METHODS

Generation of an initial list of items likely to be important to adults with CF was informed by the following processes: (1) consideration of the general literature exploring HRQoL, (2) literature dealing with the psychosocial and physiological aspects of CF, (3) examination of existing generic and respiratory specific questionnaires, (4) the outcome of interviews with 60 CF adults conducted during the course of previous research10 which outlined areas of concern to them, (5) examination of the literature exploring HRQoL in CF, and (6) discussions with multidisciplinary staff from two specialist CF centres. These suggested several areas for inclusion in the questionnaire including physical functioning, social functioning, CF specific issues (chest symptoms, treatment, and body image), interpersonal relationships, career/college issues, and concerns for the future. Fifty items were constructed to represent these areas. Responses to the questionnaire were measured by a six point Likert scale ranging from 1 = very strongly agree to 6 = very strongly disagree. The response scale was arbitrary in a bid to encourage patients to suggest the most appropriate method of responding. A final screening item asked patients to report what they found most troublesome about their CF. This was included to ascertain whether any salient issues had been omitted. This initial questionnaire was designed as a framework from which items could be removed or added to in subsequent versions.

PATIENTS AND PROCEDURE

One hundred and five patients were recruited via the outpatient clinic of a specialist adult CF unit. Patients were randomly approached during routine visits and asked for consent to participate in the development of the questionnaire. The researcher had no prior knowledge of the patients or of their medical status. There was no ‘a priori’ exclusion of any patient attending the clinic on any grounds. To encourage genuine and meaningful responses no demographic or clinical details were taken for this initial and important phase of development.11 Patients self-completed the questionnaire which took approximately 10 minutes. Following completion, each patient was asked to comment on and criticise the structure and content of the measure. Responses to this were noted in order to include any refinements in a subsequent draft.

RESULTS

Of the 105 questionnaires completed, four were discarded because of incomplete data. Principal components analysis (PCA) with varimax rotation was conducted on items 1–50. PCA with varimax rotation is a statistical technique based on correlational principles which is applied to a single set of variables (questionnaire items) in order to highlight coherent subsets within the overall data set. In questionnaire design this technique is applied to identify the presence of domains (subscales reflecting different aspects of functioning) within a larger battery of items. The solution emerging from the PCA highlighted 10 principal factors with eigen values above 1.0 which accounted for 71.1% of the overall variance within the data set. An additional method of determining the relative importance of factors is the scree plot described by Cattell.12 This technique involves the graphical display of the descending variance accounted for by all the factors extracted by the analysis. The cut off point for this method is determined by the gradient of the slope. A scree plot confirmed the factors highlighted by the eigen values.

The factors identified were (1) physical functioning, mild to moderate impairment; (2) physical functioning, severe impairment; (3) social functioning; (4) treatment issues; (5) chest symptoms; (6) body image; (7) interpersonal relationships; (8) career concerns; (9) relationships at work; and (10) concerns for the future. The two physical functioning components were merged in order to create an overall domain of physical functioning which contained mild, moderate, and severe elements. The decision to do this was supported statistically by applying Cronbach alpha analysis to the merged physical functioning domain, the outcome of which indicated very good internal reliability (table 1). A single physical functioning domain is also clinically sensible and easily interpretable. With physical functioning merged into a single domain a total of nine domains of functioning remained.

Internal reliability of each domain was initially assessed using Cronbach alpha coefficients (table 1). Internal reliability is generally acceptable for factors with a Cronbach alpha coefficient of 0.7 or above.13 The exceptions to this were relationships at work, social functioning, and body image. These factors all contained items that were reworded slightly for subsequent versions, thereby eliminating ambiguities and leading to better reliability values. Because Cronbach alpha is influenced by the number of items in a domain, additional analyses using Pearson product moment correlation coefficients were conducted for all domains that contained only two items. Moderate to strong correlation coefficients emerged for treatment issues (r = 0.68), chest symptoms (r = 0.73), body image (r = 0.52), and relationships at work (r = 0.55). Overall, items loaded cohesively onto the emergent factors. However, there were several ambiguities and, in order to maintain the integrity of the domains that the factors represented, 10 items were omitted from the next draft of the questionnaire.

The acceptability and face validity of the questionnaire was confirmed by (a) the patient's willingness to complete it, (b) positive comments regarding its content, (c) indication by the patients that they would be happy to complete and comment on subsequent versions of the measure, (d) failure of the screening item to reveal additional issues, and (e) positive reaction by all members of the multidisciplinary CF team about the content and its relationship to the overall measurement aims. The feedback provided by both patients and all members of the CF team raised an additional 15 items for inclusion in the subsequent CFQoL. This included a new section covering emotional responses to CF. Patient feedback also led to the development of new response scales. For domains related to physical functioning, social functioning, chest symptoms, treatment issues and emotional responses, a six point Likert scale was applied with a range of 1 = all of the time to 6 = never, reflecting the fluctuating nature of CF. A two week time scale was applied across these domains. For the remaining domains a modification of the original scale was applied with 1 = strongly agree to 6 = strongly disagree.

METHODS

Thirty eight CF specific items were retained from the CFQoL1, either in their original form or in the reworded version, and the 15 new items were added to the second version of the questionnaire (CFQoL2). This generated a questionnaire with 53 items across nine domains of functioning.

PATIENTS AND PROCEDURE

The CFQoL2 questionnaire was completed by 223 adults with CF. The sample comprised 103 men and 120 women with a mean age of 25.2 years (range 14–52). Percentage of predicted forced expiratory volume in one second (FEV₁) was also calculated (mean FEV₁ = 55%, range 12–118%). The CFQoL2 questionnaire was distributed to coincide with attendance at the outpatient clinic. Patients self-completed the measure and returned completed questionnaires by post. For purposes of assessing the concurrent validity of the CFQoL2, the Short Form 36 item (SF-36) Health Status questionnaire was distributed in parallel. The SF-36 is a generic health status measure that has been well tested and validated across a number of populations.14-17 It is one of the briefest measures currently in use, but is still one of the most comprehensive.14

RESULTS

Testing the structure of the CFQoL2 questionnaire was conducted by applying PCA with varimax rotation, Cronbach alpha coefficients, and item to domain correlations. The PCA indicated a solution consisting of eight factors on the basis of eigen values of 1.0 and above and the examination of the data from a scree plot. These were (1) chest symptoms, physical functioning (mild to moderate impairment) and social functioning; (2) emotional functioning; (3) interpersonal relationships; (4) concerns for the future; (5) career concerns; (6) treatment issues; (7) physical functioning (severe impairment); and (8) body image.

For those factors with an eigen value below 2.0 (career concerns, treatment issues, and body image) an additional screening measure was used. The percentage of adults with CF who responded negatively across domains provided a measure of “importance to patients”. A negative response to domains is indicated by a raw score of 3 or less and is indicative of the difficulties patients are experiencing; 43% of patients responded negatively with regard to career concerns, 19% regarding treatment issues, and 29% concerning body image, so the retention of these domains was also warranted.

Physical functioning emerged as two factors that were again merged into a single domain. The rationale for this was that physical functioning is more easily interpreted and clinically sensible as a continuum of severity. Additionally, treating mild and moderate physical functioning separately from severe physical functioning may serve to generate floor and ceiling effects across each of these domains. For example, more severely affected individuals are likely to generate floor effects in the mild to moderate range whereas healthier individuals would generate ceiling effects across the severe range of items. Merging of the items serves to buffer this effect while still being sensitive to a range of disability across individuals. Statistical review supported this decision given that the Cronbach alpha coefficients (table 1) and the item to domain correlations (table 2) indicated that the domain was structurally robust. Item to domain correlations test whether items display a stronger correlation with the total domain score from their related domain compared with unrelated domains. To prevent overfitting of the item with related domain scores, the score from the item being considered was removed from the total domain score.

Also loading on to factor 1 were items that had previously emerged as separate factors (chest symptoms and social functioning). Treating these factors as a single domain may serve to confuse interpretability. Amalgamation of these items would lead to a loss of this specificity and the ability to respond in a clinical context. Cronbach alpha values (table 1) and item to domain correlations (table 2) demonstrated the integrity of the chest symptoms and social functioning domains as independent constructs. Because only two items were included in the chest symptoms domain a correlation analysis was conducted, the outcome of which revealed a strong relationship (r = 0.72).

On the basis of the statistical and clinical interpretation of the data, the following nine final domains of functioning were identified: physical functioning, social functioning, treatment issues, chest symptoms, emotional functioning, concerns for the future, interpersonal relationships, body image, and career issues. Both Cronbach alpha testing and item to domain correlation coefficients confirmed that the internal structure of the final nine domains were robust (tables 1 and2).

CONCURRENT VALIDITY

DISCRIMINATION OF THE CFQOL2 QUESTIONNAIRE

METHOD

On the basis of the analyses described in phase 2, further minor amendments were made to the CFQoL resulting in a third and final version of the questionnaire consisting of 52 items across nine domains of functioning (appendix ). The test-retest procedure measured the stability of scores on the CFQoL over time. If the questionnaire is administered at time point 1 and then again 7–10 days later, the scores should remain relatively unchanged for patients who report no change in health status. This is important in determining whether changes measured by a questionnaire are genuine or an artefact of chance fluctuations.

PROCEDURE AND PATIENTS

To test the external reliability of the measure, the third version of the CFQoL questionnaire was posted to 200 adults with CF. A second questionnaire was sent to be completed within 7–10 days of the first. This time period was chosen to be long enough not to introduce memory confounds while being short enough to prevent fluctuations in disease status. Patients were also sent a further questionnaire alongside the second distribution of the CFQoL questionnaire which asked them to what extent they felt their health had changed since the first completion of the measure. Only patients who stated no change in health status and who completed both questionnaires within 7–10 days were included in the analysis.

RESULTS

One hundred patients completed and returned the questionnaire. Of these only 32 (17 men) fitted the inclusion criteria. The mean age of responders was 27 years (men 28 years, women 26 years). The overall age range was 16–53 years (men 18–40, women 16–53). Correlation analysis was applied to assess the degree of relationship between data completed at the two designated time points (table 3). The analysis indicated strong correlations (r >0.8) and explained 72.9–92.2% of overall variance between pairs (r ² × 100). The only exception to this was for treatment issues (r = 0.74), explaining 55.2% of the variance. This indicated that the test-retest reliability for the CFQoL was robust.

METHOD

Sensitivity testing examines whether a measure is able to detect either deterioration in health status or improvements brought about by an intervention. HRQoL using the third and final version of the CFQoL questionnaire was assessed in patients presenting with a pulmonary exacerbation. It was anticipated that improvements in pulmonary function as determined by percentage predicted FEV₁, and possibly small changes in body mass index (BMI), would have occurred by the end of the treatment period.

PATIENTS AND PROCEDURE

Patients presenting at the clinic over a four month period who were diagnosed by the attending physician as presenting with an exacerbation of chest symptoms consented to participate in the study. Data were collected from 24 patients (nine men) of mean age 23 years (range 16–48). Participants completed the CFQoL questionnaire on diagnosis (time point 1) and again at the end of a two week period of intravenous antibiotic treatment (time point 2). Also collected at each time point were FEV₁ and height and weight in order to calculate percentage predicted FEV₁ and BMI.

RESULTS

Each domain was analysed separately by applying a series of related t tests and calculating the effect size of the change between time point 1 and time point 2. The analyses revealed a range of significant differences with large effect sizes for clinical status across the majority of domains (table 4). Higher values were attained at the second time point indicating an improvement in function.