METHODS

Stages in instrument development

Table 1 lists the stages in the PsAQoL development programme.

Patients

Different patient samples were employed at each stage of instrument development and testing. Stages 1, 3, and 4 employed patients diagnosed with PsA currently attending hospital outpatient clinics in Leeds. As a larger sample was required for the second postal survey (stage 5), patients were recruited with the help of the Psoriatic Arthropathy Alliance. The questionnaire was included in a routine mailing to its 450 members. Participants in this survey were required to provide information on the history of the PsA, aspects of its current status, and current use of drugs.

Item generation

The content of the PsAQoL was derived from unstructured, qualitative interviews conducted with patients with PsA in their own homes. This method of item generation fulfils a basic requirement of QoL instruments—namely, that their content should be generated from relevant patients.²³ This is because it ensures that the content of the final instrument is relevant to the target population and that issues considered important by patients are not omitted.

The interviews, which were audio recorded, were informal, focused conversations and were designed to explore how PsA affects the lives of patients. For example, interviewees were encouraged to consider how functional limitations that they experienced influenced their ability to meet their needs.

Transcripts were produced from the interview tapes, which were then wiped clean to maintain confidentiality. Information that would allow the identification of the interviewee was also omitted from the typed transcripts.

Selection of items for the draft questionnaire

The interview transcripts were subjected to content analysis to identify potential items for the questionnaire. Items were selected for inclusion where they were relevant to the needs model, were applicable to all potential respondents, reflected a single idea, were unambiguous, and were short and simple. Duplicated, idiosyncratic, and badly phrased items were rejected and a draft questionnaire constructed from the remaining items.

Field test methodology

Field test interviews were used to check on the relevance and comprehensiveness of the draft PsAQoL and to see whether patients could understand the questionnaire and complete it without problems. This was the final stage at which the wording of items could be changed to improve comprehension. It was also possible for interviewees to suggest additional items. Interviews were conducted at rheumatology outpatient clinics. Potential participants were approached in the waiting room at the clinic and had the nature of the project explained to them. If willing to take part, they were interviewed one at a time in a quiet private office.

The questionnaire was completed in the presence of an interviewer who noted whether the participants read the instructions, had any obvious difficulties, or hesitated over particular items. On completing the questionnaire, participants were invited to comment on the suitability of questionnaire items, instructions, and response categories. They were also asked specific questions about four of the items to see how they were interpreted or to check on the most appropriate wording. After this a second draft questionnaire was produced.

Postal survey 1

This survey was conducted primarily to reduce further the number of items included in the draft PsAQoL. Patients with PsA were sent a questionnaire booklet to complete consisting of demographic questions and the PsAQoL. Rasch analysis was conducted on the resulting data to identify items that failed to fit the underlying measurement construct (QoL) and/or that worked differentially by age or disease duration (above or below the median). At this stage, only the most grossly misfitting items or those with large differential item functioning (DIF) were removed.

Postal survey 2

The purpose of the second postal survey was to identify the final set of items to be included in the PsAQoL and to determine the scaling properties, internal consistency, reliability (reproducibility), and construct validity of the revised measure.

As a larger sample was required for this survey, patients were recruited with the help of the Psoriatic Arthropathy Alliance. The questionnaire was included in a routine mailing to its 450 members.

Members of the Psoriatic Arthropathy Alliance who volunteered to participate in the survey were sent a revised questionnaire booklet consisting of the demographic questionnaire (including questions about patient perceived disease activity and severity of illness), the current version of the PsAQoL, the Nottingham Health Profile (NHP), and a reply paid envelope. The NHP is a measure of perceived distress that consists of six sections: energy level, pain, physical mobility, sleep, social interactions, and emotional reactions.²⁴ This measure was selected as the comparator measure as it is generally well accepted by patients and has better psychometric properties than the other generic health status measures.^24–26 Patients who completed and returned this pack and gave permission to be approached again, were sent a similar package timed to arrive two weeks later. It was not possible to send follow up questionnaires as a condition of the access to the society’s database was that its membership should remain anonymous unless they responded to the first mailing and specifically gave permission to be approached for the test-retest mailing.

Data from the PsAQoL were again fitted to the Rasch model to confirm that the items formed a unidimensional scale, that they each represented a different amount of QoL (hierarchical ordering), and that there was an absence of DIF. The analyses also determined whether the scale operated at the ordinal or interval level. Internal consistency, test-retest reliability, and construct validity were then determined based on the final PsAQoL item set.

Statistics

Rasch model

The application of the Rasch model ensures that the fundamental scaling properties of an instrument (for example, unidimensionality and level of measurement) are assessed in addition to the traditional psychometric assessments of reliability and external construct validity.²⁷ In the present context, the Rasch model adopts the premise that the likelihood of a person affirming a particular item depends on both the level of QoL of the person and on the level of QoL represented by the item.

One way in which unidimensionality may be compromised is through DIF.²⁸ The basis of the DIF approach lies in the item response function, the S-shaped trace of the proportion of people at the same level of QoL who affirm an item. Under the requirement that the ability under consideration is unidimensional, if the item measures the same QoL across groups then, except for random variations, the same curve is found irrespective of the nature of the group for which a function is plotted.²⁸ Thus data from the draft PsAQoL measures were fitted to the Rasch model and all items tested for DIF by age and sex. Items that did not fit the model and those displaying significant DIF were removed from the scale. Fit to the model was assessed with a significance level of 0.01 to account for multiple testing.

Internal consistency (the degree of item interrelatedness) was assessed by Cronbach’s alpha coefficients. A value of 0.70 or above was taken to be indicative of adequate internal consistency.²⁹ The reliability of the PsAQoL (an estimate of the instrument’s reproducibility over time, assuming that no change in condition has taken place) was assessed using the test-retest method. Scores on the instrument at each administration were correlated, with a Spearman rank correlation coefficient of at least 0.85 considered necessary to indicate that the PsAQoL would be suitable for use in a clinical trial or for monitoring individual patients.³⁰

Construct validity was assessed by relating scores on the PsAQoL to those on the NHP and to perceived general health, reported severity of illness, and patient perceived disease activity. It was predicted that there would be moderate associations between the PsAQoL and the NHP sections, indicating that they assess related but different outcome constructs. It was also suggested that QoL would be worse for respondents reporting poorer general health or those considering their PsA to be severe.

Ethics

Ethical approval was given by the United Leeds Hospital Trust ethics committee.

RESULTS

Findings from qualitative interviews (stage 1)

Interviews were conducted with 48 (32 male, 16 female) patients with a mean (SD) age of 46.9 (13.4) years. Thirty eight (79%) were married or cohabiting and 22 (46%) were full time employed. Time since diagnosis ranged from 1 to 40 (mean 12.1) years, and 36 (75%) of the sample were currently experiencing both psoriatic and arthritic symptoms. Twenty nine (60%) interviewees were currently experiencing a flare up of their condition.

The interviews lasted between 0.5 and 2.0 hours. PsA was reported to have a major impact on many aspects of the interviewees’ lives. The main issues raised could be categorised into four main experiences or events:

• Reaction to diagnosis: preoccupation with the illness and grieving over loss of previous lifestyle.

• Life changes: reduced ability to fulfil personal roles, social life, and work. Loss of motivation to take part in, and reduced pleasure gained from, activities.

• Adaptation and acceptance: the need to adapt day to day routines and to acknowledge that such changes are necessary.

• Concerns for the future: uncertainty over future capabilities, fear of potential worsening of the condition, and the adverse effects of treatments and loss of independence.

It was clear from the interviews that as a result of PsA several needs could not be fulfilled. These included needs for control over the illness, stimulation, independence, spontaneity, companionship, communication, to feel valued, and to have a purpose in life.

Selection of items for the draft questionnaire (stage 2)

As far as possible, items for the questionnaire consisted of wording taken from the transcripts. In some cases it was necessary to shorten the quotation, change the word order, or express them as first person statements or in the present tense. The item selection process yielded 169 potential questionnaire items. Removal of duplicate, idiosyncratic, and poorly phrased items resulted in a pool of 51 items for the first version of the questionnaire. A dichotomous (“true/not true”) response format was selected, as previous studies had indicated that this was the most appropriate for such questionnaires.³¹

Field testing for face and content validity (stage 3)

Fifteen patients (nine male, six female) were interviewed. The sample had a mean (SD) age of 48.1 (10.8) years. All interviewees were married or cohabiting and eight (53%) were working full time. There was a mean (SD) period of 14.0 (10.1) years since diagnosis of PsA and 13 (87%) of the sample were experiencing both psoriatic and arthritic symptoms. Four (27%) of the sample were not currently experiencing a flare up of their condition.

The interviewees took a median of five (range 3–10) minutes to complete the 51 item questionnaire. The PsAQoL was considered to be well presented and acceptable to patients. There did not appear to be any problems during the completion of the questionnaire and no items were determined to be inappropriate. During the interview, patients were asked if there were any issues that had been omitted from the questionnaire. Some issues were raised but these tended to be idiosyncratic, such as difficulties experienced with local public transport.

Although interviewees commented that the questionnaire was relevant and covered pertinent issues they had experienced because of their illness, problems were encountered with eight of the items and some duplication of ideas was found. Problems included some interviewee’s inability to understand the word spontaneity used in one item, vaguely expressed items, and others where there was a perceived lack of specificity to PsA. Five items were removed from the questionnaire and the wording of three more was changed as a result of the interviews, resulting in a 46 item scale which was taken forward for further testing.

Exploratory analysis—first postal survey (stage 4)

Demographic details of participants who took part in both postal surveys are shown in table 2. The table also includes patient reported information about health status.

Questionnaire packs were distributed to 120 patients and returned by 94, a response rate of 78%. Rasch analyses were performed on the data to identify items for which misfit or DIF were most problematic. While some items were found to misfit, DIF was minimal. As a result of these analyses, 11 items were removed from the PsAQoL, producing a 35 item version that was taken forward to stage 5.

Internal and external construct validity and reliability—second postal survey (stage 5)

Completed booklets were returned by 286 members of the Psoriatic Arthropathy Alliance. Of those who responded, 237 (83%) also returned a completed questionnaire at time 2. As this was a validation study, measures with one or more missing responses were excluded from the analyses. Table 3 shows descriptive scores on the main outcome measures (including the final 20 item PsAQoL—see below).

Fitting data to the Rasch model again identified misfit in this larger sample. Fifteen additional items were removed owing to misfit or DIF. The 20 item solution resulted in a good overall item fit mean (SD) of 0.183 (1.115) and person fit mean (SD) of −0.232 (0.807); and item trait interaction χ² of 96.1 (df = 80; p = 0.106). Individual item difficulty level (QoL impact) ranged from −2.02 logits to +2.8 logits, and individual item fit statistics ranged from p values of 0.954 to 0.017. A person separation index of 0.922 indicated that the scale could differentiate four or more groups of patients along the continuum. Table 4 gives the best and worst fitting items of this solution.

The 20 item solution was tested again in the retest sample. Once again this resulted in a good overall item fit mean (SD) of 0.047 (1.508) and person fit mean (SD) of −0.205 (0.704); and item trait interaction χ² of 79.9 (df = 60; p = 0.043). Individual item difficulty level (QoL impact) ranged from −2.18 logits to +2.86 logits and individual item fit statistics ranged from p values of 0.976 to 0.010. A person separation index of 0.930 indicated that once again the scale could differentiate four or more groups of patients along the continuum. This second analysis confirmed the robustness of the internal construct validity of the scale.

Reliability and internal consistency of the PsAQoL

The test-retest Spearman rank correlation coefficient was 0.89, indicating that the measure has excellent reliability, producing low levels of random measurement error (table 5). In contrast, the reliability of the Health visual analogue scale (VAS; 0.72), the QoL VAS (0.78), and that of three of the NHP sections failed to meet the standard required of instruments used in clinical trials. Internal consistency of the PsAQoL was 0.91 at both administrations, demonstrating an adequate degree of interconnectedness of items.

External construct validity

Moderately high correlations were seen between the PsAQoL and NHP section scores (table 6). The highest correlation was with the energy level section (0.75)—a predictable result reflecting the importance of fatigue on QoL in PsA. Similar levels of correlations were also found between PsAQoL scores and those on the Health and QoL VASs, suggesting that these items were measuring related but different constructs than the PsAQoL.

Scores on the PsAQoL were unrelated to sex, age, or marital status or time since diagnosis. However, as expected, employed respondents had better QoL (p<0.001). Predictably, PsAQoL scores were also significantly related to perceived current health (p<0.001) and to scores on the QoL VAS (p<0.001) and those on the Health VAS (p<0.001).

PsAQoL scores were unrelated to whether or not facial skin was affected, respondents were having a psoriasis flare up, or they were currently being treated for their psoriasis. In contrast, scores on the measure were worse when the hands were affected by psoriasis (p<0.001) and were significantly related to perceived severity of psoriasis (p<0.02). PsAQoL scores were also significantly related to whether respondents were having a flare up of arthritis (p<0.001), whether the arthritis was being treated (p<0.01), and the perceived severity of arthritis (p<0.001).

Figures 1 and 2 illustrate the associations between PsAQoL scores and demographic and health status variables.

DISCUSSION

Perhaps one reason for the lack of success of patient reported instruments used in studies of PsA is that the condition has both rheumatic and dermatological impacts on patients. No measure is available that is specific to PsA. The HAQ and AIMS are measures developed for use with rheumatic illnesses, and the SF-36 has been found to provide contradictory scores for patients with psoriasis, suggesting either that their health status is better than that of an average population¹³ or that the impact of the illness on patients is greater than that of arthritis, cancer, and myocardial infarction.¹⁴ The SF-36 is a generic measure of health status and the HAQ and AIMS are intended for use with all rheumatic conditions, implying that their content may not be particularly well focused on the issues of importance in PsA. Furthermore, these measures focus primarily on impairments and activity limitation (disability), adding relatively little information to routine clinical indicators of outcome. Consequently, they do not inform on the overall impact of the condition and its treatment on the patient’s QoL. The patients’ perspective is crucial when determining the acceptability of an intervention and associated compliance.

The PsAQoL is based on a clear, conceptual model of QoL that has been successfully used in the development of several disease-specific QoL instruments, including measures for rheumatoid arthritis, ankylosing spondylitis, and systemic lupus erythematosus. The items in the measure largely represent the words of individual patients with PsA and were derived directly from qualitative, unstructured interviews. This method fulfils the basic requirement that QoL instruments should be based on information derived directly from patients and ensures that future respondents find the instrument acceptable, comprehensive, and relevant to their condition. The PsAQoL is quick and easy to complete—taking less than five minutes, making it suitable for use in clinical settings.

In rheumatology, Rasch analysis is increasingly seen as the standard approach ensuring quality measurement.³² Fitting data from the PsAQoL to the Rasch model confirmed a unidimensional scale with good item stability over time, and minimal DIF. Scales that fit the Rasch model have the added advantage that scores can be transformed to the interval level. This allows valid arithmetic operations such as the calculation of change scores or effect sizes to be applied, an important requirement for all outcome measures in musculoskeletal disorders.³³

The test-retest reliability and internal consistency of the PsAQoL are excellent, indicating that the instrument is suitable for use in individual patients. For an instrument to have construct validity it must be based on a model of the construct assessed and be shown to have good reliability.³⁴ As these requirements are met for the PsAQoL it can be concluded that the measure provides a valid assessment of the construct defined in the model. Additional indications of the validity of the PsAQoL were provided by the moderately high correlations between PsAQoL scores and those on the NHP (perceived distress), the Health VAS and the QoL VAS. Scores on the PsAQoL were also related, as expected, to a number of variables, including perceived current health, whether the hands were affected by psoriasis, perceived severity of both psoriasis and arthritis, and whether respondents were having a flare up of their arthritis. It was important to establish that the measure was related to both the dermatological and rheumatological aspects of PsA.

It is acknowledged that the sample used in the second postal survey was self selected and might have included some subjects who did not have PsA. However, the psychometric analyses did not suggest that there were problems with the sample employed.

It also remains to be shown that the PsAQoL is responsive to changes in QoL associated with effective interventions for PsA. Such a property requires that a measure has good reliability and validity, as has been demonstrated in the case of the PsAQoL.

A criticism of disease-specific measures is that they do not allow comparisons to be made across diseases and, consequently, that they cannot form the basis of economic analyses comparing the benefits of interventions for different diseases. Modern psychometric techniques overcome this problem through item banking.³⁵ Where instruments are designed to measure the same construct it is possible to “co-calibrate” them onto the same underlying metric (logit) scale. The PsAQoL has been explicitly designed to be added to such an item bank by an item selection process that, other things being equal, selects items that are common with other disease-specific scales. This then permits common item equating of different scales and incorporation into a QoL item bank.

It is concluded that the scaling and psychometric properties of the PsAQoL indicate that the instrument can be used with confidence by researchers and clinicians. The instrument represents the first patient based outcome measure specific to PsA and (assuming that it proves to be responsive to change in QoL, as suggested by its psychometric properties) it will be a valuable tool for assessing the impact of PsA and its treatment in clinical settings, trials, and research studies. Such an instrument will allow assessment of the effectiveness of interventions from the patient’s perspective. Furthermore, the items included in the PsAQoL will be incorporated into a rheumatology item bank, allowing comparisons to be made between the QoL of patients with different rheumatic conditions.

APPENDIX

Final 20 items of the PsAQoL

The PsAQoL must not be reproduced, used or adapted into other languages without permission of the authors. UK researchers wishing to use the measure should contact Professor Alan Tennant at the Academic Unit of Musculoskeletal and Rehabilitation Medicine, The University of Leeds, 36 Clarendon Road, Leeds LS2 9NZ. UK (Email: AlanTennantcompuserve.com). Commercial organisations and overseas researchers should contact Dr Stephen P McKenna at Galen Research, Enterprise House, Lloyd Street North, Manchester M15 6SE, UK (Email: smckennagalen-research.com).

• I feel tired whatever I do

• I find it difficult to have a good wash

• It’s too much effort to go out and see people

• I feel there’s no enjoyment in my life

• I feel I am losing my independence

• I often get angry with myself

• I can’t do the things I want to do

• I feel older than my years

• I’m unable to join in activities with my friends or family

• It limits the places I can go

• I have to push myself to do things

• I am easily irritated by other people

• I have to keep stopping what I’m doing to rest

• I feel dependent on others

• It takes me a long time to get going in the morning

• I take it out on people close to me

• I can’t do things on the spur of the moment

• I feel like a prisoner in my own home

• I have to limit what I do each day

• It puts a strain on my personal relationships