ASSESSMENT OF THE RELIABILITY AND VALIDITY OF THE ARTHRITIS IMPACT MEASUREMENT SCALES FOR CHILDREN WITH JUVENILE ARTHRITIS

Transcription

1 819.~ BRIEF REPORT ASSESSMENT OF THE RELIABILITY AND VALIDITY OF THE ARTHRITIS IMPACT MEASUREMENT SCALES FOR CHILDREN WITH JUVENILE ARTHRITIS CLAUDIA J. COULTON, ELIZABETH ZBOROWSKY, JUDITH LIPTON. and ARTHUR J. NEWMAN There are few multidimensional measures of functional status in children, and none have been developed for children with juvenile arthritis (JA). This report describes an attempt to apply selected components of the Arthritis Impact Measurement Scales (AIMS), which were developed and validated for adults, to a sample of children with active JA (n = 60) or inactive JA (n = 17). Our results suggest that the Pain scale and the Physical Activity scale are the most reliable measures for children with JA; the correlations of the Pain scale, Physical Activity scale, and Dexterity scale results with the clinical measures of JA diagnostic category and joint count support the convergent validity of these scales for the active JA group; and the AIMS revised Physical Disability dimension and the Pain dimension, used to predict the children s classification in the active or inactive JA group, evidence discriminant validity. Additional studies of children with a wider range of impairments are needed to further assess the usefulness of the AIMS for children with JA. From the Northeast Ohio Multipurpose Arthritis Center, School of Medicine, the School of Applied Social Sciences, and the School of Law, Case Western Reserve University, Cleveland, Ohio. Supported by Multipurpose Arthritis Center grant AM from the National Institutes of Health. Claudia J. Coulton, PhD: Professor, School of Applied Social Sciences and Associate Director for Community Programs, No~i heast Ohio Multipurpose Arthritis Center; Elizabeth Zborowsky, PhD: Research Associate, School of Applied Social Sciences; Judith Lipton, MSW, JD: Assistant Professor, School of Law; Arthur J. Newman, MD: Associate Professor of Pediatrics, School of Medicine and Chief of Pediatric Rheumatology, Rainbow Babies and Children s Hospital, Cleveland, Ohio. Address reprint requests to Claudia Coulton, PhD, School of Applied Social Sciences, Case Western Reserve University, Cleveland, OH Submitted for publication January 31, 1985; accepted in revised form November 5, Few measures exist for measuring functional status among children (l), and none have been developed specifically for studying children with juvenile arthritis (JA). However, several functional status measures, for example, the widely used Arthritis Impact Measurement Scales (AIMS), have been developed and validated for use with adult arthritis patients (2). This raises the question of whether selected components of the AIMS are potentially useful measures of the functional status of children with JA. We investigated the reliability and validity of selected AIMS components for use in evaluating children with JA, and we report the results here. PATIENTS AND METHODS AIMS instrument. Previous studies of adult arthritis patients have shown the AIMS to be a reliable and valid measure of 3 dimensions of health status: physical disability, pain, and psychological disability (3-6). The AIMS contain 45 items which form 9 component scales, each related to 1 of 4 dimensions as follows: (a) Physical Disability dimension-mobility scale, Physical Activity scale, Dexterity scale, Activities of Daily Living scale, and Household Activities scale; (b) Pain dimension-pain scale; (c) Psychological Disability dimension-anxiety scale and Depression scale; (d) 1 other separate scale, the Social Activity scale. The AIMS dimensions and scales that were considered appropriate for children and were used in this survey of children with JA were (a) a revised Physical Disability dimension, which did not include the Household Activities scale, and (b) the Pain dimension. The component scales related to these 2 Arthritis and Rheumatism, Vol. 30, No. 7 (July 1987)

2 820 BRIEF REPORTS dimensions each contained 4-5 items with 2-6 fixedchoice responses for each item. The score for each component scale was obtained by summing the item scores and then indexing this sum to a range of scores from 0-10 (no disability to maximum disability) for each scale (3,4). The scores for the Physical Disability and Pain dimensions were calculated by summing the respective component scale scores and dividing by the number of valid scales in order to maintain a range of scores from 0-10 (no disability to maximum disability) for each dimension (5). Patients. The study sample consisted of 77 children with JA who were seen in the pediatric rheumatology clinic affiliated with the University Hospitals of Cleveland between March and December of 1983 and were available to be interviewed at the time of their clinic visit. The total number of children observed in this clinic in 1983 was approximately 250. The children in the sample ranged in age from 2-17 years, with a mean age of 9.3 years. Twenty-three were boys and 54 were girls; 75 were white and 2 were black. The predominance of girls in the sample was consistent with the results of epidemiologic studies of JA, while the ratio of blacks to whites suggested that blacks were underrepresented in this sample. The breakdown of the sample by JA diagnostic category showed that there were 23 children with pauciarticular, antinuclear antibody (ANA)-positive JA, 11 with pauciarticular, ANA-negative JA, 1 with polyarticular, seropositive JA, and 42 with polyarticular, seronegative JA. Children with systemic onset JA were not grouped separately, but were classified according to disease status (pauciarticular versus polyarticular). In the individual children, the total number of joints affected by JA on the day of the survey interview ranged from 0-32, with a mean joint count of 5.0. Disease duration ranged from <I-14 years, with a mean of 3.7 years. The sample was divided into 2 subsamples. The active JA group (n = 60) included children who had continuous JA activity over most of the previous year and/or soft tissue swelling, pain, or limitation in 1 or more joints on the day of the survey interview. The inactive JA group (n = 17) was composed of children with no evidence of continuous JA activity over most of the previous year and no joint involvement on the day of the survey interview. (In 1 child, the status of the disease was unknown; this child was included in the inactive JA group.) Data collection. The AIMS questions for the Mobility, Physical Activity, Dexterity, Activities of Daily Living, and Pain scales were embedded in a survey questionnaire, which also contained items pertaining to other research questions. The questionnaire was administered, by a social worker, to the parents of the 77 children in the sample, after the children s regularly scheduled clinic appointments. The Dexterity scale and the Activities of Daily Living scale questions were not administered to the parents of children under the age of 6 years (n = 20), because most of these items refer to tasks that children under 6 could riot be expected to perform. Additional medical information about the children was obtained from a medical record review completed by the registered nurse at the pediatric rheumatology clinic. The JA diagnostic category, number of joints with JA involvement Cjoint count), duration of JA, and continuity of disease manifestations were some of the medical record data used in this analysis. Data analysis. The methods used to assess the reliability and validity of the AIMS scales for this sample of children with JA were based on the methods used in the studies of AIMS with adult arthritis patients (3-6). To evaluate the reliability of the AIMS scales with regard to their being unidimensional and cumulative scales, a Guttman scalogram analysis (7) of the pattern of the responses to the items on each scale was peiformed. The reliability coefficients produced by the Guttman scalogram analysis were based on the number of deviations (errors) in the actual pattern of responses compared with the ideal pattern of responses for a unidimensional, cumulative scale. The criteria for acceptable scale reliability were a Guttman coefficient of reproducibility (CR) 20.9 and a Guttman coefficient of scalability (CS) 20.6, with the passfail division points for the scale items set to maximize CS (3,4). In addition, another reliability coefficient, Cronbach s alpha, a measure of the internal consistency of the items on a scale (S), was computed to further assess the unidimensionality of the AIMS scales. An alpha 20.7 was the criterion for an acceptable unidimensional scale (3,4). The convergent validity of the AIMS was assessed hy the correlation of the children s AIMS scores with 2 specific clinical measures: JA diagnostic category (coded 0 for pauciarticular JA and 1 for polyarticular JA), and joint count. According to Cassidy (9), the functional outcome for children with JA is least favorable for those with unremitting polyarthritis. Therefore, we theorized that the children s AIMS scores would be positively correlated with their JA

3 BRIEF REPORTS 82 1 Table 1. Comparison of the descriptive statistics for selected Arthritis Impact Measurement Scales (AIMS) dimensions and scales for children in the active juvenile arthritis (JA) and inactive JA groups Active JA group scores Inactive JA group scores Selected AIMS dimensions and scales Range* Mean SD Range* Mean SD t-test Revised Physical Disability dimension 0.C C t Motility scale 0.C Physical Activity scale t Dexterity scale t Activities of Daily 0.C $ Living scale Pain dimension ' * The possible range of scores was 0.C-10.0, with 0.0 = no disability and 10.0 = maximum disability. t P $ P diagnostic categories and with their joint counts. We further theorized that the AIMS revised Physical Disability dimension and Pain dimension would be more highly correlated with joint count, a more sensitive measure of current JA involvement, than with diagnostic category. A discriminant analysis (10) was also performed, to evaluate the discriminant validity of the AIMS. The revised Physical Disability dimension and the Pain dimension scores were used to predict the classification of the children into 1 of 2 groups: the active JA group or the inactive JA group. The predicted classification of the children, based on the 2 AIMS dimension scores, was subsequently compared with their actual classification. The success of the AIMS dimensions in predicting the children's active or inactive disease status was determined by the percentage of fewer classification errors that were made by the AIMS dimension predictors than would be expected from classification by random assignment (7). RESULTS Scale scores. The ranges, means, and standard deviations of the AIMS scores for the children in the active JA and inactive JA groups are shown in Table 1. The means and standard deviations indicated that the variance in the scores was very limited or nonexistent for s'ome of the scales. In the active JA group, 58 of the 60 children performed all of the activities on the Mobility scale; each of the 2 other children was not able to perform 1 activity on the Mobility scale. Similarly, 36 of the 44 children aged 6 years or older in the active JA group performed all of the items on the Activities of Daily Living scale; of the remaining 8 children, 7 needed help with 1 activity, and only 1 required help with 2 activities on the scale. In the inactive JA group, there was no variance in the scores on the Mobility scale, the Activities of Daily Living scale, or the Dexterity scale: All of the children were able to perform all of the items. The results of t-tests (Table 1) showed that with the exception of the Mobility scale scores, the mean scores for the selected AIMS components used in the study were significantly different (P ) between the active JA and inactive JA groups. Reliability of AIMS for assessing children. Table 2 displays the reliability coefficients of the selected AIMS components for children in the active JA group and the inactive JA group. Based on the criteria set for the reliability coefficients (CR 20.9, CS 20.6, and alpha 2?0.7), the Pain scale was the most reliable of the AIMS for both groups of children. The Pain scale met the Guttman coefficient criteria for a unidimensional, cumulative scale as well as the alpha criterion for a unidimensional scale. In contrast, the Physical Activity scale met the Guttman coefficient criteria in both groups of children, but did not meet the alpha criterion in either group (though the criterion was approached in the active JA group). On the Activities of Daily Living scale, the pattern of responses among children with active JA met the Guttman criteria for a perfect scale (CR = 1.00, CS = 1.00), but the alpha was very low (0.06) because the limited variance of this scale resulted in weak inter-item correlations. The other 2 AIMS scales, the Mobility scale and the Dexterity scale, were the least reliable scales in the active JA group: They failed to meet 1 of the Guttman coefficient criteria and the criterion for alpha. The reliability coefficients for 3 of the AIMS scales-the Mobility scale, the Dexterity scale, and the Activities of Daily Living scale-could not be computed for the inactive

4 822 BRIEF REPORTS Table 2. Comparison of the reliability coefficients for selected Arthritis Impact Measurement Scales (AIMS) dimensions and scales for children in the active juvenile arthritis (JA) and inactive JA groups* Selected AIMS scales No. of items Mobility scale 4 Physical Activity scale 5 Dexterity scale 5 Activities of Daily Living scale 4 Pain scale 4 Active JA group reliability coefficients CR cs abha oo 1.oo Inactive JA group reliability coefficients CR cs alpha 1.oo oo 1.oo 1.oo 1.oo 0.74 * CR = Guttman coefficient of reproducibility; CS = Guttman coefficient of scalability; alpha = Cronbach s alpha; = reliability coefficient could not be computed because the scale had zero variance (i.e., all of the children in the group had a score of 0, meaning no disability). JA group, due to the total lack of variance in the scores for these scales. Convergent validity of AIMS for assessing children. The correlation coefficients between the AIMS scores and the specific clinical measures-ja diagnostic category and joint count-in the active JA and inactive JA groups are displayed in Table 3. In the active JA group, all but I of the correlations were in the predicted positive direction (r = ), and the majority of these correlations were significant at the P i 0.05 level. The 1 exception was the weak negative correlation (r = -0.04) between the Mobility scale and JA diagnostic category. The revised Physical Disability dimension and Pain dimension scores of the children in the active JA group were more highly correlated with the more sensitive measure of joint count than with JA diagnostic category. These findings tended to support the convergent validity of the selected AIMS for the children in the active JA group. For the children in the inactive JA group, correlation coefficients could be computed only between JA diagnostic category and the revised Physical Disability dimension, the Physical Activity scale, and the Pain dimension, because of the lack of variance in the other scale scores and the joint count for this group. The correlations of the revised Physical Disability dimension and the Physical Activity scale with JA diagnostic category were relatively weak and not in the predicted direction (r = and r = -0.12, respectively). In contrast, the correlation between the Pain dimension or scale, the most reliable scale, with JA diagnostic category (r = 0.32) was in the predicted direction and sufficiently strong to tend to support the convergent validity of the Pain dimension for the inactive JA group. Discriminant validity of AIMS for assessing children. Table 4 shows the classification results of the discriminant analysis. However, the predicted group classifications shown in this table should be interpreted with caution because the predictor variables Table 3. Comparison of the pairwise correlations of the scores for selected Arthritis Impact Measurement Scales (AIMS) dimensions and scales with clinical validation measures (juvenile arthritis [JA] diagnostic category and joint count) for children in the active JA and inactive JA groups* Selected AIMS dimensions and scales Active JA groiip correlation Coefficients Inactive JA group correlation coefficients* JA diagnostic.joint JA diagnostic Joint category (n) count (n) - category (n) count Revised Physical Disability dimension Mobility scale Physical Activity scale Dexterity scale Activities of Daily Living scale Pain dimension 0.16 (60) (60) 0.24 (60)i 0.30 (44)$ 0.12 (44) 0.26 (60)i 0.32 (60)t 0.19 (60) 0.31 (60)t 0.40 (44)t 0.20 (44) 0.35 (60)t (17) (17) 0.32 (16) * = correlation coefficient could not be computed because of a lack of variance in one or both of the variables. i P $ P

5 BRIEF REPORTS 823 Table 4. Discriminant analysis results for the predicted classification of children in the active juvenile arthritis (JA) group or the inactive JA group based on their scores on the Arthritis Impact Measurement Scales (AIMS) revised Physical Disability dimension and the Pain dimension, with listwise treatment of missing data Actual group classification Predicted group classification % of cases Prior Active JA Inactive classified Group n probability* group JA group correctly 7 Active JA (93.3%) 4 (6.7%) t Inactive JA (81.2%) 3 (18.7%) Unknown 1 * Based on actual number in group. t Indicates that the predicted classification of the children in the 2 groups, based on their discriminant scores for the AIMS revised Physical dimension and Pain dimension, results in 33% fewer errors than would be expected by random assignment. were skewed toward zero in both the active JA group and the inactive JA group. Therefore, the predictor variables could not be assumed to have a multivariate normal distribution in the population of children with JA. The consequences of violating the multivariate normality assumption are that the percentages of children correctly classified may be slightly overestimated or underestimated (10). The discriminant scores, using the revised Physical Disability dimension and the Pain dimension, correctly predicted classification in the active JA group or the inactive JA group for 77.63% of the children. The errors in prediction were much more frequent for children whose disease was inactive. The predicted classification, based on the children s revised Physical Disability dimension and Pain dimension scores, represented 33% fewer errors (T = 0.33) than would be expected if the classification were done by random assignment to the 2 groups (i.e., 17 errors made in the predicted classification versus 25 errors expected by chance). This modest improvement in the number of classification errors tended to support the discriminant validity of the revised Physical Disability dimension and the Pain dimension for predicting the JA activity status of children. DISCUSSION The results of the assessment of the reliability and validity of the AIMS for this sample of children with active or inactive JA are mixed. The patients were divided into active JA and inactive JA groups because the AIMS was initially constructed for use with patients who have active disease. However, a separate analysis of the data for the total sample of children revealed that the reliability coefficients and convergent validity correlation coefficients for the total sample were similar to those reported here for the active JA group. The Pain scale has the highest degree of reliability for both the active JA group and the inactive JA group; for the active JA group, the Physical Activity scale also has an acceptable degree of reliability. The extremely limited variance in the scores for the Mobility scale and the Activities of Daily Living scale suggest that they may require some revisions for use with children. One possible explanation for the limited variance in the Mobility scale scores is that our study sample had fewer children with lower extremity JA involvement than would be expected given the clinic population. However, when the scores of the children with lower extremity involvement were examined separately, it was found that most of these children were able to perform all of the items on the Mobility scale. It is possible, therefore, that the areas in which children have mobility limitations differ from the areas of mobility limitation that are experienced by adults and are addressed in the AIMS Mobility scale. The convergent validity of the selected AIMS scales or children in the active JA group tends to be supported by the correlation coefficients between the scale scores and the children s diagnostic categories and joint counts. The weakest correlations are seen with the Mobility scale and the Activities of Daily Living scale. The low reliability and restricted range of scores for these scales contribute to the weakness of the correlations. The substantial differences in the mean scores of the active JA group and the inactive JA group for both the revised Physical Disability dimension and the Pain dimension suggest that these AIMS dimensions are good predictors of JA activity in children. The classification results of the discriminant analysis tend to support the discriminant validity of both of these dimensions in predicting JA activity status. Finally, 2 possible confounding factors should

6 824 BRIEF REPORTS be considered in this assessment of the reliability and validity of the selected AIMS components for use with children. The first factor is age, which, as an indicator of developmental stage, might affect a child s ability to perform functional activities. However, in this sample of children, age is not significantly correlated with any of the AIMS scale scores in either the active JA group or the inactive JA group. A second factor is the use of parents as proxy respondents for the children. This does introduce an additional source of variability compared with the direct administration of the AIMS to adult arthritis patients. However, parents usually are a primary source of information about a child s day-to-day functioning. Therefore, it seems logical to test the reliability and validity of the AIMS when parents are the respondents. The selected AIMS scale questions could probably be administered directly to study populations consisting solely of older children, but the results would probably be different than those for a group of children with a wide range of ages, and no information on younger children would be available. The AIMS is a functional status measurement instrument which was specifically developed for use in a population of adults with arthritis. Caution must be exercised in any effort to use the AIMS with another population, such as children with JA, because the children s areas of functional disability may differ markedly from adults areas of functional disability. However, the AIMS Pain dimension and revised Physical Disability dimension are, at the present time, probably the most reliable and valid instruments available for measuring the functional status of children with JA. Studies of children with a wider range of arthritis-related impairments are needed to further assess the usefulness of the selected AIMS components, especially the Mobility scale and the Activities of Daily Living scale, for children with JA. Acknowledgments. The authors gratefully acknowledge the assistance of Roland W. Moskowitz, MD, Director, Northeast Ohio Multipurpose Arthritis Center; Susan Holland, BSN, University Hospitals; Michael Vender, MSW, University Hospitals; and Marcia Petchers, PhD, School of Applied Social Sciences, Case Western Reserve University. REFEREES 1. Eisen M, Ware JE, Donald CA, Brook RH: Measuring components of children s health status. Med Care 17: , Liang MH, Cullen K, Larson M: In search of a more perfect mousetrap (health status or quality of life instrument). J Rheumatol 9: , Meenan RF, Gertman PM, Mason JH: Measuring health status in arthritis: the Arthritis Impact Measurement Scales. Arthritis Rheum 23: , Meenan RF, Gertman PM, Mason JH, Dunaif R: The Arthritis Impact Measurement Scales: further investigations of a health status measure. Arthritis Rheum 25: , Kazis LE, Meenan RF, Anderson JJ: Pain in the rheumatic diseases: investigation of a key health status component. Arthritis Rheum 26: , Brown JH, Kazis LE, Spitz PW, Gertman P, Fries JF, Meenan RF: The dimensions of health outcomes: a cross-validated examination of health status measurement. Am J Public Health 74: , McIver JP, Carmines EG: Unidimensional Scaling. Quantitative Applications in the Social Sciences, no. 24. Beverly Hills, Sage Publications, Cronbach LJ: Coefficient alpha and the internal structure of tests. Psychornetrika 16: , Cassidy JT: Juvenile rheumatoid arthritis, Textbook of Rheumatology, Vol. 2. Edited by WN Kelley, ED Hams Jr, S Ruddy, CB Sledge. Philadelphia, WB Saunders, 1981, pp Klecka WR. Discriminant Analysis. Quantitative Applications in the Social Sciences, no. 19. Beverly Hills, Sage Publications, 1980