INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Research suggests there are perceptual differences in quality of life and life satisfaction among persons of different socioeconomic and cultural backgrounds (1, 2). According to epidemiologic data, low-income and minority populations in the United States bear a disproportionate share of the disease burden in asthma, with higher prevalence rates, less regular sources of care, and more hospitalizations and emergency room use, which may adversely affect their health-related quality of life (HRQL) (3). Unfortunately, relatively little is known about the HRQL of these vulnerable populations. In fact, most HRQL instruments are developed and validated in well-educated, clinic-based populations. In order to understand the unique effects of asthma and its treatment and draw accurate conclusions, it is imperative that HRQL be measured accurately and validly. The purpose of this study was to evaluate the internal consistency, reproducibility, and construct and concurrent validity of the Asthma Quality of Life Questionnaire (AQLQ) (8) in a population-based U.S. sample of low-income African American and Caucasian adults with asthma.

The AQLQ is a condition-specific, quality-of-life measure developed by Elizabeth Juniper and colleagues at McMaster University, Hamilton, Ontario, Canada. Based on structured interviews with asthmatic adults, the instrument evaluates four domains of HRQL important in this population: activity limitation (11 items), symptoms (12 items), emotional function (5 items), and environmental exposure (4 items). A list of potential items for the instrument was generated through a review of the literature, discussions with Canadian chest physicians, and interviews with six patients with asthma (9). Those items rated as most important by a sample of 150 asthmatics were selected for inclusion in the questionnaire. Patients are asked to respond to each of the instrument's 32 items on a seven-point Likert scale, with one indicating maximal impairment and seven indicating no impairment. Items in the activity limitations subscale are individualized; activities to be rated are provided by the respondents themselves. Subscale scores are expressed as the average of all items within each domain, with an overall quality of life score derived by calculating the mean across all items in the instrument. Higher scores on the AQLQ indicate better quality of life. Research suggests that a minimal important within-subject change in an AQLQ subscale or overall scale score is 0.5, with changes of 1.0 considered moderate and changes of 2.0 considered large (10). A minimal important change score is the smallest change score patients perceive as beneficial and that warrants a change in management, with consideration given to side effects and cost (10).

The AQLQ has shown evidence of reliability, validity, and responsiveness to change in Canadian adults with asthma (8, 11, 12), and has been culturally adapted, translated, and psychometrically tested in Spain and the Netherlands (13). Internal consistency estimates (Cronbach's ) suggest the instrument is reliable, with values ranging from 0.90 to 0.96 in Rowe and Oxman's study of 52 Canadian patients (12); 0.78 to 0.96 in the study by Sanjuas and coworkers in Spain (14); and 0.61 to 0.90 in the evaluation of 110 Dutch patients with asthma by van der Molen and coworkers (16). Reproducibility results (intraclass correlation coefficients [ICC]) suggest the AQLQ is also stable over time in these populations with 2-wk ICCs ranging from 0.89 to 0.96 (8, 12).

Construct and concurrent validity of the AQLQ has been evaluated in cross-sectional and longitudinal studies examining the relationship between this condition-specific measure and pulmonary function (FEV₁ percentage of predicted; provocative concentration of methacholine producing a 20% reduction in FEV₁ [PC₂₀]; peak expiratory flow rate [PEFR]), symptoms, medication use, and generic HRQL measures (Medical Outcomes Study Short-Form 36 [SF-36], Sickness Impact Profile [SIP]) (8, 11, 12, 15, 16). Correlation of the total score with FEV₁ percentage of predicted has been low to moderate (0.12 to 0.47); higher correlations have been found with symptoms (0.59 to 0.91) and generic HRQL measures (0.58 with the physical domain and 0.21 with the psychological domain of the SIP; 0.19 and 0.47 with rating scale and standard gamble utilities, respectively) (8, 12, 15, 16).

Less is known about the psychometric performance of the measure in the United States (U.S.) population. Responsiveness to change was demonstrated in a clinical trial evaluating the effectiveness of zileuton (17). Internal consistency () and reproducibility (ICC) estimates from one psychometric study in the U.S. ranged from 0.80 to 0.97 and 0.81 to 0.93, respectively (18). The instrument also correlated modestly with FEV₁ percentage of predicted (0.08 to 0.21) and moderately to strongly with the Asthma Disease Severity Scale (0.43 to 0.64) and the Health Utilities Index (0.40 to 0.60) (18). The majority of subjects in this study were well-educated and Caucasian and were undergoing treatment in a university-affiliated asthma clinic, leaving questions as to the instrument's performance in low-income, ethnically diverse populations.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sampling

A population-based sampling strategy was used to recruit low-income adults with asthma who may or may not be receiving regular care for their condition. The sampling frame was an existing database of participants in a multidisease health survey (recruited through random-digit dialing and telephone interview) conducted in Baltimore County, Maryland and Atlanta, Georgia. The database was screened for participants with asthma-like symptoms (i.e., wheezing, shortness of breath, chest tightness, and/or coughing attacks with mild activity during the past 12 mo) and self-identified as African American or Caucasian with less than two-year college or technical education.

Individuals meeting these initial screening criteria were contacted via telephone and asked to participate in a brief interview to determine eligibility for the study, specifically: self-reported medical diagnosis of asthma or symptoms compatible with asthma for a minimum of 6 mo; treatment regimen compatible with asthma (inhaled or oral -agonist, inhaled or oral corticosteroid, or sustained-release theophylline), and a household income consistent with low income as defined by the United States Housing Act, Section 3 (b) (2): 1 person household:  $29,100, 2 people:  $33,300; 3 people:  $37,450; 4 people:  $41,600; 5 people:  $44,950; 6 people:  $48,250; 7 people:  $51,600; 8 people:  $54,900. These values are the income limits used to determine eligibility of applicants for Public Housing Section 8 and other programs in the Baltimore and Atlanta metropolitan areas (19).

Of the 2,109 individuals contacted during initial screening, 571 (27%) refused to participate and 1,312 (62%) failed the clinical screening and/or income criteria and were deemed ineligible. The remaining 226 individuals were scheduled for a clinic visit to confirm the medical diagnosis of asthma and participate in the study. Of these, 41 (18%) did not appear for the clinic visit, and 73 (32%) did not meet study criteria; 120 were enrolled in the study. A careful review of the data uncovered eight additional subjects who met the initial sociodemographic and clinical screening criteria, but whose household income reported at the time of the interview exceeded the study criterion. These individuals were excluded from the analyses, yielding a final sample size of 112.

Measures

Construct and concurrent validity of the AQLQ in this sample were evaluated by examining the relationship between this condition-specific indicator of HRQL and indicators of disease severity, generic HRQL, global quality of life, and health utility. Traditionally, small but statistically significant correlations have been found between indicators of disease severity and HRQL, with stronger correlations between symptoms and HRQL than between pulmonary function and HRQL. If the AQLQ is valid for use with low-income U.S. patients, similar results should be observed in this study. Because a condition-specific HRQL measure is designed to capture elements of HRQL not detected in generic and global measures, correlations among these instruments should be positive and moderate in size.

Disease severity was quantified through pulmonary function and the Asthma Disease Severity Scale (ADSS). FEV₁ percentage of predicted was used as the spirometric indicator of disease severity; Crapo's formula, which includes correction for race, age, and gender, was used to derive predicted values (20).

The ADSS is an adaptation of the Asthma Control Scale developed by Juniper and colleagues (8, 18, 21). The ADSS rating scale is an indicator of disease severity based on a composite of emergent care, spirometry, and symptoms: emergency room visits during the past 6 mo ( 1); hospitalizations during the last 6 mo ( 1); FEV₁ percentage of predicted  70%; chronic cough or chronic phlegm; chronic wheeze; chronic breathlessness; and use of medications for a breathing problem. Each positive item is assigned a score of one. ADSS scores range from 0 to 7, with higher scores indicating more severe asthma.

HRQL was assessed by the Medical Outcomes Study (MOS) Short-Form 36. The MOS SF-36 consists of eight subscales for assessing physical functioning (PF), physical role functioning (RP), bodily pain (BP), general health (GH), vitality (V), social functioning (SF), emotional role functioning (RE), and mental health (MH). The scales are scored from zero (maximum impairment) to 100 (no impairment) using a published scoring algorithm (22). Subscales can be aggregated into two composite indices: A Physical Component Summary (PCS) and Mental Component Summary (MCS) (23). The measure has been used extensively in patients with a variety of physical and mental health problems, and its reliability and validity are well-documented. Results from several studies indicate the SF-36 performs comparably in disadvantaged populations and in African American and Caucasian patients (24). The internal consistency coefficients for participants in the MOS ranged from 0.76 to 0.93 for African Americans and from 0.79 to 0.93 for Caucasians (27). The instrument has been used to evaluate HRQL in patients with asthma (25, 28) and to validate the AQLQ in a Canadian sample (8).

Cantril's Self-Anchoring Ladder of Life Satisfaction was used as a global indicator of life quality. Cantril's Ladder is an easy-to-administer, four-item, subjective measure of an individual's overall assessment of life satisfaction (29). Subjects are asked to evaluate their life at the present time, 1 yr ago, and 1 yr from now on a 10-rung ladder, with the bottom rung (zero) representing the worst possible life and the top rung (10) the best possible life. Subjects' appraisal of life at the present time was used in this study.

Health utility was measured by the Health Utilities Index (30). the HUI is a multi-attribute health utility index comprising six dimensions of life quality: sensory ability (i.e., vision, hearing, speech), mobility, emotional function, self-care, cognitive function, and pain and discomfort. It is administered as a series of 15 dimension-specific questions, each rated on a 4- to 6-point scale. A multi-attribute preference function is used to combine the levels on each dimension into a summary value or utility score, with a range of zero (death) to one (perfect health) (31).

Procedure

The study was approved by the appropriate institutional review board for the protection of human subjects. All patients gave informed consent prior to the onset of the interview. Those who met the initial screening criteria and who agreed to participate in the study were asked to report to the appropriate Innovative Medical Research Incorporated clinical research site (Baltimore and Atlanta) for medical assessment, pulmonary function testing, and interview. Patients were asked to bring any medication they take for asthma to the visit, for recording purposes, and refrain from using bronchodilators (oral or inhaled) during the 12 h preceding the clinic visit.

Participants underwent spirometric evaluation to assess the presence and severity of airflow limitation and the reversibility of obstruction with treatment. All pulmonary function tests (PFT) were conducted by an individual trained and experienced in spirometry, following the American Thoracic Society (ATS) recommendations for lung function testing (20). Tests were performed before and 30 min after the administration of an inhaled -adrenergic agonist. Subjects performed a minimum of three FVC maneuvers, with at least two yielding reproducible values. FEV₁ percentage of predicted was used as an indicator of disease severity. Reversibility was expressed as a percentage of change in FEV₁ pre- to postbronchodilator (15).

After spirometry, subjects underwent a medical evaluation to confirm the presence of asthma. The evaluation included: pulmonary function test results, as described previously, general medical history, concomitant medications, respiratory history (development of the disease, symptoms, pattern of symptoms, precipitating and/or aggravating factors, profile of typical exacerbation, smoking history, current treatment, and related comorbid conditions) and a physical examination (33). Patients did not undergo a methacholine or histamine challenge, nor were they asked to return to the clinic for further medical evaluation. Those whose diagnosis of asthma could not be confirmed under these conditions were reimbursed for their efforts and excused from the remainder of the study. Upon completion of the clinical assessment, subjects participated in the HRQL interview. Interviews were conducted by a trained interviewer in a private room located in the clinic, using structured interview procedures.

To examine the reproducibility of the AQLQ, a convenience sample of 61 subjects returned for a second interview within 12 to 16 d (mean: 14.2 d ± 1.2). Subjects were asked to evaluate the stability of their asthma symptoms and their quality of life during the 2-wk period on a 15-point Global Rating of Change Questionnaire, where the midpoint is zero (no change) and the two tails reflect intervals of improvement (0 to +7) or deterioration (0 to 7) (10). Subjects reporting no change (absolute zero) in their asthma and no change in treatment during the 2-wk interval (n = 38) were included in the reproducibility analyses. All subjects were reimbursed $50.00 per clinic visit.

Analytic Approach

Descriptive statistics of baseline demographics, clinical status, and HRQL were used to characterize the sample. These values were also compared across racial groups using Student t test and chi-square procedures. Each psychometric test was performed on data from the entire sample followed by an exploratory analysis by race. Internal consistency reliability of the instrument was estimated using Cronbach's formula for coefficient alpha with group differences compared using the test statistic W = (1  ₁)/(1  ₂) for independent samples, where W is distributed as the product of two independent central F variables and approximates a single F with (N₁  1) and (N₂  1) degrees of freedom (34). Pearson product-moment correlations were used to estimate the relationship among AQLQ subscales and between each subscale and the AQLQ overall score, with group differences compared using the formula for inferences about _xy  xz in independent samples (35). When notable differences were found, adjustments were made for attenuation (i.e., the effect of measurement error on the observed correlation) using the formula

where r₁₂ refers to the observed correlation between the two variables, r₁₁ and r₂₂ refer to the internal consistency () of each variable, and ^{^}r₁₂ refers to the correlation coefficient corrected for attenuation (36). Within-scale reproducibility (2-wk test-retest reliability) was assessed by examining intraclass correlation coefficients, supplemented by Pearson correlation coefficients and t tests to locate any systematic differences between the two observations.

Construct validity was examined by evaluating the relationship between the AQLQ and disease severity, specifically prebronchodilator FEV₁ percentage of predicted and score on the Asthma Disease Severity Scale (ADSS), using Pearson product-moment correlation coefficients. Concurrent validity was evaluated by correlating the AQLQ with the quality of life measures (SF-36 summary scores, Cantril's Ladderpresent state, and the HUI). To enable comparison of these results with those of previous studies of the AQLQ, Spearman correlation coefficients are also reported. Demographic effects were evaluated by correlating the AQLQ with age. Exploratory analyses were conducted to evaluate the extent to which the Pearson correlation coefficients varied by racial group, using the formula for inferences about _xy  xz discussed earlier. Finally, exploratory analysis of covariance (ANCOVA) procedures were performed to evaluate the effect of racial group, gender, and race by gender interaction on AQLQ subscale and overall scores controlling for disease severity (FEV₁ percentage of predicted). If the interaction effect was not significant (p < 0.05), the analysis was repeated with this term excluded from the model. Because of the relatively small sample size, these analyses must be considered exploratory.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sample

A total of 112 adults, 46 (41%) African Americans (29 women, 17 men) and 66 (59%) Caucasians (54 women, 12 men) participated in the study. Thirty-eight subjects (14 African Americans and 24 Caucasians) were included in the 2-wk reproducibility analyses. Demographic and clinical characteristics of the total sample and by race are provided in Tables 1 and .2All subjects met the low-income criterion. Two-thirds (n = 48; 69%) of those reporting a household income less than $29,100 indicated the income supported two or more people, with 23% (n = 14) indicating it supported four or more. African Americans were more likely to be male (² = 4.98, p < 0.05) and single (² = 9.18, p < 0.01); there were no additional between-group sociodemographic differences. Clinically, African Americans had a significantly lower mean FEV₁ percentage of predicted both pre- (AA: 77%; C: 86%; p < 0.05) and postbronchodilator (AA: 81%; C: 92%; p < 0.01); no racial group differences in duration of breathing problem or pharmacologic treatment were observed.

Most of the participants (n = 92; 82%) reported that they had been previously diagnosed with asthma. Pulmonary function, physician severity rating, and ADSS score indicated subjects who had not reported a previous diagnosis (n = 20) had relatively mild disease. The FEV₁ percentage of predicted of this group was 90% (± 25) compared with a mean value of 81% (± 19) in those diagnosed previously; most of the undiagnosed subjects (75%) had ADSS scores  2 and most (95%) described their breathing problem as mild to moderate. Reported duration of breathing problems was similar in the two groups (18.3 ± 14.2 in the previously diagnosed group versus 16.4 ± 22.0 yr); perception of activity limitation was consistent with expectations, with 67% in the previously diagnosed and 60% in the undiagnosed group reporting at least some limitation. No relationship was found between racial group and diagnostic history.

The majority of subjects comprising the sample (86%, n = 96) described their breathing problem as mild to moderate; one-third (33%) indicated their activity was not at all or slightly affected by their breathing problems; 41% reported some to moderate limitation. Most (76%) reported no change in their asthma during the past 2 wk. Participants rated their overall health as follows: excellent (4%), very good (27%), good (51%), fair (14%), or poor (5%). Forty-six percent of the sample had a comorbid condition, including hypertension (9%), arthritis (10%), and diabetes (5%). There were no significant racial group differences in any of these indicators of health or health perception.

Descriptive statistics for the validation measures in this sample are provided in Table 2. With the exception of the bodily pain subscale in the SF-36, there were no significant differences between racial group on any of these instruments. A unique feature of the AQLQ is the capacity to individualize the activity subscale by having patients select and rate activities most important to them. A list of the seven activities the participants selected as most important to them by racial group together with the activities identified in the original Canadian validation sample (8) are provided in the APPENDIX.

Reliability

Internal consistency reliability levels for the AQLQ in this sample were 0.88 (activity limitation), 0.91 (symptoms), 0.87 (emotional function), 0.67 (environmental exposure) and 0.96 (overall). Alpha levels by racial group for the four subscales and overall score, respectively, were as follows: African American (0.88, 0.94, 0.91, 0.74, 0.96) and Caucasian (0.86, 0.89, 0.80, 0.58, 0.95). Differences in the symptoms and the emotional functioning subscales by race were statistically significant (p < 0.05).

The relationship among AQLQ subscales and between each subscale and the AQLQ overall score is shown in Table 3. Interscale correlation coefficients were lowest for the environmental exposure-emotional function relationship (0.56) and highest for activity limitation-overall score (0.92) and symptoms-overall score (0.92) relationships. Relationships between emotional function and both activity limitation and environmental exposure were substantially lower for the Caucasian group. However, the difference was not significant and became notably smaller when corrected for attenuation. Adjusted estimates for the emotional function and activity limitation correlations were 0.90 and 0.82 for the African American and Caucasian groups, respectively; adjusted estimates for the emotional function and environmental exposure were 0.73 and 0.69, respectively.

Intraclass correlation coefficients estimating the reproducibility of the AQLQ over a 2-wk period, together with means and standard deviations at the two observations, are provided in Table 4. The difference in mean activity limitation, symptoms, and overall scores between the two observations were statistically significant but did not reach the clinically significant criterion of 0.5 proposed by Juniper and colleagues (10).

Validity

The correlation coefficients between the AQLQ and indicators of disease severity for the total sample and by race are provided in Table 5. Spearman correlation coefficients (r_s) between ADSS and AQLQ were as follows: activity limitation, r_s = 0.32 (p < 0.01); symptoms, r_s = 0.30 (p < 0.01); emotional function, r_s = 0.40 (p < 0.01); environmental exposure, r_s = 0.20 (p < 0.01); and overall score, r_s = 0.35 (p < 0.01). Spearman coefficients between pulmonary function and the AQLQ were as follows: activity limitation, r_s = 0.10; symptoms, r_s = 0.14; emotional function, r_s = 0.21 (p < 0.05); environmental exposure, r_s = 0.02; and overall score, r_s = 0.15. 

Correlation coefficients describing the relationship between the AQLQ and generic measures of quality of life (SF-36, Cantril's Ladder, and HUI) for the total sample and by racial group are also provided in Table 5. All of the coefficients were in the expected direction. The strongest correlations were between the AQLQ and the generic HRQL measure. Coefficients for African American and Caucasian groups were not significant different.

Sociodemographic Effects

The relationships between AQLQ scores and age were small (r =  0.08, 0.10, 0.04, 0.09, 0.07; overall score, activity limitation, symptoms, emotional function, and environmental exposure, respectively; p > 0.05). This held true for both racial groups.

Unadjusted means and standard deviations for the AQLQ by race and gender are shown in Figure 1. Interaction effects were found for the AQLQ symptom subscale (F = 10.43, p < 0.01) and overall score (F = 4.01, p < 0.05). A main effect for gender was found in the activity (F = 4.39, p < 0.05) and symptom (F = 4.46, p < 0.05) subscales, with men reporting better HRQL in this domain than women. Effects for race were noted in the activity (F = 7.59, p < 0.01), symptom (F = 11.58, p < 0.01), and environment (F = 4.87, p < 0.05) subscales and the overall score (F = 6.08, p < 0.05), with African Americans reporting poorer health-related quality of life in each area. FEV₁ percentage of predicted was a significant factor in the emotion subscale analysis (F = 7.80, p < 0.01).

View larger version (34K): [in this window] [in a new window]   Figure 1.   AQLQ scores by race and gender. *p < 0.05; **p < 0.01 significance of the effect evaluated in an ANCOVA model controlling for FEV1% predicted.  = race;  = gender; § = race by gender interaction.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Instrument development and validation studies of HRQL outcome measures are generally conducted in relatively homogenous samples of well-educated patients during clinic visits. The reliability and validity of the Asthma Quality of Life Questionnaire has been demonstrated in clinic-based samples in Canada, Europe, and the United States. Little is known about the performance of the measure in vulnerable, socioeconomically disadvantaged populations in the United States. The purpose of this study was to describe the internal consistency, reproducibility, and construct and concurrent validity of the AQLQ in low-income adults with asthma. As one would expect in a population-based sample of disadvantaged adults, not all of the participants had been previously diagnosed with asthma, despite a relatively long history of breathing-related problems and some degree of associated impairment. As intended, this sample was removed from the university or clinically based population. Further research is needed to determine the extent to which the AQLQ is also reliable and valid in clinically based low-income and severely disadvantaged urban, inner city, and rural populations.

In general, the psychometric characteristics of the AQLQ in this sample were consistent with guidelines for a reliable and valid instrument set forth by Nunnally and Bernstein (36). With the exception of environmental exposure, internal consistency reliability levels for all scales fulfilled the recommended criterion of 0.70 (36) for group comparisons. Furthermore, the symptom subscale and overall scale demonstrate sufficient reliability for individual score analysis (Cronbach's  > 0.90) (36). Alpha levels were lower than those reported in Leidy and Coughlin's (18) study of U.S. patients undergoing care in a university-affiliated asthma and allergy center, indicating a higher degree of measurement error in low-income subjects. If instrument reliability is generally lower in disadvantaged groups, as has been observed with the MOS SF-36 (37, 38), studies of HRQL in these subgroups will require larger samples to uncover statistical relationships or detect treatment effects (27).

In the exploratory analyses by race, the internal consistency reliability estimates for the AQLQ in the African American group tended to be higher than the Caucasian group, particularly in the symptoms and emotional functioning subscales. This translates into less measurement error in the former group. The similarity in the between-subscale correlations for African Americans and Caucasians suggests that the internal structure among AQLQ subscales is consistent across the two groups. Future research, with larger sample sizes, should examine the extent to which the measurement model, i.e., the item-subscale and subscale-overall relationships, holds across diverse populations using a confirmatory factor analytic approach.

The ICC reproducibility values observed in this sample exceeded Nunnally's criterion of 0.70 for "modest reliability" (36); two subscales and the overall scale exceeded 0.80 required for basic research. Estimates in this sample were lower than those found in a clinically stable Canadian sample (8) and in a well-educated predominantly Caucasian clinic-based U.S. sample (18). The latter study found significant improvement in activity limitation over 2 wk (M_diff = 0.32 ± 0.14, t = 2.40, p < 0.05). In the present study, significant improvements were observed in the activity limitation and symptom subscales and overall score; the changes were not considered clinically significant under Juniper's criterion of 0.5 (8). The clinically significant improvement and modest ICC in the environmental exposure subscale for the African American group suggests that the potential for measurement instability should be considered when estimating effect sizes and powering studies.

It is important to note that, consistent with accepted practice, stability was defined by global self-report evaluation. The lack of clinical and pulmonary function data at the second observation limited our ability to evaluate the extent to which self-reported appraisal of stability corresponded to clinical stability in these subjects. It is possible that these patients did not perceive a change; symptomatic fluctuations in low-income patients may be the "norm." Under this scenario, the AQLQ may have detected true changes over the 2-wk interval that were not detected through global evaluation. Perhaps the method for determining clinically significant change and/or the currently accepted value of 0.5 may not be appropriate for this population. Further research on HRQL measurement, HRQL variations, perceived and objective stability of disease, and clinically meaningful change in disadvantaged populations will contribute to a more thorough understanding of the perceptual and clinical experiences of this group.

The correlation coefficients describing the relationship between the AQLQ and FEV₁ percentage of predicted, the ADSS, and generic, global, and utility indicators of quality of life provided evidence of construct and concurrent validity of the instrument in this sample. All of the coefficients were in the expected direction and small to moderate in size (39). Values were consistent with those reported previously (8, 15, 16, 18), suggesting the psychometric properties of the AQLQ are stable across populations. Spearman rank-order correlation coefficients between the AQLQ subscales and FEV₁ percentage of predicted ranged from 0.02 to 0.21 in this sample of low-income asthmatics, 0.06 to 0.18 in Juniper and coworkers (8), 0.04 to 0.31 in van der Molen and coworkers (16), and 0.06 to 0.17 in Leidy and Coughlin (18). The Spearman correlations between the AQLQ subscales and the ADSS in the present study were moderate and lower than correlations reported by Leidy and Coughlin (18) and those reported by Juniper and coworkers (8) using a similar measure, the Asthma Control Scale. This may be due to greater measurement error in the present sample. The modest correlations across studies are consistent with results from a recent principal component analysis (40) indicating HRQL data provide unique information beyond that provided by symptoms alone.

The AQLQ correlated with both the physical and mental domains of the generic HRQL measure, the SF-36, supporting the concurrent validity of the instrument in this sample. Juniper and coworkers' study (8) found that the AQLQ correlated substantially with the physical and, more moderately, with the mental component of the Rand General Health Survey. Rowe and Oxman (12) reported the same pattern of relationship with the physical and psychosocial components of the Sickness Impact Profile. As expected, relationships between the AQLQ and the global quality of life and health utility indices were in the predicted direction, but weaker, indicating the AQLQ provides information on the experiences of asthma not captured in broader instruments.

Evidence of the AQLQ's validity was similar for African Americans and Caucasians. As expected, the relationships between spirometric values and the AQLQ were relatively weak in both groups. FEV₁ percentage of predicted was a stronger correlate of emotional function and overall score for African American subjects. Environmental exposure plays a stronger role in patients' global assessment of quality of life (Cantril's Ladder) for Caucasians than for African Americans.

Age was not a confounding factor in AQLQ score, for the sample as a whole or for each racial group. To some extent, this is a function of the sample characteristics, i.e., the constrained variance in age. Few studies have examined HRQL in older adults with asthma; the extent to which age, gender, race, and comorbidity are confounding factors in AQLQ score is an area for further research.

The exploratory ANCOVA analyses performed here suggest there may be a race by gender interaction for the symptom and overall AQLQ scores. That is, the relationship between race and symptom or overall score varies with gender, and vice versa. Race effects were noted for the activity and environmental subscales; African Americans reported poorer HRQL in each of these areas, controlling for any differences in airway obstruction (FEV₁ percentage of predicted). Weiss and coworkers (7) have suggested that social, cultural, and physical environments of minority and socioeconomically disadvantaged groups influence symptom recognition, exposure to irritants, therapy compliance and care-seeking patterns, each of which may contribute to the greater disease burden experienced by these populations. The fact that there were no differences in generic HRQL appraisal (SF-36 scores), with the exception of bodily pain, suggests the AQLQ is tapping perceptual differences that are unique to the asthma experience. Although the results presented here should be considered exploratory, they suggest further study of the HRQL experience by racial group and gender is warranted.

Mean AQLQ values for this sample compared with published means from clinic-based samples suggest that economically disadvantaged adults with asthma experience poorer health-related quality of life. Mean values in this sample were substantially lower than those reported by Leidy and Coughlin (18) (differences as large as 1.0 point). Studies conducted by Juniper and coworkers (11), Boulet and coworkers (41) and van der Molen and coworkers (16) similarly report higher baseline AQLQ scores (4.9 to 5.3, 4.6 to 5.0, and 5.3 to 6.3 for the three studies, respectively) for their samples. Although Apter and coworkers (42) reported slightly lower baseline AQLQ scores (4.0 to 4.3) than the present study, their subjects also had more severe asthma (FEV₁ percentage of predicted < 80%). Clearly, further study of HRQL in socioeconomically diverse samples is needed.

In summary, the results of this study suggest the AQLQ is a useful condition-specific measure of HRQL for low-income asthmatics in the United States. Overall score reliability levels exceeded recommended standards; subscale internal consistency and reproducibility levels suggest investigators should consider the potential for greater measurement error and adjust sample size estimates accordingly. The relationships between the AQLQ and pulmonary function, symptoms, and alternative indicators of HRQL supported the validity of the measure in this sample.