The views expressed in this manuscript are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs.
Validity, factor structure, and clinical relevance of the AUSCAN osteoarthritis hand index†
Version of Record online: 30 JAN 2006
Copyright © 2006 by the American College of Rheumatology
Arthritis & Rheumatism
Volume 54, Issue 2, pages 551–556, February 2006
How to Cite
Allen, K. D., Jordan, J. M., Renner, J. B. and Kraus, V. B. (2006), Validity, factor structure, and clinical relevance of the AUSCAN osteoarthritis hand index. Arthritis & Rheumatism, 54: 551–556. doi: 10.1002/art.21615
- Issue online: 30 JAN 2006
- Version of Record online: 30 JAN 2006
- Manuscript Accepted: 31 OCT 2005
- Manuscript Received: 5 AUG 2005
- Merit Review Entry Program award from the Department of Veterans Affairs, Veterans Health Administration, Health Services Research and Development Service
- National Center for Research Resources. Grant Number: M01-RR-30
The Australian/Canadian (AUSCAN) Osteoarthritis Hand Index is a self-report assessment of pain, stiffness, and function in patients with hand osteoarthritis (OA). Small studies have confirmed the reliability, construct validity, and responsiveness of this measure, but the factor structure has not been examined. In this study, we examined the clinimetric properties and clinical relevance of the AUSCAN index in a large sample of patients with familial hand OA.
The study group comprised 700 patients (80% female, mean age 69 years) who were part of a study on the genetics of generalized OA. All patients had radiographic hand OA bilaterally. The analyses examined internal consistency, factor structure, and relationships of the subscales to grip and pinch strength and a single-item pain measure.
Internal consistency was high for the total AUSCAN index and the subscales (Cronbach's α = 0.93–0.96). The AUSCAN function subscale had the strongest correlation with grip and pinch strength, and the pain subscale had the strongest correlation with the single-item pain measure, thus supporting the construct validity of these subscales. Factor analysis showed that all pain and function items clearly loaded on the subscale they were intended to measure. Each 1-unit increase in the AUSCAN function subscale was associated with a clinically relevant decrease in hand strength.
The results of this study strongly confirm the clinimetric properties of the AUSCAN index, including the validity of specific subscales. Results indicate that the AUSCAN index can measure meaningful changes in pain, stiffness, and function.
Pain, stiffness, and function are key outcomes in osteoarthritis (OA). The Australian/Canadian (AUSCAN) Osteoarthritis Hand Index is a self-report assessment of these outcomes and was designed specifically for use among patients with hand OA (1, 2). Studies have shown that this measure has acceptable reliability, construct validity, and responsiveness (1–4). However, this is a relatively new scale, and published data on the measurement properties of the AUSCAN index still involve relatively few patients (1–3). Further examination of this scale in larger patient groups is important for confirming its clinimetric properties and clinical relevance.
Prior studies also have not examined the factorial validity of the AUSCAN index. This scale comprises 3 subscales, pain, stiffness, and function, that can be used individually or summed to create a composite score. For measures including multiple subscales, it is important to examine whether each of the subscales measures a discrete domain corresponding to the attribute it proposes to assess. Recent studies have reevaluated the original proposed factor structure of another OA measure, the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) (5–10). Factor analyses suggest that the WOMAC is composed of 5 factors that are related to types of activities, rather than the 3 proposed subscales of pain, stiffness, and function (6–8). Although the overall WOMAC score is valid and responsive to change (11), those studies underscore the importance of examining the factor structure of multiscale measures, particularly if the subscales are to be used independently (5).
The purpose of this study was to examine the clinimetric properties of the AUSCAN index in a large cohort of patients with hand OA. Specifically, this study assessed the internal consistency and construct validity of the AUSCAN subscales as well as the factor validity of the scale. This study also assessed the clinical relevance of the AUSCAN scale by examining its association with changes in hand strength.
PATIENTS AND METHODS
All participants were enrolled in the Genetics of Generalized Osteoarthritis (GOGO) study. The primary objective of the GOGO study is to identify OA susceptibility genes through genotyping and linkage analysis of OA-affected sibling pairs. The GOGO study involves a consortium of 7 clinical research sites in the US and the UK (12–14). Participants included in our analysis were from the 2 GOGO sites where grip and pinch strength testing was conducted, Duke University and the University of North Carolina at Chapel Hill (UNC).
All families recruited for the GOGO study had at least 2 siblings with bilateral hand OA, defined as bony enlargement of at least 1 distal interphalangeal joint (DIP) and bony enlargement of 2 or more other interphalangeal joints or carpometacarpal joints. For these analyses, we also required participants to have radiographic evidence of Kellgren/Lawrence (15) grade ≥2 OA in at least 1 DIP joint. We used this criterion because bony involvement of a DIP joint is a component of the American College of Rheumatology criteria for hand OA (16), and because a previous twin study (17) showed that a single Heberden's node was strongly inherited. All radiographs were read by a single reader (JBR). Participants were excluded if they had self-reported or radiographic evidence of arthritis from rheumatoid arthritis, systemic lupus erythematosus, psoriasis, or gout of the hands, hips, or knees. Hemochromatosis was also an exclusion criterion, but none of the patients met the serologic criterion of a serum iron:total iron-binding capacity ratio >55%. Among 878 eligible participants, 700 completed grip and pinch strength tests; the remaining 178 patients were not included in analyses involving strength measurements. The majority of patients who did not complete strength tests were enrolled at the beginning of the study, prior to the initiation of strength testing.
AUSCAN OA hand index.
The AUSCAN index is a 15-item scale measuring pain (5 items), stiffness (1 item), and function (9 items) during the preceding 48 hours. All items are rated on a scale of 0 (none) to 4 (extreme). The AUSCAN index was developed through an interactive process involving expert opinion from health care providers (rheumatologists, physiotherapists, orthopedic surgeons) and interviews with patients. Items retained for this scale were those that had a prevalence of >60% in the sample population and a mean importance rating of >2.0 (on a 1–5 scale). Internal consistency of the subscales was excellent (Cronbach's α = 0.90–0.98). The test–retest reliability for each of the subscales was also acceptable (intraclass correlation coefficient 0.70–0.86). Construct validity was confirmed against a variety of measures, including Dreiser's function index (18, 19).
Grip and pinch strength.
We examined the associations of the AUSCAN subscales to pinch and grip strength. Strength measurements were performed on both hands. Grip strength was measured with a Jamar hydraulic hand dynamometer (Sammons Preston, Bolingbrook, IL) and was reported in kilograms. Pinch strength was measured with a Jamar hydraulic pinch gauge (Sammons Preston) and was also reported in kilograms. One site (UNC) conducted 3 trials for grip and pinch strength measurements, and an average of the results from the 3 trials was calculated. The other site (Duke) conducted 1 trial for each measure. There were no differences in mean grip strength measures between the sites, and results of study analyses did not differ between sites.
We also evaluated the associations of the AUSCAN subscales to a single-item pain measure for each hand. Participants were asked, “On most days of any 1 month in the last year, did you have pain, aching, or stiffness?” Possible responses were none, mild, moderate, and severe, and each response was assigned a value of 0–3.
We first examined the internal consistency of the AUSCAN scale as well as the pain and function subscales. Cronbach's alpha was used to examine internal consistency. Next, we examined correlations among the AUSCAN subscales. Based on previous research (8), we expected moderate correlations among the subscales, but very high correlations may indicate overlap between the scales.
To examine construct validity of the subscales, we assessed the correlations (and 95% confidence intervals [95% CIs]) of all 3 AUSCAN subscales to hand strength (grip and pinch strength for both right and left hands) and the single-item pain measure (for each hand). To support the construct validity of the AUSCAN subscales, the function subscale should have the highest correlation with hand strength, and the pain subscale should have the highest correlation with the single-item pain measure. We also examined the relationships of the AUSCAN subscales to hand strength and pain according to hand dominance.
To further assess the construct validity of the pain and function subscales, which were of particular interest, we examined the partial correlations (and 95% CIs) of these subscales with hand strength and the single-item pain measure. This analysis allowed us to test the associations of each subscale with strength and pain while controlling for the other subscale. To support the construct validity of these subscales, the function subscale should remain significantly associated with strength when controlling for the pain subscale, and the pain subscale should remain significantly associated with the single-item pain measure when controlling for function. We examined these partial correlations in right and left hands as well as dominant and nondominant hands. The AUSCAN stiffness subscale (1 item; 1–5 scale) and the single-item hand pain measure (0–3 scale) were not normally distributed. Therefore Spearman's rank correlation coefficient was used for any correlations involving either of these 2 variables. All other associations were examined using Pearson's correlation coefficient.
To assess the factor validity of the AUSCAN index, we first conducted an exploratory factor analysis using SAS Proc Factor (20, 21). We used an oblique rotation (SAS option = Promax), because we expected the subscales to be correlated. We used the scree test to determine the number of factors to be retained (22). The scree test plots eigenvalues against factors. The eigenvalue for a given factor measures the variance in all variables that is accounted for by that factor. In a scree test, the point where the plot (of eigenvalues versus factors) changes slope indicates the number of factors that should be retained. We then conducted a second factor analysis, specifying the number of factors retained, based on the first exploratory factor analysis. Items were considered to load on a factor if its coefficient was >0.4. Because it is not possible to have a single-item factor, the AUSCAN stiffness subscale was excluded from this analysis.
To examine the clinical relevance of the AUSCAN function subscale, we used regression analysis to determine the change in grip and pinch strength associated with a 1-unit change in the function subscale. We performed this analysis for both the right and left hands, as well as the dominant and nondominant hands. We were not able to perform an analysis for the pain subscale, because the other pain measure used in this study is a single item and therefore was not suitable for regression analysis.
The sample consisted of 878 participants from 387 families. Characteristics of the sample are presented in Table 1. The mean ± SD age of the sample was 69 ± 9 years. The majority of subjects (80%) were female and right-handed (93%). The mean grip strength values were ∼50 kg for right hands and 47 kg for left hands, and the mean pinch strength value for both right and left hands was ∼12 kg. The mean score for the single-item pain measure was ∼1.2, corresponding to mild-to-moderate pain. The mean ± SD total AUSCAN score was 33.6 ± 12.7 (SE = 0.44), and the mean ± SD AUSCAN subscale scores were as follows: for pain, 11.6 ± 4.7 (SE = 0.16); for stiffness, 2.4 ± 1.0 (SE = 0.03); and for function, 19.6 ± 7.9 (SE = 0.27).
|Age, years||68.6 ± 8.9|
|Female sex, %||80.3|
|Hand dominance, %|
|AUSCAN total (scale 15–75)||33.6 ± 12.7|
|AUSCAN pain (scale 5–25)||11.6 ± 4.7|
|AUSCAN stiffness (scale 1–5)||2.4 ± 1.0†|
|AUSCAN function (scale 9–45)||19.6 ± 7.9|
|Grip strength in right hand, kg||49.7 ± 23.7|
|Grip strength in left hand, kg||46.6 ± 23.9|
|Pinch strength in right hand, kg||12.4 ± 6.7|
|Pinch strength in left hand, kg||11.8 ± 7.2|
|Pain severity in right hand (scale 0–3)||1.3 ± 0.9‡|
|Pain severity in left hand (scale 0–3)||1.2 ± 0.9§|
Internal consistency was high for the overall AUSCAN scale (Cronbach's α = 0.96) as well as the pain and function subscales (Cronbach's α = 0.93 and 0.94, respectively). Table 2 shows correlations among the AUSCAN subscales, as well as the relationships of the subscales to hand strength and self-reported pain. The correlations between subscales were fairly high (0.63–0.79), and the strongest association was between the pain and function subscales. The AUSCAN function subscale had moderate, statistically significant correlations with right and left hand grip strength (r = −0.42 and r = −0.40, respectively; P < 0.001). The pain and stiffness subscales were also significantly associated with grip strength (P < 0.001), but the correlations were weaker (r = −0.22 to −0.34). Correlations of the AUSCAN subscales with pinch strength were weaker than those for grip strength but were still statistically significant (P < 0.01). The function subscale had stronger correlations with right and left hand pinch strength compared with either the pain or stiffness subscale. All AUSCAN subscales were significantly correlated with the single-item pain measure for both right and left hands (P < 0.001). Correlations were strongest for the pain subscale (r = 0.55–0.58) but were only slightly weaker for the other 2 subscales.
|Variable||Pain subscale||Stiffness subscale||Function subscale|
|AUSCAN stiffness||0.63 (0.58, 0.66)||1.00|
|AUSCAN function||0.79 (0.76, 0.81)||0.64 (0.60, 0.68)||1.00|
|Grip strength, right||−0.34 (−0.40, −0.27)||−0.23 (−0.30, −0.16)||−0.42 (−0.48, −0.36)|
|Grip strength, left||−0.30 (−0.37, −0.23)||−0.22 (−0.29, −0.14)||−0.40 (−0.46, −0.33)|
|Grip strength, dominant||−0.33 (−0.40, −0.26)||−0.23 (−0.30, −0.16)||−0.42 (−0.48, −0.36)|
|Grip strength, nondominant||−0.31 (−0.37, −0.24)||−0.22 (−0.29, −0.15)||−0.40 (−0.46, −0.33)|
|Pinch strength, right||−0.16 (−0.23, −0.08)||−0.17 (−0.25, −0.10)||−0.23 (−0.30, −0.15)|
|Pinch strength, left||−0.11 (−0.18, −0.03)||−0.14 (−0.21, −0.06)||−0.16 (−0.23, −0.08)|
|Pinch strength, dominant||−0.16 (−0.23, −0.08)||−0.19 (−0.26, −0.11)||−0.24 (−0.31, −0.16)|
|Pinch strength, nondominant||−0.09 (−0.17, −0.02)||−0.13 (−0.20, −0.05)||−0.14 (−0.22, −0.07)|
|Pain severity, right||0.58 (0.54, 0.63)||0.57 (0.52, 0.61)||0.53 (0.48, 0.58)|
|Pain severity, left||0.55 (0.50, 0.59)||0.52 (0.48, 0.58)||0.51 (0.45, 0.55)|
|Pain severity, dominant||0.59 (0.54, 0.63)||0.57 (0.53, 0.62)||0.54 (0.49, 0.59)|
|Pain severity, nondominant||0.55 (0.50, 0.60)||0.52 (0.47, 0.57)||0.50 (0.45, 0.55)|
All of the AUSCAN subscales were significantly correlated with strength and pain in both dominant and nondominant hands (Table 2). These associations were slightly higher for the dominant hand. Because of the high proportion of right hand–dominant patients, associations for the dominant and nondominant hands closely paralleled those for right and left hands, respectively.
Analysis of partial correlations with grip strength showed that when controlling for the AUSCAN pain subscale, the association with the function subscale was still statistically significant (r = −0.28 and r = −0.27 for the right hand and left hand, respectively; P < 0.001) (Table 3). In contrast, when controlling for the function subscale, the pain subscale no longer had a statistically significant correlation with grip strength. Results were similar for pinch strength. Analyses of partial correlations with the single-item pain measure showed that both the pain and function subscales retained statistically significant associations when controlling for the other subscale. However, the correlation was stronger for the pain subscale than for the function subscale (Table 3). Partial correlations of the AUSCAN subscales for dominant and nondominant hands (results not shown) were similar to the partial correlations for right and left hands, respectively.
|Variable||Pain subscale||Function subscale|
|Spearman's rank correlation coefficient (95% CI)||P||Pearson's correlation coefficient (95% CI)||P|
|Grip strength, right hand||−0.01 (−0.09, 0.07)||0.822||−0.28 (−0.35, −0.21)||<0.001|
|Grip strength, left hand||0.01 (−0.07, 0.08)||0.843||−0.27 (−0.34, −0.20)||<0.001|
|Pinch strength, right hand||0.03 (−0.04, 0.11)||0.387||−0.17 (−0.25, −0.10)||<0.001|
|Pinch strength, left hand||0.03 (−0.05, 0.10)||0.514||−0.12 (−0.20, −0.05)||0.016|
|Pain severity, right hand||0.33 (0.27, 0.38)||<0.001||0.15 (0.08, 0.21)||<0.001|
|Pain severity, left hand||0.30 (0.23, 0.36)||<0.001||0.14 (0.08, 0.21)||<0.001|
The initial exploratory factor analysis of the AUSCAN pain and function subscale items indicated that 2 factors should be retained. Specifically, the scree plot (which plots the magnitude of the eigenvalues for the factors) leveled off after 2 factors, indicating that 2 is the maximum number of factors to extract (22). Results of the second factor analysis (with oblique rotation and 2 factors specified) are presented in Table 4. This analysis showed that all pain and function items clearly loaded on the factor/subscale they were intended to measure.
|Item||Factor 1||Factor 2|
|Physical function difficulty|
|Carrying a full pot with one hand||0.663||0.201|
|Peeling vegetables or fruit||0.707||0.157|
|Picking up large, heavy objects||0.566||0.238|
|Wringing out washcloth||0.609||0.293|
Our final analysis examined the change in hand strength associated with a 1-unit change in the AUSCAN function subscale. Regression analysis showed that for each 1-unit increase in the function subscale, there was a corresponding decrease of 1.31 kg (SE = 0.11) in right hand grip strength and 0.19 kg (SE = 0.03) in right hand pinch strength. Results were similar for the dominant hand (for grip, decrease of 1.30 kg [SE = 0.11]; for pinch, decrease of 0.20 kg [SE = 0.03]). For the left hand, each 1-unit increase was associated with a decrease of 1.21 kg (SE = 0.11) in grip strength and 0.15 kg (SE = 0.04) in pinch strength. Results were similar for the nondominant hand (for grip, decrease of 1.20 kg [SE = 0.11]; for pinch, decrease of 0.13 kg [SE = 0.04]).
This study examined clinimetric properties of the AUSCAN index in a large sample of patients with hand OA (n = 878). To our knowledge, this is the largest study examining the measurement properties of the AUSCAN index and is the first study examining the factor structure of the AUSCAN index. Overall, our results support the measurement properties of the AUSCAN index. First, we observed that internal consistency was high for each of the subscales, providing evidence that each of these subscales measures a single characteristic. Internal consistency was also very high for the AUSCAN index as a whole. Massy-Westropp et al also reported high internal consistency for the total AUSCAN scale and suggested that this indicated that the 3 AUSCAN scales do not measure discrete aspects of hand function (3). We did observe a very high correlation between the AUSCAN pain and function subscales specifically (r = 0.79), which is likely attributable to overlap in some of the activities queried in the 2 subscales. However, results of other analyses in this study indicated that the AUSCAN subscales do measure 3 discrete domains.
Results of this study also support the construct validity of the AUSCAN subscales. Specifically, function had the strongest association with grip and pinch strength, and the pain subscale had the strongest association with the single-item pain measure. Analysis of partial correlations provided additional support for the construct validity. When function was controlled for, pain was no longer significantly associated with grip strength. Although partial correlations with the single-item pain measure were statistically significant for both the AUSCAN pain and function subscales, the relationship of the pain subscale was stronger.
We separately examined associations of the AUSCAN subscales with hand strength and pain in both dominant and nondominant hands. This is important, because the AUSCAN index was originally designed for use in the dominant hand, and little is known about its construct validity in the nondominant hand. Our results showed that the AUSCAN index was significantly associated with both strength and pain in both dominant and nondominant hands. These associations were somewhat stronger in the dominant hand, but there were no substantial differences according to handedness. Thus, our results provide support for the validity of the AUSCAN index in the context of outcomes in the nondominant hand, but additional research is needed to examine its sensitivity to change in the nondominant hand.
Factor analysis strongly confirmed the subscale structure of the AUSCAN index. All pain items clearly loaded on one factor, and all function items clearly loaded on another factor. Although some items are similar between the pain and function subscales (e.g., turning [on both subscales], lifting [pain subscale]/carrying [function subscale], squeezing [pain subscale]/wringing [function subscale]), results indicate that scale items group together as intended and not according to activity type, which may be the case for the WOMAC (8). This is particularly important because the investigators who developed the scale have endorsed the use of the subscales individually as well as the total AUSCAN score (4, 23).
In addition to examining the clinimetric properties of the AUSCAN index, we were interested in assessing the clinical relevance of this measure. Results showed that for every 1-unit increase in the AUSCAN function scale, right hand grip strength decreased 1.31 kg. For this sample, 1.31 kg is ∼2.5% of mean peak grip strength. For right hand pinch strength, every 1-unit increase in the AUSCAN function scale corresponds to a decrease of 0.19 kg, which corresponds to ∼1.5% of mean peak pinch strength for this sample. A previous study showed that among older adults, the mean force applied when opening various types of containers ranged from 1.0 kg to 4.4 kg (24). Therefore, these results indicate that even relatively small changes in the AUSCAN function subscale correspond to strength changes that may impact on some daily activities, particularly among older adults.
There are some limitations to this study. This sample was composed of primarily female patients (80%) and included individuals with familial OA. Therefore, these results may not be generalizable to the entire population of patients with OA, and additional study is needed to examine the measurement properties of the AUSCAN index in other patient groups.
In summary, the results of this study confirmed the internal consistency, construct validity, factor structure, and clinical relevance of the AUSCAN Osteoarthritis Hand Index. Our results indicate that the AUSCAN index should be able to successfully detect changes in both pain and function, even if pain and loss of function do not occur concurrently. However, it would be useful to conduct additional prospective studies using the AUSCAN index in situations in which pain and function would not be expected to change in parallel (5). This type of study design will further test the ability of the AUSCAN index to detect changes in key outcomes among patients with hand OA.
We thank Dr. Nicholas Bellamy for providing the AUSCAN index for use in this study. This study would not have been possible without the aid of the following GOGO Consortium study personnel: Ann Riley, Gabor Varju, Norine Hall (Duke University Medical Center), Patricia Cummins, Deborah MacDonald, Amy Cohen, Janice Woodard, Kelly Neal, and Sarah Williams (UNC). We also thank the Greensboro Hand Clinic for invaluable assistance. Most especially we wish to thank the GlaxoSmithKline sponsors: Drs. Allen Roses and Lefkos Middleton and the GOGO Scientific Coordinators (past and present), Dr. Scott Sundseth and Dr. Uzma Atif, for their invaluable assistance.
- 8Exploring the factorial validity and clinical interpretability of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC). Physiother Can 2003; 55: 160–8., , , , , .
- 13The Genetics of Generalized Osteoarthritis (GOGO) Study: frequencies of large joint involvement [abstract]. Osteoarthritis Cartilage 2004; M30: S12., , , , , , et al.
- 14The Genetics of Generalized Osteoarthritis (GOGO) Study: patterns of hand osteoarthritis [abstract]. Osteoarthritis Cartilage 2004; P285: S115., , , , , , et al.
- 19Validation of an algofunctional index for osteoarthritis of the hand. Rev Rheum Engl Ed 1995; 6 Suppl 1: 43S–53S., , , , .
- 20Using multivariate statistics. 3rd ed. New York: HarperCollins College Publishers; 1996., .
- 21A step-by-step approach to using SAS for factor analysis and structural equation modeling. Cary (NC): SAS Press; 1994..
- 23PARIS sectrogram: a method for weighting and aggregating the AUSCAN Osteoarthritis Hand Index [abstract]. Arthritis Rheum 1998; 41 Suppl 9: S145., , , , , .