Objective Delegates at the Outcome Measures in Rheumatology (OMERACT) 10 conference (Borneo, 4–8 May 2010) questioned how the new seven-domain Rheumatoid Arthritis Impact of Disease (RAID) score performs as a global measure. Score distributions and associations between the RAID score and patient-reported outcomes (PROs) and demographic variables were examined in a large sample of rheumatoid arthritis (RA) patients.
Methods 1086 patients in the Oslo RA Register responded to a postal survey with commonly used PROs. Bivariate associations between the RAID score and other measures are reported as Pearson correlation coefficients.
Results The mean RAID was 3.37±2.17. The distribution of the RAID score showed a slight floor effect: 17.5% had a score between 0 and 1, and 14.4% between 1 and 2, whereas only 1.0% and 0.3% had scores between 8 and 9, and 9 and 10, respectively. Correlations between the RAID score and the patient global assessment, Rheumatoid Arthritis Disease Activity Index, Short-Form (SF)-6D and EQ-5D were 0.82, 0.82, −0.77 and −0.73, respectively. Strong correlation was also seen between RAID and pain, the domain with highest weight, whereas correlations to measures of other RAID domains were moderate. The RAID score was higher in women than men (3.49 vs 2.95, p=0.001).
Conclusion The RAID score was correlated more strongly to other global measures than to PROs, reflecting single health domains.
Statistics from Altmetric.com
The new Rheumatoid Arthritis Impact of Disease (RAID) score meets some timely requirements: it is feasible, patient-derived and provides a profile as well as a score.1 An examination of the psychometric properties of the RAID score has been performed in a sample of 570 patients (about 50 patients from each of the participating European League Against Rheumatism (EULAR) countries, working with the RAID score).1 Both patients, clinicians and researchers at the Outcome Measures in Rheumatology (OMERACT) 10 conference (Borneo, 4–8 May 2010) asked how the RAID score correlates with other, established global measures. They also discussed whether the profile of the RAID score represented a more differentiated expression of a global measure. Thus, there is a need to perform a wider assessment of the profile as well as the score in larger populations of patients with rheumatoid arthritis (RA), and in particular, to examine the correlation of the score with other frequently used patient-reported outcomes (PROs).
Our objective was to examine the performance of the RAID score within a large representative population of patients with rheumatoid arthritis (RA). In particular, we wanted to examine the distribution of the RAID score and its associations with other PROs and demographic variables to complement the primary multinational validation in selected patient groups.2
Patients and methods
The Oslo RA register (ORAR) longitudinally assesses PROs in a representative RA population.3 A postal survey to individuals in the ORAR was performed in 2009. The 1086 responding patients with RA (response rate 60.6%, mean age 61.7±14.0 years, 77.1% female, mean disease duration 14.1±11.1 years, 56.6% rheumatoid factor (RF)-positive) completed questionnaires including 100-mm visual analogue scales (VAS) on pain, fatigue, patient global assessment of disease activity and a variety of other health status measures. We computed scores for RAID,1 Rheumatoid Arthritis Disease Activity Index (RADAI),4 Short-Form (SF)-36,5 utility measures (SF-6D6 and European Quality of Life 5-Dimensions (EQ-5D)7), Health Assessment Questionnaire (HAQ),8 a subscale of the Medical Outcomes Study (MOS) sleep disturbance questionnaire,9 self-efficacy scales for function, pain and symptom,10 and the short-form Rheumatology Attitude Index (RAI-5).11
RAID measures seven domains, each with 0–10 numeric rating scales (NRS) that are perceived by patients to be particularly important for their health. Each domain has the following weight: pain 0.21, functional disability 0.16, fatigue 0.15, sleep problems 0.12, emotional well-being 0.12, physical well-being 0.12 and coping 0.12.1 The score has a range from 0 to 10 (10 worst health). A more detailed description of the PROs with relevant references is available in the online supplementary material.
The scores are presented as mean±SD and median with IQR. Bivariate associations between RAID score and other PROs, age and disease duration are reported as Pearson correlation coefficients, but analyses with Spearman's r were also performed. To display correlations between health status measures we also constructed matrix plots. All analyses were performed with SPSS/PASW version 17.0. Data collection in the ORAR is approved by the regional ethical committee and patients signed written consent when they answered the questionnaires. Correlations below 0.40 were considered weak, between 0.40 and 0.70 moderate, and strong if ≥0.70.
RAID score could be computed in 1041 patients (95.9%). The mean RAID was 3.37±2.17. Table 1 and figure 1 show the distribution of the RAID score. A weak floor effect was observed, that is, aggregation of scores at the lower end of the scale: 17.5% had a score between 0 and 1, and 14.4% between 1 and 2, whereas only 1.0% and 0.3% had scores between 8 and 9, and 9 and 10, respectively. Distribution of the seven individual NRS is also shown in table 1 and histograms are shown in online supplementary figures S1A–G. Floor effect was most pronounced for the sleep and coping components.
Correlations between RAID score and other measures are presented in table 2. Strongest correlations were seen to other global measures with the exception of SF–36 general health and to pain scales. Correlations to other measures of domains of the RAID score were generally moderate (table 2). Matrix plots are presented as online supplementary figures. Figure S2A supports the strong correlation between RAID and other global measures and figure S2B–G show that associations as expected were stronger between related than unrelated dimensions of health.
RAID score was higher in women than men (3.49 vs 2.95, p=0.001). This sex difference did not seem to be driven by any particular domain(s), since all NRS had higher scores in women than men. Sex-related differences in the same direction were also observed for other PROs, highest for HAQ Disability Index (DI) (female 1.01, male 0.67, p<0.001). The RAID score was similar in RF-positive/negative patients (3.35/3.28, p=0.62). RAID score had weak correlation to age/disease duration (0.21/0.19) (table 2) and on the same level as other global measures: patient global VAS 0.20/0.18, RADAI 0.21/0.16, SF-6D −0.26/−0.17, EQ-5D −0.15/−0.17 and HAQ-DI 0.36/0.33. The RAID score and each component were higher in patients with disease duration exceeding 10 years than in subgroups with short (<1 year) or intermediate disease duration (1 to <5 and 5 to <10 years).
By opening the consultation with “How are you?”, clinicians frequently assess their patients globally. It is, however, unclear what patients themselves emphasise in response to this question. The RAID instrument provides a weighted profile of seven health domains of major importance to patients. Splitting the global question into such a short and feasible profile could lead to more targeted suggestions for care and treatment. However, the main purpose of RAID is to provide a single score that can be used in clinical research and daily practice.1
Our results support the feasibility of RAID score. It could be calculated in 95.9% of the patients, and missing values were found in only 2–3% of the patients for each of the seven NRS (table 1). However, distribution differed between the scales with lowest floor effect for the pain scale (figure S1A–G), which is consistent with the perceived importance of pain to patients with RA.12 A weak floor effect was also observed for the RAID score (table 1, figure 1). Floor effect is a general limitation of most health status measures and for comparison we present histograms of patient global VAS, RADAI and HAQ, which also display considerable floor distribution of these scores (see online supplementary file, figure S3A–C). The utility instruments SF-6D and EQ-5D have different distributions as also shown previously in data from ORAR (figure S3D,E).13
We considered an alternative unweighted RAID score as a mean of the seven NRS. This version might be more feasible than a weighted score, but has less face validity since the influence from the patients' perception of impact of the domains would not be incorporated. However, we examined the performance of an unweighted RAID score in exploratory analyses and found that the weighted versus unweighted version had slightly higher correlations to global measures and less floor effect.
It is a possible concern that the RAID score was significantly higher in women than men. This difference was more strongly statistically significant for the RAID score than for other PROs with the exception of HAQ-DI (1.01 vs 0.67, p<0.001). The ten patients who identified the initial major dimensions in the first elaboration step were all women. Both sexes were represented in the following data-driven steps in the elaboration of the RAID score, but as many as 82.5% were women in the weighting procedure.1 Further studies may be needed to clarify whether the RAID score has a bias towards selection of health problems of special importance to women.
The RAID score is considered to give an informative expression of the general status of the patient in domains of major importance to people with RA. Our results support the concept of a global measure since RAID correlated strongly to other global measures of health (patient global VAS, RADAI, EQ-5D and SF-6D) and moderately to six of the seven domains. The strongest correlation was as expected to pain, since pain has the highest weight in the RAID score (table 2).
Disability, measured with HAQ, correlated moderately to RAID score (0.68). The HAQ-DI score was more than RAID associated to age and disease duration, which is consistent with lower responsiveness of HAQ-DI in established versus recent onset RA.14 HAQ-DI measures physical functioning, which only constitutes one of the seven domains in the RAID score.
The current results are reported from a postal survey in ORAR. This register has been shown to represent 85% of the underlying patient population aged between 20 and 79 years.3 Thus, data from ORAR can be considered to be representative for an RA population within a geographical area and our results are complementary to the more stringent recent validation of the RAID score.2 One strength in our study was the comprehensive data collection with an opportunity to present consistent results across different instruments capturing similar domains (table 2). A postal survey in this study was considered to be the most appropriate design, allowing participation of a large number of patients, but prevented clinical and laboratory examination. Thus, RADAI was the only measure for disease activity.
Correlation analyses were examined with bivariate Pearson correlation coefficient, but non-parametric correlation analyses gave correlation coefficients of the same magnitude as Pearson correlation coefficients. All correlations were strongly statistically significant, as expected, taking into account the large patient population.
Other PRO composite scores have been used in clinical research and practice. Patient Activity Scale15 and Routine Assessment of Patient Index Data (RAPID)16 are the most widely used, but these measures only include pain, function and global assessment. The RAID score has a broader focus on other health domains including fatigue, sleep, well-being and coping.
The RAID score has been developed and translated across several countries; it is free of charge and short, making it feasible and widely applicable. Our results support RAID being a global composite measure and captures and combines information about domains of health that are important to patients with RA. Thus, it is an attractive PRO when the patient perspective on disease impact is needed. This new score has undergone a successful validation,2 but further data are needed, especially on responsiveness in intervention studies and on associations with joint damage.
Patient consent Obtained.
Ethics approval This study was conducted with the approval of the South-Eastern Regional Committee for Research Ethics in Medicine and Health.
Competing interests Professor Johannes Bijlsma was the handling Editor
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.