Pursuing Quality in the Application of Bladder Cancer Quality of Life Research

Modeling concerns: What model of HRQOL to choose?

Because HRQOL is a complex, multifaceted construct, a variety of conceptual models were used to guide HRQOL research. A conceptual model is a schematic representation of a theory that acts as a heuristic map to provide a better understanding of a phenomenon (e.g., HRQOL) by depicting interrelationships among concepts or domains (e.g., physical, psychological, social, and emotional functioning) [12–13]. Models are important in HRQOL research because they determine how the construct is conceptualized, and can be assessed, andevaluated.

Bakas et al. [10] examined 100 English language papers published between January 1, 1999and August 31, 2010 pertaining to HRQOL modelsin different patient populations. These papers included literature reviews, instrument-development studies, descriptive or correlational studies, and intervention-based studies. Their study revealed that most of the HRQOL models described in these papers focused primarily on the influence of symptoms of various index conditions, rather than their management. Their study also revealed a large variationin the terminology and assessment of HRQOL, thus contributing to the difficulty in cross-studycomparisons.

To address these limitations, Bakas et al. utilized a set of ten criteria [22] to develop recommendations for three distinct HRQOL models. Two of these three models apply primarily to individuals’ HRQOL (see Wilson and Cleary [23], and Ferrans et al. [24]); the third model is a generic QOL model that applies to individuals, families, communities, populations and cultures (see World Health Organization International Classification of Functioning, Disability, and Health model; WHO ICF [25]).

Bakas et al., recommend the individual model of HRQOL developed by Wilson and Cleary as it includes five well-defined domains: biological, symptoms, function, general health perception and overall QOL. Two less well-defined domains also exist in this model (i.e., individual and environmentalcharacteristics). According to Bakas et al., the domains of this model relate to one another, reciprocal relationships among the different domains may exist, and both environmental and individual factors are associated with outcomes, thereby affecting total HRQOL. They also recommend Ferrans et al.’s model as a revision of the Wilson and Cleary model, as it retains the five major domains of the original model, makes explicit the individual and environmental characteristics, and simplifies the Wilson and Cleary model by removing non-medical factors. Finally, they recommend the WHO ICF model because its main concepts are well-defined (i.e., encompassing perception of general health, health-related domains of well-being, with general health and health-related domains further conceptualized in terms of functioning within the model). However, according to Bakas et al., the WHO ICF’s model is more appropriate as a mapping and classification framework than as a guide for hypothesis generation in HRQOL research compared to the models proposed by Wilson and Cleary [23] andFerrans [24].

Although the Bakas et al. study reflects ten years of research on HRQOL, the authors concluded that further refinement of HRQOL models used in research is needed. According to Bakas et al., analyzing uniquely derived HRQOL models may provide distinctive domains that might further inform commonly used models. Likewise, while the United Kingdom’s National Institute for Health and Care Excellence (UKNIHCE) recommends the use of global HRQOL measures, the UKNIHCE concludes that data may not be available or appropriate for all health conditions [26]. This is an important caveat because the use of global HRQOL measures implies that, across diseases, changes in HRQOL scores reflect changes in how various patients respond to the items on a HRQOL questionnaire. Additionally, the inherent ways to weigh different items in researcher-assessed HRQOL may further bias the interpretation of these measures especially when patients’ specific values or symptoms change overtime [27].

Patients’ responses to items on a HRQOL measure may vary greatly given the nature of a disease or a treatment when the HRQOL measure incorporates a variety of symptoms. For example, if some patients treated with chemotherapy experience fatigue but no pain, whilst other patients treated with surgery have no fatigue but do experience pain, these two items might be uncorrelated with each other — or even negatively correlated — and yet both items could be strong predictors for HRQOL [28]. These issues emphasize the importance of having “bolt on,” disease-specific subscales attached to a global HRQOL measure when studying a specific disease or treatment option [29]. If not assessed, clinically important HRQOL signs and symptoms may go unreported— and, in a clinical setting, untreated [30]. Thus, a failure to characterize a HRQOL model adequately, beginning at the point of its definition may lead to a mismatch between the conceptual model and its application in clinical settings [28].

In sum, models are important in PRO research as they determine how the HRQOL constructs are conceptualized. Accurately representing the breadth of clinically significant symptoms and changes in functioning following MIBC and NMIBC treatment are necessary so that indicators of low levels of HRQOL are not overlooked by health careproviders.

Instrument concerns: How well the instrument of choice function within a selected mode?

Having settled on a model of HRQOL in clinical bladder cancer care, the next step is to develop or select a measurement instrument. Placing PRO-based variables at the center of research and clinical trials makes accurate measurement a cornerstone of HRQOL in clinical care [30, 31]. There are a multitude of PRO measures of HRQOL and some useful guidelines for using reviews to choose among the available instruments as well as recommendations for working with patients with low levels of health-literacy, socially excluded, and non-English speaking patients (see Boynton et al. 2004) [30]. Table 1 depicts examples of HRQOL measures currently used in bladder cancer research.

One method for choosing a HRQOL measurement instrument is to make a selection from among a set of instruments previously vetted by researchers. The Clinimetrics working group of the EMGO Institute of Health and Care Research of the VU University Medical Center in Amsterdam developed COSMIN, a checklist created to reconcile approaches to evaluating measurement instruments [31–32]. Beginning with a four-round Delphi study of 57 international experts (psychologists, epidemiologists, statisticians and clinicians), the COSMIN group developed a standardized tool that appraises three measurement domains: 1) reliability (internal consistency, reliability, and measurement error); 2) validity (content validity, construct validity and criterion validity),and; 3) responsiveness (the ability to detect change over time) [31–33]. In addition, the COSMIN checklist includes items related to interpretability and generalizability of HRQOL measurements. Both classical and item response test theories can be assessed with this checklist. The COSMIN group also provided a database of systematic reviews of studies that utilized HRQOL instruments in different populations which is available at the COSMIN web-site http://database.cosmin.nl/.

Another approach to evaluating HRQOL measures comes from a study conducted by the International Society for Quality of Life Research (ISOQOL) [34]. Measurement standards in this approach include: 1) documentation of the conceptual and measurement model; 2) evidence for reliability, validity (content validity, construct validity, responsiveness); 3) interpretability of scores; quality translation, and; 4) acceptable patient and investigator burden.

Besides attending to the measurement properties of a preexisting bladder cancer HRQOL assessment instrument, making a choice among the available options requires an awareness of the assumptions measurement experts make in their own model selection process – especially when selecting a disease-specific measure. There are more than 50 published studies of HRQOL in bladder cancer using over a dozen instruments [3–5, 35]. Different methods for selecting items for a disease-specific (i.e., MIBC or NMIBC) HRQOL questionnaire produce different instruments. Homogeneous sets of items produced by psychometric strategies can produce scales that do not include important patient concerns [36, 37]. Scales that include a variety of uncorrelated items may capture a greater number of patient concerns. Such “clinimetric” scales combine a variety of items that may not relate to each other statistically [38].

Interpretation concerns: How interpretation of the resulting data fits the model?

Translating research from the findings of aggregate sample and population studies requires a change in perspective from groups to the individual. Osoba et al. put the issue succinctly: “Small numerical differences in mean scores derived from health-related quality-of-life … assessment instruments may reflect statistically significant results when large samples of subjects are involved, but the clinical interpretation of the meaning of small numerical differences is uncertain” [41].

One way to achieve translation of group data to clinic-based individual assessment is to combine group data and clinical assessments. Cella et al. used item-bank responses collected from 840 cancer patients to develop symptom vignettes across a range of symptom severity [42]. They then had clinical experts rank order the vignettes by severity, eventually arriving at a consensus regarding which two vignettes were at the upper and lower boundaries of normal and mildly symptomatic for each symptom. Using this process repeatedly, they identified cut scores separating mildly from moderately symptomatic, and moderately from severely symptomatic scores. The study yielded T-score thresholds that differentiated levels of symptom severity for pain interference, fatigue, anxiety, and depression based on a combination of clinical judgment and self-reported patient severity scores.

Other methods of turning group data into information for ranking individual HRQOL scores involve boot-straping information from large groups of patients, while using that information to adapt or revise a HRQOL instrument to a clinical need. This is a process similar to that used by Cella et al. [42] but on a larger scale.

Hjollund et al. describe a generic PRO system, WestChronic, that boot-straps large quantities of individual data into new HRQOL areas. Developed in Demark in 2004 for epidemiological research, this system allows clinicians to access PRO data on symptoms, functional status, and other indicators of HRQOL [7]. In a recent report on the system, 22 PRO projects existed for 18 diagnostic groups, and available data included 59,232 questionnaires from 30,174 patients. Although none of the reported projects included bladder cancer patients, projects existed for lung and prostate cancer. In line with recommendations on the design and evaluation of outcome measurements [30], the authors reported a four-step process for the development of specific-disease measurements: 1) identifying HRQOL content – specific content is identified through literature reviews, interviews with patients, and input from content experts such as clinicians and nurses; 2) questionnaire development– questionnaire items are written, again based on information from literature reviews, patient interviews and content experts; 3) patient interviews – testing the draft questionnaire determines if the meaning and intention of each question is understood by patients, and; 4) testing – the questionnaire is administered to a large sample of patients and analyzed to test the statistical properties, including validity, reliability, sensitivity and responsiveness. Using this process it is possible to enable clinical access to PRO data on symptoms, functional status, and health-related quality of life of individual patients. This allows clinicians to flag symptoms that need further attention during a patient visit. In summary, combining group data with clinical assessments provides a work-around for the difficulties inherent in translating research from the findings of aggregate sample and population studies into a format useful for clinical care. This process requires a systematic approach to creating HRQOL instruments, a large amount of data and a good deal of expense.

Utility concerns: How to derive some utility from the chosen model?

Alston et al. surveyed a nationally representative sample of 1,068 patients who had seen at least one health care provider in the previous 12 months, and found four issues patients face regarding medical communication [43]: a) patients’ desire to engage in shared health-care decision making, b) there is a gap between patients’ desire for engagement in their health care and the usual process of decision making, communication, and patients’ preferences, c) patients’ needs to experience more coordinated health care, d) and engaged patients are more likely to report a better healthcare experience.

It is these types of concerns that POR and HRQOL measures seek to address [44]. However, if themodeling, instrument and interpretation issues are not addressed by the HROL measures used, these goals cannot be met. Research on bladder cancer patients’ HRQOL is making progress, but if the goal is to create a standard procedure for addressing engagement and outcome in clinical practice, then we are not there yet. Studies of MIBC do not provide sufficient information to clearly assist patients in choosing one form of urinary diversion over another following radical cystectomy, for example [40]. Studies in NMIBC rarely find any effects on HRQOL, despite suggestions in the literature that effects exist on mental health and sexual function [29].

To better understand how bladder cancer affects patients’ HRQOL, we suggest returning to our conceptualizations of bladder cancer HRQOL models and instruments. The Bladder Cancer Index is effective in showing differences in patients managed with endoscopic, intravesical or radical surgical treatments [29]. Further enhancing this the model may improve understanding of HRQOL across the MIBC and NMIBC disease spectrum.

Schmidt et al. found that psychological distress increased when patients learned of their bladder cancer diagnosis. The impact of receiving the diagnosis lasted at least six months, but apparently not twelve months [38]. This suggests that a bladder cancer measure must be responsive enough to measure changes in psychological well-being over a period of at least six months. Adding a few more items to measure a psychological health domain, for example, might increase the responsiveness of the BCI [29] to both MIBC and NMIBC. This would mean having a new model of bladder cancer HRQOL: one that reasonably includes psychological as well as physiological impacts. Furthermore, the model would have to assume that psychological distress does not occur at extreme endpoints. Life events such as divorce or sudden unemployment following bladder cancer may introduce additional distress. Hassles or daily stressors (e.g., daily use of catheters and stoma appliances) experienced may be significant contributors to emotional distress for a period of time before the patient masters skills required for self-care.

Similarly, adding other constructs to our model of bladder cancer HRQOL may provide a better measure of the breadth of issues associated with this disease. Research in other cancer populations reported significant associations between fear of cancer recurrence or progression and HRQOL [45]. Although not well documented in the bladder cancer population, fear of cancer recurrence or progression is likely to be triggered by frequent follow-up cancer and morbidity surveillance, thus further decreasing the patient’s HRQOL. Therefore, a comprehensive model of bladder cancer HRQOL should address these psychological parameters.

Another component of the breadth of model issue lies in the unmet needs of bladder cancer patients. Unmet needs are problems experienced by a patient that are not addressed by the health-care provider, health care system or the patient’s support network. Unmet needs may include the informational needs (e.g., information about treatment options and side effects, follow-up care), supportive care needs (e.g., physical, emotional, sexual, practical, and social needs) or spiritual needs a patient perceives as necessary to achieve optimal well-being [45]. Research with cancer populations other than bladder cancer shows high levels of unmet need related to provision of information, psychosocial support, practical assistance, and sexual issues, resulting in increased anxiety and poorer HRQOL. Very few studies examined these issues in bladder cancer patients [45].

Understanding potential associations between unmet needs and HRQOL in bladder cancer patients is very important as these associations could be influenced by several patient factors. A patient’s low health literacy level for example could influence patient-provider communication, thus, leading to increased unmet informational needs and difficulties with self-care and symptom management. Similarly, a patient’s cultural beliefs, expectations, and values could influence a patient’s treatment decision making leading to potentially avoidable treatment side effects and dissatisfaction with health care (e.g., opting for a neobladder to avoid altered body image associated with an ileal conduit). Thus, understanding a patient’s needs, treatment-related objectives, values, and cultural/religious beliefs is pivotal in reducing the patient’s unmet needs and improving his/her HRQOL and satisfaction with health care.

In developing the concept of individualized HRQOL, Huebner et al. [46] asked patients to determine which objectives besides survival are relevant and should be achieved by treatment. Integrating this concept into bladder cancer HRQOL research points out the importance of understanding patient-specific values and preferences (i.e., clinimetric measures) for the development of disease-specific bladder cancer HRQOL measurement. Equally important is the need to identify a comprehensive HRQOL model of bladder cancer that takes into account the differential impact of the disease and treatment outcome on patients’ functioning and wellbeing. The development of a new bladder cancer HRQOL metric should follow recommended psychometric standards [30]. Establishing the theoretical model of perceived health or symptoms of interest is the first and most important step in developing bladder cancer models and HRQOL measures [47].

Bladder cancer models will not be comprehensive without addressing informal caregivers’ emotional adjustment and unmet needs. Despite the fact that cancer presents a crisis for patients and their families, [48] research in bladder cancer HRQOL has focused almost exclusively on the patient’s response and adjustment to the disease. Research in other cancer populations has shown that informal caregivers play a significant role in helping patients manage their disease and treatment [49]. Most informal caregivers assume their role with little or no formal preparation and training and have low self-efficacy for caregiving [50, 51]. Caregivers also frequently encounter communication difficulties with patients, which can adversely affect the coordination of care as well as the quality of their relationship and QOL [52, 53]. For example, even well-meaning caregivers may offer assistance in ways that are controlling or over-protective rather than supportive [54] and can engage in maladaptive communication (e.g., criticism, nagging) [55, 56] that can undermine the patient’s emotional well-being and abilities to achieve self-care goals (e.g., learning skills needed for stomacare).

In addition to undertaking complex care tasks and providing emotional support, informal caregivers must cope with the grief and loss associated with their loved one’s life threatening illness [57]. Often, they are reluctant to discuss their fears and concerns because they worry that doing so will distress the patient [58–60]. This not only exacerbates patient and caregiver distress, [58] but it may also affect the caregiver’s ability to provide care (e.g., assist with the use of stomal appliances) [61]. Poorly managed symptoms have been attributed at least in part to inadequate coping and exhaustion on the part of the family caregiver in other cancer populations [62, 63]. Thus, understanding and addressing caregiver unmet needs for information and support in bladder cancer is not only essential for improving caregiver health and well-being [64–66], but also for improving the quality of patient care – particularly for bladder cancer patients managing symptoms, treatment side effects, and self-care requirements associated with cystectomy and urinary diversions [67–68].