A systematic review of the application of Wilson and Cleary health-related quality of life model in chronic diseases

Background A conceptual model approach to clarify the elements of health-related quality of life (HRQL), their determinants and causal pathways is needed to aid researchers, health practitioners and policy makers in their bid to improve HRQL outcomes in patients. The aim of this systematic review was to appraise empirical evidence on the performance of the Wilson and Cleary Model of HRQL. Methods We conducted a search of MEDLINE, Science Direct, PsyARTICLES and CINAHL databases to identify articles that used Wilson and Cleary model to examine HRQL in chronic diseases. A narrative synthesis was employed in the review of the articles. Results Evidence supports linkages between adjacent concepts and between non-adjacent concepts of the Wilson and Cleary model indicating that in practice there is a need to examine relationships among constructs - or to consider interventions in terms of - both with and without mediators. Symptoms status has the highest magnitude of relative impact on health-related quality of life. Conclusion The Wilson and Cleary model demonstrated good features suitable for evaluating health-related quality of life in chronic diseases.


Background
Health-related quality of life (HRQL) is an important outcome measure in clinical trials and health research. The concept includes measures of patient and social perspectives regarding the impact of illness in order to improve treatment efficacy, safety and shared decisionmaking [1][2][3]. Examining the construct of HRQL has become important because it focuses on components of well-being, which are affected by progressive changes in health status, health care and social support [4].
Patients living with chronic illnesses such as sickle cell disease, HIV/AIDs, chronic obstructive pulmonary disease, arthritis and obesity have been found to have impaired HRQL [5]. HRQL has thus become an important construct in the evaluation of the effect of a disease and its management. However, there has been lack of agreement on the definitions and dimensions of HRQL which could guide standard measurements and research that could lead to a more detailed understanding of the concept. Nonetheless, there are three identified areas of consensus, that first, HRQL is a multidimensional construct encompassing symptoms of diseases, treatment side effects, general perception of health status and life satisfaction [6]; second, the assessment of HRQL is subjective based on self-report termed patientreported outcomes (PROs) [3,7] and third, research on HRQL should be based on conceptual models [8,9]. Such models would enhance the understanding of the relationships and linkages among dimensions of HRQL which in turn could facilitate the design of protocols for optimal care.
Wilson and Cleary [10] have proposed a conceptual model to integrate clinical and psychosocial approaches to health care. Their model links the biological and physiological (objective health) variables to the measure of HRQL or subjective health constructs. This link was made to move research on HRQL from the traditional descriptive methods to models, so that causal relationships among the components could be investigated and clarified. Knowing the proximate causes of HRQL in a disease population would help target rather than just monitor the improvement of HRQL in clinical trials.
The Wilson and Cleary model is the most widely cited conceptual framework of HRQL [8,9,11]. In a systematic review of health-related quality of life models, Bakas and colleagues [8] showed that the Wilson and Cleary model was unique to HRQL, adequate, clear and consistent and could be applied to all individuals irrespective of age, health and disease conditions as well as culture. They further showed that the Wilson and Cleary model could generate hypotheses to provide clinicians with a broader view of HRQL beyond just biological factors and symptoms, and that the model 'makes sense' for real world application. The model focusses on relationships among different domains of health by proposing a linear sequence of causal links along a causal pathway which begins with the bio-physiological level moving along the causal pathway outward to the subjective level and the interaction of the individual as a social being.
The Wilson and Cleary model presents a taxonomy of patient outcomes categorised into five underlying health concepts and proposes specific causal links between these health concepts. Their underlying assumptions are that understanding relationships among these concepts will inform the design of optimally effective clinical interventions ( [9] The five health concepts described in the model are biological and physiological factors, symptoms status, functioning, general health perceptions and overall quality of life.
The biological and physiological factors focus on the functioning of cells, organs and organ systems. The clinical factors include factors that generally affected health but are mediated by changes in cells, organs or organ systems functions. The next point on the continuum is symptoms status which has been described as a patient's perception of an abnormal physical, emotional or cognitive state [10]. The complexity of relationship between biological and physiological factors and symptoms is underlined by the fact that some physiological abnormality may not immediately produce symptoms while some symptoms such as depression may not be clinically traceable to physiologic abnormality [10]. Following symptoms is functional status which is reflected in the ability of the individual to perform specific tasks such as climbing the stairs. The next link is the general health perceptions, a subjective rating that integrates all the previously mentioned health concepts and others such as mental health and is followed by the overall or global health-related quality of life at the end of the continuum [10]. Arrows in the model depict dominant causal associations. Reciprocal relationships are implied but not shown. The possibility of bidirectional relationship has also been suggested [10] but not indicated.
Empirical evidence from studies that have used the Wilson and Cleary model is needed to establish patterns of relationship and their consistency. Integrating results of empirical studies onto the model will reveal the features and performance of the model and enhance our understanding of patterns of relationships and effects of mediators thereby increasing the information available to health researchers and practitioners. Furthermore, understanding the relative importance of each of the concepts with respect to their effects on the overall quality of life may be useful in future research. This paper reports a systematic review of literature on the application of Wilson and Cleary's model in chronic diseases to examine the paths and pattern of relationships of the concepts as well as determine their relative importance. We aimed to answer three important research questions:

Methods
This study followed the format of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement [12]. The electronic databases searched consisted of Science Direct, MEDLINE, CINAHL and PsyARTICLES. The search term used was "Wilson and Cleary" (free text). Further related search terms such as, "Wilson and Cleary model", "Wilson and Cleary conceptual model", "(Healthrelated quality of life OR HRQL OR HRQOL) AND (Wilson and Cleary OR Wilson and Cleary model)", were also used, but did not yield any additional studies. The search covered a period from 1995 (when the model was published) to December 2016.

Inclusion criteria:
Chronic disease Articles published in English language HRQL measured with validated instruments Empirical study Wilson and Cleary model was used or tested Peer-reviewed articles with full-text accessible.

Exclusion criteria:
Articles based on instrument development Articles that did not apply the model The titles and abstracts of retrieved articles were reviewed by AO for eligibility and selected based on the inclusion criteria. The selection was validated by JL. Full-texts of articles were reviewed for inclusion by AO, JL and YG validated these and also agreed on the five articles that were hand-searched for inclusion in the study.

Quality assessment of selected articles
The Quality Assessment Tool for Quantitative Studies designed by the Effective Public Health Practice Project (EPHPP) was used to evaluate the quality of included articles [13]. The EPHPP tool was designed to assess quality of observational and clinical studies. The tool was used to rate each article on a three-point scale (strong, moderate and weak) in six components: selection bias, study design, confounders, blinding, data collection methods, and withdrawal and drop-outs. A global rating was allocated to each study.

Data extraction, synthesis and analysis
Standardised data extraction form was used to extract data from the included studies by AOO, these were reviewed independently by YG and JL. Information extracted from each article included author, year of publication disease, study design, measures of latent factors, study aim, outcome of study and percentage of variance explained by the model. Articles selected had the primary objective of testing or applying the Wilson and Cleary model in the disease population. Most of the study designs were cross sectional (77%), heterogeneity was not formally calculated as meta-analysis was not performed.
Three research questions proposed for this study were to evaluate linearity and non-linearity of relationships and the effects of predictor variables on HRQL. Linearity was evaluated based on direct causal links between the concepts along the path of continuum proposed by Wilson and Cleary [10]. Non-linearity was evaluated based on significance of paths between non-adjacent variables -that is we establish that there is non-linearity if the effects between non-adjacent variables were significant so that paths between adjacent and non-adjacent variables were allowed [14]. The magnitude of the influence of each of the variables was also examined to evaluate their relative effects on HRQL.

Results
The initial search yielded a total of 2018 full text peer reviewed articles (Fig. 1). Duplicates were removed and articles were screened on titles and abstracts. The full-texts of the selected 78 articles were screened. Of these, 59 articles were excluded: 14 because they were based on instrument development, 18 did not apply the model, 14 were not empirical studies and 11 could not be categorised as focusing on chronic disease. Five additional articles were added through searching of reference lists of the selected studies. The total number of articles reviewed was 26 . The flow chart of the included studies is displayed in Fig. 1.
Thirteen different types of disease were studied: heart failure/surgery (n = 5), HIV/AIDS (n = 6), coronary artery Full-text article selected for further review for inclusion in study (78) Full-text article excluded with reasons (n = 57) Not a disease or chronic disease (11) Generic or disease-specific instruments (14) Not Empirical study (14) Not Wilson and Cleary model (18) Studies included in the review (26) Fig. 1 Flow chart of study selection procedure disease (n = 3), oral health disease (n = 2), obesity (n = 2), chronic obstructive pulmonary disease (n = 2) and asthma, diabetes, Hodgkin's lymphoma, kidney, Pompe disease, generalised anxiety disorder and stroke (all n = 1). Study designs were either cross sectional (n = 20) or longitudinal (n = 6). The total number of participants was 11,849, with a mean age of 50.5 years; 43.7% were female.

Quality assessment
Six studies were rated as strong (23%), 19 studies had moderate rating (73%) and one study had a weak rating (Table 1).

Disease-specific instruments for HRQL
Disease-specific HRQL instruments used in the studies included the MacNew Heart Disease Quality of Life [24], Minnesota Living with Heart Failure Questionnaire (MLFHQ) and the New York Heart Association (NYHA) classification to measure the global HRQL in heart failure [23,26,38]. The HIV/AIDS-Targeted Quality of Life (HAT-QoL) instrument [29][30][31] for HIV/AIDs populations while the Quality of Life, Obesity and Dietetics (QOLOD) [18], Oral Health Impact Profile (OHIP-14) [17,33] were used in diabetes and oral health populations respectively.

Analytical tools
In modelling the data (Table 2), different analytical tools were used: stepwise/hierarchical regression (n = 3), linear  • Perceived asthma control at baseline was affected by asthma symptom, physical activity, self-efficacy and smoking • Perceived asthma control at follow-up was predicted by asthma symptom, physical activity, self-efficacy and baseline perceived asthma control.
• Perceived asthma control was indirectly predicted by emotion status through selfefficacy and physical activity      • Social function is low in patients with more symptoms of anxiety.
• General health is negatively related to anxiety and depression but positively related to physical and social functions • Better overall QOL is associated with less symptoms and depression but related negatively to social function Wilson and Cleary [10] hypothesised that there existed direct causal links between biological and physiological factors, symptoms, functional status, general health perceptions and HRQL. Symptoms mediate between physiological factors and functional status, while functional status mediates between symptoms and general health perceptions, and general health perceptions mediates between functional status and overall HRQL. Eleven studies supported the direct causal link proposition between biological and physiological factors and symptoms (Fig. 2). Markers of biological and physiological variables were found to associate with worse symptoms in the patients with HIV/AIDs, xerostomia, coronary artery disease, Hodgkins lymphoma and generalised anxiety disorder [17,37,38,40,46]. The next level of the model associates symptoms with functioning and mediates between functioning and biological/physiological variables. This has been established in 20 studies (e.g., [16,22,29]). More symptoms predicted a greater impact on everyday functioning, with symptoms status explaining 49% of functional health in HIV/AIDs patients [37]. Functional status was found to have direct links to general health perception and mediated between general health perception and symptoms in 16 studies. Worse functioning indicated low perceived health. For example, worse functioning was associated with lower global oral health perception in Hodgkin's lymphoma [46]. More symptoms and less functional health may lead to a perceived decrease in perceived general health. The hypothesised effect of general health perception on overall HRQL was established in 12 studies (Fig. 2).
Research question 2: Does the Wilson and Cleary model strictly follow a linear unidirectional path? Linkages between non-adjacent concepts We examined the links between non-adjacent variables to establish whether empirical data show that the model allows non-linear, indirect paths.
Biological and physiological variables were directly associated with functional status, general health perception and overall HRQL in nine, two and three studies respectively (Fig. 2). For example, Kanters et al. [25] showed that enzyme activity, a biological marker, was significantly associated with HRQL in adult Pompe disease. Direct links were established between symptom status and, general health perception and HRQL in nine and seven studies respectively. Furthermore, functional status was associated directly with overall HRQL in seven studies. In coronary artery disease, physical functioning showed high positive significant effect on HRQL (β = 0.36) indicating that a reduction in functional health may reduce HRQL [24]. The studies assumed non-reciprocal relationships except Mathisen et al. [27] who attempted to model reciprocal relationship between general health perception and HRQL. This did not take into consideration possible effects between other concepts. Hence, we could not establish the possibilities of bidirectional relationships between the abstract concepts in this study.
Research question 3: What is the relative effect of each variable?
The relative effects of the variables were measured in terms of the magnitude of their influence on HRQL (Fig. 3). The causal links were labelled 0 M, 1 M, 2 M and 3 M to signify the number of mediators between constructs that were bypassed. 0 M was a direct link between the concepts with the proposed mediating variable signifying that no mediator was bypassed in the link, 1 M was an indirect link with one mediator bypassed, 2 M with two mediators bypassed and 3 M with three mediators bypassed. 0 M, 1 M and 2 M revealed symptoms status as a consistently important factor that affected HRQL, followed by functional status. In 0 M all four concepts; biological and physiological, symptom status, functioning status and general health perception were compared with respect to the effect of each on the adjacent variable. Clinical variables had the lowest magnitude of effect followed by general health perception, function status and symptoms status in order of increasing magnitude. Clinical factor was however on the same level with functional status when only the immediate mediator was bypassed.

Brief summary
The findings of this systematic review support the model of HRQL as proposed by Wilson and Cleary [10] and establish the conceptualised relationships and mediation. Of the articles reviewed, 74% found symptom status a significant predictor and critical mediator making it the most important predictor of HRQL by indirect effects through functional status and general health perceptions and by direct effects. More symptoms implied impaired functioning with consequently worse general health perception and lower HRQL. Most of the studies (73%) were of moderate quality; this was because the quality assessment criteria [13] gives low ratings to study designs that are not experimental or longitudinal in nature.

Comparison with previous studies
Both symptoms and general health perception were found to account for 38.2% of variance in global HRQL [26] and studies also showed that general health perception alone mediated the relationship between symptom status and HRQL [16,28,35,37,38]. One of the most important symptoms was depression which strongly associated with physical functioning (β = −0.32) and general health perceptions [24]. Two studies found no association between the clinical factors, and any of the health constructs in heart failure and diabetes [23,35]. This may be due to other profound non-clinical factors that responsible for impaired HRQL in these populations. The non-adjacent links among the health concepts showed that the model was non-linear also the effects of the variable were not fully mediated by their proposed mediators. For example, the direct link between symptom status and overall HRQL indicated that both functional status and general health perception did not fully mediate the effects of symptom burden on HRQL.

Possible explanations and implication
The findings of Sullivan et al. [14] in coronary artery disease patients supported our findings of a direct link between biological and physiological variables, and functional status. Further studies of a longitudinal nature will be required to establish possible bidirectional relationships among the concepts and whether the factors exert reciprocal influence on each other. For example, while the symptoms of pain may reduce functioning in patients with sickle cell disease, the inability to function as expected may lead to depression which may further

Limitations
We identified some potential limitations to this study. The study focused on several chronic diseases which have different clinical statuses, prognoses and levels of disability, which restricts our ability to generalise based on the lack of homogeneity of symptom status and functional status of the patients. There is also the potential limitation due to publication bias as only published articles were used in this study. In addition, different instruments were used to measure HRQL in the included studies; while some are generic, some are diseasespecific [47]. As there is no instrument that is a "gold standard", researchers often select instruments sensitive to the health state they are investigating ( [48] rather than a general measure of HRQL. Moreover, there are also variations in clinically important differences across groups of patients defined by diseases, conditions, severity level, socio-economic status and nationality [49].

Conclusion
Our findings show that the Wilson and Cleary model demonstrates a good fit and proved useful in identifying relationships among the health constructs, and predictors of HRQL in the studied disease populations. The model explained between 22.9% and 72% of the variance in overall quality of life indicating that, in some cases, the model may require modification to capture factors not specified in the model but that may be important determinants of overall quality of life.
The findings supported the robustness of the Wilson and Cleary model as a conceptual framework to characterise predictors of HRQL in chronic diseases and to aid understanding of the relationship between clinical and psychological outcomes for patients with chronic illness. Our understanding of specific directions of influence will aid healthcare practitioners and researchers to develop appropriate care protocols that will address psychosocial variables alongside clinical factors in chronic disease management. This study has demonstrated that symptoms are a major determinant of HRQL in patients with chronic disease, thus a clinical approach to reduce symptoms may help improve HRQL. Furthermore, in treating patients with chronic diseases, clinicians and healthcare practitioners should be alert for signs of depression because this study has highlighted depression as a major issue in HRQL.
Further work is needed to examine bidirectional relationships. Studies so far have focused on an assumption of no reciprocal relationship but low health perception or low HRQL might also worsen disease conditions and responses to treatment. Further studies on evaluation of the Wilson and Cleary model should be compared to the findings of this study.