Background: The Health Assessment Questionnaire Disability Index (HAQ-DI) is the most widely used measure of function in rheumatoid arthritis (RA).
Objective: To evaluate individualised forms of the HAQ-DI and thus enhance the incorporation of patients’ views in outcome assessment.
Patients and methods: HAQ-DI data were prospectively obtained from 370 outpatients with RA treated with leflunomide over a 6-month period. At baseline and final visits, patients had to rate the importance they attached to each activity addressed by the 20 HAQ-DI items, and to select the five activities they considered the most important. Different individualised scales were evaluated: scales preserving all domains, in which the score for each item is multiplied by or added to its importance; and scales involving, for each patient, only the five most important items. The psychometric properties of these scales were compared with those of the HAQ-DI.
Results: For each HAQ-DI item, severity and importance scores were weakly correlated. Scores for all individualised scales were highly correlated with the HAQ-DI score (rs>0.75). All scales had a good internal consistency (Cronbach’s α 0.87–0.88). Compared with the HAQ-DI, individualised scales did not have better sensitivity to change (standardised response mean 0.64–0.69 vs 0.74).
Conclusion: Individualised scales have similar properties to the HAQ-DI. However, individualised questionnaires measuring importance gave complementary information to the measure of disability. Individualisation is probably not needed for group assessment in all randomised controlled trials but, the use of individualised questionnaires may be clinically relevant for individual patients with RA.
Statistics from Altmetric.com
Outcome assessment in rheumatoid arthritis (RA) was primarily based on doctor-reported outcomes such as measures of disease activity1 2 3 or structural progression.4 5 However, symptoms such as pain, fatigue and disability are relevant and their assessment is recommended for evaluation of response to treatment.6 The Health Assessment Questionnaire Disability Index (HAQ-DI) is a valid, reliable, and widely used tool to measure functional status in patients with RA.7 8 9 10 It is a standardised questionnaire, identical for all patients, investigating patients’ ability to perform some activities, as well as the importance of being able to perform each activity. However, these issues may vary widely among patients. Furthermore, enhancing patients’ perspectives in RA outcome assessment was noted as a major focus for research.6 11
For these reasons, patient-specific, or individualised, scales have been developed to deal with each patient’s priorities in outcome assessment in RA.12 13 14 15 Some have been shown to be more sensitive to change than classical instruments.16 17 18 However, most of them involve semistructured interviews requiring 15–40 min to complete.19 Hewlett et al developed an individualised questionnaire derived from the modified HAQ,20 21 where each item is weighted by the importance it has for each patient. However, responsiveness of the scale has not been evaluated.
We designed this prospective study to develop, validate and evaluate psychometric properties of patient-specific scales, assessing the level of disability in patients with RA, derived from the HAQ-DI according to several methods of individualisation.
Patients and methods
Data were obtained from a 6-month, prospective, open-label study involving 378 outpatients with RA treated with leflunomide.22 Patients included had RA as defined by the American Rheumatism Association revised criteria,23 with active disease defined by a doctor global assessment of disease activity ⩾2 on a five-point Likert scale (from 1 to 5). Two visits were scheduled, one at baseline before initiation of leflunomide and one 6 months later.
At baseline and final visits, patients completed:
The French version of the HAQ-DI,24 termed the severity questionnaire. It comprises 20 items, referring to ability in performing daily life activities, involving eight domains and includes a “help and device” component. For each item, score ranged from 0 to 3. The score of each domain is the highest item score of that domain, unless help or a device is required. Dependence on help and device increases a lower score to 2. The overall score is the mean of the eight domain scores and varies from 0 to 3.
The importance questionnaire: patients had to rate the importance of each HAQ-DI item, by answering the question “In your daily life, the following activity is…”. Importance was rated by a four-point Likert scale, ranging from 0, “not important” to 3,“extremely important”.
The preference questionnaire: patients had to select and order the five activities (ie, HAQ-DI items) they considered the most important in their daily lives.
The swollen and tender joint counts, erythrocyte sedimentation rate (ESR), level of C-reactive protein (CRP) were also collected to obtain the 28 joint count Disease Activity Score (DAS28).
Methods of individualisation
Several individualised scales were obtained:
Scales based on the importance questionnaire
With multiplicative, or additive, methods: for each item, the severity score was multiplied by, or added to, the importance score.
The score for each domain was the highest score obtained after combining the severity and importance scores of items of that domain. The total score was the mean score of all eight domains. An individualised scale, with the multiplicative method, preserving all 20 items was also obtained.25
Scales based on the preference questionnaire.
Top five HAQ activities: involving only the five most important activities for each patient.
Weighted top five HAQ activities: involving only the five most important activities, in which the severity score is weighted by the rank of preference for this activity (highest weight referred to the most important activity).
In these scales, the resulting items are not the same for all patients. The total score was the sum of the scores, weighted or not, of the five most important items.
Initial scores were obtained by combining data from the baseline visit. Final scores were obtained by combining baseline importance or preference ratings and final severity scores. Because the rating of importance has not been applied to the help and device component, individualised scales were developed without this component. Each scale was linearly transformed to a scale of 0 to 3; 0 indicating no disability and 3 indicating maximal possible disability.
All patients who had completed the original HAQ-DI with no missing data at baseline were involved in the development of individualised forms. Psychometric properties of each individualised scale were evaluated, and compared with those of the HAQ-DI:
Construct validity was assessed by Spearman rs correlation between the score of each individualised scale and that of the HAQ-DI. We also examined divergent validity using Spearman correlation between scores of individualised scales and other measures of disease activity (swollen and tender joint count, ESR, CRP level and DAS28).
Internal consistency was assessed, when estimable (ie, for scales with identical items), by Cronbach’s α coefficient.26 Estimation of confidence intervals and comparisons of Cronbach’s α coefficients involved the bootstrap method, with 1000 replications.27
Responsiveness was assessed by the standardised response mean (SRM), which is the mean change in score between baseline and final visit divided by the standard deviation of the change in score.28 SRM values can be considered large (>0.8), moderate (0.5–0.8) or small (<0.5).29 30 SRM confidence interval estimations and SRM comparisons involved the bootstrap method with 1000 replications.27
Statistical analysis involved use of SAS, release 9.1 (SAS, Cary, North Carolina, USA), and R, release 2.2.1 (The R Foundation for Statistical Computing, Vienna, Austria).
Of the 378 outpatients with RA enrolled in the study, 370 (98%) completed the HAQ-DI without missing data at baseline (table 1). Among them, 360 (95%) and 281 (74%) patients correctly completed the importance and preference questionnaires, respectively.
For each item, severity and importance scores were weakly or not significantly correlated (rs ranging from 0.01 to 0.29) (table 2). Plots of severity by importance scores showed that patients with the highest severity scores for items did not necessarily grade the items as important and patients with the lowest severity score for items did not necessarily consider the items as not important (fig 1 and online supplementary fig 1). As for item “get on and off the toilet”, 71% (124/175) of the non-disabled patients (severity score of 0) considered this activity as important or extremely important. At the baseline, based on the preference questionnaire, the three most important items were “dress yourself” (85%), “wash and dry entire body” (63%), and “walk outdoors on flat ground” (55%). Data collected by the importance questionnaire also showed that each of these three activities was considered important or extremely important by more than 75% of the patients. At the final visit, no change in selection of the five most important items was seen for 17% of patients, whereas 34% (61/182) and 66% (121/182) of patients selected at least four and three, respectively, of the initial five items they chose.
All individualised scale scores were highly correlated with the HAQ-DI score (rs⩾0.75; table 3). Lower correlations were observed with measures of disease activity: tender and swollen joint counts (rs ranging from 0.21 to 0.39) and DAS28 (rs ranging from 0.38 to 0.47). The lowest correlations were observed with biological features of disease activity, such as ESR and CRP level (rs ranging from 0.10 to 0.18).
For scales with eight domains, Cronbach’s α coefficient ranged from 0.87 to 0.88 (table 3). Scales preserving all 20 items had higher Cronbach’s α coefficients (α = 0.94).
The SRM of the HAQ-DI was 0.74, compared with 0.69 without help and device component (p<0.05). SRMs of individualised scales ranged from 0.64 to 0.69 (table 3). SRM of scales preserving all 20 items was 0.66.
This study evaluated the interest of individualising the HAQ-DI to measure disability in patients with RA. Individualised scales have similar psychometric properties to those of the parent HAQ-DI, whatever the individualisation method used. However, adding a measure of “importance” to a measure of “severity” for each activity provided complementary information that could be clinically useful in decision-making for individual patients.
Many studies have shown divergent views of a patient’s health between patients and clinicians.31 32 33 So far, the use of patient-reported outcomes and inclusion of patients’ opinions in their development has gradually increased.34 One usual approach to incorporate patients’ opinion in development of patient-reported outcomes is to determine, in a group of patients, the importance of each item and to weight that item with the mean importance score at group level. However, such a group approach assumes concordance of all patients’ views. To facilitate patient input into the assessment process, individualised measures have been proposed. Two key practices are recommended for individualisation35: (a) incorporation of patients’ rating of importance in addition to severity for each item, and/or (b) augmentation of the instrument with supplemental items added by patients. Our study relies on the first recommendation, and used two methods. The first method combined, for each item, the rating of “severity” and “importance”. In the second method, each patient selected a limited number of items he considered the most important. Moreover, both methods were combined, by ranking the importance of these most important items. Regarding the second recommendation, our study unfortunately, did not offer patients the opportunity to provide supplemental relevant items, not contained in the HAQ-DI.31
Arguing that importance of an activity is closely related to level of disability in performing that activity, some authors wonder whether assessment of disability alone might be sufficient.36 37 Against this hypothesis, our resulted showed that for most items, rating of importance and level of disability were weakly correlated. Moreover, as previously shown, ratings of the importance were extremely variable among patients with similar levels of disability, but were also similar among patients with diverse levels of disability.20 38 So, an individualised approach may be useful in clinical practice to help the doctor to prioritise care on what is really meaningful to each patient.
In our previous study evaluating individualised forms of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) function subscale, ratings of the importance and severity of each item were strongly correlated38; furthermore, the individualised scale retaining the five most important activities was more sensitive to change than the parent scale. In that study, patients were asked to define activities that were most in need of improvement. In the present study, the wording was slightly different, and patients had to define the importance of activities of their daily lives, even though they had no difficulty performing them. As shown for the item “get on and off the toilet”, with this wording, it is likely that important items for a patient may be those that are relevant to him, but perhaps easier to do. Therefore, if these items are initially close to normal, improvement is not possible. Items are therefore less sensitive to change. This difference might explain the discrepancies between results of these studies.
As previously proposed,25 39 we also used the HAQ-DI without the help and devices component, but found it less sensitive to change than the parent HAQ-DI. Because the rating of importance was not applied to the help and devices component, individualised scales did not include this component. This omission might have explained in part why the individualised scales had lower responsiveness than the parent HAQ-DI.
Individualised scales derived from the HAQ-DI did not have better responsiveness than the parent HAQ-DI. This might be explained by the wording of the individualised questionnaire, which did not focus on activities in need of improvement, and by the scoring method of the HAQ-DI, somewhat already individualised, considering for each domain the activity for which the patient is most disabled. However, the measure of importance of an activity gave complementary information to the measure of disability. Therefore, even if individualisation is probably not needed for group assessment in all randomised controlled trials, the use of individualised questionnaires could be clinically relevant in decision-making for individual patients.
We thank all the doctors who participated in this study.
Web Only Data 69/1/97
▸ An additional figure is published online only at http://ard.bmj.com/content/vol69/issue1
Funding This study was supported by a by an unrestricted grant from Sanofi-Aventis Laboratories, France.
Competing interests None.
Ethics approval Ethics approval from local ethics committees.
Patient consent Patient consent received.
Provenance and Peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.