Dr. Redmond is Lecturer for the UK Arthritis Research Campaign in Rheumatological Podiatry, University of Leeds.
Development of a foot impact scale for rheumatoid arthritis
Article first published online: 2 JUN 2005
Copyright © 2005 by the American College of Rheumatology
Arthritis Care & Research
Volume 53, Issue 3, pages 418–422, 15 June 2005
How to Cite
Helliwell, P., Reay, N., Gilworth, G., Redmond, A., Slade, A., Tennant, A. and Woodburn, J. (2005), Development of a foot impact scale for rheumatoid arthritis. Arthritis & Rheumatism, 53: 418–422. doi: 10.1002/art.21176
- Issue published online: 2 JUN 2005
- Article first published online: 2 JUN 2005
- Manuscript Accepted: 6 DEC 2004
- Manuscript Received: 30 APR 2004
- Rheumatoid arthritis;
- Foot problems;
To develop a new foot impact scale to assess foot status in rheumatoid arthritis (RA) using established qualitative methodology and the latest item response techniques (Rasch analysis).
Foot problems in RA were explored by conducting qualitative interviews that were then used to generate items for a new foot impact scale. Further validation was undertaken following postal surveys and Rasch analysis.
Analysis of the first postal survey (n = 192 responses) produced a 63-item binary response, 4-subscale instrument. The 4 subscales covered the domains impairment, activities, participation, and footwear. Following test–retest postal surveys and additional analysis, the instrument was reduced to a 2 subscale, 51-item questionnaire covering the domains of impairments/shoes and activities/participation. Initial results of these subscales indicate good psychometric properties, external validity, and test–retest reliability.
A foot impact scale to assess the impact of RA and to measure the effect of interventions has been developed. The 2 scales comprising the instrument demonstrate good psychometric properties.
In rheumatoid arthritis (RA), pain, stiffness, deformity, and functional loss are among the major determinants of foot-health–related quality of life (1, 2). These disease-related components are measurable using several generic (3–5) and a single RA-specific (6), foot-health instruments. Satisfactory psychometric properties have been described for all of these self-administered instruments. The instruments differ largely in the foot-health concepts measured, including pain (3–6), disability (4, 6), function (3), activity limitation (6), footwear (3), and general foot health (3). The instruments use different scoring methods, including Likert scales (3–5), visual rating scales (7), and visual analog scales (6), for both the intensity and/or frequency of the measured constructs. Only the Foot Function Index (FFI) has been validated for use in non-English languages (7). However, there are 2 major concerns with the FFI: item selection for the pain, disability, and functional limitation constructs was conducted by expert panel (nonpatient) agreement, and the subscale scores are evaluated as if based on continuous variables (6, 8–12). The former has resulted in a questionnaire largely consisting of disease-related items in the pain and disability constructs. Through patient interviews, other studies have found that nondisease factors such as footwear, participation restriction, body image, and psychological impact are important constructs (3–5); however, these are excluded from current instruments.
The development of specialist RA foot clinics has provided health professionals with an opportunity to gain a greater understanding of the impact of RA on the foot and ankle. Generic foot-health instruments and the FFI have aided health professionals in measuring impact, but, as shown above, the latter has flaws in its psychometric properties. Therefore, there are 3 main reasons why we developed a new foot-health impact scale for RA. First, the high prevalence of foot involvement in RA suggests that there is a clinical need to measure the impact of the disease at this specific site. Second, an instrument is needed that through carefully conducted interviews includes both disease- and nondisease-related constructs that are relevant to the target population and, through psychometric testing, meets current standards for instrument properties, including validity, repeatability, and responsiveness. Third, treatment of foot-related conditions often requires more than systemic drug therapy, calling for podiatry, physiotherapy, and orthopedic services. Therefore, one of the main uses of a new instrument will be the evaluation of the efficacy of multidisciplinary foot-health care in RA, both for routine clinical purposes and for research (9, 11).
Ethical approval was obtained from the Local Research Ethics Committee for both the Leeds and Bradford Teaching Hospitals. Patients were recruited from rheumatology outpatient departments at Leeds General Infirmary, Leeds, UK and St. Luke's Hospital, Bradford, UK.
PATIENTS AND METHODS
Patients were recruited who had been clinically diagnosed with RA by the consultant in charge of their care and had “problems with their feet.” The study was divided into 6 stages as outlined in Figure 1. Patients with significant comorbidity, such as primary mental health problems, and patients who were unable to read and understand English, were excluded.
Stage 1. Qualitative interviews.
The content of the RA foot-health measure was derived from qualitative interviews with patients, conducted by 2 experienced qualitative interviewers. Thus, specific statements included within the scale were derived from the patient's own comments, and often in their own words. This approach ensures that the content of the final instrument is relevant to the target population and that issues considered important by patients are not omitted.
Thirty patients were selected for the interview by use of a sample frame, which was constructed with patients stratified according to age (<50 years or ≥50 years), sex, and disease duration (<2 years, 2–6 years, or >6 years). The interviews, which were audio-taped, took the form of informal, focused conversations. Patients were asked to talk about the problems they experienced with their feet due to RA. No attempt was made to suggest possible difficulties or particular problems. The interviews lasted 30–60 minutes.
Stage 2. Transcript analysis and selection of items for the draft questionnaire.
A sample of the transcripts were read by 2 members of the research team to identify agreement on the main themes that emerged. Detailed content analysis of all the interview transcripts was then carried out with the assistance of NUD*IST computer software (QSR International, Bundoora, Victoria, Australia). There were 8 main themes or nodes (symptoms, mobility, footwear, affects on others and relationships, restrictions other than mobility problems, foot appearance, treatment, and emotions) each of which had a number of subthemes. The transcripts were analyzed to identify potential items for the questionnaire. Items selected fulfilled the following criteria: relevant to patients with RA who were experiencing foot problems; applicable to all potential respondents, irrespective of age or sex; and reflected a single idea and were unambiguous.
Four members of the research team identified potential items. Whenever possible, items selected for the questionnaire were direct quotations from the transcripts. However, the actual words used by interviewees were altered for some items so that they were expressed in the first person and/or in the present tense. A total of 131 potential items were identified at this stage.
A draft questionnaire was prepared using a yes/no format and the 131 items identified. There was agreement within the research team that the items could be grouped into 4 dimensions. Three dimensions corresponded with the International Classification of Functioning, Disability and Health dimensions of activity (37 items), impairments (56 items), and participation (24 items) (http://www3.who.int/icf). A fourth dimension concerning factors related to footwear/shoes (14 items) was also apparent.
Stage 3. Field-testing for face and content validity (cognitive debriefing).
This stage was designed to test whether patients with RA were able to understand and complete the draft questionnaire, and to ensure that the content was relevant and that important issues had not been omitted. The 131 items were divided, as the total number was too lengthy for every patient to complete and comment on. A total of 9 patients with RA completed the first draft of the measure. Patients were asked to complete part of the draft Foot Impact Scale (FIS) in the presence of an interviewer and then to comment on its ease of completion and on the appropriateness of the instructions, items, and response format. As a result it was agreed that the items were clearer with a true/not-true response choice than with a yes/no format. In addition, 5 items were thought not to apply to all patients and were discarded, and 1 item was divided into 2 items for clarity. A second draft of the measure was subsequently prepared that comprised 127 items with a true/not-true response format. This was tested with an additional 6 patients, and no further problems were encountered.
Stage 4. Initial postal survey.
The initial postal survey tested the scaling properties of the draft FIS, facilitated item reduction, and provided preliminary evidence of construct validity. A random sample of 288 patients with RA selected from a computerized database were asked to complete and return a questionnaire booklet that included demographic questions, the initial draft FIS (FISv1), the Health Assessment Questionnaire (HAQ) (13), the FFI (6), and the Manchester Foot Disability Questionnaire (MFDQ) (4). Data from the FISv1 were then fitted to the Rasch measurement model (14) (see below) to facilitate item reduction.
Stage 5. Second postal survey (test–retest).
A second random sample of 188 patients with RA were selected from the database and sent a revised questionnaire (FISv2) together with the HAQ, the FFI, and the MFDQ. Patients who completed and returned this booklet were sent a second booklet 2 weeks later. Eighty-five of these patients provided paired data.
The Rasch measurement model.
The Rasch model (14) is the current standard for the development of unidimensional scales (e.g., level of foot impairment) that deliver metric quality outcomes in health care (15). Thus, data from the items comprising the FIS in both the first and second postal surveys were fitted to the dichotomous Rasch model following a protocol described in detail elsewhere (16, 17). Previous studies constructed short form scales using this method without any loss of validity, and found it to be a better method for item reduction than more classic forms of item reduction such as principal components analysis or multiple linear regressions (18).
Briefly, data collected from the questionnaires are tested against the expectations of the model. A variety of fit statistics determine how well the data fit the model (19). Also, within the framework of Rasch measurement, the scale should work in the same manner, irrespective of which group (based on age, sex, or disease duration) is being assessed (20). For example, when assessing participation, older and younger age groups should have the same probability of affirming an item if they have the same level of participation. If, for some reason, the older group did not affirm the level of participation on a particular item, then this item would be deemed to have differential item function (DIF) and would violate the requirement of unidimensionality (20, 21). Consequently, every item was checked for DIF by age, sex, and, for the current analysis, by duration of disease. Data were fitted to the Rasch model using the RUMM2010 software (22).
First postal survey.
An initial postal questionnaire sent to 288 patients elicited 192 responses (67%) after one reminder. The 192 responses yielded 148 patients (34 men and 114 women, mean age 61.7 years [range 28–89 years], mean disease duration 15.4 years [range 1–55 years), mean HAQ score 1.5 [range 0–3]) who reported an arthritis-related problem with their feet. After the first postal survey, items that failed to fit the Rasch model, or displayed DIF by age or sex, were removed, leaving a 63-item scale in 4 dimensions.
Second postal survey.
A revised questionnaire was sent to a second sample of 188 patients with RA, and then sent again 2 weeks later. A total of 135 responses (72%) were obtained from the first mailing; and 85 (45%) responded to the second mailing. Sixty-one women and 24 men (mean age 67.1 years [range 28–92], mean disease duration 17.3 years [range 2–57], mean HAQ score 1.6 [range 0–3]) completed both questionnaires. Further Rasch analysis of the responses to the 63 items across 4 dimensions revealed DIF over time and significant ceiling effects. As a result, the decision was made to test whether the 4 subscales could be combined into 2 subscales, a 32-item impairment/shoes scale and a 31-item activities/participation scale.
The 32 items from the impairments and shoes subscales were combined and tested against the Rasch model in the test–retest data. Some of these items were found to display DIF across time; therefore, the scale was reduced further. One item showed DIF by sex (“I can't wear my choice of shoe”). A final subscale of 21 items was then fitted to the original postal questionnaire data and found to have a good fit to the Rasch model, with an overall mean ± SD item of −0.25 ± 1.07, a mean ± SD person of −0.34 ± 0.71, and item-trait interaction chi-square with degrees of freedom (df) of χ2 = 45.07, 42 df; P = 0.35, indicating no significant deviation from model expectation across the trait. There was no DIF for age, sex, and disease duration, and the best and worst fitting items are shown in Table 1. Person separation was 0.808, indicating that the scale was able to discriminate groups of patients. The test–retest analysis of these 21 items showed an intraclass correlation coefficient (ICC) of 0.84 (95% confidence interval [95% CI] 0.75–0.90). The limits of agreement between time 1 and time 2 (calculated from the mean difference and 1.96 × the SD of the difference) were −0.47 ± 4.42.
|Item description||Location||SEM||Fit residual||χ2||Probability|
|Feet throb at night||0.306||0.202||0.483||0.055||0.813785|
|Pebbles in shoes||0.875||0.197||−0.428||0.139||0.708808|
|Pain tension in feet||−1.231||0.251||−1.647||3.257||0.071121|
|Wider fit shoes||−1.188||0.249||0.771||5.234||0.022152|
The 31 items from the activities and participation scales were combined and fitted to the Rasch model from the test–retest data. One item (“I can't walk barefoot”) was found not to fit, and there was no DIF by time. After removing the item, the final 30-item scale was tested on data from the original postal questionnaire and showed a good fit to the Rasch model. The overall mean ± SD item was −0.38 ± 0.93 and the mean ± SD person was −0.29 ± 0.80, with a chi-square with df interaction of χ2 = 64.4, 60 df; P = 0.18, showing no significant deviation from model expectation. Once again there was no DIF, and the best and worst fitting items are shown in Table 2. A person separation of 0.908 indicated the ability to discriminate more than 4 groups of patients along the trait. Test–retest analysis showed an ICC of 0.96 (95% CI 0.93–0.98). The limits of agreement between time 1 and time 2 (calculated from the mean difference and 1.96 × the SD of the difference) were −0.54 ± 4.59.
|Item description||Location||SEM||Fit residual||χ2||Probability|
|Can't go for walks||0.401||0.219||0.143||0.078||0.961902|
|Nervous of curb edge||0.938||0.208||1.601||0.079||0.961089|
|Take longer to do things||−1.503||0.317||0.902||4.454||0.107828|
|Have to plan||1.135||0.209||−1.506||5.521||0.063272|
|Need help climbing stairs||2.492||0.227||0.318||5.546||0.062483|
Relationship to other scales.
The scores for the MFDQ, FFI, and HAQ were divided into 4 quartiles, and for each category, the corresponding mean scores for the FIS impairments/shoes subscale and FIS activities/participation subscale were compared using the Kruskal-Wallis nonparametric test for k independent variables. In each case, scores on the impairment/shoes and activities/participation subscales demonstrated appropriate progression with highly significant differences between groups for each of the measures.
Foot involvement is almost universal in RA, yet little is known of the impact of foot involvement on the patient in terms of impairment, disability, and quality of life. The study described here has produced 2 disease-specific scales that can be used to assess the impact of RA of the foot in terms of footwear, impairments, activity limitation, and participation. Both scales demonstrated good repeatability and appropriate incremental change in scores with reference to both a generic disability measure (the HAQ) and to other instruments designed to assess pain and disability due to foot problems.
Are these new scales an improvement on existing instruments? Although the MFDQ was developed from patient-generated statements, the patients were recruited from a variety of backgrounds, including community podiatry clinics, and therefore the statements may not reflect the severity of the problems experienced by patients with RA. This is reflected by the ceiling effect found with the MFDQ in patients with established rheumatoid disease (23). In contrast, the FFI was developed as a means to establish the benefit of orthotic intervention, specifically in RA; however, a major concern is the fact that the statements used in the instrument were adopted by expert, nonpatient, panel discussion. In addition, parametric statistics are often used inappropriately, thus creating, at least in theory, the potential for error in statistical comparisons. Therefore, the new instrument has at least 2 major advantages: it is based on statements derived from appropriate patients, and it fits the Rasch model so that it has, with the consequent transformation to a linear metric, the potential for parametric analysis.
The new instrument contains impairment items in the sensory pain domain, and many of these will be recognizable to experienced clinicians. For example, many patients compare the symptoms related to metatarsophalangeal joint disease with “walking on pebbles or marbles,” and this item was included in these words as derived from patient interviews. Foot pain in RA also has burning and throbbing qualities, symptoms made worse by prolonged daily activities, that relate to both the underlying disease process and the mechanical stresses placed on deformed joints. These sensory aspects are captured in the new instrument along with temporal aspects such as worsening symptoms towards the end of the day. We were not surprised to see items emerge related to both the affective and cognitive elements of foot pain. Foot pain is often undervalued and not fully assessed in the clinical setting, and it is often difficult to treat even when inflammation is quiescent. The new instrument contains items where the sense of frustration, annoyance, and fear associated with foot pain in RA can be adequately assessed.
Joint stiffness and deformity are key features of the rheumatoid foot and are associated with frustrating everyday footwear problems related to availability, choice, and fit, and, when ill-fitting, associated with discomfort and exacerbation of symptoms. The value attached to this domain of foot problems will be of no surprise to podiatrists and orthotists who regularly encounter these problems in daily practice. The new instrument contains a series of items in the impairment/footwear scale to assess these problems.
Foot problems are common in RA, and patients are often referred for podiatry treatment at an early stage. The study of podiatric interventions in RA has lagged behind studies of other interventions, notably those of new pharmaceuticals. However, the evidence base for podiatry is now expanding (11, 24). This study has produced a reliable, disease-specific scale to measure the outcome of interventions for further studies in this field.
- 14Probabilistic models for some intelligence and attainment tests. Chicago: University of Chicago Press; 1960..
- 16Dimensionality of an early childhood scale using rasch analysis and confirmatory factor analysis. J Outcome Meas 1997; 1: 56–85., , .
- 19Fit analysis in latent trait measurement models. J Appl Meas 2000; 1: 199–218..
- 20DIF detection and description: Mantel-Haenszel and standardisation. In: HollandPW, WainerH, editors. Differential item functioning. Hilldale (NJ): Lawrence Erlbaum Associates; 1993. p. 36–66., .
- 21Assessing and adjusting for cross-cultural validity of impairment and activity limitation scales through differential item functioning within the framework of the Rasch model: the PRO-ESOR project. Med Care 2004; 42 Suppl 1: I37–48., , , , , , et al.
- 22RUMM2010. Perth: RUMM Laboratory, 2000., , , .
- 24Custom-designed foot orthoses modulate plantar pressures in rheumatoid arthritis [abstract]. Arthritis Rheum 2001; 44: S371., .