Symptom Status Predicts Patient Outcomes in Persons with HIV and Comorbid Liver Disease

Persons living with human immunodeficiency virus (HIV) are living longer; therefore, they are more likely to suffer significant morbidity due to potentially treatable liver diseases. Clinical evidence suggests that the growing number of individuals living with HIV and liver disease may have a poorer health-related quality of life (HRQOL) than persons living with HIV who do not have comorbid liver disease. Thus, this study examined the multiple components of HRQOL by testing Wilson and Cleary's model in a sample of 532 individuals (305 persons with HIV and 227 persons living with HIV and liver disease) using structural equation modeling. The model components include biological/physiological factors (HIV viral load, CD4 counts), symptom status (Beck Depression Inventory II and the Medical Outcomes Study HIV Health Survey (MOS-HIV) mental function), functional status (missed appointments and MOS-HIV physical function), general health perceptions (perceived burden visual analogue scale and MOS-HIV health transition), and overall quality of life (QOL) (Satisfaction with Life Scale and MOS-HIV overall QOL). The Wilson and Cleary model was found to be useful in linking clinical indicators to patient-related outcomes. The findings provide the foundation for development and future testing of targeted biobehavioral nursing interventions to improve HRQOL in persons living with HIV and liver disease.


Introduction
Since the introduction of highly active antiretroviral therapy (HAART), survival of persons living with human immunodeficiency virus (HIV) has dramatically improved.
Five-year survival rate among patients following World Health Organization (WHO) standard therapies reaches approximately 75% [1]. The result is that persons living with HIV are more likely to suffer significant morbidity and mortality from other disorders such as liver disease (LD) and its related complications (anemia, end stage liver disease, lipodystrophy, and hepatocellular carcinoma) than from HIV [2]. Because of the toxic and metabolic effects of antiretroviral medications on the liver and coinfection with LD, the number of persons living with HIV and LD is increasing [3][4][5][6].
Hepatitis B (HBV) and C virus (HCV) infections are prevalent among HIV-infected individuals with different epidemiologic profiles, modes of transmission, natural histories, and treatments [7]. It is estimated that 3 to 6 million people are infected with HIV and chronic HBV worldwide, which is approximately 10% of HIV-positive persons. In the USA, where HBV and HIV are most often acquired by sexual transmission or injection-drug use and HBV prevalence is low, the prevalence of HIV-HBV coinfection in HIVpositive population is generally less than 10%. However, the prevalence is up to 50% among injection-drug users with HIV [8]. Generally, one-third of persons with HIV also have 2 AIDS Research and Treatment chronic HCV infection. In the USA, approximately 300 000 individuals are living with HIV coinfected chronic HCV [9]. HIV and HCV coinfection increases up to 70-90% of hemophilia and 60-80% of injection-drug users who have high risk of blood exposure [10].
In addition, HIV infection accelerates the natural course of HBV and HCV infections, including death, histological fibrosis/cirrhosis, decompensated liver disease, and hepatocellular carcinoma [11][12][13]. More recently, the evidence shows that HIV itself and immunosuppression contributes to the liver injury. Also, antiretroviral therapy (ART) attenuates the progress of HBV and HCV [12]. Furthermore, among certain populations, such as the homeless or incarcerated individuals with HIV, the prevalence of liver disease reaches 69% [14]. Nonalcoholic steatohepatitis (NASH) is also increasing in HIV-positive populations [12]. A recent study showed that fatty acid production increases in HIV and HCV coinfection [15].
Treatment advances have improved survival rates for HIV-infected individuals, although not always with a good quality of life (QOL). Deteriorated liver conditions have shown to have a significant negative effect on persons' healthrelated quality of life (HRQOL) [16]. In addition, HIVpositive individuals with liver disease are likely from socially vulnerable groups, such as injection-drug users, the homeless, and incarcerated people. Therefore, it is important to identify interventions that have the potential to improve QOL in people living with HIV and liver diseases (LD). An overarching goal of Healthy People 2020 is to increase life expectancy and promote QOL of individuals of all ages [17]; the challenge for researchers and practitioners is to first determine what aspects of individuals' HRQOL are affected when they live with multiple comorbid conditions.
One framework for examining HRQOL is the model proposed by Wilson and Cleary (1995) [18]. The primary aim was to test the null hypothesis, wherein, the hypothesized model would hold true with the components (biological/physiological factors, symptom status, functional status, general health perceptions, and overall quality of life) in persons living with HIV without LD and in persons living with HIV and LD. The null hypothesis was that there would be no difference in the model in persons living with HIV without LD and in persons living with HIV and LD. The secondary aim of the study was to test the relationships proposed within Wilson and Cleary's directional model of HRQOL between biological/physiological factors, symptom status, functional status, general health perceptions, and overall quality of life among persons living with HIV without LD and persons living with HIV and LD. The knowledge gained from persons with HIV and LD in this study may provide a way to support people living with this complex illness.

Participants.
HIV-positive individuals who were currently treated with antiretroviral drugs, at least 18 years of age, and had telephone access were included in the study. All participants gave written informed consent.

Study Design and Data
Collection. The parent study was a randomized controlled trial supported by the National Institute of Nursing Research testing interventions to improve medication adherence in persons living with HIV. Participant data was collected between April 1999 and March 2007 from western Pennsylvania and eastern Ohio community hospitals, university-based clinics, and comprehensive HIV care centers. A history of HIV and LD was ascertained by self-reported data from the Comorbidity Questionnaire, Center for Research in Chronic Disorders (CRCD), University of Pittsburgh School of Nursing (1999), and medical record review. The type of LD was confirmed through medical record review by a registered nurse and from selfreport. LD included but was not limited to hepatitis (A, B, C, and other), cirrhosis, steatosis, and hepatocellular carcinoma. The precise nature of LD was not available in 5% of cases.

Wilson and Cleary Health-Related Quality of Life
Framework. HRQOL was defined as the state of perceived health and its effect on the person [18][19][20][21][22][23][24][25]. The Wilson and Cleary model of HRQOL includes health-related and social factors. As a causal model, Wilson and Cleary allows for identification of potential causal factors in the overall HR-QOL paradigm.
Each of the components in the Wilson and Cleary model (biological/physiological factors, symptom status, functional status, general health perceptions, and overall QOL) has been addressed separately in numerous studies. The researchers reviewed those studies that were relevant to HIV and liver disease and to the operationalized Wilson and Cleary model of HRQOL (Figure 1). The definition of each parameter of Wilson and Cleary's model is described below.
Biological/Physiological Factors. Biological/physiological factors are any measurable function of the cells or organs of an individual. This component includes other clinical indicators such as measures of change in the function of the cell, organ, or organ system. In this study, HIV viral load and CD4 counts were used as biological and physiological factors that were assessed. Symptom Status. Symptom status was described theoretically as any psychophysical, emotional, or cognitive state that influences the individual [18]. Often depressive symptoms and mental health are included in this definition. Symptom status was operationally assessed using each subject's Beck Depression Inventory II (BDI-II) score and the Medical Outcomes Study HIV Health Survey (MOS-HIV) mental function summary score [26].
Functional Status. Functional status was defined as one's ability to perform specific tasks such as going to work or making and keeping medical appointments [18]  summary score [26] and missed clinic appointments during the prior 6 months. The MOS-HIV physical function summary score was used to report self-perceived functionality [26]. Missed clinic appointments were quantified using selfreport or chart review data from the medical record review and were categorized as missed or not missed.
General Health Perceptions. General health perceptions were theoretically defined as how individuals perceive their own health, based on the integration of biological/physiologic factors, symptom status, and functional status combined with the effect of the particular disease or organ state on the individual [18]. General health perceptions were measured by the Perception of Illness Visual Analog Scale [27] and the single item MOS-HIV health transition score [26].
Overall Quality of Life. Theoretically, overall quality of life was described as how satisfied individuals are with all aspects of their life [18]. Overall quality of life was measured with the Satisfaction with Life Scale [28] and a single MOS-HIV item assessing overall HRQOL [26].
Characteristics of the Individual. Characteristics of the individual were specific descriptors of the person [18]. In this study, age (measured in years), sex (categorized as male or female), ethnicity/race (categorized as white and nonwhite and further categorized based on self-report), and number of years of education were included. This demographic information was collected using the CRCD Socio-demographic Questionnaire, University of Pittsburgh, School of Nursing (1999) or via medical record review.
Characteristics of the Environment. The characteristics of the environment theoretically incorporated all of the individual's surroundings, including tangible or intangible available resources [18]. An intangible resource that is often associated with health outcomes is social support. Social support was limited to one measure of subjective and one measure of objective social support. The subjective measure of the extent to which individuals feel that they have interpersonal resources available to them was ascertained from the total score of the Interpersonal Support Evaluation List (ISEL) [29]. Income, as a potentially tangible supportive influence from an individual's environment, was measured by annual gross household income.

Data Analysis.
Data were compiled from the baseline data collection and medical record review from the parent study. All persons living with HIV who self-reported a "liver problem" were included in the HIV and LD group regardless of available objective medical record data. All persons living with HIV who had an LD noted in their baseline medical record were also included in the LD group. All others were considered persons living with HIV without LD.
Each variable was examined for its distribution, range, mean, median, mode, and standard deviation. Assessment was done to test for normality specific to the type of variable. For dichotomous variables, frequencies were explored to identify whether cell sizes were relatively equal. Chi-square and the independent samples t-test were used to examine differences between the initial and continuation studies and the two groups: persons living with HIV and LD and persons living with HIV without LD. Correlations between the variables were analyzed for systematic entry into the structured equation model (SEM). SPSS version 15.0 (SPSS Inc., Chicago, IL, USA) was used to further manage and analyze the data.
A multigroup structural equation modeling (SEM) was used to test the fit of the Wilson and Cleary (1995) model of HRQOL in the two groups. A maximum likelihood estimation method used EQS software package version 6.1 (Multivariate Software, Inc., Encino, CA, USA) to perform the statistical analyses [30].
In addition, multiple pathways were added to SEM model to identify whether there were different relationships between characteristics of individuals and environment between and among persons living with HIV without liver disease and persons living with HIV and LD. Model paths  included: paths from each of the four selected demographic characteristics to the four endogenous variables (symptom status, functional status, general health perceptions, and overall QOL) (16 paths); correlations among the four selected demographic characteristics (6 paths) and paths from each of the four endogenous variables to the two characteristics of the environment (household income and social support), along with the correlation between those two characteristics (9 paths). These paths are depicted graphically in Figure 2. Age, gender, ethnicity/race, and number of years of education were included as characteristics of the individual and social support (measured with the mean score on the Inventory Support Evaluation List (ISEL)), and household income (measured as total gross annual household income) as characteristics of the environment.
For determination of adequacy of sample size the observation per parameters (N:q) was calculated. Generally, if more than five observations are present per parameter the sample size for the model is deemed adequate. In this study, the number of observations (sample size) was 532 and the number of the largest parameters included (covariates in the model) was 36. Therefore, N:q for model testing was 14.8, which is greater than five. Thus, the sample size was deemed sufficient for the model [31].

Results
A total of 532 individuals living with HIV (305 with HIV and 227 with HIV and LD) were included in the study. There were no significant differences between the groups with regard to gender, race, employment status, and household income. However, subjects with HIV and LD were significantly older and less educated than the HIV group without LD (Table 1). Participants had a mean CD4 count of 455 cells/mm 3 (range 44-1540 cells/mm 3 ) and 59% of the overall sample had an undetectable HIV viral load ( Table 2). The classifications of types of comorbid LD are noted in Table 3. All others without evidence of LD were classified as HIV.
The baseline hypothesized Wilson and Cleary model was assessed with SEM and found to hold true in both groups, persons living with HIV without LD and in persons living with HIV and LD, when a Lagrange multiplier modification was applied. The baseline model parameters and subsequent model modifications are noted in Table 4. Model 1 parameters were found after releasing model constraints and allowing additional pathways from symptom status to general health perceptions and overall quality of life, and from biological/physiological factors to general health perceptions. Model 2 allowed for correlations between the biological/ physiological factors (CD4 count and HIV viral load) and       There was a significant direct effect of symptom status on functional status in both groups. However, categorized missed appointments ("yes" or "no") was not related to the study measures and was therefore not included in the SEM/path analysis. The MOS-HIV mental summary score, as a measure of symptom status, significantly predicted functional status, as measured with the MOS-HIV physical summary score, in both the HIV (B = .669, z = 11.52, P < .001) and HIV and LD groups (B = .644, z = 11.52, P < .001).
The BDI-II, as a measure of symptom status, did not predict functional status, as measured by the MOS-HIV physical summary score.
There were four significant additional paths identified by using the Lagrange multiplier linking symptom status, as measured by the BDI-II and the MOS-HIV mental summary score, to distal components of the Wilson and Cleary model of HRQOL. Three of the identified paths linked symptom status directly to general health perceptions and one linked symptom status directly to overall QOL. There was no significant direct effect of functional status, as measured with the MOS-HIV physical summary score, on general health perceptions, as measured with either the perception of illness visual analogue score or the MOS-HIV health transition score, in persons with HIV without liver disease or in persons living with HIV and LD.
There was no significant direct effect of general health perceptions, as measured by either the perception of illness visual analogue score or the MOS-HIV health transition score, on overall QOL, as measured by the satisfaction with life scale, in either group. The significant retained SEM modeled pathways are depicted in Figure 3.
Of the 36 covariance pathways that were added to the good-fitting model, 11 individual parameters were found to be significant. There were four significant pathways found
Other possible factors influencing HRQOL, such as characteristics of the individual and of the environment, were investigated. Additionally, 36 parameters were assessed simultaneously. Social support was associated with both general health perceptions and overall QOL. Social support was also significantly related to mental symptom status. Total gross annual household income was related to MOS-HIV mental and physical summary scores. Thus, characteristics of the environment, as others have shown, have relevance to the components of HRQOL [43][44][45].
There are several limitations in this study. First, a longitudinal study would be necessary with data gathered from different time points to assess the causal relationships. The data used in this study were cross-sectional and therefore the causal relationships were not able to be assessed. Second, the higher proportion of non-white participants and lower income levels, or specific populations such as homeless and injection-drug users, may aid in making this study more generalizable as these differences are more reflective of the overall U.S. population of persons living with HIV. In addition, not having access to a telephone may have further biased the sample recruited.

Conclusions
The findings of this research show that both health-related factors, such as CD4 count and HIV viral load, and social factors, such as self-reported mental health and depressive symptoms, are important indicators of HRQOL in this sample of persons living with HIV (N = 532). These findings imply that symptom status, specifically depressive symptoms and altered mental function, is a key issue in determining HRQOL in persons with HIV without LD and in persons with HIV and LD.
There is no cure for HIV. Therefore, persons living with HIV with or without LD are no longer just trying to survive day to day; rather, they are seeing the future of living with HIV as a chronic disease with debilitating longterm consequences. Direction for a future study needs to include better measures of biological/physiological factors that specifically assess current disease status. Specifically, work needs to focus on a specific liver disease that may significantly impact persons living with HIV such as the hepatitis C virus or on specific symptoms that affect the individual most, such as depressive symptoms.
Depressive and mental symptoms had the strongest relationship to the other model measures suggesting that a focus for clinical intervention would be to more closely address these issues in persons living with HIV and LD. These findings also suggest that a more complete understanding of the symptom experience in persons living with HIV and LD is fundamental to achieve optimal patient outcomes. It may be that specific symptoms, such as depressive symptoms, need to be controlled for model testing. Future research should consider a symptom-specific tool to look for clusters of symptoms in both HIV and liver disease. This study also found that race, social support, and income were important covariates. Therefore, another direction for future research would be to assess differences between racial and socioeconomic status on HRQOL in persons with HIV and LD.

Ethical Approval
The study was approved by the Institutional Review Board of the University of Pittsburgh and the National Institutes of Health Office of Human Subjects Research.