Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Determinants of Perceived Health in Patients With Left Ventricular Dysfunction: A Structural Modeling Analysis RAYMOND C. ROSEN, PHD, RICHARD J. CONTRADA, PHD, LARRY GORKIN, PHD, AND JOHN B. KOSTIS, MD Objective: Few studies have evaluated the determinants of perceived health in patients with chronic illness. The present study was designed to evaluate the role of biomedical, demographic, and psychosocial influences on global subjective health by means of a structural equation modeling approach. Method: A conceptual model of perceived health was tested in a subsample of patients {N = 146) from the multicenter Studies of Left Ventricular Dysfunction (SOLVD) trial. Domain-specific quality of life constructs (emotional status, social support, and physical functioning), were assessed by means of multiple indicators. These latent (mediating) variables, along with six single-indicator biomedical and background variables, were modeled as predictors of a composite index of perceived health. Results: A satisfactory fit was obtained for the proposed model, with practical fit indices ranging from .89 to .95. High levels of perceived health were associated with low levels of emotional distress and high levels of physical functioning. Social support was positively correlated with physical functioning and negatively associated with emotional distress. Among the background variables, no direct associations were observed between any of the single-indicator variables and perceived health. Several background variables (eg, age, income, walk-test scores) had indirect effects via associations with the latent variables of physical functioning and emotional distress. Conclusions: These findings support the use of a structural modeling approach in assessing determinants of perceived health in patients with congestive heart failure. Further research is needed to evaluate the utility of the model in other patient populations. Key words: perceived health, left ventricular dysfunction. INTRODUCTION Global subjective assessments of health status have been strongly linked to mortality (1-6). Furthermore, when the effects of age, gender, and education are statistically controlled for, perceived health has emerged as a consistently better predictor of long-term mortality than objective measures of physical health (3,7). Whereas previous research has shown modest effects of physical health and psychosocial factors on self-assessment of health status (4, 8] there has been little effort to develop a multivariate model of the determinants of global subjective health in selected patient groups (9). The purpose of this study was to evaluate a path-analytic model of perceived health in patients with left ventricular dysfunction (LVD). From the University of Medicine and Dentistry, Robert Wood Johnson Medical School (R.C.R., J.B.K.), Rutgers-The State University of New Jersey (R.J.C.), and Pfizer Inc. (L.G.). Address reprint requests to: Raymond C. Rosen, PhD, Department of Psychiatry, Robert Wood Johnson Medical School, 675 Hoes Lane, Piscataway, NJ 08854 Received for publication October 17, 1995; revision received April 9, 1996. Psychosomatic Medicine 59:193-200 (1997) 0033-3174/97/5902-0193S03.00/0 Copyright © 1997 by the American Psychosomjtic Society Perceived health and quality of life are significantly affected in patients with congestive heart failure (10-12) Many of these patients have limited physical capacity, restricted social interactions, and dysphoric mood, in addition to a poor overall prognosis. Although significant advances have been made in the treatment of heart failure in recent years (13-16), it is unclear whether patients' perceptions of their health status have been affected accordingly (17). Furthermore, the determinants of subjective health assessments in symptomatic or asymptomatic LVD patients have not been investigated, nor has the predictive value of those assessments been examined. For the present study, we evaluated medical and psychosocial determinants of global subjective health in a subgroup of patients participating in the multicenter Studies of Left Ventricular Dysfunction (SOLVD) Trial. The SOLVD investigation consisted of two, largescale double-blind, placebo-controlled trials designed to evaluate the effects of pharmacological therapy (enalapril) on the mortality and morbidity of patients with left ventricular dysfunction and overt heart failure (treatment trial) or covert heart failure (prevention trial) (18, 19). To assess the impact of treatment on quality of life and other psychosocial factors in congestive heart failure (CHF), a sub-study 193 R. C. ROSEN et al. was conducted at five participating clinical centers (University of Alabama at Birmingham, University of Florida, Brown University, UMDNJ-Robert Wood Johnson Medical School, and West Roxbury Veterans Administration Medical Center). Details of the design and methodology of the sub-study are reported elsewhere (20). In the present study, a structural-modeling approach was utilized to assess physical and psychological determinants of perceived health before randomization in patients participating in the SOLVD trial. Analyses were guided by a conceptual model of perceived health that included a set of relatively distal, background factors, such as physical disease, social-demographic characteristics and cognitive-intellectual functioning. A second set of more proximal determinants of perceived health ratings reflect the person's appraisal of quality of life in various domains, and consists of emotional status, perceived social support, and self-assessment of physical functioning. Each of these domains was represented in the final model. Structural modeling provides simultaneous evaluation of a set of measurement models and structural path coefficients. In a measurement model, latent constructs, such as global subjective health or domain-specific self-assessments, are defined as combinations of two or more measured variables. Structural path coefficients reflect associations between latent constructs or between single-indicator predictors and a latent construct. It was anticipated that cross-sectional examination of this model would facilitate longitudinal assessment of the effects of LVD treatment on global subjective health, and the relationship between perceived health and longterm survival. These latter comparisons will be reported in a separate publication. METHOD Subjects A total of 318 patients were enrolled in the SOLVD quality-oflife sub-study. The major criteria for enrollment in the SOLVD trial were an ejection fraction s 35%, and no myocardial infarction within 30 days before enrollment. Patients over the age of 80 years were excluded, in addition to those with unstable angina pectoris, severe pulmonary disease, or any other disease likely to affect long-term participation in the trial. Patient screening was conducted independently by each of the study centers, and included a 1-week, single-blind lead-in period to assess possible adverse drug reactions. Uniform screening criteria were employed at each center, which resulted in a similar distribution of age, racial characteristics, and gender (20). All patients provided informed consent prior to participation in the sub-study. 194 Measures Measures included in the study were divided into five separate categories: a) background characteristics; b) physical functioning; c) emotional status; d) social support; and e) subjective health. A complete description of the rationale and psychometric properties of the measurement battery has been reported elsewhere (20, 21). For purposes of the present analysis, assessment of background characteristics included three primary demographic variables, age, education, family income; and two biomedical indices, ejection fraction (EF) and the 6-Minute Walk Test (22). The latter test has patients walk as far as possible along a level track for 6 minutes. The distance walked, in meters, is used as an objective measure of functional status (22, 23). Ejection fraction and Walk Test Scores served as proxy variables for prior illness (eg, history of myocardial infarction). Also included among the background variables was verbal intelligence, which was assessed using the Vocabulary Test from the WAIS (24). Background variables were selected for inclusion based upon availability of complete data for each subject, adequate baseline distribution of scores, and reported association with perceived health in previous studies (4, 5, 8, 9). Domain-specific quality of life measures were included to assess emotional status, perceived social support, and physical functioning. Emotional status was assessed by means of the Rand Corporation's Mental Health Inventory (25). This is a 38-item self-report instrument that comprises four primary sub-scales: positive affect, behavioral control, depression, and anxiety. Each of these scales shows a high degree of internal consistency. In addition, factor analysis indicates that correlations among these scales reflect a single, underlying mental health dimension (25, 26). Perceived social support was measured by means of the Rand Social Support Scale (27). This 19-item questionnaire includes sub-scales measuring positive interactions, tangible support, emotional support, informational support, and affection. Internal consistency (Cronbach's a) for the full scale was .97 in a previous report (20). Physical functioning was assessed using two measures. The first consisted of 7 items from the Functional Status Scale (28), which assesses physical and role limitations. The second consisted of three items from the Dyspnea Scale (29). This scale measures breathlessness according to the magnitude and effort of tasks involved. The third major category of measures was global perceived health. Previous research has demonstrated that items requiring subjects to rate the overall quality of their physical health show high levels of internal consistency and unidimensionality (30). Three Likert-scale items were used for assessment of global subjective health in the present study (31): "In general, would you say your health is: very good, good, fair, poor"; "I am somewhat ill: l = definitely true, 5=definitely false"; and "I am as healthy as anybody I know in my condition: 1 = definitely true, 5 = definitely false". The three items were combined into a single factor, reflecting global perceived health. This factor correlated highly with individual item scores (r <85), and was internally consistent. Analytic Strategy The analyses reported below reflect a general approach to structural equation modeling (SEM) that is described by Joreskog and Sorbom (32). An SEM analysis has two major components. The measurement component involves associations between a set of Psychosomatic Medicine 59:193-200 (1997) DETERMINANTS OF PERCEIVED HEALTH measured variables and a smaller number of latent constructs that are the hypothesized determinants of the measured variables. The measurement model also includes estimates of the associations between latent constructs. Although this feature of SEM requires that at least one of the constructs under study is measured using multiple indicators, it is possible to include in the full SEM model both single- and multiple-indicator variables. The second component of an SEM analysis is the structural component, which involves directional effects among constructs. These effects may be direct, or indirect. Indirect effects involve mediating pathways whereby the influence of one predictor on the dependent variable of interest is partially or completely accounted for by another, more proximal determinant of the outcome. Following a strategy outlined by Anderson and Gerbing (33), we examined the core measurement component of the model before the structural component. This involved an evaluation of the hypothesis that responses to the domain-specific quality of life measures reflect three latent constructs, namely, emotional status, social support, and physical functioning. A measurement model was defined for these constructs by permitting each of the relevant test items or scale scores to load on a single factor representing the latent construct it was hypothesized to measure. These loadings were estimated as free parameters, and all other factor loadings were constrained to equal zero. Correlations among the latent constructs were unconstrained and also were estimated as free parameters. After evaluating the measurement model, we examined the full SEM model. Structural path coefficients were estimated to reflect the influence of the six, single-indicator, background variables on each of the three domain-specific quality of life constructs, as well as on the general subjective health construct. In addition, path coefficients were estimated to reflect the influence of each of the three domain-specific quality of life constructs on the general subjective health construct. All analyses were conducted using the LISREL VIII computer program (32). The complete SEM model submitted for analysis is shown in Figure 1. RESULTS Characteristics of the Final Sample Testing of the model was conducted on 146 subjects for whom complete data were available on all variables included in the analysis model. Reasons for missing data included patient non-compliance or unavailability for testing, staff failure to administer specific items or scales, and data entry errors or omissions. While patients included in the trial were no different on the major demographic variables (see Table 1), small, but significant differences were noted on indices of cardiovascular function. In particular, patients included in the trial had slightly Fig. 1. Standardized maximum likelihood parameter estimates for a final structural model of the determinants of perceived health in LVD patients. Pos. = positive; Behav. = behavior; Ctl .= Control; Int. = interaction. Psychosomatic Medicine 59:193-200 (1997) 195 R. C. ROSEN et al. TABLE 1. Demographic & Health Characteristics of Included Versus Excluded Study Patients Variable Age (yrs) Gender (% male) Race (% white) Income (1-5) Education (1-5) EF (%) NYHA(1-4) Walk Test (meters) Included patients (n = 146) Excluded patients (n = 172) p Mean SD Mean 58.4 91.0% 95.8% 3.15 3.21 28.3 1.44 408.7 10.1 60.4 91.2% 91.7% 3.06 3.09 27.0 1.66 345.3 1.0 0.8 5.9 0.5 138.9 SD 9.2 0.7 0.6 6.4 0.7 160.4 NS NS NS NS NS .05 .001 .001 EF = ejection fraction; NYHA = New York Heart Association classification of heart failure (1 = least severe, 4 = most severe); NS = not significant. higher mean ejection fractions (28.3 vs. 27.0] and lower mean scores on the New York Heart Association classification of heart failure (1.44 vs. 1.66). Six-Minute Walk scores were similarly higher in the included compared to excluded subjects (408.7 vs. 345.3). Overall, the sample was predominantly middle-aged (58.4 ± 10.1), male (91.0%), and white (95.8%). The Measurement Model Table 2 presents the correlational matrix for all variables examined in this study. The data array was converted to a standard variance-covariance matrix before structural modeling analyses were undertaken. Tests of normality were performed, and data were found to be relatively normally distributed. Table 3 presents goodness-of-fit indices examined in this study. Model 1 is a measurement model pertaining to the domain-specific quality-of-life measures. Model 1 hypothesized that responses to the 11 domain-specific quality-of-life measures reflect three latent constructs involving emotional distxess, perceived social support, and self-assessed physical functioning. The null hypothesis associated with Model 1 assumed that there is no covariation among the variables measuring domain-specific quality of life. Null models are used as standards against which to evaluate the adequacy of substantive empirical models (32). A significant *2 for Model 1, /(41) = 104.63, p < .0001, indicated that this model is less than perfect in the sense that the covariance matrix derived from the model differs significantly from the observed data. However, the #2 difference test indicated that Model 1 fit the data better than the null model, X2(14) = 1,431.6, p < .0001. Moreover, Model 1 fit the data well in absolute terms as reflected by practical fit indices ranging from .93 to .96 (see Table 3, Model 1). These values are analogous to estimates 196 of the proportion of variance in the observed data that is accounted for by the model and should be .90 or higher in a well-fitting model. The next stage in the analysis was to evaluate the full SEM model that is depicted in Figure 1. The x* test indicated that this model also showed less than perfect fit, x2(131) = 223.9, p < .001. However, the X2 difference test showed that Model 2 fit the data better than its corresponding null model, x2(59) = 1,822.1, p < .0001. Practical fit indices ranged from .89 to .95, indicating a satisfactory degree of fit (see Table 3, Model 2). Parameter estimates derived for the final model are presented in Figure 1, These coefficients represent completely standardized, maximum likelihood estimates of factor loadings and structural path coefficients. Not shown in Figure 1 are structural path coefficients that failed to approach statistical significance at p < .05. Measurement models for the three, domain-specific quality of life constructs generated substantial, statistically significant factor loadings in the expected direction. Strictly speaking, zero-order correlations among these constructs are not part of the SEM model, because they are statistically controlled for in the calculation of structural path coefficients. However, they are made available in the LISREL output, and indicate moderate negative associations between emotional distress and social support (r = — .40), emotional distress and physical functioning (r = —.48), and a positive association between social support and physical functioning (r = .30). The measurement model for global health perceptions also indicated substantial, statistically significant factor loadings in the expected direction. Structural path coefficients representing predictive relationships between domain-specific and global health self-assessments indicated that high perceived health ratings reflected lower levels of emotional distress (|3 = -0.28, p < .0001), and Psychosomatic Medicine 59:193-200 (1997) DETERMINANTS OF PERCEIVED HEALTH I I I i i i i I I I I higher levels of physical functioning (j3 = 0.68, p < .0001). In addition, a slight trend was observed for higher social support to be associated with lower levels of perceived health (j8 = —0.15, p < .05). As shown in Figure 1, a number of significant associations were observed between single-indicator background predictors and domain-specific quality of life constructs. Age showed a modest inverse relationship with emotional distress (/3= -0.21, p < .01). Not surprisingly, the 6-Minute Walk Test was predictive of the level of physical functioning (|3 = 0.50, p < .001) and also showed a modest negative association with emotional distress (/3 = -0.20; p < .02) and positive association with social support (/3 = 0.24; p < .01). There was a nonsignificant trend for educational status to be positively associated with physical functioning (/3 = 0.16, p < .10). Family income showed an inverse relationship with emotional distress Q3 = -0.27, p < .002) and a positive association with physical functioning (j8 = 0.26; p < .05). Neither ejection fraction nor verbal intelligence showed significant associations with any domain-specific quality of life constructs. It is noteworthy that in no case was there a statistically significant structural coefficient reflecting the effects of any of the background variables on perceived health. The only such effect to approach significance showed a direct, positive association between family income and perceived health (j3 = 0.14, p < .10). Inspection of standardized residuals and modification indices provided in the LISREL output indicated that overall goodness of fit of the model depicted in Figure 1 could be improved slightly if parameters were added to estimate correlations between error terms for the social support measures, rather than setting these correlations to zero as was done in the analysis described above. This strategy was not pursued since post hoc model adjustments of this type may capitalize on chance variations. DISCUSSION S S S S ^ - o i S a ; I I H 1 £ f IJ Psychosomatic Medicine 59:193-200 (1997) — S 3 I 3 33 This is the first study of its kind to examine the pattern of associations between global subjective health and a broad array of objective and subjective health indices in patients with left ventricular dysfunction. In the model presented, background factors such as age, left ventricular function, intelligence, and socioeconomic status were viewed as influencing perceived health via the mediating effects of latent subjective variables, including emotional distress, physical functioning, and perceived social 197 R. C. ROSEN et al. TABLE 3. Goodness of Fit Tests Model ModeM Model 2 Goodness of Fita Hypothesized Null Difference 104.63* (41) 233.91* (131) 1,536.24* (55) 2,045.97* (190) 1,431.61* (14) 1,8322.06* (59) .93 .94 .93 .96 .95 " The first three columns contain x* values, with corresponding degrees of freedom given in parentheses. Values in columns 4 through 6 are normed fit indexes (NFI), nonnormed fit indexes (NNFI), and comparative fit indexes. For columns 1 and 2, statistical significance indicates that the model fails to account for the data. For column 3, statistical significance indicates that the hypothesized model provides improvement in fit by comparison with the corresponding null model. The models are described in the text. * p < .0001. support. The latent variables in this model were found to have a high degree of internal consistency, and the overall model provided adequate predictive power in accounting for the effects of both distal and proximal determinants of perceived health. Furthermore, the use of a structural modeling approach permitted simultaneous evaluation of the relationships between different classes of independent or predictor variables, in addition to the association of these variables with perceived health. Overall, these findings provide strong support for the use of structural equation modeling in quality-of-life research, and for the specific value of this approach in assessing determinants of global subjective health. Additionally, the model shows which elements or parameters patients used psychologically in arriving at their rating of global subjective health. Among the latent variables examined in the present study, physical functioning was most strongly related to perceived health. In patients with left ventricular dysfunction, it thus appears that perceived health is closely linked to the ability to perform tasks of daily living and the relative absence of dyspnea. This finding is consistent with results of other studies indicating a positive relationship between these variables and severity of CHF (20, 34, 35). However, it should be noted that most patients in the present study evidenced relatively mild degrees of heart failure (Class I and II) and had only modest levels of physical disability associated with the disease. It is therefore likely that other factors, such as patients' knowledge and expectations regarding long-term effects of CHF on physical functioning, might partially account for the strength of the association with global subjective health. Indeed, this association might have been even stronger had patients with more advanced heart failure been included in the study. In any event, these findings underscore the importance of physical functioning as a major determinant of perceived health in patients with CHF. Emotional distress was negatively associated with 198 global subjective health, although the strength of this relationship was relatively modest (fi = -0.28). Similarly, a weak negative association was noted for the dimension of social support (j3= -0.15). In accounting for the negative effects of emotional distress on subjective health, it should be noted that the majority of patients in the present study (76.8%) had previous myocardial infarctions (MI). Given the growing evidence of depression and other emotional disorders in post-MI patients (36-38), it is not surprising that emotional distress was evident in our sample, and that it contributed significantly to perceived health. On the other hand, social support has also been linked to survival in post-MI patients (39-41) and the lack of positive association with perceived health warrants explanation. It is possible that the measure of social support used in this study did not adequately capture the dimensions relevant to this population, or that social support may affect survival through other means. Few direct effects on perceived health were observed for the independent or background variables, including age, ejection fraction, 6-minute walk, income, education, and verbal intelligence. In fact, the only association to attain even marginal significance was that between income level and perceived health (/3 = 0.14). Additional effects of the independent variables on global subjective health were shown indirectly via association with the latent variables. Thus, age and walk test scores were both negatively associated with emotional distress, whereas income, education, and walk scores all correlated positively with physical functioning. The lack of direct association between age and either perceived health or physical functioning is noteworthy, given evidence from other studies that age contributes negatively to both physical functioning and perceived health (9, 42). It is possible that the limited age range of our patients may have obscured this relationship. On the other hand, a negative association was observed in our study between age and emotional distress (ie, older patients rated themselves as less distressed), Psychosomatic Medicine 59:193-200 (1997) DETERMINANTS OF PERCEIVED HEALTH which is consistent with the findings of an earlier study of psychosocial factors in chronic illness (43]. In contrast to the highly significant association between walk test scores and physical functioning, left ventricular ejection fraction failed to predict any latent or dependent variables in the study. This finding is consistent with results of other studies showing little or no relationship between ejection fraction and measures of functional status or quality of life (20, 44, 45). Several factors might account for the lack of association between this well-established measure of disease severity and functional status, although the most likely explanation is the relatively truncated range of ejection fraction levels in most multicenter clinical trials of heart failure (20). A major limitation of the present study was the homogeneity of the patient sample, which included few minority or female subjects. According to Johnson and Wolinsky (9), the determinants of perceived health in older adults vary according to both gender and race. Therefore, it would be necessary to replicate the current findings in women or minority patients with LVD if the model is to be generalized to these populations. As noted above, individuals with more advanced degrees of heart failure or a greater degree of disability also were not included in the present study. Inclusion of these individuals may have strengthened the relationship between physical functioning and global subjective health. Alternatively, emotional distress may have played a greater role in determining perceived health if more disabled patients were included in the sample. Despite the advantages of the structural modeling approach for theory development and testing, certain weaknesses should also be noted. First, the analyses were conducted using only subjects for whom complete data were available, which represented less than half of the total sample studied. Since the present study was also based upon a secondary analysis of the SOLVD quality-of-life substudy data, weaknesses and omissions in the original data set could not be corrected. The number of subjects included in the analysis may have been less than optimal considering the number of independent variables in the study (46), although other authors have argued for the value of structural modeling analyses with even smaller sample sizes (47, 48). The limitation in sample size may have implications for the reliability of the model, however, which warrants replication in a larger sample. Additionally, the model was developed and tested solely on cross-sectional baseline data, which limits the degree to which the results may reflect causal associations. Thus, for example, global subjective Psychosomatic Medicine 59:193-200 (1997) health estimates may have influenced subjects' perceptions of physical limitations or emotional wellbeing. Finally, the maximum likelihood procedure used to generate parameter estimates makes certain assumptions regarding the structure of the data (eg, multivariate normality), and it is possible that some alternative method would have yielded superior estimates. Several clinical implications are apparent from the findings of the present study. First, it would appear that intervention programs aimed at improving global health perceptions in patients with heart failure should be directed at both increasing physical functioning and relieving emotional distress. In a recent study in our laboratory examining drug and nondrug treatments for CHF (9), we have noted marked improvements in both objective and subjective measures of health in patients receiving a combination of structured exercise training and groupbased cognitive therapy. Older patients with CHF should be targeted for such interventions as much as younger patients, given that physical functioning appears to be an especially important determinant of perceived health in these patients. REFERENCES 1. Mossey JM, Shapiro E: Self-rated health: A predictor of mortality among the elderly. Am J Public Health 72:800808,1982 2. Kaplan G, Barell V, Lusky A: Subjective state of health and survival in elderly adults. J Gerontol 43:Sll4-Sl20,1988 3. Idler EL: Self-assessed health and mortality: A review of studies. In S Maes, H Leventhal, M Johnston (eds), International Review of Health Psychology, New York; Wiley, 1992, 33-54 4. Idler EL, Kasl, SV, Lemke JH: Self-evaluated health and mortality among the elderly in New Haven, Connecticut, and Iowa and Washington Counties, Iowa, 1982-1986. Am J Epidemiol 131:91-103, 1990 5. Wannamethee G, Shaper AG: Self-assessment of health status and mortality in middle-aged British men. Int J Epidemiol 20:239-245, 1991 6. Wolinsky FD, Johnson RJ: Perceived health status and mortality among older men and women. J Gerontol 47:S304S312, 1992 7. Idler EL, Kasl, SV: Health perceptions and survival: Do global evaluations of health status really predict mortality? J Gerontol Soc Sci 46:555-565, 1991 8. Kaplan GA, Camacho T: Perceived health and mortality: A nine-year follow-up of the Human Population Laboratory cohort. Am J Epidemiol 117:292-304, 1983 9. Johnson RJ, Wolinsky FD: Gender, race, and health: The structure of health status among older adults. Gerontologist 34:24-35, 1994 10. Dracup K, Walden JA, Stevenson LW, Brecht M: Quality of life in patients with advanced heart failure. J Heart Lung Transplant 11:273-279, 1992 199 R. C. ROSEN et al. 11. Rogers WJ, Johnstone, DE, Yusuf S, et al: Quality of life among 5,025 patients with left ventricular dysfunction randomized between placebo and enalapril: The Studies of Left Ventricular Dysfunction. J Am Coll Cardiol 23:393-400, 1994 12. Schron E, Gorkin L, Garg R: Evaluating quality of life in congestive heart failure. In GT Kennedy, MH Crawford (eds)., Congestive Heart Failure: Current Clinical Issues, New York, Futura Publishing, 1994, 121-135 13. Hobbs RE, Czerska B: Congestive heart failure: Current and future strategies to decrease mortality. Postgrad Med 96:167172, 1994 14. Kubo SH, Cohn JN: Approach to treatment of the patient with heart failure in 1994. Adv Intern Med 39:485-515, 1994 15. White M, Rouleau JL: Overview of clinical trials in congestive heart failure. Can J Cardiol 9:629-634, 1993 16. Uren NG, Lipkin DP: Exercise training as therapy for chronic heart failure. Br Heart J 67:430-433, 1992 17. Wenger NK, Furberg CD: Cardiovascular disorders. In Spilker B (ed). Quality of Life Assessments in Clinical Trials, New York, Raven Press, 1990, 335-446. 18. The SOLVD Investigators: The Studies of Left Ventricular Dysfunction (SOLVD)-rationale, design and methods: Two trials that evaluate the effects of enalapril in patients with reduced ejection fraction. Am J Cardiol 66:315-322, 1990 19. The SOLVD Investigators: Effect of Enalapril on survival in patients with reduced left ventricular ejection fractions and congestive heart failure. N Engl J Med 325:293-302, 1991 20. Gorkin, L, Schron EB, Brooks MM: Psychosocial predictors of mortality in the Cardiac Arrhythmia Suppression Trial (CAST). Am J Cardiol 71:263-267, 1993 21. Diener E: Subjective well-being. Psychol Bull 95:542-575, 1984 22. Guyatt GH, Sullivan MJ, Thompson, PJ et al: The 6-minute walk: A new measure of exercise capacity in patients with chronic heart failure. Can Med Assoc J 132:919-923, 1985 23. Bittner V, Weiner DH, Yusuf S, et al: Prediction of mortality and morbidity with a six-minute walk test in patients with left ventricular dysfunction. JAMA 270(14):1702-1707, Oct 1993 24. Wechsler D: The WAIS-R Manual. New York, Psychological Corporation, 1981. 25. Veit CT, Ware JE: The structure of psychological distress and well-being in general populations. J Consult Clin Psychol 51:730-742, 1983 26. Hays RD, Stewart AL: The structure of self-reported health in chronic disease patients. J Consult Clin Psychol 2:22-30,1990 27. Sherbourne CD, Stewart AL: The MOS social support survey. Soc Sci Med 32:705-714, 1991 28. Stewart AL, Ware, JE, Brook RH: Advances in the measurement of functional status: Construction of aggregate indexes. Med Care 19:433-438, 1981 29. Mahler DA, Weinberg DH, Wells CK, Feinstein AR: The measurement of dyspnea: Contents, interobserver agreement, and physiological correlates of two new clinical indexes. Chest 85:751-758, 1984 30. Hays RD, Marshall GN, Wang EY, Sherbourne CD: Four-year, cross-lagged associations between physical and mental health in the Medical Outcomes Study. J Consult Clin Psychol 62:441-449, 1994 200 31. Ware JE, Karmos AH: Development and validation of scales to measure perceived health and patient role propensity. Vol.11 of final report. VirginiaTechnical Information Services, NTIS Publication No. PB-288-231, 1976 32. Joreskog KG, Sorbom D: LISREL VIH: Structural equation modeling with the SIMPLIS command language. Chicago: Scientific Software International, 1993 33. Anderson JC, Gerbing DW: Structural equation modeling in practice: A review and recommended two-step approach. Psychol Bull 103:411-423, 1988 34. Feinstein AR, Fisher MB, Pigeon JG: Changes in dyspneafatigue ratings as indicators of quality of life in the treatment of congestive heart failure. Am J Cardiol 64:50-55, 1989 35. Dracup K, Walden JA, Stevenson LW, Brecht M: Quality of life in patients with advanced heart failure. J Heart Lung Transplant 11:273-2791992 36. Frasure-Smith N, Lesperance F, Talajic M: Depression following myocardial infarction: Impact on 6-month survival. JAMA 270:1819-1825, 1993 37. Fielding R: Depression and acute myocardial infarction: A review and reinterpretation. Soc Sci Med 32:1017-1027,1991 38. Schleifer SJ, Macari-Hinson MM, Coyle DA, et al: The nature and course of depression following myocardial infarction. Arch Intern Med 149:1785-1789, 1989 39. Case RB, Moss AJ, Case N, et al: Living alone after myocardial infarction: Impact on prognosis. JAMA 267:520-524, 1992 40. Gorkin L, Schron EB, Brooks MM: Psychosocial predictors of mortality in the Cardiac Arrhythmia Suppression Trial (CAST). Am J Cardiol 71:263-267, 1993 41. Ruberman W, Weinblatt E, Goldberg JD, Chaudhary BS, Psychosocial influences on mortality after myocardial infarction. N Engl J Med 311:552-559, 1984 42. Liang J: Self-reported physical health among aged adults. J Gerontol 41:248-260, 1986 43. Cassileth BR, Lusk EJ, Strouse TB, et al: Psychosocial status in chronic illness: A comparative analysis of six diagnostic groups. N Engl J Med 311:506-511, 1984 44. Franciosa JA, Park M, Levine TB: Lack of correlation between exercise capacity and indexes of resting left ventricular performance in heart failure. Am J Cardiol 47:33-39, 1981 45. Willens HJ, Blevins RD, Wrisley D, et al: The prognostic value of functional capacity in patients with mild to moderate heart failure. Am Heart J 114:377-382, 1987 46. Boomsma A.: On the robustness of LISREL against small sample size and nonormality. Unpublished doctoral dissertation. University of Groningen, Groningen, Sweden. 47. Tanaka JS: "How big is big enough?" Sample size and goodness of fit in structural equation models with latent variables. Child Dev 58:134-146, 1987 48. Gebring DW, Anderson JC: The effects of sampling error and model characteristics on parameter estimation for maximum likelihood confirmatory factor analysis. Multivariate Behav Res 20:255-271, 1985 49. Kostis JB, Rosen RC, Cosgrove NM, et al: Nonpharmacologic therapy improves functional and emotional status in congestive heart failure. Chest 106:1001, 1994 Psychosomatic Medicine 59:193-200 (1997)