| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the Cardiovascular Research Institute (Dr. Chen), Division of Occupational and Environmental Medicine (Drs. Blanc and Eisner), and Division of Rheumatology (Drs. Katz and Yelin), University of California, San Francisco, San Francisco, CA.
Correspondence to: Hubert Chen, MD, MPH, 350 Parnassus Ave, Suite 609, San Francisco, CA 94143-0924; e-mail: hubert.chen{at}ucsf.edu
Abstract
Background: The Airways Questionnaire 20 (AQ20) is a concise measure of health-related quality of life (HRQL) in obstructive airway disease; however, its original format may underestimate impairment due to the complete cessation of certain activities.
Methods: We revised seven items of the original AQ20 (revised AQ20 [AQ20-R]), adding response options for inability to perform certain activities. We assessed the performance of the AQ20-R among 352 adults with various airway conditions identified through a random telephone sample. Concurrent validity of the AQ20-R was assessed relative to the Short Form-12 (SF-12) physical component summary (PCS), FEV1, and medication use. Predictive validity was assessed relative to health-care utilization among 278 subjects studied longitudinally.
Results: Twenty-one of 352 subjects were unable to perform at least one activity. These subjects demonstrated higher AQ20-R scores (p < 0.001) indicating worse HRQL. Mean (± SD) AQ20-R scores differed significantly (p < 0.001) among subjects with COPD (8.9 ± 5.2), asthma (6.7 ± 5.0), and chronic bronchitis (4.7 ± 4.2). At baseline, the AQ20-R correlated with the SF-12 PCS (r = – 0.55, p < 0.001) and FEV1 (r = – 0.43, p < 0.001), and was associated with the use of respiratory-specific therapies (p 
0.001). In multivariate models, the AQ20-R was an independent predictor of outpatient visits (odds ratio, 2.2; 95% confidence interval, 1.6 to 3.1), emergency department visits (odds ratio, 2.9; 95% confidence interval, 1.9 to 4.6), hospitalization (odds ratio, 2.8; 95% confidence interval, 1.6 to 4.9), and ICU admission (odds ratio, 3.0; 95% confidence interval, 1.2 to 7.3) during the following year.
Conclusions: The AQ20-R is a valid respiratory-specific HRQL measure that accounts for activity cessation among the most impaired and can be used across various airway conditions.
Key Words: airway disease • Airways Questionnaire 20 • asthma • COPD • quality of life • validation
A growing number of health status measures have been developed to assess disease-specific health-related quality of life (HRQL) among persons with respiratory impairment. The majority of these instruments are limited for use in a specific condition, particularly asthma.12 Although more broadly applicable measures do exist, they tend to be lengthy and less adaptable to certain interview situations, especially telephone surveys.3 The Airways Questionnaire 20 (AQ20) is a simple instrument composed of only 20 items intentionally developed for use in populations with various types of airway disease.4
Validity of the AQ20 was first established by Barley et al5 in 90 adults with asthma; AQ20 scores correlated well with other established measures such as the 76-item St. George Respiratory Questionnaire and the 32-item Juniper Asthma Quality of Life Questionnaire. Additionally, the AQ20 has been studied in subjects with COPD (excluding asthma), in whom it correlated with both the St. George Respiratory Questionnaire and the 20-item Chronic Respiratory Disease Questionnaire.67 To date, the AQ20 has not been applied in a population with mixed respiratory conditions, a situation reflective of most clinical case mixes. Furthermore, little is known about its predictive validity, as follow-up in prior studies567 has been limited to 6 months.
To make the AQ20 easy to administer and score, all item responses were originally formatted as "yes," "no," or "not applicable" (with not applicable scored equivalent to no). One potential drawback to this response format is the ambiguity created for persons who avoid activities due to their respiratory condition or are too impaired to perform them all together. Under this scenario, response misclassification may occur in a subset of AQ20 items. For example, one item of the AQ20 poses the following question: "Because of your chest trouble, do you feel breathless maintaining the garden?" If a person no longer gardens due to respiratory limitation, a "no" or "not applicable" response may be elicited so as to yield a lower total score, thereby underestimating that person’s actual impairment. In fact, the AQ20 often demonstrates a skewed distribution toward the milder end of the scale, which could be compounded by such misclassification.689 The possibility of systematic misclassification is further supported by a pilot study10 of 102 adults with asthma, wherein those with the greatest disability were more likely to respond not applicable to key AQ20 activities.
In order to address this potential limitation, we modified seven activity-based items of the original AQ20 to include an "unable" response. We evaluated the characteristics of the modified scale in different respiratory conditions, including asthma, chronic bronchitis, and COPD, assessing its validity relative to other measures of health status, disease severity, lung function, medication use, and future health-care utilization.
Materials and Methods
Overview
We tested our revised AQ20 (AQ20-R) among 352 adults in the United States with obstructive airway disease. We evaluated concurrent validity of the AQ20-R by assessing its cross-sectional relationship with health status, disease severity, lung function, and medication use. Among 278 of these subjects, we evaluated predictive validity of the AQ20-R in relation to health-care utilization ascertained 1 year later. Approval for this study was obtained from the Committee on Human Research at the University of California, San Francisco, and was judged to be in accordance with the Helsinki Declaration of 1975.
Subject Selection
Subjects were drawn from a larger, prospective study11 of adults in the United States with obstructive airway disease. Survey methods have been previously described in detail.11 In brief, the 352 subjects included in this analysis were selected from a sample of 2,113 English- and Spanish-speaking adults, aged 55 to 75 years, originally recruited by random-digit dial telephone interviews conducted in 2001. Approximately one half of the sample (n = 1,001) was randomly identified among residents in the 48 contiguous states. The remainder of the sample (n = 1,112), which included a subset enriched for disease, was recruited from persons residing in geographic "hot spots" with the highest COPD mortality.12
During the initial recruitment interview, subjects responded to four yes/no items assessing whether they had ever received a physician’s diagnosis of COPD, emphysema, asthma, or chronic bronchitis. Subjects were included in our analysis if they reported any one of these four chronic respiratory conditions, and successfully completed a second telephone interview conducted approximately 1 year later (which serves as the baseline for this analysis). Details regarding subject participation at this second interview have been previously reported.13 For 278 subjects, a third telephone interview was completed (1 year after baseline for this analysis, 2 years after initial recruitment) during which longitudinal data on health-care utilization from the preceding year were obtained.
AQ20-R
In our analysis, we assessed HRQL using a modified version of the AQ20. We changed the original AQ20 in two substantive ways. First, we substituted the term chest trouble with breathing problem, which is more appropriate to speakers of American English. The questionnaire was also translated and administered in Spanish to six subjects. Secondly, we modified the response format to account for the complete cessation of certain activities. The original AQ20 allows for three possible responses: yes, no, and not applicable. Yes responses are scored as 1, and both no and not applicable responses are scored as 0, resulting in a possible range from 0 to 20. We modified seven activity-based items (see Appendix) by adding the response option unable. We scored unable as 1 (equivalent to yes); other responses were scored as described. The range of our modified scale remains identical to the original AQ20, with higher scores indicating worse HRQL. Because the AQ20 is inversely scaled, clustering that occurs at the lower end (zero being the best possible score) has been previously referred to as a ceiling effect.68
Short Form-12
The Short Form-12 (SF-12) is a 12-item version of the Short Form-36, a generic measure of health status. The reliability, validity, and responsiveness of the SF-12 has been demonstrated in numerous disease states.14 A physical component summary (PCS) score can be calculated that represents general physical health status. SF-12 PCS scores range from 0 to 100, with a normative population mean of 50. Higher scores represent better health status.
Geriatric Depression Scale
To assess mental health status we used the Geriatric Depression Scale–Short Form (GDS-SF), which is composed of 15 yes/no items designed to assess depressive symptoms over the past week.15 The GDS-SF, originally developed for use among the elderly, has also been validated in younger populations.16 Scores range from 0 to 15, with higher scores reflecting more depressive symptoms.
Disease Severity
We calculated two disease-specific severity scores, one for asthma and another for COPD. The asthma severity score used is a previously validated measure that includes respiratory symptoms, asthma-related medication use, systemic corticosteroid use, and prior health-care utilization.1718 For this study, we modified the scoring algorithm to account for slight differences in the available survey data (adding 1 point for leukotriene modifier use rather than scoring an additional point for high-dose inhaled steroids; substituting history of allergic rhinitis for rhinitis medication use). These modifications preserved the range of the original scale from 0 to 28. The COPD severity score used is a newly developed measure based on the asthma severity score, but also accounts for home oxygen and antibiotic use.19 The COPD severity score ranges from 0 to 32. Higher scores represent greater disease severity on both scales.
Medication Usage and Health-Care Utilization
Medications were assessed at baseline using specific prompts (yes/no) listing drugs by generic and trade names. Health-care utilization was assessed both at baseline (for severity scores) and 1 year later (for predictive modeling). For outpatient visits, we ascertained the number of visits to a medical office, health clinic, or hospital outpatient center specifically for a breathing problem. Number of visits was then dichotomized for analysis using two or more outpatient visits as the cut-off value (defined a priori based on the median value). Respiratory-specific emergency department (ED) visits, hospitalizations, and ICU admissions were each assessed individually (yes/no).
Lung Function
We employed two methods for obtaining lung function measurements. Details have been previously reported.19 Briefly, we abstracted lung function data for 61 of 170 subjects who had undergone pulmonary function testing during the past 5 years. The mean (± SD) time between pulmonary function testing and assessment of HRQL was 17 ± 21 months (median, 10 months; interquartile range, 2 to 30 months). Subjects without lung function data were recontacted and sent peak flow meters by mail. Using this method, we obtained telephone-directed peak expiratory flow rate (PEFR) measurements for an additional 67 patients. The mean time between assessment of HRQL and PEFR measurement was 4 ± 2 months (median, 4 months; interquartile range, 3 to 6 months).
Statistical Analysis
All statistical analyses were performed using software (SAS version 8.1; SAS Institute; Cary, NC). Internal consistency of the AQ20-R was evaluated using Cronbach 
and by examining item-total correlations. To compare subject characteristics, we used t test or Mann-Whitney U test for continuous variables, and 
2 test or 2 test for trend for categorical data.
To compare health status and lung function among the different respiratory conditions, subjects were categorized into mutually exclusive diagnostic groups. The COPD group included subjects with a diagnosis of either COPD (n = 70) or emphysema (n = 101) regardless of other diagnoses. The asthma group included subjects with a diagnosis of asthma (n = 135), excluding those with concomitant COPD or emphysema. The chronic bronchitis group included subjects with a diagnosis of chronic bronchitis alone (n = 82), excluding those with concomitant COPD, emphysema, or asthma. Histograms were generated to compare frequency distributions of the AQ20-R for each diagnostic group. Kruskal-Wallis test was used to compare health status and lung function among these groups. Mann-Whitney U test was then used for further pairwise comparisons. Bonferroni-adjusted p values were applied when determining statistical significance for the five Kruskal-Wallis tests (p < 0.01), and 15 Mann-Whitney pairwise comparisons (p < 0.003) performed.
Concurrent validity of the AQ20-R was assessed in relation to general health status, lung function, and disease severity scores using Spearman rank correlations. We also assessed concurrent validity relative to medication use reported at baseline. Comparison of AQ20-R scores between those with and without use of respiratory-specific medications were made with Mann-Whitney U test. Again, Bonferroni-adjusted p values were applied when interpreting the 11 Spearman rank correlations (p < 0.0045) and the six Mann-Whitney comparisons (p < 0.008) performed.
Predictive validity of the AQ20-R was assessed relative to health-care utilization at longitudinal follow-up. Crude and adjusted odd ratios (ORs) were calculated using logistic regression and expressed per SD change in AQ20-R score. We first evaluated univariate predictor models, one for each health-care utilization variable. We then evaluated multivariable models that included covariates adjusting for respiratory condition and medication use. Because subjects may have reported more than a single diagnosis, we used four separate dichotomous variables, one for each possible respiratory condition: COPD, emphysema, asthma, and chronic bronchitis (on which the previously defined diagnostic categories were based). Similarly, to adjust for medication use, we used six separate dichotomous variables, one for each type of medical therapy reported: daily bronchodilator, home nebulizer, inhaled steroid, oral steroid, antibiotics, and home oxygen.
For comparison, we repeated this same set of analyses substituting SF-12 PCS as the predictor variable. Finally, we evaluated a comprehensive model in which we included both the AQ20-R and SF-12 PCS together, along with covariates for respiratory condition and medication use as delineated above.
Results
Performance of the Modified AQ20 Items
The AQ20-R demonstrated strong internal consistency with a standardized Cronbach = 0.88. As shown in Figure 1
, AQ20-R scores were skewed toward the milder end of the scale indicating better HRQL. This "ceiling effect" was most prominent among the chronic bronchitis group with 11 subjects (13%) scoring zero (the best possible score).
|
As shown in Table 1 , subjects with at least one unable response were more likely to be female (p = 0.05) and possibly nonwhite (p = 0.06), and had poorer physical health status (lower SF-12 PCS; p < 0.001) as well as more depressive symptoms (higher GDS-SF; p = 0.004). AQ20-R scores among the unable responders were considerably higher than those of the remaining subjects indicating worse HRQL (12.4 ± 4.1 vs 6.7 ± 5.0; Mann-Whitney; z = 4.6, p < 0.001). Excluding the modified items from the analysis, subjects who responded unable scored higher on the remaining 13 items than those without any unable responses (7.7 ± 2.8 vs 4.6 ± 3.4; Mann-Whitney; z = 4.0, p < 0.001; data not shown).
|
ranging from 0.86 to 0.88). As shown in Table 2
, AQ20-R scores varied by condition (Kruskal-Wallis; degrees of freedom [df] = 2, p < 0.001). In pairwise testing, each diagnosis group differed significantly from one another (Mann-Whitney; p < 0.003 in all cases), with the COPD group demonstrating the highest mean AQ20-R score (worst HRQL) and the chronic bronchitis group having the lowest. Mean scores of SF-12 PCS and GDS-SF also differed significantly by condition (Kruskal-Wallis; df = 2, p < 0.01 in both cases). In pairwise testing, the COPD group had lower SF-12 PCS scores (poorer health status) than those with either asthma (Mann-Whitney; z = 3.7, p = 0.0002) or chronic bronchitis (Mann-Whitney; z = 3.5, p = 0.0004); however, there was no statistical difference between asthma and chronic bronchitis. In addition, the COPD group had higher GDS-SF scores (greater depression) than those with asthma (Mann-Whitney; z = 3.2, p = 0.001), but there was no difference observed between chronic bronchitis and either COPD or asthma. Similar differences in lung function were found, with FEV1 and PEFR being lowest for the COPD group.
|
|
Association Between AQ20-R Score and Medication Use
Figure 2
shows the mean AQ20-R scores for those who reported use of a given respiratory therapy compared with those who did not. AQ20-R scores were higher for those reporting current use of a daily bronchodilator (9.6 ± 4.9 vs 6.4 ± 5.0), home nebulizer (10.7 ± 5.3 vs 6.4 ± 5.0), inhaled corticosteroid (8.7 ± 5.1 vs 6.6 ± 5.1), or home oxygen (10.3 ± 5.0 vs 6.7 ± 5.0), or use of antibiotics (8.1 ± 5.2 vs 5.9 ± 5.0) or oral corticosteroids (9.0 ± 5.2 vs 6.6 ± 5.1) within the past 12 months (Mann-Whitney; p 0.001 in all cases). The mean difference in scores ranged from 2.1 (inhaled corticosteroid) to 4.3 (home nebulizer).
|
2 (2 > 3.84 for 12 – 11 = df of 1). To assess the effect of a different cut-off value for outpatient visits, we also reanalyzed the data based on a definition of one or more visits compared to no visits. Using this different cut-off value did not substantively effect the results (data not shown).
|
|
Our findings support the utility and validity of the AQ20-R across a range of airway conditions. The AQ20-R thus has the advantage of a disease-specific HRQL measure in that it assesses health effects relevant to the condition of interest, while at the same time not being overly focused on a single diagnostic category. This property of the AQ20-R makes it well suited for use in populations with mixed airway diseases or in groups in which diagnostic homogeneity cannot be ensured.
In this study, we made minor, although substantive, modifications to the AQ20 to accommodate individuals who are no longer able to do certain activities. Our results suggest that the AQ20 could be vulnerable to systematic bias in which the impact of severe disease on quality of life might be underrated; however, this could not be tested directly since the AQ20 was not administered in its original format concurrently.10 In the present analysis, the scores most likely to be affected were those of a small subset of subjects with the worst disease. Because our population-based sample had a mild-to-moderate case mix, little impact was observed in the performance characteristics of the instrument overall. Nonetheless, if the unmodified AQ20 were to be administered in a population with greater disease severity (as might be the case in a clinical trial or pulmonary rehabilitation setting), we anticipate that these differences could be magnified substantially.
We studied a population with an array of conditions associated with airway obstruction. While previous studies have examined the performance characteristics of the AQ20 in homogeneous populations, either asthma (excluding COPD)89 or COPD (excluding asthma),67 none have simultaneously evaluated both types of airway disease for comparison. In addition, we included individuals with a diagnosis of chronic bronchitis alone (excluding asthma or COPD), which has not been previously reported. Incorporating different types of airway disease allowed us to integrate our findings across diverse respiratory conditions, while also making direct pairwise comparisons between the groups. For example, in our study population, those with COPD demonstrated poorer physical health status, greater depression, and worse lung function than those with asthma or chronic bronchitis alone. Importantly, the scores of the AQ20-R paralleled these findings, supporting the discriminative properties of our revised instrument among different airway conditions.
The concurrent validity of the AQ20-R is further supported by its positive association with disease-specific severity scores and with the use of various respiratory-specific therapies, including inhalers, nebulizer treatments, corticosteroids, antibiotics, and home oxygen. Prior studies6789 have focused primarily on comparisons between the AQ20 and other disease-specific scales. The only other analysis to include medication use was the original validation study,5 which showed a weak-to-moderate correlation between the AQ20 and frequency of bronchodilator use. Our results using the AQ20-R expand significantly on these earlier findings.
To evaluate the predictive validity of the AQ20-R, we also analyzed its longitudinal relationship with subsequent health outcomes. The AQ20-R was an independent predictor of future outpatient visits, ED visits, hospitalization, and admission to an ICU in multivariate models accounting for both respiratory condition and medication use. Not only was the AQ20-R a better predictor of health-care utilization than the SF-12 but, as a disease-specific measure of HRQL, it provided additional explanatory power over and above general health status measured by the SF-12. Although similar results have been reported with cross-sectional data,7 this is the first prospective demonstration of this relationship.
We recognize that our study design has certain limitations. We did not separately administer the original AQ20 in conjunction with our revised instrument, an approach that we feel would have been repetitive and tedious for our participants. Therefore, we cannot know with certainty how subjects unable to perform certain activities would have chosen to respond if not provided with the unable option. It is possible that some respondents would have interpreted the question as hypothetical and chosen yes if the unable option was not available.
Overall, the mean values we obtained using the AQ20-R are consistent with values previously reported for COPD (5.9 to 9.9)67 and asthma (3.1 to 8.1)8 using the original AQ20. While only a minority of subjects in this study endorsed the unable response, we feel that our modifications represent an improvement of the original response format, and create less confusion for the respondents as well as investigators. The strong performance characteristics of the AQ20-R, as described above, support our modifications, while reinforcing the validity of the AQ20 as a whole. Nonetheless, replicating these findings in a larger group of subjects more likely to endorse the unable response would be important to gauge the true impact of our modifications.
As with any longitudinal analysis relying on self-reported events, loss to follow-up and recall bias could affect the results. Minor differences were observed in baseline AQ20-R scores between subjects who were unavailable for follow-up and those who were not, but these differences were not statistically significant. To minimize variation in responses, our surveys were conducted using highly structured interviews consistent with those used by the National Health Interview Survey.20 Health service research supports the use of self-reported events, although suggests that older persons may underreport utilization of certain health resources.212223 Such an effect would bias our results toward the null. In light of our positive findings, however, this would have only caused us to underestimate the true effect size.
Overall, these results demonstrate that the AQ20-R is a valid and reliable measure of respiratory-specific quality of life across different airway conditions. Moreover, the AQ20-R addresses potential problems with the response format of the AQ20 as originally conceived. We believe that this instrument could prove extremely useful for gauging HRQL in multiple airway diseases in both clinical and research settings.
Appendix
Activity-Based Items of the AQ20 as Used in the AQ20-R With Modified Response Option
All occurrences of the term chest trouble in the original English version were substituted with breathing problems in our revised questionnaire. Modified response options for items mentioned are as follows: yes = 1, no = 0, unable = 1, not applicable = 0.
Item 3: Because of your breathing problems, do you feel breathless when gardening?
Item 4: Do you worry when going to a friend’s house that there might be something there that will set off an attack of breathing problems?
Item 10: Because of your breathing problems, are there times when you have difficulty getting around the house?
Item 11: Because of your breathing problems, do you suffer from breathlessness when performing activities at work?
Item 12: Do you feel breathless walking upstairs because of your breathing problems?
Item 13: Because of your breathing problems, do you suffer from breathlessness doing housework?
Item 14: Because of your breathing problems, do you go home sooner than others after a night out?
Footnotes
Abbreviations: AQ20 = Airways Questionnaire 20; AQ20-R = revised Airways Questionnaire 20; df = degrees of freedom; ED = emergency department; GDS-SF = Geriatric Depression Scale-Short Form; HRQL = health-related quality of life; PCS = physical component summary; PEFR = peak expiratory flow rate; SF-12 = Short Form-12
Financial support was provided by National Institutes of Health grant R01 HL607438 from the National Heart, Lung, and Blood Institute, and from the Flight Attendants Medical Research Institute CoE2001. Dr. Chen was also supported by National Institutes of Health grant F32 HL077994.
Received for publication May 21, 2005. Accepted for publication November 23, 2005.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
