Quality of Life Research (2006) 15: 725737 DOI 10.

1007/s11136-005-3975-4

Springer 2006

Putting Wilson and Cleary to the test: analysis of a HRQOL conceptual model using structural equation modeling
Karen H. Sousa1 & Oi-Man Kwok2 1 College of Nursing, Arizona State University, Tempe, AZ, USA; (E-mail: karen.sousa@asu.edu); 2 Department of Educational Psychology, Texas A & M University, College Station, TX, USA
Accepted in revised form 8 October 2005

Abstract Wilson and Cleary (1995) proposed a conceptual model of health-related quality of life (HRQOL) that integrates both biological and psychological aspects of health outcomes. There are ve dierent levels in their model, namely, physiological factors, symptom status, functional health, general health perceptions, and overall quality of life. Their model has been widely applied to dierent populations, including patients living with cancer, Parkinsons disease, arthritis, and HIV+/AIDS. However, their conceptual model has only been partially examined. That is, the ve major concepts have not been examined simultaneously. Using structural equation modeling (SEM), the Wilson and Cleary HRQOL model was validated in patients living with HIV from the AIDS Time-Oriented Health Outcomes Study. The results showed that the HRQOL model t the data adequately, and the relationships between the constructs were all signicant (at p<0.05 level). Based on the modication indexes, an alternative model linking symptom status directly with general health perceptions and overall quality of life was specied. Implication and limitation of the ndings are discussed. Key words: Functional health, General health perception, HIV, HRQOL conceptual model, Quality of life, Structural equation modeling, Symptom status

Wilson and Cleary [1] presented a conceptual model focused on relationships among aspects of health. Their model linked physiological variables, symptom status, functional health, general health perceptions, and overall quality of life. Wilson and Cleary suggested this health-related quality of life (HRQOL) conceptual model could be used to unify the biomedical and social science paradigms. The biomedical paradigm focuses on etiologic agents; pathological processes; and biological, physiological, and clinical outcomes. It integrates both reductionism, the philosophic view that complex phenomena are ultimately derived from a single primary principle; and mindbody dualism, the doctrine which views a separation of mental

from somatic [2]. The ultimate goal of the biomedical paradigm is to understand causal relationships and to classify patients into groups with specic prognostic or therapeutic meaning. In contrast, the social science paradigm focuses on functioning and overall well-being. It takes into account the patient, the social context in which he or she lives, and the complementary system devised by society to deal with the disruptive eects of illness. The social science paradigm evaluates all the factors contributing to illness; with a primary focus on the way numerous social structures and institutions inuence individuals [2]. The integration of these two perspectives makes the Wilson and Cleary conceptual model a potentially useful model for health care providers.

726 A signicant feature of the Wilson and Cleary model is its theoretical approach. For the most part, the indicators identied in the majority of HRQOL frameworks have little theoretical basis. An atheoretical approach to conceptualizing HRQOL, as a multidimensional construct, has resulted in a list of variables with no hypotheses about the associations between them. The relationship patterns implied consist of multiple rstorder associations with no means of indicating the relationship patterns among the parts. According to Haase and Braden [3] an atheorectical approach fails in several ways: (1) one cannot assess if or how domains are related to one another; (2) there is no way to interpret the meaning of relationship patterns; and (3) there is no basis for specifying whether the dimensions are moderated or mediated by the person, the disease, treatment-related factors, or all three, relative to cost and quality of care outcomes. Measuring HRQOL without reference to a conceptual model has hindered the development of a knowledge base for HRQOL research. A conceptual model places concepts in a context and guides the development of new theories [4]. The use of theoretically based conceptual models will enhance the applicability of the concept as a reliable and valid outcome measure [5]. A valid HRQOL conceptual model will help researchers understand the relationships among the concepts, providers learn about conditions with the greatest impact on patients lives, evaluate the relative importance of dierent approaches to patient care [1], and consequently translate the clinical relevance of HRQOL. The purpose of this analysis was to estimate the primary pathways of the Wilson and Cleary HRQOL conceptual model using structural equation modeling (SEM; Figure 1). Parts of this HRQOL conceptual model have been widely applied to dierent populations, including patients with cancer, Parkinsons disease, heart disease, and HIV/AIDS [617]. While these studies have yielded important results, the ve levels have never been examined simultaneously to determine the overall conceptual soundness of the model. SEM is a comprehensive statistical approach to test hypotheses about relationships among measured and latent variables. It permits the simultaneous estimation of two types of relationships that were of interest. First, a variant of SEM known as conrmatory factor analysis was used to estimate the latent factors that were hypothesized to represent the underlying set of measured items (measurement models). Second, SEM permitted the estimation of relationships between the latent factors (structural model) simultaneously.

The conceptual model The Wilson and Cleary HRQOL conceptual model links physiological variables, symptom status, functional health, general health perceptions, and overall quality of life. The model also links individual and environment characteristics, which were not part of this analysis. The arrows in Figure 1 represent the hypothesized linkages between the dimensions tested in this analysis. It is theorized that physiological variables inuence symptom status, symptom status inuences functional health, functional health inuences general health perceptions, and general health perceptions inuences overall quality of life. In the model, the evaluation of physiological variables centers on cells, organs, and organ systems, while the assessment of symptom status shifts to the organism as a whole [1]. Functional health has been dened as the ability of an individual to perform and adapt to ones environment, measured both objectively and subjectively over a given period [18]. Symptom status and functional health have been included in many HRQOL outcome studies [1927]. General health perceptions represent an integration of all the health concepts previously reviewed, plus others such as mental

Figure 1. The main pathway in the Wilson and Cleary (1995) HRQOL conceptual model.

727 health [1]. They are by denition subjective ratings. Although health perceptions are personal beliefs, overall quality of life has been described as the discrepancy between a persons expectations or hopes and their present experiences [28]. Sample The sample for these analyses comes from data collected as part of the AIDS Time-Oriented Health Outcomes Study (ATHOS). The ATHOS is a longitudinal, observational database of persons with HIV-associated illness cared by communitybased providers. Data were collected from two general sources: medical records and patient questionnaires. Researchers collected these data from three private practices in the San Francisco Bay Area, two private practices in Los Angeles, and ve community clinics in San Diego [29]. The ATHOS database includes eight dimensions: disability, energy, general health, pain, cognitive functioning, mental health, social functioning, and health distress. Several of the scales were adapted from the Medical Outcome Study and were evaluated in patients living with HIV [30, 31]. The reliability and validity of the scales used in the ATHOS database are extensively discussed elsewhere [3237]. The sample included 917 patients with complete responses on the items associated with the four constructs of interest in the HRQOL conceptual model: symptom status, functional health, general
Table 1. Characteristics of the sample Characteristic Age Months since HIV diagnosis Ethnicity White Latino Black Other Education <12 years 12 years 1315 years 16 years Mean 30.35 50.63 Frequency 761 92 34 30 40 110 366 188 213 SD 8.13 28.28 Percentage 83.0 10.0 3.7 3.3 4.4 12.0 39.9 20.5 23.2

health perceptions, and overall quality of life (Figure 1). The descriptive information from the 917 patients is shown in Table 1. Of the 917 patients, 395 also had a CD4 count recorded within 100 days from their rst completed responses on the items of interest. Hence, a two-step procedure was applied to examine the HRQOL conceptual model. First we examined the relations between the four constructs using the total sample of 917 patients. Then the 395 patients were used to examine the full HRQOL conceptual model that included the physiological variable represented by the CD4 count. This strategy was used in part because of the complexity of the model and concerns about the sample size. Measurements Physiological variables The CD4 count, used as a proxy for the physiological variable, focuses on the function of cells. CD4 counts were originally abstracted from outpatient and hospital records [33]. HIV targets the CD4 receptor on the lymphocytes, and the CD4 cell count has been shown to be a predictor of disease progression [38]. The collection of the CD4 count within 100 days of the rst completed questionnaire seemed reasonable given the characteristics of the CD4 count and its chronicity nature. The mean CD4 count in this sample was 329.74 (SD=252.46). Symptom status The ATHOS database contained a checklist of AIDS-specic symptoms. The patients were asked to check any of the symptoms that applied to their health during the previous 7 days. The measurement model for the symptom status has been reported elsewhere (Sousa et al., submitted and was based on the Sign and Symptom Checklist for Persons with HIV disease (SSC-HIV; [39]). It was hypothesized that the relationships among the measured symptoms could be represented by a second-order factor structure, which consisted of one higher order general factor (symptom status) and six lower order factors (malaise/weakness/ fatigue, confusion/distress, fever/chills, gastrointestinal discomfort, shortness of breath, and nausea/vomiting). Composites of the six-order factors were used for this analysis. Malaise/Weakness/Fa-

728 tigue was a composite of muscle pain, muscle weakness, joint pain, fatigue, and dry mouth. Confusion/Distress was a composite of diculty concentrating, depression, trouble remembering, nervousness, and confusion. Fever/Chills was a composite of fever, chills, and night sweats. Gastrointestinal Discomfort included diarrhea, indigestion, abdominal pain, and abdominal cramps. The Shortness of Breath factor was a composite of dyspnea, asthma, and dry cough. The Nausea/ Vomiting factor included nausea, vomiting, and anorexia. Functional status The Health Assessment Questionnaire-Disability Index (HAQ-DI; [40]) was used to measure functional health. The HAQ-DI is a self-reported questionnaire measuring functional health over the previous week by asking a total of 20 questions in 8 dierent categories: activities, reach, grip, eating, dressing, hygiene, walking and arising. Each response was scored on a 4-point scale anchored by without any diculty to unable to do. The highest score in each of the 8 categories were added to form a total score (range 024); this was then divided by 8 to provide a 03 continuous score for each category. The individuals in the sample had a relatively high level of functional health (Table 2), and so it was decided to dichotomize their responses. The scores were, therefore, recoded as 0 (without any diculty) and 1 (with some diculty, with much difculty, and unable to do). Several studies have shown the HAQ-DI to be sensitive to change and valid as a measure of functional health [33, 4143]. General health perceptions General health perceptions were measured by a double-anchored, 100 mm visual analogue scale (from 0 [very poor health] to 100 [very healthy]) and an ordinal scale rating from 1 (excellent) to 5 (poor). These items were recoded to be in the same direction and then standardized to be on the same scale. A higher score was representative of a more positive perception of their general health. Overall quality of life The measurement model for the quality of life dimension was derived from general health status scales [33, 4447]. The development of a quality of life measurement model has been described in Sousa
Table 2. Description data of the HAQ Instrument Category Activities Without any diculty With some diculty With much diculty Unable to do Reach Without any diculty With some diculty With much diculty Unable to do Grip Without any diculty With some diculty With much diculty Unable to do Eating Without any diculty With some diculty With much diculty Unable to do Dressing Without any diculty With some diculty With much diculty Unable to do Hygiene Without any diculty With some diculty With much diculty Unable to do Walking Without any diculty With some diculty With much diculty Unable to do Arising Without any diculty With some diculty With much diculty Unable to do Frequency Percentage

647 184 51 35 777 116 14 10 829 75 10 3 858 47 8 4 806 94 11 6 815 78 12 12 740 138 33 6 780 121 14 2

70.6 20.1 5.6 3.8 84.7 12.6 1.5 1.1 90.4 8.2 1.1 0.3 93.6 5.1 0.9 0.4 87.9 10.3 1.2 0.7 88.9 8.5 1.3 1.3 80.7 15.0 3.6 0.7 85.1 13.2 1.5 0.2

and Chen [48]. Wilson and Cleary [1; p. 62] suggested that overall quality of life should be assessed with general measures of how happy and/or satised one is with their life as a whole. Therefore, composite scores of two subscales (mental health and health worry) from the measurement model tested by Sousa and Chen were used to represent overall quality of life. The mental health questions were designed to measure mental health in terms of

729 psychological distress and well-being [49]. The ve mental health questions were (1) do you feel calm and peaceful? (2) do you feel downhearted and blue? (3) do you feel very happy? (4) do you feel very nervous? and (5) do you feel so down in the dumps that nothing could cheer you up? Health worry is the extent to which health problems cause people to worry or be greatly concerned about their health [50]. The composite questions were (1) are you frustrated about your health? (2) are you afraid because of your health? and (3) is your health a worry in your life? Analysis Structural equation modeling, an inclusive statistical approach for testing hypotheses about relationships among measured and latent variables, was used to estimate the Wilson and Cleary HRQOL conceptual model. The initial step was to develop measurement models to represent symptom status, functional health and overall quality of life. A measurement model was not examined for general health perceptions because it contained only two measured items that had a high correlation of 0.81. Next measurement models (Figure 2) representing associations between symptom status, functional health, general health perceptions, and overall quality of life were estimated. The structural model was then specied by indicating the relationships between each of the latent variables in the model (Figure 3). Finally, the full HRQOL conceptual model including the physiological variable, CD4 count, was examined. All of the models were tested against the data using Mplus (v. 2.01, [51]). To account for the categorical data used to represent
Malaise/Fatigue
1.55 1.0 .86 .54 1.39 .46 .80 .35 .45

Estimator: WLSM Chi-sq (129) = 336.47 p < .0001 RMSEA = .042 CFI = 1.00

Confusion/Distress Fever/Chills Gastrointestinal Pain Shortness of Breath Nausea/Vomiting

Symptom
Status

.41 .46 .36 .43

.86

Activities
1.0 .97 .96 .90 .94 -.86

.04

Reach Grip Eating

.09 .22 .15 .10 .10 .10 .14

Functional Health

.97 .97 .97

Dressing Hygiene Walking

Arising

-.55 -.69 .87 -.33

.95

Health Perceptions
.44 .57

1.0

Current Health Relative Health

.13

.94 .24

1.0

Disease Worry

.27

Overall Quality of Life

.85

Mental Health

.17

Figure 2. Measurement model with unstandardized estimates of the parameters (n=917).

730
Estimator: WLSM Chi-sq (132) = 811.41 p < .0001 RMSEA = .075 CFI = .99
Malaise/Fatigue
1.0 1.54 .85 .55 1.41 .46 .80 .35 .45

Confusion/Distress Fever/Chills Gastrointestinal Pain Shortness of Breath Nausea/Vomiting

Symptom Status

.41 .47 .36 .43

.60

Activities
1.0 .97 .39 .90 .94

.06

Reach Grip Eating Dressing Hygiene Walking Arising

.10 .24 .17 .11 .11 .10 .15

Functional Health

.97 .97 .98 .95

-.77

.26

Health Perceptions
.55

1.0 .95

Current Health Relative Health

.18

.27

.34

1.0

Disease Worry

.26

Overall Quality of Life

.83

Mental Health

.18

Figure 3. Structural model (Model 1) without physiological variables and with unstandardized estimates of the parameters (n=917).

functional health, analyses were conducted using mean-adjusted weight least square estimation (WLSM) [52, 53]. WLSM can estimate models that include both continuous and categorical data. The t of the models was evaluated by examining various statistics. Chi-square statistic was initially used to assess the magnitude of the discrepancy between the sample and tted covariance matrices. A signicant test indicates a poor t. Chi-square tests are sensitive to large sample size, and therefore a small discrepancy may lead to rejection of the model, even though the model may t the data well. Consequently, the following goodness of t indices were used to assess the model t: 1. The Comparative Fit Index (CFI; [54]). CFI ranges from 0 (indicating poor t) to 1.00

(indicating a perfect t) and is derived from the comparison of a restricted model (i.e., one in which structure is imposed on the data) with a null model (i.e., one in which all observed variables are uncorrelated with each other). The CFI provides a measure of complete covariation in the data; a value greater than 0.90 indicates a psychometrically acceptable t to the data. 2. Root Mean Square Error of Approximation (RMSEA; [55]). RMSEA is a measure of discrepancy between the observed and model implied covariance matrices adjusted for degrees of freedom. Browne and Cudeck [55] suggest that values of RMSEA of 0.05 or less indicate close t, and less than 0.08 indicates a fair t.

731 3. Standardized Root Mean Square Residual (SRMR; [56]). SRMR is a measure of the average of the standardized tted residuals. It ranges from 0.00 to 1.00, and a value of less than 0.05 indicates a good t. Mplus does not calculate the SRMR if the WLSM estimator is used for an analysis containing both continuous and categorical measured variables, which was the case in this analysis (Mplus discussion forum retrieved 4/15/05: http://www.statmodel.com/ discussion/messages/23/26/html?1109463167). general health perceptions, 0.81; and overall quality of life, 0.70. The hypothesized measurement model (Figure 2) correlating symptom status, functional health, general health perceptions, and overall quality of life t the data adequately, v2(129)=336.47, p<0.0001; CFI=1.00; RMSEA=0.042). The structural model with the specied relationships between each of the constructs (Figure 3) was examined. The model had reasonable t, v2(132)=811.41, p<0.0001; CFI=0.99; RMSEA =0.075. Compared with the measurement model, the structural model was a more restrictive model with fewer free parameters to estimate and more degrees of freedom. An increase of the Chi-square value in the structural model (i.e., v2 structural model v2 measurement model [DF=3]=474.94) implied some crucial relationships between the constructs were missing. Hence, the structural model was modied in conjunction with the modication index (i.e., the expected change/decrease of the Chi-square value when relaxing a specic parameter) and theoretical considerations. In addition to the magnitude of the Chi-square change, the interpretability of relaxing a specic parameter was also considered [58]. According to the modication index, a substantial change of the expected Chi-square value occurred when the structural path from symptom status to general health perceptions was relaxed (v2 Change [freeing path from symptom status general health perceptions]=432.79 with DF=1). The modied model with an additional path from symptom status to general health perceptions t the data better also (v2(131)=389.70, p<0.0001; CFI=1.00; RMSEA=0.046). Further improvement of the model was acquired by relaxing the structural path from symptom status to overall quality of life (v2 Change [free path from symptom status overall quality of life]=56.15 with DF= 1). The nal modied model, alternative model 1, is shown in Figure 4 with a close t (v2(130)= 40.16, p<0.0001; CFI=1.00; RMSEA=0.042). The t of

Results The corresponding measurement model t statistics of symptom status, functional health, and overall quality of life are presented in Table 3. According to the model t statistics, the measurement models t the data adequately. As previously reported elsewhere, two constructs, symptom status and overall quality of life, had a second-order measurement structure. Continuing to estimate the original second-order measurement model in the hypothesized structural HRQOL model would have lead to a very low ratio of parameters to participants [57]. Consequently, unit-weighted observed composites for each rstorder factor were created (i.e., malaise/weakness/ fatigue, confusion/distress, fever/chills, gastrointestinal discomfort, shortness of breath, and nausea/vomiting for symptom status; mental health and health worry for overall quality of life). The composites were used as indicators for symptom status and overall quality of life in the hypothesized structural HRQOL model. The descriptive statistics and correlations of the composites are presented in Table 4. All composites were signicantly correlated with each other (p <0.001). The ranges of the correlations (n=917) for the composites of each dimension were: symptom status, 0.270.56; functional health, 0.770.94;

Table 3. The t statistics for each measurement model (n=917) v2 Symptom status Functional health Quality of life 493.42 36.99 128.44 DF 225 20 19 CFI 0.99 1.00 0.97 RMSEA 0.036 0.030 0.079 SRMR 0.061 0.025 0.033

732

Table 4. Zero-order correlations, means and standard deviations for all variables in the Measurement Model 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Scale

1. Malaise/Fatigue 2. Confusion/Distress 3. Fever/Chills 4. Gastrointestinal pain 5. Shortness of breath 6. Nausea/Vomiting FH 7. Activities 8. Reach 9. Grip 10. Eating 11. Dressing 12. Hygiene 13. Walking 14. Arising HP 15. Current health 16. Relative health QOL 17. Health worry 18. Mental health Mean Standard deviation (SD) 0.36 0.37 0.39 0.41 0:47 0:44 0:46 0:43 0:42 0:38 0:41 0:45 )0.43 )0.42 )0.40 )0.57 1.23 1.59 0.32 0.41 0.45 0:46 0:42 0:35 0:33 0:39 0:37 0:40 0:39 )0.43 )0.41 )0.28 )0.27 0.51 0.85 0.27 0.45 0:39 0:36 0:28 0:26 0:36 0:32 0:36 0:40 )0.39 )0.37 )0.26 )0.27 0.84 1.06 0.38 0:44 0:43 0:41 0:38 0:41 0:37 0:45 0:42 )0.40 )0.37 )0.24 )0.28 0.46 0.74 0:44 0:41 0:37 0:32 0:45 0:39 0:40 0:40 )0.45 )0.41 )0.27 )0.31 0.53 0.86 0.93 0.88 0.88 0.91 0.94 0.92 0.89 0:70 0:67 0:42 0:40 0.29 0.45 0.80 0.89 0.91 0.92 0.91 0.86 0:65 0:63 0:35 0:35 0.15 0.36 0.89 0.78 0.77 0.83 0.82 0:62 0:56 0:30 0:33 0.10 0.30 0.86 0.86 0.86 0.89 0:62 0:54 0:23 0:25 0.06 0.24 0.93 0.89 0.87 0:70 0:65 0:39 0:38 0.12 0.32 0.90 0.87 0:67 0:63 0:32 0:30 0.11 0.31 0.90 0:69 0:65 0:37 0:38 0.19 0.39

SS

0.56 0.50 0.47 0.50 0.51 0:67 0:59 0:53 0:50 0:57 0:54 0:63 0:60 )0.59 )0.57 )0.38 )0.41 1.30 1.45

0:63 0:59 0.81 0:35 0.51 0.47 0:38 0.48 0.42 0.70 0.15 )0.01 )0.02 3.01 3.36 0.36 1.00 1.00 0.92 0.76

Note: n=917. All correlations are signicant at p <0.001 (two-tailed). SS: Symptom Status; FH: Functional Health; HP: Health Perception; QOL: Overall Quality of Life. The underlined correlations are biserial correlations (i.e., correlation between a dichotomous variable and a continuous variable); the bolded correlations are tetrachoric correlations (i.e., correlation between two dichotomous variables). The means of the functional health items are the proportion of patients who endorsed the items (e.g., mean of activities=0.29, which indicated that 29% of the patients reported to have experienced at least some diculty on their daily activities. The standard deviations of the p functional health items are equal to: p1 p, where p is the proportion of the patients who endorsed a specic functional health item.

733
Malaise/Fatigue
1.0 .53 1.38 .46 .80 .35 .45

Estimator: WLSM Chi-sq (130) = 340.16 p < .0001 RMSEA = .042 CFI = 1.00

1.56

.85

Confusion/Distress
Fever/Chills

Symptom Status

.41 .46 .36 .43

Gastrointestinal Pain Shortness of Breath

Nausea/Vomiting

.55

Activities
1.0 .97 .49 .90 .94

.04 .09 .22 .15 .10 .10 .10 .14

Reach
Grip

Functional Health

Eating Dressing
Hygiene

.97 .97 .97 .95

Walking Arising

-.20

-.33

-.42

.35 .08

Health Perceptions

1.0 .94

Current Health

Relative Health
.35 .26 1.0

.19

Disease Worry

.27

Overall Quality of Life

.85

Mental Health

.17

Figure 4. Alternative model 1 without physiological variables and with unstandardized estimates of the parameters (n=917).

this model is better than model 1. The RMSEA improved to 0.042 indicating a close t as opposed to the reasonable t for model 1. The alternative model 1 (Figure 4) indicated a signicant relationship from symptom status to functional health (!=0.56), such that patients with more symptoms were likely to perceive a decrease in their functional health. The R2 for functional status equaled 0.490, which suggested that symptom status explained 49.0% of the variance in functional health. On the other hand, the signicant relation between symptom status and general health perceptions (! 0:33), and functional health and general health perceptions (! 0:42) reected that patients who experienced more symptoms and less functional health perceived a decrease in their general health. The R2 for general health perceptions (R2=0.625) suggested that

both symptom status and functional health accounted for 62.5% of the variance in general health perceptions. The statistically signicant relationships between symptom status and overall quality of life (! 0:20) and between general health perceptions and overall quality of life (! 0:26) indicated that patients who experienced fewer symptoms and perceived their general health as good were more likely to experience a better overall quality of life. Both symptom status and general health perceptions accounted for 38.2% of the variance in overall quality of life (R2=0.382). The full HRQOL model with the physiological variable (CD4 count) included and the modied relationships between the four constructs was examined (Figure 5). CD4 count was added as an observed variable with a direct path to symptom status. This model also t the data adequately,

734
CD4 Count Estim ator: WLSM Chi-sq (147) = 287.89 p < .0001 RMSEA = .049 CFI = .99
Malaise/Fatigue
-.20 1.07 1.0 .91 .55 1.08 .38 .76 .29 .32

Confusion/Distress Fever/Chills Gastrointestinal Pain Shortness of Breath Nausea/Vomiting

Symptom Status

.37 .41 .33 .40

.64

Activities
1.0 .96 .52 .85 .94

.04

Reach Grip Eating Dressing Hygiene Walking Arising

.12 .30 .16 .14 .10 .10 .28

Functional Health

.95 .97 .97 .86

-.18

-.34

-.30

.23

Health Perceptions
.26

1.0 .95

Current Health

.12

Relative Health
.24

.14

1.0

Disease Worry

.30

Overall Quality of Life

1.0

Mental Health

.13

Figure 5. Alternative model 2 with physiological variables and with unstandardized estimates of the parameters (n=395).

v2(147)=287.89, p<0.0001; CFI=0.99; RMSEA =0.049. The CD4 had a negative relation with symptom status (! 0:20, p<0.05). Patients with higher CD4 levels were more likely to experience fewer symptoms. This supports the hypothesized relationships.

Implications SEM, as a reliable data analytic tool, aords great precision in testing theories and evaluating construct validity. It is important to note, however, that SEM in and of itself cannot be used to imply causation. The use of theory to guide this analysis

was essential, but the validity of the results was also inuenced by the data and how SEM was used [59]. Three conditions are necessary to infer causation. First, there must be an association between the variables. Second, there must be temporalprecedence such that X (independent variable) must precede Y (dependent variable) in time. Finally the relationship between X and Y must be nonspurious. X must contain unique information about Y, with all other causes partialled out [60, 61]. Because this was a cross-sectional analysis, the temporal-precedence condition was violated. Longitudinal analysis combining autoregressive and latent curve models [62] is

735 necessary to further support the causal relationships implied within the Wilson and Cleary [1] HRQOL conceptual model. The analysis presented implies that the relationships depicted in the gures were supported by the data. This is an initial step in the comprehensive testing of the Wilson Cleary conceptual model. A fundamental paradigm shift has been occurring in health care that has necessitated a redenition of health and therefore of points of care. The previous paradigm emphasized disease, focused on medical care, and measured disease primarily on pathophysiologic disturbances. The new view emphasizes health, functioning, and quality of life and focuses on health care. Health care providers continue to struggle with identifying and measuring appropriate patient outcomes that capture quality patient care. This HRQOL conceptual model can be used to identify, measure, and improve quality patient care. It thereby challenges researchers, administrators, and clinicians to be accountable and responsible for the consequences of their actions in response to changing roles and perceptions about what constitutes HRQOL. This paradigm shift has also prompted a change in thinking about the concept of health, with health now encompassing more than the absence of disease [63]. The Wilson and Cleary HRQOL conceptual model reects this paradigm shift. The model implies that symptom status, functional health, general health perceptions, and overall quality of life are dimensions of HRQOL, and the present analysis supports this conceptual model. Wilson and Cleary [1] stated that the absence of arrows between nonadjacent levels does not imply that relationships do not exist. The pathways from symptom status to general health perceptions and overall quality of life are therefore consistent with the intent of their original model. The Wilson and Cleary [1] HRQOL conceptual model provides a theoretical approach to conceptualizing HRQOL as a multidimensional construct. This model places concepts in a context and will be useful to guide the development of new theories. This model, as described and tested, could be used as a tool to assess interventions and organizational performance within the new paradigm. Acknowledgements This project was partially funded by Grant #R01 NR04817 from the National Center for Nursing Research, National Institutes of Health. Thanks to J. F. Fries for sharing data from the ATHOS databank and to Steve West, PhD, for sharing his expertise. References
1. Wilson IB, Cleary PD. Linking clinical variables with health-related quality of life: A conceptual model of patient outcomes. J Am Med Assoc 1995; 273: 5965. 2. Engel GL. The need for a new medical model: A challenge for biomedicine. Science 1977; 196: 129136. 3. Haase JE, Braden CJ. Guidelines for achieving clarity of concepts related to quality of life. In: King CR, Hinds PS (eds.), Quality of Life: From Nursing and Patient Perspectives. Sudbury, MA: Jones & Bartlett, 1998. 4. Fawcett J, Downs FS. The Relationship of Theory and Research. Philadelphia: FA Davis, 1992. 5. Vallerand AH, Breckenridge DM, Hodgson WA. Theories and conceptual models to guide quality of life research. In: King CR, Hinds PS (eds.), Quality of life: From Nursing and Patient Perspectives. Boston: Jones & Bartlett, 1998: 3751. 6. Bennett SJ, Perkins SM, Lane KA, et al. Social support and health-related quality of life in chronic heart failure patients. Qual Life Res 2001; 10(8): 671682. 7. Chrischilles EA, Rubenstein LM, Voelker MD, et al. Linking clinical variables to health-related quality of life in Parkinsons disease. Parkinsonism Relat Disord 2002; 8(3): 199209. 8. Cosby C, Holzemer WL, Bakken-Henry S, et al. Hematological complications and quality of life in hospitalized AIDS patients. AIDS Patient Care STD 2000; 14(5): 269279. 9. Cunningham WE, Crystal S, Bozzette S, et al. The association of health-related quality of life with survival among persons with HIV infection in the United States. J Gen Intern Med 2005; 20(1): 2127. 10. Hays RD, Cunningham WE, Sherbourne CD, et al. Healthrelated quality of life in patients with human immunodeciency virus infection in the United States: Results from the HIV cost and services utilization study. Am J Med 2000; 108(9): 714722. 11. Lee DTF, Yu DSF, Woo J, Thompson DR. Health-related quality of life in patients with congestive heart failure. Eur J Health Fail 2005; 7(3): 419422. 12. Masoudi FA, Rumsfeld JS, Havranek EP, et al. Age, functional capacity, and health-related quality of life in patients with heart failure. J Card Fail 2004; 10(5): 368373. 13. Sousa KH, Holzemer WL, Henry SB, Slaughter R. Dimensions of health-related quality of life in persons living with HIV disease. J Adv Nurs 1999; 29(1): 178187.

736
14. Straits TK, Fields JA, Wilkinson SB, et al. Health-related quality of life in Parkinsons disease after pallidotomy and deep brain stimulation. Brain Cognition 2000; 42(3): 399416. 15. Wettergren L, Bjorkhom M, Axdorph U, Languis-Eklof A. Determinants of health-related quality of life in long-term survivors of Hodgkins lymphoma. Qual Life Res 2004; 13(8): 13691379. 16. Wilson IB, Cleary PD. Clinical predictors of functioning in persons with acquired immunodeciency syndrome. Med Care 1996; 34(6): 610623. 17. Wilson IB, Cleary PD. Clinical predictors of declines in physical functioning in persons with AIDS: Results of a longitudinal study. J Acq Immun Def Synd 1997; 16(5): 343349. 18. Lipkin M. Functional Status Measurement in Primary Care. New York: Springer-Verlag, 1988. 19. Alonso J, Anto JM, Gonzalez M, Fiz JA, Izquierdo J, Morera J. Measurement of general health status of nonoxygen-dependent chronic obstructive pulmonary disease patients. Med Care 1992; 30: MS125MS135. 20. Bombardier C, Raboud J. The Auranon Cooperating Group. A comparison of health-related quality-of-life measures for rheumatoid arthritis research. Control Clin Trials 1991; 12: 243S256S. 21. Cleary PD, Greeneld S, McNeil BJ. Assessing quality of life after surgery. Control Clin Trials 1991; 12: 189S203S. 22. Cleary PD, Fowler FJ, Weissman J, et al. Health-related quality of life in persons with acquired immune deciency syndrome. Med Care 1993; 13: 569580. 23. Gelber RD, Lenderking WR, Cotton DJ, et al. Qualityof-life evaluation in a clinical trial of zidovudine therapy in patients with mildly symptomatic HIV infection. Ann Intern Med 1991; 116: 961966. 24. Laupacis A, Wong C, Churchill D. The use of generic and specic quality-of-life measures in hemodialysis patients treated with erythropoietin. Control Clin Trials 1991; 12: 168S179S. 25. Lenderking WR, Gelber RD, Cotton DJ, et al. Evaluation of the quality of life associated with zidovudine treatment in asymptomatic human immunodeciency virus infection. New Engl J Med 1994; 330: 738743. 26. Nerenz DR, Repasky DP, Whitehouse FW, Kahkonen DM. Ongoing assessment of health status in patients with diabetes mellitus. Med Care 1992; 30: MS112MS124. 27. Wachet T, Piette J, Mor V, Stein M, Fleishman J, Carpenter C. Quality of life in persons with human immunodeciency virus infection: Measurement by the Medical Outcomes Study instrument. Ann Intern Med 1992; 116: 129137. 28. Calman KC. Quality of life in cancer patients A hypothesis. J Med Ethics 1984; 10: 124127. 29. Lubeck DP, Fries JF. Assessment of quality of life in early stage HIV-infected persons: Data from the AIDS TimeOriented Health Outcome Study (ATHOS). Qual Life Res 1997; 6: 494506. 30. Wu AW, Rubin HR, Mathews WC, et al. A health status questionnaire using 30 items from the Medical Outcomes Study: Preliminary validation in persons with early HIV infection. Med Care 1991; 29: 786798. 31. Wu AW, Mathews WC, Brysk LT, et al. Functional status and well-being in a placebo-controlled trial of zidovudine in early symptomatic HIV infection. J AIDS 1993; 6: 452458. 32. Lubeck DP, Fries JF. Changes in quality of life among persons with HIV infection. Qual Life Res 1992; 1: 359366. 33. Lubeck DP, Fries JF. Health status among persons infected with human immunodeciency virus. Med Care 1993; 31: 269276. 34. Lubeck DP, Fries JF. Changes in health status after one year for persons at risk for and with HIV infection. Psychol Health 1994; 9: 7992. 35. Lubeck DP, Shi H, Bloch DA, Fries JF. Clinical correlates of quality of life among males with HIV infection. Qual Life Res 1995; 4(5): 456457. 36. ODell MW, Lubeck DP, Matsuno S. Validity of the Karnofsky performance status in an HIV-infected sample. J AIDS 1995; 10: 350357. 37. ODell MW, Hubert HB, Lubeck DP, ODriscoll P. Physical disability in a cohort of persons with AIDS. AIDS 1996; 10: 667673. 38. Strathdee SA, OShaughnessy MV, Montaner JSG, Schechter MT. A decade of research in the natural history of HIV infection: Part 1. Markers. Clin Invest Med 1996; 19: 111120. 39. Holzemer WL, Henry SB, Nokes KM, et al. Validation of the signs and symptom check-list for persons with HIV disease (SSC-HIV). J Adv Nurs 1999; 30: 10411049. 40. Singh G, Fries JF, Williams CA, Zattarain E, Spitz P, Bolch DA. Toxicity proles of disease modifying antirheumatic drugs in rheumatoid arthritis. J Rheumatol 1991; 18: 188194. 41. Milligan SE, Hom DL, Ballou SP, Persse LJ, Svilar GM, Boulton CJ. An assessment of the Health Assessment Questionnaire functional ability index among women with systemic lupus erythematosus. J Rheumatol 1993; 20: 972976. 42. Ramey DR, Fries JF, Singh G. The health assessment questionnaire 1995 Status and review. In: Spiker B. (ed.), Quality of Life and Pharmacoeconomics in Clinical Trials, 2nd edn. Philadelphia: Lippincott-Raven, 1996: 227237. 43. Wolfe F, Kleinheksel SM, Cathey MA, Hawley DJ, Spitz PW, Fries JF. The clinical value of the Stanford Health Assessment Questionnaire functional disability index in patients with rheumatoid arthritis. J Rheumatol 1988; 15: 14801488. 44. Hays RD, Stewart AL. The structure of self-report health in chronic disease patients. Psychol Assessment. J Consult Clin Psychol 1990; 2: 2230. 45. McHorney CA, Ware JE, Raczek AE. The MOS 36-item short-form health survey (SF-36): II Psychometric and clinical tests of validity in measuring physical and mental health constructs. Med Care 1993; 31: 247263. 46. Stewart AL, Ware JE. Measuring Functioning and Wellbeing: The Medical Outcomes Study Approach. Durham, NC: Duke University, 1992. 47. Ware JE, Davies-Avery A, Brook RH. Conceptualization and Measurement of Health for Adults in the Health Insurance Study: Vol. 6. Analysis of Relationships among Health Status Measures. Santa Monica, CA: RAND, 1980Publication no R-1987/6-HEW.

737
48. Sousa KH, Chen FF. A theoretical approach to measuring quality of life. J Nurs Measure 2001; 10: 4758. 49. Fries JF, Spitz P, Kraines RG, Holman HR. Measurement of patient outcome in arthritis. Arthritis Rheumatism 1980; 23: 137145. 50. Stewart AL, Hays RD, Ware JE. Methods of constructing health measures. In: Stewart AL, Ware JE (eds.), Measuring Functioning and Well-being: The Medical Outcomes Study Approach. Durham NC: Duke University, 1992: 6785. 51. Muthen LK, Muthen BO. Mplus Users Guide. Los Angeles: Muthen & Muthen, 1998. 52. Kaplan D. Structural Equation Modeling: Foundations and Extensions. Thousand Oaks, CA: Sage, 2000. 53. Muthen BO, du Toit SHC, Spisic D. Robust inference using weighted least squares and quadratic estimating equations in latent variable modeling with categorical outcomes. Psychometrika (in press). 54. Bentler PM. Comparative t indices in structural equation models. Psychol Bull 1990; 107: 238246. 55. Browne MW, Cudeck R. Alternative ways of assessing model t. In: Bollen KA, Long JS (eds.), Testing Structural Equation Models. Newbury Park, CA: Sage, 1993: 136162. 56. Hu LT, Bentler PM. Fit indices in covariance structure modeling: Sensitivity to underparameterized model misspecication. Psychol Methods 1998; 3(4): 424453. 57. Bentler PM, Chou C. Practical issues in structural modeling. Sociol Method Res 1987; 16: 78187. 58. Joreskog KG. Testing structural equation models. In: Bollen KA, Long JS (eds.), Testing Structural Equation Models. Newbury Park, CA: Sage, 1993: 294316. 59. Pearl J. Causality: Models, Reasoning, and Inference. Cambridge: Cambridge University Press, 2000, pp. 133 171. 60. Granger CWJ. Developments in the study of cointegrated economic variables. Oxford Bull Econ Stat 1986; 48: 213 228. 61. Granger CWJ, Newbold P. Forecasting Economic Time Series. Orlando, FL: Academic Press, 1986. 62. Curran PJ, Bollen KA. The best of both worlds: Combining autoregressive and latent curve models. In: Collins LM, Sayer AG (eds.), New Methods for the Analysis of Change. Washington, DC: American Psychological Association, 2002: 107135. 63. Greeneld S, Nelson EC. Recent developments and future issues in the use of health status assessment measures in clinical settings. Med Care 1992; 30: MS23MS41. Address for correspondence: Dr K.H. Sousa, College of Nursing, Arizona State University, P.O. Box 872602, Tempe AZ 852872602, USA Phone: +1-480-965-5047; Fax: +1-480-965-6887 E-mail: Karen.Sousa@asu.edu