Arthritis & Rheumatism Official Journal of the American College of Rheumatology

AIMS2 The Content and Properties of a Revised and Expanded Arthritis Impact Measurement Scales Health Status Questionnaire ROBERT F. MEENAN, JOHN H. MASON, JENNIFER J. ANDERSON, ANDREW A. GUCCIONE,and LEWIS E. KAZIS surement performance was tested in 408 subjects: 299 with rheumatoid arthritis (RA) and 109 with osteoarthritis (OA); 45 of these subjects completed a second AIMS2 within 3 weeks. Internal consistency coefficients for the 12 scales were 0.72-0.91 in the RA group and 0.74-0.96 in the OA group. Test-retest reliability was 0.78-0.94. All withm-scale factor analyses produced single factors, except for mobility level in OA. Validity analyses in both the RA and the OA groups showed that patient designation of an area as a problem or as a priority for improvement was significantly associated with a poorer AIMS2 scale score in that area. Reliability, factor analysis, and validity results were consistent in age, sex, and education subgroups. Satisfaction was moderately correlated with level of function in the same health status area, and the satisfaction items formed a reliable scale. Responses to the arthritis attribution items showed that most dysfunction in this sample was due to arthritis. Conclusion. The AIMS2 is a revised and expanded health status questionnaire with excellent measurement properties that should be useful in arthritis clinical trials and in outcomes research.

Objective. The goal of this project was to develop a more comprehensive and sensitive version of the Arthritis Impact Measurement Scales (AIMS). Methods. AIMS scale items were revised, and 3 new scales were added to evaluate arm function, work, and social support. Sections were also added to assess satisfaction with function, attribution of problems to arthritis, and self-designation of priority areas for improvement. The new instrument was designated the -2. A pilot test of format and content and a performance test of reliability and validity were carried out. Results. Questionnaire completion times in a pilot study of 24 subjects averaged 23 minutes, and evaluations were positive regarding the instrument’s length and ease of completion, and the subjects’ willingness to complete serial forms and return them by mail. MeaFrom the Arthritis Center, the School of Medicine, and the School of Public Health, Boston University, and the Departments of Medicine, Boston City and University Hospitals, Boston, Massachusetts. Supported by NIH Multipurpose Arthritis and Musculoskeletal Diseases Center grant AR-20613 from the NIAMS. Robert F. Meenan, MD, MPH, MBA: Professor of Medicine, Boston University School of Medicine; John H.Mason, PhD: Senior Research Associate, Boston University School of Medicine; Jennifer J. Anderson, PhD: Associate Research Professor of Medicine (Biostatistics), Boston University School of Medicine; Andrew A. Guccione, PhD, MSPT: Assistant Research Professor of Medicine (Sociology), Boston University School of Medicine; Lewis E. Kazis, ScD: Associate Professor of Public Health, Boston University School of Public Health. Address reprint requests to Robert F. Meenan, MD, MPH, MBA, Boston University Arthritis Center, 80 East Concord Street, Boston, MA 02118. Submitted for publication February 20, 1991; accepted in revised form August 19, 1991.

The use of questionnaire measures of health status has become an important approach to assessing outcome in the rheumatic diseases (1-4). There are a number of conceptual and practical advantages in using these questionnaires as arthritis outcome measures ( 5 ) . Conceptually, health status questionnaires broaden the notion of outcome to include aspects of patient status that are not addressed by traditional

Arthdtls and Rheumatism, Vol. 35, No. 1 (January 1992)

1

2 clinical measures of disease. Questionnaire-based health status parameters, such as mobility and mood, are particularly relevant because they assess aspects of outcome that most concern the individual patient. The major practical advantage of health status questionnaires is their relatively low cost. Since they are usually self-administered, they can provide information, in both clinical research and clinical practice settings, with a minimal investment of professional time. Furthermore, since these questionnaires can be administered by mail or even by telephone, their use is compatible with a variety of study designs. The Arthritis Impact Measurement Scales (AIMS) was one of the first questionnaires specifically designed for the purpose of assessing health status in subjects with rheumatic diseases. The original AIMS questionnaire contained 45 items grouped into 9 scales (6). These scale scores could in turn be combined to produce overall models of health status, with 3 or 5 components (7,8). The measurement properties of the AIMS, including its reliability, validity, and sensitivity to change, have been reported (9-12). The AIMS questionnaire has been used by a number of investigators to study arthritis health status in a variety of settings (13-15), and it has been translated into several different languages (16-18). Despite its strong measurement properties and widespread acceptance, we decided to undertake a major revision of the original AIMS questionnaire (AIMS1). There were three main reasons for this decision. First, we recognized that the items and scales in the AIMS1 could be improved by deleting weak items and by using a more consistent format for the remaining items. Second, we realized that there were arthritis-relevant aspects of health status that were not well addressed in the AIMS1. Finally, other investigators have raised concerns about approaches to arthritis health status assessment that were not part of the original AIMS approach. Pincus and colleagues have shown that satisfaction with health status is a relatively distinct aspect of patient outcome (19). Hughes et a1 have pointed out the potential problems in attributing all health status change to musculoskeletal disease in subjects with arthritis (20). Tugwell and coworkers have stressed the potential advantages of focusing outcome assessment on those aspects of health status that are of most concern to the individual patient (21,22). Our overall goal was to produce a revised and expanded AIMS that would maintain the strengths of the original questionnaire while incorporating substan-

MEENAN ET AL

tial changes. We describe herein the development of the AIMS2, a revised and expanded version of the original Arthritis Impact Measurement Scales. The development of the AIMS2 content and format are detailed. The scores and measurement properties of the AIMS2 scales in a large group of subjects with rheumatoid arthritis (RA) and osteoarthritis (OA) are reported, as are the results in major demographic subsets. Initial results of the AIMS2 approach to measuring satisfaction, arthritis attribution, and problem prioritization are reported, and suggestions are made on how scores from the AIMS2 can be used in the assessment and reporting of health status in arthritis.

METHODS Revisions to the AIMS1. The format of all AIMS1 items was altered to produce a standardized item format for the AIMS2 version. One of 2 standard sets of 5 response options was used for each item. One set ranged from all days to no days; the other ranged from always to never. The wording for all AIMS1 items was adjusted to accommodate the phrasing of these response options. All yeslno responses were eliminated. In order to standardize the time frame of the responses, the phrase “During the past month. . . .” was added to the beginning of each AIMS2 scale. In addition to these general format changes, specific changes were made in some of the 45 individual items in the AIMS!. Thirty-five AIMS1 items were left unchanged, 4 were revised, and 6 were deleted. Decisions on deletion and revision were based on an evaluation of our large AIMS1 database, which allowed us to examine the effect of individual items on the measurement properties of each scale. Three items were deleted from the household activity scale, and 1 each from the pain, anxiety, and depression scales. Three new items were added to the original scales: a mobility item to assess the ability to do errands in the neighborhood, a pain item to measure difficulty with sleeping, and a social activity item to assess attendance at church, club, or other group activities. As a result of these changes, all AIMS2 scales contain 4 or 5 items, a number that combines good content coverage with strong reliability. The AIMS! item that assessed the respondent’s overall health status was changed from a visual analog format to a discrete response format, with 5 options ranging from very poor to very well. Additions in the AIMSZ. Three new scales were added in the revision of AIMS1 to produce the core AIMS2 questionnaire. These new scales, arm function, work, and support from family and friends, were designed to cover aspects of health status that were not assessed by the AIMS!. The items for the arm function scale included items designed to measure elbow and shoulder motion. The work scale included 5 items. The first was designed to indicate the main form of work done by the respondent. Respondents who were employed outside the home, students, or homemakers were asked to answer 4 other questions designed to assess work disability and the impact of arthritis on the

AIMS2 performance of work. Respondents who were retired, disabled, or unemployed were instructed to skip these items. The 5-item social support scale was added to supplement the social activities scale by measuring the qualitative aspect of social interactions (23,24). New names were given to all the revised scales in the AIMS2, in order to avoid confusion with the similar scales in the AIMS1 version. This was also done so that the name of the scale alone would indicate the AIMS version being used. Names were chosen that clearly indicate the content of the scale, so that the content could be recalled by the subject when the name alone was used in other sections of the AIMS2 questionnaire. New assessments. Three new sections were added to the AIMS2 questionnaire to address specific issues in arthritis health status assessment. In each of these 3 sections, the 12 AIMS2 scale labels and an example of 1 typical item from each scale were used to query the subject. Items in the first new section assessed satisfaction with the current level of function in each of the 12 AIMS2 health status areas. The 5 response options for this section ranged from very satisfied to very dissatisfied. The second new section was designed to assess more clearly the specific impact of arthritis on the subject’s health status in each of the 12 AIMS2 areas. The response options in this section allowed subjects to indicate whether they had a problem in each area and, if so, to choose 1 of 5 options to indicate if the problem was entirely or largely due to arthritis, entirely or largely due to other causes, or due partly to both. The third new section listed the 12 scales and asked the subject to check up to 3 areas in which he or she would most like to see improvement. New items to assess general perceptions of current and future health were included in AIMS2, but those items are not considered in this report. Data collection. A pilot test of the AIMS2 questionnaire was conducted at the arthritis clinics of Boston City Hospital and University Hospital. AIMS2 questionnaires were completed by 24 subjects, who also completed a short debriefing questionnaire and interview. The measurement properties of the AIMS2 were studied in a performance test sample of 797 potential subjects with RA or OA whose names were submitted by participating physicians at 13 clinical practice sites across the country. A single mailing to these potential subjects yielded 408 returns, for a response rate of 51%. In order to estimate test-retest reliability, repeat AIMS2 questionnaires were sent within 2 weeks to the first 50 subjects who returned the initial mailing. Fortyfive of these subjects returned the second questionnaire. Data analyses. All AIMS2 items and scales were scored such that low scores indicate good health status while high scores indicate poor health status. Responses for certain items were recoded to comply with this convention. After recoding, scale scores could be calculated by summing the responses for all items in a particular scale. All scale scores were then standardized by a simple mathematical transformation, so that the best possible scale score was 0 and the worst possible score was 10. Work scale scores could be calculated only for subjects who indicated that they were not disabled, retired, or unemployed; this resulted in a substantially lower sample size for the work scale in all scale-level analyses.

3 The reliability of the AIMS2 scales was estimated by calculating the internal consistency coefficient alpha (25) using the scores for all responding subjects. The data from the 45 repeat questionnaires were used to estimate testretest reliability, using the intraclass correlation coefficient for ordinal measures (26) and the kappa statistic for categorical measures (27). Alpha coefficients 20.70 and test-retest coefficients 20.80 are generally regarded as evidence of acceptable reliability. Reliability was also assessed by performing principal component factor analyses on individual scales to determine if the constituent items loaded on a single factor. Failure to load on a single factor suggests that the items in a scale do not all estimate the same aspect of health status. An eigenvalue criterion of 1.0 was used for these factor analyses (28), and the results are given in terms of the percentage of variance in the scale score explained by the principal factor. The validity of the 12 AIMS2 scales was studied by examining the relationship between the scale scores and the responses obtained concurrently from subjects on other items in the AIMS2 questionnaire. One estimate of validity was the extent to which the scale scores of subjects who indicated that they had a problem in a particular area were significantlyworse than the scores of subjects who indicated that they did not have a problem in that area. A second estimate of validity involved comparing scale scores for those who indicated that a particular scale was a priority area for improvement with the scores of those who did not so indicate. Both types of validity analysis were done using t-tests with a significance level of P < 0.05, not adjusted for multiple comparisons. All reliability and validity analyses were also done in major disease and demographic subgroups, including subgrouping by diagnosis, sex, age, and education. A pattern of consistent results in the subgroup analyses was sought as additional evidence of reliability and validity. Responses to the satisfaction items were analyzed to see if they would form a reliable scale using the alpha coefficient and the test-retest intraclass correlation coefficient. Satisfaction scores for individual scales were correlated with the related health status scale scores, using Spearman’s coefficient, to determine if these measures were estimating the same phenomenon. The reliability of the arthritis attribution items was estimated using the intraclass correlation coefficient of the test-retest data. The validity of the attribution responses was estimated by examining the effects of attribution-based adjustments on different AIMS2 scale scores. The stability of the prioritization responses was checked using test-retest data from the repeat questionnaire subsample.

RESULTS Pilot tests. The pilot test sample included 24 subjects, most of whom were female, had at least a high school education, and had RA. It took these subjects an average of 23 minutes (range 13-37 minutes) to complete the questionnaire. Twenty of the 24 subjects omitted 3 or fewer of the 78 items in the

MEENAN ET AL

4

Table 1. AIMS2 scale scores in rheumatoid arthritis and osteoarthritis subject groups

Rheumatoid arthritis (n = 299) AIMS2 scale Mobility level Walking and bending Hand and finger function Arm function Self-care Household tasks Social activities Support from family and friends Arthritis pain Work? Level of tension Mood

Actual score range* 0-10

0-10 0-10 0-10 0-10 0-10

0-9 0-10 0-10

0-8.8 0-9.5 0-8

Mean

f SD

1.69 f 2.28 4.58 f 2.83 3.16 f 2.58 2.01 f 2.26 0.73 f 1.53 1.67 f 2.43 4.91 f 1.74 1.84 f 2.03 4.69 f 2.52 3.80 f 3.39 4.05 f 1.82 2.63 f 1.59

Osteoarthritis (n = 109) Actual score range 0-7.5 0-10 0-10

0-5 0-10

0-8. I 0-9.5 0-7.5 0-9.5 0-7.5 0-8.5

0-7.5

Mean

f

SD

1.35 rt 1.97 4.24 f 3.22 1.75 f 2.14 0.68 rt 1.17 0.39 f 1.47 0.91 f 1.76 4.88 f 1.97 1.83 f 2.03 4.69 f 2.31 3.48 f 3.38 4.19 f 1.95 2.48 2 1.48

* Potential range of scores on all scales is 0-10. tn

=

126 for the rheumatoid arthritis group and 43 for the osteoarthritis group.

questionnaire. No single item or group of items was omitted with notable frequency. In response to the debriefing questionnaire, 79% of the subjects answered that they found it either very easy or somewhat easy to complete the AIMS2. Most (63%) answered that the questionnaire was neither too long nor too short, while 25% believed it was somewhat long. Almost everyone (96%) believed that the AIMS2 covered all or nearly all the major areas in which arthritis affected their health, and 58% said that the available response options always or usually allowed them to describe their situation. The open-ended portion of the debriefing questionnaire did not indicate consistent problems with any particular item. Therefore, there were no substantive changes made in the draft questionnaire as a result of the pilot test. Scale reliability and validity. In the performance test sample of 408 subjects, the mean ? SD age of the respondents was 60.7 5 13.3 years, and the mean ? SD duration of arthritis was 14.0 & 11.2 years. Seventy-eight percent of the respondents were female, 73% had RA, and 84% had at least a high school education. Sixty-six percent of the group had no comorbidities, 25% had hypertension, and 8% had another medical condition. The ranges and means of the AIMS2 scale scores for subjects in the performance test sample, grouped by diagnosis of RA or OA, are presented in Table 1. Mean scores ranged from a good health status score of 0.39 on self-care in subjects with OA to a poorer health status score of 4.91 on social activities in subjects with RA. The mean scores for RA and OA subjects did not differ substantially except for the

expected poorer scores on hand and finger function and arm function in the RA group. Although the mean scores for all scales were 4 . 0 , subjects did use the full range of response options, as demonstrated by the ranges of the calculated scale scores. The reliability results are shown in Table 2. The reliability of all 12 scales scores as measured by the internal consistency (alpha) of their component items exceeded a threshold of 0.70 in both the RA and the OA subsamples, with most coefficients in the range of 0.80-0.89. Test-retest reliability, as measured by the intraclass correlation coefficient of scores from two administrations of the instrument, equalled or exceeded 0.80 for all scales except work (n = 17) and exceeded 0.90 for 6 of the 12. Within-scale principal component factor analysis revealed that the items in each of the 12 scales loaded on a single major factor in each disease group, except for the mobility level scale items, which loaded on 2 factors in the OA group. The percentage of the scale score variance explained by the principal factor equalled or exceeded 50% in all cases. The reliability of the AIMS2 scales by disease was also examined in 3 major demographic subsets of the study sample: sex, age (1. Factor 2 explained an additional 23% of the variance. 7 n = 126 for the rheumatoid arthritis (RA) group and 43 for the OA group in the internal cosistency analysis and the factor analysis; n = 17 in the test-retest reliability analysis.

The validity of the AIMS2 scale scores was tested using internal standards based on the subject’s responses to other items in the questionnaire. These results are displayed in Table 3. They show that the scale scores consistently indicated significantly (P < 0.001) worse health status in subjects who indicated on the related attribution question that they had a prob-

demographic subgroups). The alpha level exceeded 0.70 in 96% of these estimates (104 of 108), with the work scale demonstrating weak reliability in 3 groups that had 50 or fewer subjects. Scale items loaded on a single factor in 94% (101 of 108), with mobility level loading on 2 factors in 3 subgroups and work in 2 subgroups. Table 3. Validity of the AIMS2 health status scales

Subject attributes as health status problem area (n = 408) Yes

Subject designates as priority area for health status improvement (n = 408)

AIMS2 scale

No.

Score*

No.

Score*

Mobility level Walking and bending Hand and finger function Arm function Self-care Household tasks Social activities Support from family and friends Arthritis pain Work$ Level of tension Mood

229 281 254 194 129 249 146 128 337 87 267 243

2.42 5.63 3.80 2.86 1.39 2.08 5.74 3.20 5.03 3.67 4.78 3.26

161

0.40t 1.47t 0.94t 0.51t 0.23t 0.257 4.40t 1.19t 2.14” 1.26t 2.55t 1.49t

110

140 204 267 143 247 262 45 77 118 148

No

Yes

No

Score*

No.

Score*

71

2.75

201 170 59 16

5.80

324 195 229 344 388 2% 366 379 127 149 304 336

1.34t 3.14t 2.07t 1.37t 0.57t 1.24t 4.83t 1.67t 3.62t 2.52 3.76t 2.38t

No.

108

30 23 259 19 95 66

3.73 3.31 2.30 2.10 6.05 4.59 5.21 2.60 5.14 3.67

* Mean scale score on the AIMS2 (possible 0 [best health status] to 10 [worst health status]). t P < 0.001 by t-test, versus mean score in subjects who answered “yes.” t. n = 169.

MEENAN ET AL

6

Table 4.

Frequency of non-arthritis attribution and effect on AIMS2 scale scores RA group (n max = 261)* AIMS2 scale

Mobility level Walking and bending Hand and finger function Arm function Self-care Household tasks Social activities Support from family and friends Arthritis pain Work5 Level of tension Mood

Proportion adjusted?

% reduction

26

7 8 3 5 6 8 12 22 2

19 5

14 14 24 47 45 7

25 64 62

in score$

OA group (n max = 92)* Proportion adjusted t

% reduction

23 17 2 I1

20 8

14

II 12 8 17 7 3 23 23

in score$

26 48 57

10

10 31

22 21

64 60

1 11

* n max = the number of subjects who indicated that they had a problem in a given health status area. t Proportion of subjects who indicated that they had a problem that was partly, mostly, or entirely due to causes other than arthritis. $ Percentage reduction in the mean score for all subjects, after adjustment. 5 n = 139 for the rheumatoid arthritis (RA) group and 36 for the osteoarthritis (OA) group.

lem in that health status area. Similarly, except for the work scale, the AIMS2 scores were significantly worse in subjects who designated a particular scale as 1 of their 3 priority areas for health status improvement. These validity analyses were also carried out in the disease and demographic subgroups. All scale scores differed significantly by problem area designation in the disease, sex, age, and education subsets, with the exception of self-care scores in male subjects ( P = 0.09) and self-care scores (P= 0.07) and pain scores ( P = 0.06) in those with less than a high school education. Satisfaction section. Average scores on the 12 satisfaction items ranged from I .7 for social support to 3.0 for arthritis pain, on a scale of 1 (very satisfied) to 5 (very dissatisfied). The alpha reliability coefficient for the satisfaction items as a scale was 0.91 in the RA group and 0.85 in the OA group. The test-retest reliability coefficient for the satisfaction scale was 0.89. In terms of validity, mean satisfaction scale scores ranged from 2.5 to 3.3 for those who indicated a health status problem in a particular area, while the comparable scores ranged from 1.2 to 1.7 for those who did not indicate a problem (P < 0.001 for all 12 comparisons). The correlations between levels of satisfaction and health status for each of the 12 scales ranged from 0.50 to 0.69 in the RA group and from 0.42 to 0.68 in the OA group. The overall satisfaction scale score correlated with the single-item measure of arthritis impact at the level of 0.63 in RA subjects and 0.57 in OA subjects. These results indicate that the

satisfaction items and scale are measuring a somewhat distinct element of health status in a reliable and valid manner. Attribution section. In the attribution section of the AIMS2, the number of subjects who indicated that they had a problem in a particular health status area ranged from 128 who had a problem with support from family and friends to 337 who had a problem with arthritis pain (Table 3). The test-retest reliability (kappa statistic) for this dichotomous problemho problem designation in 45 subjects exceeded 0.75 for all but 1 of the 12 attribution items (dexterity, 0.57). The testretest reliability for the 6 category attribution responses ranged from 0.35 for support from family and friends to 0.72 for household tasks ( P < 0.05 in each case). Table 3 also shows that subjects who designated an AIMS2 scale as a health status problem area consistently had significantly worse scores on that scale. To derive a clearer picture of the health status effects of arthritis in this sample, the AIMS2 scale scores were adjusted based on the subjects’ responses to the arthritis attribution items. Scale scores were multiplied by a factor of 0.50 for those who indicated they had a problem that was due partially to arthritis and partially to other causes, and multiplied by a factor of 0.25 for those who indicated they had a problem that was entirely or mostly due to other causes. The scores for respondents who chose 1 of the other 3 options were not adjusted. The proportion of

AIMS2

7

scores adjusted and the effects of adjustment on the mean AIMS2 scores in the RA and OA groups are shown in Table 4. These adjustments had relatively little effect on health status scores in this sample, in which the average age was 60.7 and only 8% of the subjects had a comorbidity other than hypertension. Score adjustments based on attribution of problems to causes other than arthritis produced mean reductions averaging 10.8% in the AIMS2 scale scores, with the largest changes, for both RA and OA subjects, occurring in the social and psychological scales and the smallest changes in the pain and physical function scales. Prioritization section. Focusing on areas that the individual patient considers a priority for improvement may produce more sensitive measures of therapeutic benefit. Table 5 summarizes the frequency with which the subjects with RA and OA designated each of the 12 AIMS2 health status scales as 1 of their 3 priority areas for improvement. Subjects with RA were much more likely to rank hand and finger function and arm function as priority areas than were subjects with OA, while walking and bending and arthritis pain were cited by 49% or more of the subjects in both groups. As indicated above, subjects who designated an area as a priority for improvement had significantly worse health status scale scores in that area than those who did not. If a health status area was chosen as a priority on the initial AIMS2 questionnaire, it was chosen Table 5. Frequency with which each AIMS2 scale was designated as 1 of 3 priority areas for improvement ~~~

% listing priority area

on initial test AIMS2 scale Mobility level Walking and bending Hand and finger function Arm function Self-care Household tasks Social activities Support from family and friends Arthritis pain Work* Level of tension Mood

% listing again on second test

RAgroup (n = 299)

OAgroup (n = 109)

RAandOA (n = 45)

18

17 55 31

38 52 48

8

71 25

49

47

17 4 30 8 7 62 9 23 7

4 17

40

6 3

50 50

81 19 27 15

83 50 40 62

* = 126 for the rheumatoid arthritis (RA)gorup osteoarthritis (OA) group.

43 for the

again on the second questionnaire by approximately half of the 45 repeat respondents, indicating that these designations may not be particularly stable.

DISCUSSION The AIMS2 is a revised and expanded version of the original Arthritis Impact Measurement Scales health status questionnaire. As was the case with the AIMSl, the primary purpose of the AIMS2 is to measure health status in subjects with arthritis, with a particular emphasis on measuring improvements produced by therapeutic interventions. The revisions and expansions incorporated into the AIMS2 were made with that purpose in mind, although the cross-sectional nature of this study cannot directly assess how sensitive the AIMS2 will be as a measure of health status change. The AIMS2 versions of the original 9 scales have reliability levels comparable with those reported for the original version (6,9). Given that these scales represent revisions rather than new scales, their validity has not been reassessed in a comprehensive manner. Nevertheless, all scale scores vary as expected based on the presence of a health status problem in that area or based on the designation of that area as a priority for health status improvement. Three new scales have been incorporated into the AIMS2 for the purpose of assessing aspects of health status that were not adequately addressed in the original questionnaire. The reliability characteristics of these scales-arm function, support from family and friends, and workare generally comparable with those of the revised scales. When the validity data on the revised and the new scales are assessed in a similar manner, the arm, work, and social support scales are again comparable with the original scales. The absence of a work scale in the original AIMS was a particularly important omission that has been corrected in the expanded AIMS2. The format of this scale allows work to be considered as a dichotomous measure in which the work variable is positive for subjects who are employed outside the home, students, or full-time homemakers. Alternatively, a 4-item work scale can be used to assess changes in the quantity and quality of work in those individuals. The arm function and support from family and friends scales do not introduce new aspects of health status, but rather complement the hand and finger function and social activities scales in those respective areas. The original AIMS questionnaire has been used

8

in a variety of settings with subjects who have differing demographic and disease characteristics. The results of this initial testing of the AIMS2 indicate that the reliability and validity of the revised and expanded questionnaire are generally stable across 2 major disease groups (RA and OA) and 3 major demographic dimensions (age, sex, and education). These results indicate that the AIMS2, like its predecessor, has good generalizability . The comprehensiveness of the AIMS2 has also been increased by the addition of items to measure satisfaction with health status. These items measure a relatively distinct aspect of health status that can be assessed independently of the related health status scale. Pincus et al have shown that satisfaction varies substantially among subjects with the same level of health status (19). Our results confirm that satisfaction is correlated with function but is not identical to it. Although each satisfaction item can be assessed and considered separately, we recommend that the 12 items be used to estimate an overall scale of health status satisfaction that can be used as a distinct aspect of outcome in arthritis intervention studies. In revising and expanding the AIMS, we wanted to make it more accurate and more sensitive to change, and to achieve this in a logistically simple manner. Many people with arthritis have other diseases that can independently affect their health status (29), and Hughes et al have reported that arthritis health status can be estimated more accurately if subjects are asked to attribute health status problems to arthritis or to other causes (21). Their GERI-AIMS questionnaire accomplishes this by including an arthritis attribution question in each and every item of the AIMS scales that assess physical function. This is a cumbersome approach that lengthens the questionnaire and can increase the burden on the respondent. It is also inconsistent in that it does not address arthritis attribution on the social and psychological scales. We therefore added a section to the expanded AIMS2 that seeks arthritis attribution information at the level of each of the 12 scales. Our results indicate that subjects can respond to these scale-level arthritis attribution items in a reliable and valid way. We recommend the inclusion of this type of scale-level attribution assessment in all health status instruments. This additional assessment is particularly important in subject groups that include a substantial proportion of elderly patients who are more likely to have other diseases, and also in groups in which subjects with important comorbidities are not

MEENAN ET AL

routinely eliminated by the eligibility requirements of the study protocol. We have proposed a particular way in which the AIMS scale scores may be adjusted using the arthritis attribution results, but other approaches may be more appropriate for other subject groups or for other questionnaires. We also recognize that attribution questions are conceptually different from traditional health status questions and that causal attribution is a complex undertaking. Further study of attribution is needed before its role in arthritis outcome assessment can be established. As noted, a major purpose of revising and expanding the AIMS was to produce a health status questionnaire that would be more sensitive to improvements produced by arthritis therapy. In the AIMS2, we have made 2 major changes from AIMS1 that were designed to address this issue. The first was to provide 5 response options for each item on every scale. This should allow for the detection of smaller degrees of improvement than would be possible with the yesho options used in a number of the original AIMS scales. We also included a separate section that allows subjects to designate 3 areas of health status in which they would most like to see improvement. The general arguments in favor of using prioritized outcome measures have been advanced by Tugwell and colleagues (22). Their approach collects this information using a semi-structured interview approach, rather than a self-assessment questionnaire. Although this allows the subject to designate highly specific areas of concern (such as gardening), the interview approach lacks the logistic advantages of a questionnaire format. The AIMS2 is designed to collect information on patient priorities at the scale level while retaining the questionnaire approach to health status assessment. Our results indicate that subjects can understand and respond to a question that asks them to designate their top 3 priority areas for improvement. These designations, in turn, correlate with the severity of their problems as measured by the health status scale score in that area. Using the AIMS2 approach, the 3 designated areas can be targeted as primary health status variables for assessing response to therapy. Although the sensitivity of this approach cannot be directly assessed using our current data, Tugwell and coworkers’ study of the use of the interview technique with patients in a methotrexate trial has demonstrated the potential benefits of problem prioritization (22). The major limitation to this prioritization

9

AIMS2 approach may be the relative instability of the patient’s designation of areas needing improvement. Concerns have been raised in the past about the length and complexity of the AIMS. Although a number of items have been added to the original questionnaire, as a result of design changes the instrument is only 1 page longer. In our pilot study, it took subjects -20 minutes on average to complete the questionnaire. Most did not consider this a particularly burdensome length of time, and it is likely to be quite tolerable if the questionnaire is completed during what would otherwise be waiting time or if it is completed at home at the patient’s leisure. The reading ability required to complete the original AIMS scales has been measured at the fifth grade level or lower for all scales except household activities (Heiffer CA: personal communication). The AIMS2 is slightly more complex due to inclusion of the satisfaction, attribution, and prioritization sections. However, we did not detect any evidence in our pilot or performance tests that subjects were having difficulty understanding or completing the AIMS2. Nevertheless, a lengthy and relatively complex questionnaire such as the AIMS2 must be used with caution in certain groups, such as the elderly and those with low average education level. Our research group will provide arthritis investigators with materials and directions for using either the AIMSl or the AIMS2. Those who are currently using the AIMSl may continue with that version or switch to the AIMS2. Since we have not directly compared the original and revised questionnaires, we cannot be certain that the scale scores from the 2 versions will be exactly comparable. Therefore, if longitudinal data are being collected or if AIMS results are to be compared across studies, it may be best to continue to use the AIMSl until such time as score comparability is documented. Many health status investigators have emphasized the need for very brief health status assessment instruments (30,31). In fact, Wallston and colleagues have described a short version of the original AIMS questionnaire (32). We based our decision to expand, rather than shorten, the AIMS on 2 important considerations. First, we believe the primary purpose of a health status assessment instrument is to provide a comprehensive and highly sensitive measure of outcome. Brief health status questionnaires are not as comprehensive or as sensitive as their longer versions. Second, the original AIMS questionnaire was developed primarily as a tool for clinical research. Since this continues to be the primary purpose for the AIMS2,

length is not as important a consideration as it would be if the questionnaire were to be used in clinical practice. Nevertheless, given the growing interest in outcomes research and the need to collect outcome data on a more routine basis in clinical practice, we plan to develop a short version of the AIMS2 once we have change data that will allow us to identify the most sensitive items. In summary, we have revised and expanded the Arthritis Impact Measurement Scales, a widely used arthritis health status questionnaire, to produce the AIMS2. The goal of this effort was to make the AIMS more comprehensive, accurate, and sensitive by using a standard format, by incorporating 3 new scales, and by adding 3 new sections to assess health status satisfaction, attribution of health status problems to arthritis, and designation of patient-selected priority areas for improvement. Results from a pilot study and a large cross-sectional performance test indicate that the AIMS2 has acceptable qualities of reliability and internal validity and that these properties are generalizable across major disease and demographic subgroups. The AIMS2 measures health status in a comprehensive manner that is acceptable to patients. Although its sensitivity has not yet been documented, the AIMS2 can be recommended as an outcome measurement tool for the assessment of various arthritis interventions.

REFERENCES 1. Bell MJ, Bombardier C, Tugwell P: Measurement of

2.

3.

4. 5.

6.

7.

functional status, quality of life, and utility in rheumatoid arthritis. Arthritis Rheum 33591-601, 1990 Pincus T, Callahan LF, Vaughn WK: Questionnaire, walking time, and button test measures of functional capacity as predictive measures for mortality in rheumatoid arthritis. J Rheumatol 14:24&257, 1987 Deyo RA, Inui TS, Lenninger J, Overman SS: Measuring functional outcome in chronic disease: a comparison of traditional scales and self-administered health status questionnaires in patients with rheumatoid arthritis. Med Care 21:18&191, 1983 Meenan RF, Pincus T: The status of patient status measures. J Rheumatol 14:411414, 1987 Fries JF: Toward an understanding of patient outcome measurement. Arthritis Rheum 26:697-704, 1983 Meenan RF, Gertman PM, Mason JH: Measuring health status in arthritis: the Arthritis Impact Measurement Scales. Arthritis Rheum 23:146-152, 1980 Brown JH, Kazis LE, Spitz PW, Gertman PM, Fries JF, Meenan RF: The dimensions of health outcomes: a cross

10

validated examination of health status measurement. Am J Public Health 74:159-161, 1984 8. Mason JH, Anderson JJ, Meenan RF: A model of health status for rheumatoid arthritis: a factor analysis of the Arthritis Impact Measurement Scales. Arthritis Rheum 31:714-720, 1988 9. Meenan RF, Gertman PM, Mason JH, Dunaif R: The Arthritis Impact Measurement Scales: further investigations of a health status measure. Arthritis Rheum 25: 1048-1053, 1982 10. Potts MK, Brandt KD: Evidence of the validity of the Arthritis Impact Measurement Scales. Arthritis Rheum 30:93-96, 1987 11. Meenan RF, Anderson JJ, Kazis LE, Egger MJ, AltzSmith M, Samuelson CO Jr, Willkens RF, Solsky MA, Hayes SP, Blocka KL, Weinstein A, Guttadauria M, Kaplan SB, Klippel J: Outcome assessment in clinical trials: evidence for the sensitivity of a health status measure. Arthritis Rheum 27:1344-1352, 1984 12. Anderson JJ, Firschein H E , Meenan RF: Sensitivity of a health status measure to short-term clinical changes in arthritis. Arthritis Rheum 322344450, 1989 13. Coulton CJ, Hyduk CM, Chow JC: An assessment of the Arthritis Impact Measurement Scales in three ethnic groups. J Rheumatol 16:I 1 10-1114, 1989 14. Joyce K, Berkebile C, Hastings C, Yarboro C, Yocum D: Health status and disease activity in systemic lupus erythematosus. Arthritis Care Res 2:65-69, 1989 15. Coulton CJ, Zborowsky E, Lipton J, Newman AJ: Assessment of the reliability and validity of the Arthritis Impact Measurement Scales for children with juvenile arthritis. Arthritis Rheum 30:819-824,1987 16. Sampalis JJ, Pouchot J, Beaudet F, Carette S, Gutkowski A, Harth M, Myhal D, Seneca1 JL, Yeadon C , Williams JI, Esdaile JM: Arthritis Impact Measurement Scales: reliability of a French version and validity in adult Still’s disease. J Rheumatol 17:1657-1661, 1991 17. Hendricson WD, Russell IJ, Prihoda TJ, Jacobson JM, Rogan A, Bishop GD, Castillo R: Development and initial validation of a dual-language English-Spanish format for the Arthritis Impact Measurement Scales. Arthritis Rheum 32:1153-1 159, 1989 18. Taal E, Jacobs JW, Seydel FR, Wiegman 0, Rasker JJ: Evaluation of the Dutch Arthritis Impact Measurement Scales (Dutch-AIMS) in patients with rheumatoid arthritis. Br J Rheumatol 28:487-491,1989

MEENAN ET AL 19. Pincus T , Summey JA, Soraci SA Jr, Wallston KA, Hummon NP: Assessment of patient satisfaction in activities of daily living using a modified Stanford Health Assessment Questionnaire. Arthritis Rheum 26:13461353, 1983 20. Hughes S L , Edelman P, Chang RW, Singer RH, Schuette P: The GERI-AIMS: reliability and validity of the Arthritis Impact Measurement Scales adapted for elderly respondents. Arthritis Rheum 34:856-865, 1991 21. Tugwell P, Bombardier C, Buchanan WW, Grace E, Goldsmith CH, Hanna B: The MACTAR patient preference disability questionnaire: an individualized functional priority approach for assessing improvement in physical disability in clinical trials in rheumatoid arthritis. J Rheumatol 14M6-451, 1987 22. Tugwell P, Bombardier C, Buchanan WW, Goldsmith C, Grace E, Bennett KJ, Williams HJ, Egger M, Alarcon GS, Guttadauria M, Yarboro C, Polisson RP, Szydlo L, Luggen ME, Billingsley LM, Ward JR, Marks C: Methotrexate in rheumatoid arthritis: impact on quality of life assessed by traditional standard item and individualized patient preference health status questionnaires. Arch Intern Med 150:59-62,1990 23. House JS, Landis KR, Umberson D: Social relationships and health. Science 241540-545, 1988 24. Procidano ME, Heller K: Measures of perceived social support from friends and from family: three validation studies. Am J Community Psychol 1l:l-24, 1983 25. Novick MR, Lewis C: Coefficient alpha and the reliability of composite measures. Psychometrika 32:1-13,1967 26. Shrout PE, Fleiss JL: Intraclass correlations: uses in assessing rater reliability. Psychol Bull 86:420-428,1979 27. Fleiss JL: Statistical Methods for Rates and Proportions. New York, John Wiley & Sons, 1987 28. Kim JO, Mueller CW: Factor Analysis: Statistical Methods and Practical Issues. Beverly Hills, Sage, 1978 29. Berkanovic E, Hurwicz ML: Rheumatoid arthritis and comorbidity. J Rheumatol 17:888-892,1990 30. Stewart AL, Hayes RD, Ware JE: The MOS short form general health survey: reliability and validity in a patient population. Med Care 26:724-735, 1988 31. Nelson EC, Berwick DM: The measurement of health status in clinical practice. Med Care 27:S77-S90, 1989 32. Wallston KA, Brown GK, Stein MJ, Dobbins CJ: Comparing the short and long versions of the Arthritis Impact Measurement Scales. J Rheumatol 16:1105-1109, 1989

AIMS2. The content and properties of a revised and expanded Arthritis Impact Measurement Scales Health Status Questionnaire.

The goal of this project was to develop a more comprehensive and sensitive version of the Arthritis Impact Measurement Scales (AIMS)...
975KB Sizes 0 Downloads 0 Views