at best. The results shown in Table 3 support the notion that the PSFS should not be used to make comparisons among patients. Correlations of follow-up, scores between the PSFS and the physical function, bodily pain, and social function dimensions of the SF-36 were demonstrated, whereas no correlations were noted for the other dimensions of the SF-36. The within-patient validity analysis at a given point in time showed that the PSFS was able to differentiate among activities of varying difficulty at both the initial and follow-up evaluations. Moreover, the PSFS was able to detect varying amounts of change between activities within patients. These findings support the validity of the PSFS with respect to within-patient decision making.

There are limitations imposed when assessing change in a functional status measurement in the absence of a gold standard of change. A popular method is to correlate the measure under investigation with a patient or clinician rating of change.[17] Although we used this approach in our study, there are several limitations. The change measured in either the scale in question or the GRC may be due to inherent error in the instrument(s) or true functional change. Errors on the PSFS may be correlated with errors on the GRC, and if so, this will upwardly bias the estimate of the overall correlation between these measures. In addition, the GRC requires patients to recall their initial state and compare it with their current state. Ross[23] suggests that judgments are based on an implicit theory of change. This theory specifies that estimates of change are derived by beginning with the present state and working backward.[23] Thus, patients may place greater emphasis on their current state, which may bias recall of their initial state. Concurrent examination of several constructs for change, such as the easy-hard pairing reported here, is desirable in light of these potential limitations in the GRC.

Our work shows that at least two health dimensions of the SF-36, physical function and bodily pain, can be used to detect change in patients with knee dysfunction. Our results demonstrate that a patient-specific measure is more adept at assessing change over time compared with generic health status measures.[9] A comparison among correlation coefficients showed that the PSFS correlation with the GRC was better than that of the generic SF-36 subscale measures with the GRC. The likely reason for this finding is that all activities identified by a given patient on a patient-specific measure represent disabilities that are important to the individual. On a generic measure, all items or subscales may not be relevant to the patient. Thus, items that are not limited or important initially would not be valid indicators of improvement over time.

A potential limitation of the PSFS is its possible floor effect. A floor effect exists when there is little or no space available on the scale for patients to demonstrate deterioration. In our study, the mean initial score was 3.2 PSFS points, and this value is consistent with that found in a previous study on patients with low back pain (3.6 PSFS points).[12] Patients tend to identify activities where substantial disability exists; therefore, there is little range available on the scale for the patient to describe increased disability. Moreover, the choice of the lower anchor point on the scale, "unable to perform activity," seemingly makes it impossible to be worse off than this with respect to the identified activity. For this reason, other strategies for detecting deterioration need to be explored when the PSFS is to be used on patients where deterioration is a likely, outcome. One strategy for dealing with the floor effect that is now incorporated into the PSFS is extending the number of activities and asking patients to also state several activities that they are having "just a little bit of difficulty with" at the time of the initial assessment.

The PSFS may assist clinicians in planning treatment and making decisions on continuation of treatment. Two examples of activities indicated by patients with knee dysfunction were "marching with my tuba" and "going up and down the riverbank to fish." These activities were the highest-priority functional limitations reported by each of these patients. Using this information and regular follow-up with the PSFS, intervention and outcome measurements could be targeted at achieving these critical tasks. Ideally, we see the PSFS as a measure that would supplement condition-specific or generic health status measures currently being used by clinicians to enhance clinical decision making when the goal is to measure change on an individual patient. The PSFS may provide an efficient and practical tool for documenting change and outcome on individual patients in cases where clinicians are not yet using either condition-specific or generic health status measures as routine outcome assessment tools.

Conclusions

Disability and change in disability are important aspects of a patient's assessment. Interviewer-assisted and self-report questionnaires are popular methods for achieving this goal. The reliability, validity, and sensitivity to change over time of the PSFS have been studied in two patient groups (outpatients with knee and low back dysfunction). The results to date suggest that the PSFS is a time-efficient and appropriate tool when the goal is the assessment of change in disability. Further investigation is needed to determine the extent to which the PSFS can be applied across a variety of conditions and age groups. The PSFS complements other condition-specific health status measurement scales. In the clinical setting, when the goal is to measure change on individual patients, the PSFS may be an important component of the functional assessment.

References

[1] Bergner M, Bobbitt RA, Carter WB, Gilson BS. The Sickness Impact Profile: development and final revision of a health status measure. Med Care. 1981;19:787-805.

[2] Ware JE, Sherbourne CD. The MOS 36-Item Short-Form Health Survey (SF-36), I: conceptual framework and item selection. Med Care. 1992;30:473-483.

[3] McHorney CA, Ware JE, Raczek AE. The MOS 36-Item Short-Form Health Survey (SF-36), II: psychometric and clinical tests of validity in measuring physical and mental health constructs. Med Care. 1993;31: 247-263.

[4] McHorney CA, Ware JE, Lu R, Sherbourne CD. The MOS 36-Item Short-Form Health Survey (SF-36), III: tests of data quality, scaling assumptions, and reliability across diverse patient groups. Med Care. 1994;32:40-66.

[5] Bombardier C, Melfi CA, Paul J, et al. Comparison of a generic and a disease-specific measure of pain and physical,function after knee replacement surgery. Med Care. 1995;33:AS131-AS144.

[6] Kantz MW, Harris MI, Levitshy K, et al. Methods for assessing condition-specific and generic functional status outcomes after total knee replacement. Med Care. 1992;30:MS240-MS252.

[7] Roland M, Morris R. A study of the natural history of back pain, part I: development of a reliable and sensitive measure of disability in low-back pain. Spine. 1983;8:141-144.

[8] MacKenzie CR, Charlson ME, DiGioia D, Kelley K. A patient-specific measure of change in maximal function. Arch Intern Med. 1986;146:1325-1329.

[9] Tugwell P, Bombardier C, Buchanan WW, et al. The MACTAR patients preference disability questionnaire: an individualized functional priority approach for assessing improvement in physical disability in clinical trials in rheumatoid arthritis. J Rheumatol. 1987;14:446-451.

[10] McHorney CA, Tarlov AR. Individual-patient monitoring in clinical practice: Are available health status surveys adequate? Qual Life Res. 1995;4:293-307.

[11] Ruta DA, Garratt AM, Leng M, et al. A new approach to the measurement of quality of life: the Patient-generated Index. Med Care. 1994;32:1109-1126.

[12] Stratford PW, Gill C, Westaway M, Binkley JM. Assessing disability and change on individual patients: a report of a patient-specific measure. Physiotherapy Canada. 1995;47:258-263.

[13] Stratford PW, Binkley JM, Solomon P, et al. Defining the minimum level of detectable change for the Roland-Morris Questionnaire. Phys Ther. 1996;76:359-365.

[14] Feinstein AR, Josephy BR, Well CK. Scientific and clinical problems in indexes of functional disability. Ann Intern Med. 1986;105:415.

[15] Guyatt GH, Berman LB, Townsend M, Taylor DW. Should study subjects see their previous responses? J Chronic Dis. 1985;38:1003-1007.

[16] Katz JN, Harris TA, Larson MG, et al. Predictors of functional outcomes after arthroscopic partial meniscectomy. J Rheumatol 1992; 19:1938-1942.

[17] Stratford PW, Binkley JM, Solomon P, et al. Assessing valid change over time in patients with low back pain. Phys Ther. 1994;74:528-533.

[18] Stratford PW, Binkley JM, Riddle DL. Health status measures: strategies and analytic methods for assessing change scores. Phys Ther. 1996;76:1109-1123.

[19] Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86:420-428.

[20] Ottenbacher KJ, Johnson MB, Hojem M. The significance of clinical change and clinical change of significance: issues and methods. Am J Occup Ther. 1988;42:156-163.

[21] Streiner DL, Norman GR. Health Measurement Scales: A Practical Guide to Their Development and Use. 2nd ed. New York, NY: Oxford University Press Inc; 1995:164.

[22] Williams JI, Naylor CD. How should health status measures be assessed? cautionary notes on procrustean frameworks. J Clin Epidemiol. 1992;45:1347-1351.

[23] Ross M. Relation of implicit theories to the construction of personal histories. Psychol Rev. 1989;96:341-357.