© 2004 by Oxford University Press
2004 © Oxford University Press
Article |
A Health Outcomes Framework for Assessing Health Status and Quality of Life: Enhanced Data for Decision Making
Correspondence to: Pennifer Erickson, PhD, Pennsylvania State University, 1316 Deerfield Drive, State College, PA 16803 (e-mail: pae6{at}psu.edu).
 |
ABSTRACT |
|---|
 TopNotes Abstract IntroductionHealth Outcomes FrameworkFeatures of the Core...Added Value of the...From Formation to ImplementationReferences |
|---|
Background: Currently, information to improve health status and quality of life is derived from independently designed data systems that range from population-based health surveys to health records used in managing individual patient care. But there is no coherent strategy for using these data sources in concert across diverse applications. Thus, it is frequently difficult to compare or combine results across studies to provide population-level inferences based on findings from specific subpopulations. Objective: The Health Outcomes Framework is an analytic structure to provide more comprehensive information about the health status and quality-of-life impact of disease and its treatment. This framework consists of three components: (1) a core set of health, lifestyle, and economic questionnaires that collect data from an individual's perspective; (2) applications that range from population to patient care levels; and (3) time. Although health is the outcome of interest, lifestyle behaviors and economic and political factors are important determinants of health, which also need to be studied using standardized procedures; thus, they are included in the core. This article focuses on the nature and application of a core health status and quality-of-life instrument. To be useful across a range of applications, such a core instrument needs to have three conceptual characteristics: (1) a theoretical model that regards health as a continuum of states; (2) domains that represent policy-relevant aspects of impairment, disability, and handicap; and (3) a set of societal preferences for these domains. In addition, the core needs to address three practical concerns: (1) brevity, (2) methods of administration that are suitable for respondents with diverse capabilities, and (3) documentation that is publicly available. These features are discussed using examples from currently available, multi-domain generic instruments, each of which has some, but not all, of the necessary features of the core instrument. Conclusion: The Health Outcomes Framework is intended to be a cooperative effort. It is proposed that the National Cancer Institute and the Centers for Disease Control and Prevention take leadership roles not only by adopting a core health status and quality-of-life instrument for use in current and future cancer data collection activities but also in encouraging industry and academic investigators to implement this core instrument in their cancer studies. Having a vertically integrated core instrument can lead to more representative data for informing decision making and ultimately for obtaining a more equitable distribution of health.
|
INTRODUCTION |
|---|
|
TopNotesAbstractIntroductionHealth Outcomes FrameworkFeatures of the Core...Added Value of the...From Formation to ImplementationReferences |
|---|
Cancer imposes a significant burden on the U.S. population, both collectively and individually. According to recent data, malignant neoplasms were the second leading cause of death in the United States, accounting for almost one-fourth of all deaths (1). The economic burden of cancer is also high. In 2002, the total cost of cancer was estimated to be $189.5 billion, approximately two-thirds of which was indirect cost, including lost work productivity (2). Another reflection of society's concern about the burden imposed by all types of cancer is the amount spent on cancer research; approximately 20% of the National Institutes of Health's (NIH) budget is allocated to the National Cancer Institute (NCI) (1). At the individual level, breast cancer has been found to be the disease that women fear the most (3), even though heart disease is the leading cause of death. In addition, cancer diagnoses and treatments are known to affect an individual's quality, as well as quantity, of life during both the treatment and remission phases of disease (4).
Information to enhance health status and quality of life, encourage healthy lifestyles, evaluate therapeutic alternatives, and allocate resources is derived from various data sources ranging from general population health surveys that record self-assessments of health to patient-derived data in clinical settings (5-8). Increasingly, individuals make treatment decisions that affect their health and require information about trade-offs between risks and benefits (9-10). Thus, reports from individuals of their health and well-being are becoming used more frequently to inform health decisions across a range of applications from managing care in clinical practice settings to implementing programs to improve the health of the general population (11-13).
Although person-reported health data exist for informing policy, they have come from data systems that traditionally have been designed and maintained according to sponsor. For example, surveys of the general population are usually supported by the U.S. government as part of a national accounting system. In addition, the U.S. federal government maintains special topic registries, such as those supported by the Surveillance, Epidemiology and End Results (SEER1) program, and conducts clinical trials through the National Institutes of Health (NIH) (14-15). Collection of clinical data is also funded in large part by pharmaceutical companies in the process of gaining Food and Drug Administration approval for new therapeutic compounds, with the data being maintained within each sponsoring company. Government-sponsored data systems are readily available for public use (16); whereas, data that supply detailed patient information, including treatment response, are generally privately owned and are not publicly available.
Even if availability were not a problem, data collected for different purposes are not easily compared and used to inform decision making across a wide range of treatment alternatives and population groups because independently designed studies generally use different measures of health outcome. If the measures were coordinated across various types of studies, this common information could be used for comparing specific and general populations, as well as for modeling to obtain estimates of, for example, the health effects of a new intervention on individuals in the general population
This article presents the Health Outcomes Framework (HOF) as an analytic structure to provide more comprehensive and comparable information about the health status and quality-of-life impact of disease and its treatment. This framework encompasses applications that range from the overall population to individual patient care levels. They are linked through a core set of questionnaire-based health outcomes instruments that collect data from the individual's perspective. Use of the same instruments across all applications would permit the vertical integration of information, thereby enhancing interpretation of findings. As presented below, the framework can be used with any disease or set of diseases, such as those within the mission of the National Institutes of Health. Next, we discuss the added value of this approach within the cancer arena. By including applications that range from clinical studies to general population surveys, the framework can use information about the outcomes of health interventions, gathered from highly selected study populations, to generalize to all members in society who are likely to benefit from a given intervention. With the capacity to have more complete, individual-based information on overall benefits than is currently available, decision makers will be better able to assess the trade-offs between efficiency and equity in allocating resources to health as well as to evaluate how the allocation fits with their tolerance for inequalities in health.
|
HEALTH OUTCOMES FRAMEWORK |
|---|
|
TopNotesAbstractIntroductionHealth Outcomes FrameworkFeatures of the Core...Added Value of the...From Formation to ImplementationReferences |
|---|
The HOF expands the scope of previous models (17-18) by incorporating (1) individually reported health status and quality-of-life data, not just mortality and morbidity, and (2) population surveillance and clinical practice applications. In addition, the HOF incorporates economic data that can be explicitly combined with health status and quality-of-life data to evaluate health care programs and allocate resources to health. Thus, the HOF, with its key components—namely, the core set of health, lifestyle, and economic questionnaires, range of applications, and time—is intended to guide the collection, analysis, and interpretation of data to facilitate decision making about health care as well as the behavioral and environmental interventions that affect the health of individuals as well as of population groups. Because data are used as inputs into decision processes, rather than as ends in and of themselves, it is necessary to introduce the HOF within the context of information for health decision making, as depicted in Fig. 1.
|
Key Components of the Health Outcomes Framework
Because the key components of the HOF interact to give decision makers information to improve population health, any core instrument consistent with this framework has to be appropriate for use in all types of studies included in the framework. Time is included for two reasons. First, this acknowledges that different applications will have different durations, or periodicities, of data collection. Second, the explicit inclusion of the time dimension serves as a reminder that effects of a health intervention, which may be observable over a relatively short period in clinical research applications may take years, rather than months, to be observed as a change in population health; in this context, time is important from a societal perspective.
The core instruments, shown as the vertical columns in Fig. 1, are necessary to analytically relate, or vertically integrate, health status and quality-of-life data in all applications to make inter-study comparisons and to model generalizability of interventions, broadly defined. The idea of a single instrument for multiple uses builds on the concept of a core questionnaire that is used in the National Health Interview Survey (20). The HOF, however, expands on this by incorporating instruments that gather subjective evaluations of health status and quality of life and by advocating that the core instrument be used not only in general population surveys but also in a range of study designs. Although health is the outcome of interest, individual characteristics, such as lifestyle and health behaviors, and also environmental characteristics, such as economic and political factors, can be important determinants of health (18). Because these characteristics likewise need to be analyzed using standardized data and methods, they are included in the core. The remainder of this article, however, focuses on the nature and application of the health status and quality-of-life instrument.
Categories of applications, or studies, that produce data about the contributions of health care and other factors that influence health are represented by the horizontal bars in Fig. 1. For each application, the length of the horizontal bar along the general-to-specific axis represents the level of detail in health data collected. Use of generic instruments to compare individuals and groups defined in terms of socio-demographic, disease, or treatment characteristics is represented by the non-core general portion of the bar to the left of the core. Shorter segments indicate the inclusion of an instrument that collects a narrower range of health information, such as the SF-36 Health Survey (SF-36) (21), compared with longer segments that indicate the use of an instrument that collects a broader range of information, such as the Sickness Impact Profile (SIP) (22).
Use of specific instruments, whether disease- or treatment-specific, is represented by the bar to the right of the core (Fig. 1). Nearest to the core would be profiles such as the EORTC QLQC30 and FACT-G (23-24), which are cancer-specific, although general in content. Although some tumor site-specific instruments have been designed as modules for use with a general cancer measure (25-27), most have been designed to be used independently, for example, the Breast Cancer Questionnaire (28), the UCLA Prostate Cancer Index (7), and the Performance Status Scale for Head and Neck Cancer (29). Use of site- or treatment-specific measures would appear to the right of a general cancer measure.
Time, shown along the third axis, represents the dynamic nature of health. Operationally, information on changes in health may be derived from studies that incorporate either repeated cross-sectional or longitudinal designs; for example, a health outcome instrument may be administered monthly as part of a clinical trial. Instruments are generally designed with an explicit respondent recall period, for example, the past week. They often are geared to detect particular short-run time effects such as change in status since last measurement—for example the evaluation of an anti-emetic to treat nausea associated with a chemotherapeutic regimen among women with breast cancer (30). Assessment of long-term effects, for example, the effects of an antismoking campaign on reducing lung cancer mortality (31), is more likely to use transition matrices or life table methodology (32).
The arrows in Fig. 1 show some of the many paths that information (when used for decision making) can take in the course of having an impact on population health. Arrows on the left indicate that changes in population health, for example, an increase in the number of deaths due to breast cancer, influence the search for determinants of and new treatments for this disease. In turn, as new chemotherapies and surgical procedures that are identified through biomedical research spread into clinical practice, wider application of these interventions may generate evaluation studies to determine the effectiveness of these interventions, for example, the evaluation of the costs and benefits associated with cancer screening in high-risk populations (19). The reverse information flow, from clinical practice to the general population, represents how the use of new treatments and knowledge of relationships between lifestyle and health findings might affect population health over time.
Health Status and Quality of Life Data Across the Range of Applications
The range of applications of outcomes measures is categorized into five types of studies that are traditionally used to inform health decision making; these applications are identified in Fig. 1. For these applications, the purpose of the study determines the form of the health status and quality-of-life instrument used, as described below.
Population surveillance studies monitor progress against disease burden using data collected from representative samples of the general population. Thus, items in health questionnaires are applicable to a wide range of socio-demographic groups as well as to groups defined in terms of specific diseases, with a large component of the health data being of a generic rather than disease-specific nature. In the United States, data for population monitoring are collected using repeated cross-sectional or longitudinal panel surveys; for example, the National Health Interview Survey (NHIS), National Health and Nutrition Examination Survey (NHANES), and the Medical Expenditures Panel Survey (MEPS) (6,8,33). In addition to the generic content, specific modules, such as the Cancer Module in the 2000 NHIS, can be included to provide population-based information on special topics (34).
Typically, population health data are presented as indicators, either singly or in clusters, to identify disparities between groups defined by age and ethnicity, to examine geographic patterns of health, and to allocate resources. For example, self-assessed health status and limitation of activity from the NHIS are used to monitor health status over time as presented in Department of Health and Human Services' annual report to Congress on the state of health of the U.S. population (1). These two direct measures of health have been combined with mortality to form a summary measure, the Healthy People 2000 Years of Healthy Life, which has been used to monitor the healthy life span of the U.S. population since 1990 (35,36).
Epidemiologic investigations permit the evaluation of distributions of diseases and other health conditions, patterns of health care use, and effects of interventions on outcomes of interest, as well as the behavioral and societal determinants of these outcomes. These findings may be the results from observational studies, natural experiments, specially designed surveys or controlled experiments, or mathematical modeling (37-39). Such investigations have the capability to examine disparities between population groups and to allow these to be related to utilization of care, lifestyle, and individual behaviors. Epidemiologic studies of persons with cancer include patterns-of-care analyses and quality-of-care monitoring. Although health data collected for epidemiologic purposes may be more specific than data collected in general population surveys, epidemiologic data are generally less specific than those collected in biomedical research or clinical practice settings (see relative lengths of the horizontal bars in Fig. 1).
Although single indicators have traditionally been common, many recent studies employed multiple indicators that use the same method of scoring, i.e., a profile to show relationships between health status and disease factors. For example, when the SF-36 was included in the Nurses Health Study, a diagnosis of breast cancer was found to be associated with lower health status in all of its subscales, except for mental health. The declines were greatest for physical and social functioning (40). If the SF-36, which is a widely used health status instrument, had been included in a national survey, then additional inferences about the relevance of these findings to the general population might be drawn from this study. Even more information might have been available if a designated core instrument, as proposed by the HOF, had been included in the Nurses Health Study. For example, not only could the general population serve as a comparison group but also the epidemiologic findings might be integrated with those from clinical studies to provide more comprehensive information for examining relationships between lifestyles, disease factors, and health status and quality of life.
Biomedical research, including clinical trials and longitudinal observational studies, are used to evaluate both safety and efficacy of diagnostic, rehabilitative, and therapeutic interventions. Although these studies are designed to collect disease-specific information on health status and quality of life, with primary and secondary endpoints being carefully monitored to determine treatment effect on health outcome, many include brief generic health measures (41-45). If the purpose of the trial is also to provide information about the relationship between costs and health outcomes, then an instrument that allows combining mortality and morbidity into a summary score, for example, the EQ-5D [also known as EuroQol (46)], may be included (47).
Data from clinical trials are generally presented as profiles of scores or clusters of indicators that show how treatment affects selected domains of health. The core data might also include a preference-based summary measure if an economic component was part of the clinical trial. For example, Uyl-de Groot and colleagues (46,47) used a battery of measures, including the preference-based EQ-5D instrument, to evaluate treatment outcomes. In addition to making between-regimen comparisons, the EQ-5D patient scores were also compared with population scores, thus illustrating the use of normative comparisons to gain insights about generalizability of findings that the core instrument can provide within a clinical trial research design.
Evaluation studies incorporate health outcomes assessment to establish priorities, examine the effectiveness of health policies and programs, and allocate resources, using methods that range from specially collected data in a variety of study types, including clinical trials and other longitudinal studies, to mathematical modeling, including simulation analyses, decision trees, and other forms of optimization models (18). The primary purpose of such investigations is to address issues associated with allocating various health care interventions that are designed to improve health.
Especially when used to inform societal resource allocation, these studies require a summary measure that includes not only mortality and health status and quality of life but also data on societal preferences for health (48). For example, the Population Health Model (POHEM) has been used to evaluate the population-based impact of cancer control interventions in terms of cost per quality-adjusted life year gained (49,50).
Clinical practice applications assess patient status over time, especially to determine treatment response for managing patient care. These uses may be based on either individual patients or groups of patients that are usually identified by characteristics such as disease or source of care. Whether individual or group level, the data are generally collected at the place of care. At this level, health outcomes data will most likely be specific to the patient's health condition and include a range of clinical measures, physician evaluations, and patient self reports. The core data may be the only generic patient-reported outcomes needed.
To manage patient care, outcomes data are generally presented as clusters of indicators and, more recently, profiles of scores. For example, in a study of men with and without prostate cancer in a managed care setting, health status and quality-of-life profile data and disease-specific symptom data were compared between the two groups to learn about changes in their health (51). Both generic and specific measures were important for assessing outcomes of care. If vertical integration had been incorporated into the study design, data on patient behaviors might have been analyzed together with those from a related epidemiologic study to understand potential health gains from the changing lifestyles and health habits of the patients.
Traditional presentation of data as summary measures as well as indicators and profiles of scores is evidence that the core instrument needs to assess multiple aspects of health, i.e., domains, and to generate both an overall score as well as scores for each of the separate domains. Further, using the same instrument in diverse study designs, e.g., population surveys and clinical trials, and with a variety of respondents differentiated by characteristics such as age and severity of dysfunction, requires a core instrument with domains that are meaningful across all socio-demographic and functional categories. This instrument also needs to be capable of being administered using a variety of methods. In addition, multiple applications require ease of access to documentation by all researchers whether publicly or privately sponsored, which details the core instrument's use and interpretation.
|
FEATURES OF THE CORE HEALTH STATUS AND QUALITY-OF-LIFE INSTRUMENT |
|---|
|
TopNotesAbstractIntroductionHealth Outcomes FrameworkFeatures of the Core...Added Value of the...From Formation to ImplementationReferences |
|---|
Central to the Health Outcomes Framework is the idea of a core instrument that can be used in all applications (although additional data elements will certainly be collected within each application to accommodate the full range of study objectives). Thus, the core instrument needs to possess three conceptual characteristics: (1) a theoretical model that regards health as a continuum of states; (2) domains and subdomains that represent policy-relevant aspects of impairment, disability, and handicap; and (3) a set of societal preferences for these domains. These three features are discussed below using examples from currently available, multi-domain generic instruments, each of which has some, but not all, of the necessary features of the core instrument (instruments are summarized in Table 1).
|
To satisfy all applications of the Health Outcomes Framework (32, 52), a theoretical model arrays health states along a continuum ranging from death to optimal health and incorporates societal preferences for health that are appropriate for use in resource allocation. In this model, death is scored as 0 and optimal health as 1, with other health states having scores between these two endpoints. Instruments based on this model, for example, the EQ-5D, Health and Activity Limitation Index (HALex), Quality of Well-being Scale (QWB), and Health Utilities Index (HUI-3) (Table 1), produce summary scores that may also be disaggregated to produce subscale scores (46,53,54). The current trend toward using both multiple indicators and summary measures in large-scale surveys as well as in biomedical research applications demonstrates the feasibility of including an instrument based on this continuum-of-health model for the range of applications in the Health Outcomes Framework. (46, 55-58).
Instruments that produce a profile—that is, multiple scores that are based on the same scoring metric and are not designed to be combined into a summary score—may be based on a model first articulated for the development of the SF-36. In this model, domains are viewed as "concentric circles," with the innermost circle representing health characteristics intrinsic to the individual, for example, disease, and the outermost representing those that are external to the individual, for example, social functioning (52). Although a profile of scores provides information to understand the impact of an intervention on individual domains, this theoretical approach lacks a preference structure that (1) specifies relationships among domains, and (2) explicitly incorporates society's values for a range of health states into the decision making process. Further, without a preference-based summary score, profile-type measures are unsuitable for use in evaluation studies that are designed to allocate resources using the principle of health optimization.
The SF-6D is one of several approaches that have been proposed to convert data collected from instruments that produce multiple scores into a summary score based on the health continuum model (59-62). Because the SF-6D uses data collected by the SF-36, which is the most widely used multi-domain instrument, it has attracted considerable attention. Although modeling a preference-based score from a descriptive instrument has been reported in the literature, no single approach for making this conversion has been found acceptable to both the research and decision-making communities.
Domains and subdomains can range from impairments, such as signs and symptoms, to handicap, such as societal reactions to one's health and ability to cope with stress (32). For use in the Health Outcomes Framework, the core needs to include, at the least, the domains of physical, social, and mental function that are generally recognized as basic to being able to function independently in society. Independence is a concern of most people and is amenable to improvement as a result of social policy. The core might also include health perceptions, as in the COOP Chart System (COOP), HALex, SF-36, and SF-12 Health Survey (SF-12) (Table 1), to capture individual sentiments about health and well-being.
Preferences, or utilities, for health states convey the importance that an individual or a society places on domains of health status and quality of life, with societal rather than individual preferences being required for the allocative functions included in the HOF. Although these data may be based on revealed preferences derived from observational studies of administrative records, the more widely used methods for determining societal preferences involve collecting data from representative samples of persons from the relevant population(s) using either time trade-off, standard gamble, or category scaling methodology. Although the standard gamble has been recommended for cost-effectiveness analysis (32,48,63), category scaling methods are more likely to be amenable for collecting preference data in a large-scale general population survey (48).
For making informed decisions, preferences as well as domains need to be meaningful to individuals across all age groups, as well as those of different socio-economic and ethnic/cultural backgrounds. The ability to measure the health of all individuals using the same metric, or yardstick, is essential for understanding the implications of decisions that affect diverse programs, such as a smoking prevention program aimed at young people and a lung cancer treatment regimen among middle-aged and older adults. The on-going work of the World Health Organization's World Health Survey has important implications for understanding cross-cultural similarities and differences in preferences for health (64).
Three practical concerns are also important for the implementation of the core instrument: (1) brevity, (2) methods of administration that are suitable for respondents with diverse capabilities, and (3) documentation that is publicly available. The significance of each of these three features to the Health Outcomes Framework is discussed below. Again, examples are drawn from the currently available, multi-domain instruments, each of which has many but not all of the features needed for the core instrument shown in Table 1.
Brevity of the core involves both the number of items that each respondent is required to answer and the amount of time it takes for completion of the instrument. The core needs to be relatively short and easy to use so that it adds little, if any, additional burden in terms of both respondent time to complete and administrative resources for assembling, editing, and tabulating these responses. Minimal additional burden is essential for widespread adoption of the Health Outcomes Framework. Ideally, the instrument would be approximately 10 items or fewer and require an average of 3 minutes, or less, of the respondent's time to complete. Because time to administer varies greatly according to respondent characteristics, with more dysfunctional persons taking more time to respond than less dysfunctional persons, and because comparable data on time are not readily available for all of the instruments summarized in Table 1, the number of items may be used as a proxy for time; this is a reasonable substitution because time to administer and number of items are directly related. As indicated by the COOP, EQ-5D, HALex, QWB, HUI, and SF-6D (Table 1), it is possible to have a multi-domain instrument that includes a parsimonious set of items, thereby assuring low respondent and administrative burden.
Methods of administration involve the use of alternative instrument formats, such as in-person or telephone-interviewer or self- or postal-administered questionnaires. Depending on the study group, different methods of administration may be required; for example, persons with low vision or linguistic difficulties may provide more valid responses if in-person interviews rather than self-administered questionnaires are used. In addition, it is unlikely that the same method would be appropriate across all applications; for example, telephone interviews might be an efficient procedure for collecting population surveillance data but considered too impersonal for use with patients in clinical practice settings. Through carefully conducted validation studies, comparable results can be obtained from multiple modes of administration, as indicated by the experience with the EQ-5D, QWB, SF-36, and SIP (Table 1).
Documentation needs to be clearly written and publicly available so that core data are collected in a standardized manner across all of the applications; the National Health Interview Survey (NHIS) website (34) illustrates the feasibility of making this information widely accessible. Good documentation would state that any change in instrument wording, format, scoring, or other procedures will lead to less than optimal results from vertical integration. In addition, the documentation needs to (1) specify the scoring algorithm; (2) give instructions for calculating scores when non-standard situations that frequently arise when collecting health outcomes data, such as missing data, are encountered; (3) supply computer programs for calculating scores; and (4) present normative data for interpretation. Inclusion of the core instrument in a general population survey, such as the NHIS, will assure that the norms are available for diverse population groups and regularly updated. Specification of the rationale for the use of the core instrument and provision of technical support are crucial for obtaining the value added proposed by vertical integration, an analytic concept that is central to the Health Outcomes Framework.
|
ADDED VALUE OF THE HEALTH OUTCOMES FRAMEWORK APPROACH |
|---|
|
TopNotesAbstractIntroductionHealth Outcomes FrameworkFeatures of the Core...Added Value of the...From Formation to ImplementationReferences |
|---|
Although the above discussions have been largely generic, the added value of the Health Outcomes Framework within the cancer arena can be illustrated through the use of the following thought experiment. Examples of instruments are taken from Table 1 to motivate this discussion although none of these instruments satisfy all of the criteria that have been set forth for the core instrument.
Suppose that we have a core instrument that satisfies the six components of the above operational definition and that this core was included in the National Health Interview Survey. If investigators were to add this instrument to a clinical trial designed to study two alternative interventions for adults with colorectal cancer in which the Functional Assessment of Cancer Therapy for Patients with Colorectal Cancer (FACT-C) (26) was also used, the core would have minimal impact on the patient's time to complete the questionnaires; for example, using the SF-12 as the core would add 2 minutes or less (21). Because documentation on how to administer the instrument, and how to score, analyze, and interpret the findings, would be publicly accessible, preferably via the internet, adding the core questionnaire would require little, if any, additional resources.
One of the advantages of the core-plus-specific-instrument design is the possibility that the core could contain one or more domains that are not part of the specific instrument, thereby supplying additional information. For example, if the core were an instrument such as the HALex, or SF-12, the health perceptions domain would complement information obtained from the FACT-C.
In addition to providing trial-specific information about perceived health, self-assessed health data could provide insights about long-term treatment effects because ratings of fair or poor health have been shown to predict mortality (65). For example, suppose that the colorectal cancer patients on intervention 1 had better response as measured by FACT-C but lower levels of self-assessed health on the core than did those on intervention 2. This would suggest that intervention 1 was the more effective treatment during the trial period but it may have a less favorable impact on long-term survival than would intervention 2. These differences in short- and long-term impacts illustrate a gain in information. In such an example, use of the analytical design proposed by the Health Outcomes Framework could generate hypotheses in addition to supporting immediate conclusions about treatment performance.
Further, outcomes based on the core instrument can be expressed in terms of quality-adjusted time. For studies with relatively short time horizons, this metric may be expressed in terms of months; for longer time horizons, in terms of years. For example, quality-adjusted life years (QALYs) for men and women with colorectal cancer can be compared with the QALYs of the general population to gain an expression of the time-related impact of colorectal cancer. Moreover, disaggregating health status and quality-of-life domains from the life expectancy component can indicate the relative impact of disease, whether it shortens the life span with, perhaps, relatively high quality of life, or lengthens life but with lower health status and quality of life. In addition, expressing health in terms of time, which is a familiar and intuitively meaningful metric, analogous to the number of life years, should be more readily understandable to patients and other stakeholders than are scores that lack a biological meaning.
The core instrument would also allow trial data to be compared with normative health scores for general population groups with the same socio-demographic and geographic characteristics (47,66). For example, to gain a better understanding of the impact of the disease and its treatment, the average level of health for patients with colorectal cancer who are more than 65 years of age might be compared with a similar group in the general population The greater the extent to which the patient and general population samples can be further stratified—for example, by ethnicity, years of schooling, and marital status—the more meaningful the comparisons. Inferences could be even more targeted if, for example, data about a diagnosis of colorectal cancer were included as part of a cancer-specific module included in the NHIS.
The on-going nature of surveys such as NHIS also allows investigators to understand changes in health that might be due to health care interventions and those that might be due to factors external to the health care system. In an analysis of population health from 1984 to 1994, the decline in health levels in 1991 and 1992 were found to be attributable, in part, to the economic recession that occurred in those years, suggesting that health levels are responsive to factors that are external to the health care system (67). Thus, the ability to compare trial outcomes with contemporaneous normative data is essential for interpretation of findings.
Being able to integrate clinical trial findings with data from a nationally representative sample, as included in the NHIS, enhances understanding of the population-based implications of the allocation of alternative colorectal cancer interventions. Without the ability to generalize to the overall population, the health effects of widespread use of an intervention can only be approximated using findings from narrowly selected groups, such as clinical trial participants or members of managed care organizations. If the selected study data happened to indicate a greater improvement in health outcomes than would actually be observed in the general population, the allocation would be less efficient than the policy maker intended. Further, an allocation that is based on less than full information may draw resources away from competing interventions, thereby having adverse consequences not only on efficiency but also on equity and inequality. On the other hand, using core data to generalize health outcomes findings for selected groups, including clinical trial data, to the overall population may result in more thorough understanding of treatment impact and lead to an allocation of resources to health that meets decision makers' goals for efficiency and equity while fitting within an acceptable level of health inequality.
|
FROM FORMATION TO IMPLEMENTATION |
|---|
|
TopNotesAbstractIntroductionHealth Outcomes FrameworkFeatures of the Core...Added Value of the...From Formation to ImplementationReferences |
|---|
The above thought experiment illustrates the value of adopting the Health Outcomes Framework. The movement from the formation of an idea to full implementation, however, will take time for both technical and political reasons. Thus, we propose that the use of a core instrument and analytic methods that incorporate vertical integration be addressed incrementally. Initially, resolution of some technical issues can proceed by using existing data. Also, by allocating resources now for the development of the necessary new methodologies needed for vertical integration, the progress toward fully implementing the HOF will be accelerated.
In the short run, currently available data can be used to empirically examine and quantify the nature of the information gained from vertical integration. For example, the self-assessed health status question "Would you say your health in general is excellent, very good, good, fair, or poor?" (which is now included in the NHIS questionnaire as well as in other data sources, including those that use the SF-36) might be used as a proxy for the core instrument. Although this is a single item, rather than a multi-domain instrument, and is not preference weighted, a sufficient number of publicly available data sets include this item that could be used for making comparisons between socio-demographic and disease-specific groups in the general population and similar groups in specific populations. Also, these analyses could help to identify methodologic and analytic issues that need to be addressed in order to use the core instrument more effectively across various types of studies.
Further, the self-assessed health status categories could be assigned scores to approximate interval scaling, in much the same way that preference weighting has been modeled to form overall scores from questionnaires that were designed to produce domain scores only (59-62). For example, mean HUI-3 scores for each of the self-assessed health status categories can be calculated using Statistics Canada data. Because these scores would be derived from a preference-based instrument, they could be used to calculate QALYs across the range of applications for which self-assessed health status data are available. Evaluating these QALYs in a variety of simulated policy situations could serve to inform a wide range of decision makers on the meaning of change in QALYs, thereby moving the use of this measure from the research community into more common usage in the policy arena. This focus on interpretability is essential for overcoming policy makers' lack of familiarity with a quality-adjusted summary measure that includes both quantity and quality of life.
In the longer run, experience from these analyses as well as from advances in instrument development that have been made in the last 10-15 years can guide the development of a new instrument for use as the core. For example, the formulation of the conceptual formulation of the instruments summarized in Table 1 was established at least 15 years ago. Since that time, new methods of instrument development have emerged, notably the use of focus groups to identify health domains that are important to the target population. Also, the development of item banks for use with computer adaptive testing offers a promising approach to collecting data that is targeted to each individual while maintaining minimal respondent burden. Although adaptive testing has been shown to be acceptable in educational testing, it has only recently begun to be applied to health outcomes measurement (68-69). Use of focus groups, adaptive testing and other methods of soliciting input from the general population are all essential if the core instrument is to reflect an individual's health concerns as inputs into the decision making process. Collecting preferences for health states from a representative sample of the general population is another way of including input from all members of society (and not a selected group that might be defined by disease or health insurance status) into the decision making process.
To get the most information from vertical integration, it is likely that new statistical designs and methods of data collection will be needed. For example, a key requirement of the HOF is the ability to apply health outcomes findings from clinical studies to the general population using national surveys. Yet, these surveys, typified by the NHIS, currently have a cross-sectional design. Those that do have a longitudinal design (such as the Medical Expenditures Panel Survey) often have sample sizes too small to permit relevant disease-specific comparisons. An alternative to the use of repeated cross-sectional data to approximate a longitudinal survey design might be to include a longitudinal panel survey as a subset of the NHIS. One would collect data from the same individuals over a minimum of two points in time, thereby allowing a more accurate estimate of variation associated with change in health than can be obtained from repeated cross-sectional data. The variance estimate from the longitudinal subset might then be applied to analyses using the full sample, especially when change in outcome is needed, e.g., to estimate marginal gain in health.
Integrating findings across studies that are conducted by public and private sponsors raises concerns about respondent privacy as well as product confidentiality, especially in the case of clinical trials sponsored by the pharmaceutical industry. Thus, methods may need to be developed to safeguard information, whether personal or proprietary and to ensure compliance with requirements of the Health Insurance Portability and Accountability Act (1996). This can foster full participation across all of the applications and is essential society is to fully gain the information that becomes possible when applying the Health Outcomes Framework to decision making.
The Health Outcomes Framework is intended to be a cooperative effort. At the national level, the NCI and Centers for Disease Control and Prevention need to exert a leadership role by including a core health status and quality-of-life instrument into all new cancer treatment and prevention trials and surveys and by inserting the core instrument into existing data collection activities as soon as possible. In addition, these federal agencies should actively encourage industry and academic investigators involved in cancer research to contribute by (1) including the core instrument in all of their studies; (2) analyzing the resultant data, taking vertical integration into account; and (3) reporting detailed findings to facilitate comparisons within and across studies and to enable modeling of treatment effects.
The integration of patient-reported data on individual health status and quality-of-life and societal-level data has the potential to inform decision making across all stakeholders. Having an instrument based on domains and preferences of the general population and using it in an ongoing national survey as well as in more specific studies can ensure that the information is truly population based—and thus broadly representative of racial and ethnic groups, socioeconomic variation, and the distribution of health conditions in society. This enfranchisement of all members of society, not just those with health insurance for example, can lead to more comprehensive and representative data for informing policy and ultimately for obtaining a more equitable distribution of health.
|
NOTES |
|---|
This article has benefited from the many helpful comments supplied by Joseph Lipscomb and Molla Donaldson.
1 Editor's note: SEER is a set of geographically defined,
population-based, central cancer registries in the United States, operated by
local nonprofit organizations under contract to the National Cancer Institute
(NCI). Registry data are submitted electronically without personal identifiers
to the NCI on a biannual basis, and the NCI makes the data available to the
public for scientific research. 
|
REFERENCES |
|---|
|
TopNotesAbstractIntroductionHealth Outcomes FrameworkFeatures of the Core...Added Value of the...From Formation to ImplementationReferences |
|---|
1 National Center for Health Statistics (NCHS). Health, United States, 2002, with chartbook on trends in the health of Americans. Hyattsville (MD): National Center for Health Statistics; 2002 .
2 National Heart, Lung, and Blood Institute (NHLBI). Fact book: fiscal year, 2002. NHLBI, National Institutes of Health, 2003. Available at www.nhlbi.nih.gov/about/02factbk.pdfp.53.
3 Bayer. Press release. www.bayeraspirin.com/press/release/2000/womencvd.pdf.
4 Rothstein MA, Kennedy K, Ritchie KJ, Pyle K. Are cancer patients subject to employment discrimination? Oncology 1995; 9: 1303 -6.[Medline]
5 Coates A, Porzsolt F, Osoba D. Quality of life in oncology practice: prognostic value of EORTC QLQ-C30 scores in patients with advanced malignancy. Eur J Cancer 1997; 33: 1025 -30.[CrossRef][Web of Science][Medline]
6 Benson V, Marano M. Current estimates from the National Health Interview Survey, 1995. National Center for Health Statistics. Vital and Health Statistics Series 10(199), 1998. DHHS 98-1527. National Center for Health Statistics, Hyattsville, MD.
7 Litwin MS, Hays RD, Fink A, Ganz PA, Leake B, et al. The UCLA Prostate Cancer Index: Development, reliability, and validity of a health-related quality of life measure. Med Care 1998; 36: 1002 -12.[CrossRef][Web of Science][Medline]
8 National Center for Health Statistics (NCHS). Plan and operation of the Third National Health and Nutrition Survey 1988-1994. DHHS 98-1308. NCHS, Hyattsville, MD.
9 Viale PH. Direct-to-consumer advertising of prescription medications: Implications for patients with cancer. Oncol Nurs Forum 29(3): 505-13, 2002 .[Medline]
10 Wei JT, Uzzo RG. Shared decision-making strategies for early prostate cancer. Semin Urol Oncol 20(1): 74 -8, 2002.[CrossRef][Medline]
11 Lipscomb J, Gotay CC, Snyder C. Outcomes assessment in cancer: measures, methods, and applications. Cambridge (UK): Cambridge University Press; 2004. In press.
12 Lipscomb J, Snyder C. The outcomes of cancer outcomes research. Focusing on the National Cancer Institute's Quality-of-Care Initiative. Med Care 40(6 Suppl): III3 -10, 2000.
13 Osoba D. A Taxonomy of the uses of health-related quality-of-life instruments in cancer care and the clinical meaningfulness of the results. Med Care 40(6 Suppl): III31 -38, 2000.
14 Surveillance, Epidemiology and End Results maintained by the National Cancer Institute. Available at www.seer.cancer.gov.
15 Clinical trials registry maintained by the National Institutes of Health. Available at http://clinicaltrials.gov.
16 Links to Statistics and Statistical Agencies. Available at www.fedstats.gov.
17 Fryback DG, Thornbury JR. The efficacy of diagnostic imaging. Med Decis Making 1991.
18 Hernandez LM, Durch JS, Blazer DG, Hoverman IV. Gulf War veterans: measuring health. Washington (D.C.): National Academy Press, 1999 .
19 Goldie SJ, Kuntz KM, Weinstein MC, Freedberg KA, Palefsky JM. Cost-effectiveness of screening for anal squamous intraepithelial lesions and anal cancer in human immunodeficiency virus-negative homosexual and bisexual men. Am J Med 2000;108(8): 674 -5.[CrossRef][Web of Science][Medline]
20 Kovar MG, Poe GS. The National Health Interview Survey Design, 1973-1984, and Procedures, 1975-1983. Series 1, No. 18. DHHS 85-1318. National Center for Health Statistics, 1985. National Center for Health Statistics, Hyattsville, MD.
21 Ware JE, Jr, Snow KK, Kosinski M, Gandek B. SF-36 Health survey: manual and interpretation guide. Boston: Health Institute. 1993 .
22 Bergner M, Bobbitt RA, Carter WB, Gilson BS. The Sickness Impact Profile: development and final revision of a health status measure. Med Care 1981;19: 787 -805.[Web of Science][Medline]
23 Cella DF, Tulsky DS, Gray G, Sarafian B, Linn E, et al. The
Functional Assessment of Cancer Therapy Scale: development and validation of
the general measure. J Clin Oncol 1993; 11: 570
-9.
24 Aaronson NK, Ahmedzai S, Bergman B, Bullinger M, Cull A, et al. The
European Organization for Research and Treatment of Cancer QLQ-C30: a quality
of life instrument for use in international clinical trials in oncology. J Natl Cancer Inst
1993; 85(5): 365
-76.
25 Sprangers MAG, Cull A, Groenvold M, Bjordal K, Blazeby J, Aaronson NK. The European Organization for Research and Treatment of Cancer approach to developing questionnaire modules: an update and overview. Qual Life Res 1998;7: 291 -300.[Web of Science][Medline]
26 Cella DF. FACIT MANUAL: manual of the Functional Assessment of Chronic Illness Therapy (FACIT) Measurement System, Version 4. Evanston (IL): Evanston Northwestern Healthcare and Northwestern University, 1997 .
27 Cella DF, Bonomi AE. The Functional Assessment of Cancer Therapy (FACT) and Functional Assessment of HIV Infection (FAHI) quality of life measurement system in quality of life and pharmacoeconomics in clinical trials. 2nd ed. Spilker B, editor. Philadelphia (PA): Lippincott-Raven Publishers, pp. 203-214, 1996 .
28 Levine MN, Guyatt GH, Gent M, De Pauw S, Goodyear MD, Hryniuk WM, et al. Quality of life in stage II breast cancer: an instrument for clinical trials. J Clin Oncol 1988; 6: 1798 -810.[Abstract]
29 List MA, Ritter-Sterr C, Lansky SB. A performance status scale for head and neck cancer patients. Cancer 1990; 66: 564 -9.[CrossRef][Web of Science][Medline]
30 Tsavaris NB, Koufos C, Katsikas M, Dimitrakopoulos A, Athanasiou E, Lanardaki G. Antiemetic prophylaxis with odansetron and methylprednisolone vs metoclopramide and methylprednisolone in mild and moderately emetogenic chemotherapy. J Pain Symptom Manage 1999; 18(3): 218 -22.[CrossRef][Web of Science][Medline]
31 Grunbaum JA, Kann L, Kinchen SA, Williams B, Ross JG, Lowry R, et al. Youth Risk Behavior Surveillance—United States, 2001. MMWR Surveill Summ 51(4): 1 -62, 2002.
32 Patrick DL, Erickson P. Health status and health policy: quality of life in health care evaluation and resource allocation. New York (NY): Oxford University Press; 1993.
33 Cohen JW, Monheit AC, Beauregard KM, Cohen SB, Lefkowitz DC, Potter DE, et al. The Medical Expenditure Panel Survey: a national health information resource. Inquiry 1996-1997; 33(4): 373 -89.[CrossRef]
34 Information about general population surveys from the National Center for Health Statistics. Available at www.cdc.gov/nchs.
35 Erickson P, Wilson R, Shannon, I. Years of healthy life: Statistical Note 7. Hyattsville (MD): National Center for Health Statistics, Centers for Disease Control and Prevention, 1995 .
36 Molla M, Wagener DK, Madans JH. Summary measures of population health: methods for calculating healthy life expectancy. Statistical Notes 21, National Center for Health Statistics, 2001 . National Center for Health Statistics, Hyattsville, MD.
37 Fletcher RH, Fletcher SW, Wagner EH. Clinical epidemiology: the essentials. Baltimore (MD): Williams and Wilkins, 1988 .
38 Lilienfeld AM. Foundations of epidemiology. New York: Oxford University Press, 1976.
39 MacMahon B, Pugh TF. Epidemiology: principles and methods. Boston (MA): Little, Brown and Company, 1970 .
40 Michael YL, Kawachi I, Berkman LF, Holmes MD, Colditz GA. The persistent impact of breast carcinoma on functional health status: prospective evidence from the Nurses' Health Study. Cancer 2000; 89(11): 2176 -86.[CrossRef][Web of Science][Medline]
41 Bernhard J, Hurny C, Coates AS, Peterson HF, Castiglione-Gertsch M, Gelber RD, et al. Factors affecting baseline quality of life in two international adjuvant breast cancer trials. International Breast Cancer Study Group (IBCSG). Br J Cancer 1998; 78(5): 686 -93.[Web of Science][Medline]
42 Berzon R. Understanding and using health-related quality of life instruments within clinical studies in quality of life assessment in clinical trials: methods and practice. Staquet MJ, Hays RD, Fayers PM, editors. Oxford (U.K.): Oxford University Press, pp. 3 -15, 1998.
43 Guyatt GH, Jaeschke R, Feeny DH, Patrick DL. Measurements in clinical trials: choosing the right approach in quality of life and pharmacoeconomics in clinical trials. Spilker B, editor. Philadelphia (PA): Lippincott-Raven Publishers, pp. 41-8, 1996 .
44 Osoba, D. Guidelines for measuring health-related quality of life in clinical trials in quality of life assessment in clinical trials: methods and practice. Staquet MJ, Hays RD, Fayers PM, editors. Oxford (U.K.): Oxford University Press, pp. 19-35, 1998 .
45 Potosky AL, Harlan LC, Stanford JL, Gilliland FD, Hamilton AS,
Albertsen PC, et al. Prostate cancer practice patterns and quality of life:
the Prostate Cancer Outcomes Study. J Natl Cancer Inst 1999; 91(20): 1719
-24.
46 Kind P. The EuroQoL instrument: an index of health-related quality of life in quality of life and pharmacoeconomics in clinical trials. 2nd ed. Spilker B, editor. Philadelphia (PA): Lippincott-Raven Publishers, pp. 191-201, 1996 .
47 Uyl-de Groot CA, Hagenbeek A, Verdonck LF, Lowenberg B, Rutten FF. Cost-effectiveness of ABMT in comparison with CHOP chemotherapy in patients with intermediate- and high-grade malignant non-Hodgkin's lymphoma (NHL). Bone Marrow Transplant 1995; 16(3): 463 -70.[Web of Science][Medline]
48 Gold MR, Siegel JE, Russell LB, Weinstein MC, editors. Cost-effectiveness in health and medicine . New York: Oxford University Press, 1996.
49 Berthelot JM, Will BP, Evans WK, Coyle D, Earle CC, Bordeleau, L.
Decision framework for chemotherapeutic interventions for metastatic
non-small-cell lung cancer. J Natl Cancer Inst 2000; 92(16): 1321
-29.
50 Will BP, Berthelot JM, Nobrega KM, Flanagan W, Evans WK. Canada's Population Health Model (POHEM): a tool for performing economic evaluations of cancer control programs. Eur J Cancer 2001; 37(14): 1797 -804.[CrossRef][Web of Science][Medline]
51 Litwin MS, Hays RD, Fink A, Ganz PA, Leake B, Leach GE, et al.
Quality-of-life outcomes in men treated for localized prostate cancer. JAMA
1995;273(2): 129
-35.
52 Ware JE. Methodology in behavioral and psychosocial cancer research. Conceptualizing disease impact and treatment outcomes. Cancer 1984;53 (10 Suppl): 2316 -23.[Web of Science][Medline]
53 Bush JW. General Health Policy Model/Quality of Well-Being (QWB) Scale in Assessment of quality of life in clinical trials of cardiovascular therapies. Wenger NK, Mattson ME, Furberg CD, et al., editors. New York: Le Jacq Publishing, pp. 189-199, 360-1, 1984.
54 Torrance GW, Furlong W, Feeny D, Boyle M. Multi-attribute preference functions: Health Utilities Index. Pharmacoeconomics 1995; 7(6): 503 -20.[Web of Science][Medline]
55 Roberge R, Berthelot JM, Wolfson M. The Health Utility Index: measuring health differences the Ontario by socioeconomic status. Health Rep 1995;7(2): 25 -32.[Medline]
56 Healthy People 2000. Washington (DC): Department of Health and Human Services, 1991.
57 Cooper JK, Kohlmann T, Michael JA, Haffer SC, Stevic M. Health outcomes. New quality measures for Medicare. Int J Qual Health Care 13(1): 9 -16, 2001.
58 World Health Organization. World health surveys. Available at www3.who.int/whs/.
59 Brazier J, Roberts J, Deverill M. The estimation of a preference-based measure of health from the SF-36. J Health Econ 21(2): 271 -92, 2002.
60 Erickson P, Kendall EA, Anderson JP, Kaplan RM. Using composite health status measures to assess the nation's health. Med Care 1989; 27(3 Suppl): S66 -76.[CrossRef][Web of Science][Medline]
61 Fryback DG, Lawrence WF, Martin PA, Klein R, Klein BE. Predicting
Quality of Well-being scores from the SF-36: results from the Beaver Dam
Health Outcomes Study. Med Decis Making 1997; 17(1): 1
-9.
62 Chancellor JV, Coyle D, Drummond MF. Constructing health state preference values from descriptive quality of life outcomes: mission impossible? Qual Life Res 1997; 6: 159 -68.[Web of Science][Medline]
63 Drummond MF, O'Brien B, Stoddart GL, Torrance GW. Methods for the economic evaluation of health care programmes 2nd ed. Oxford (U.K.): Oxford University Press, 1998.
64 King G, Murray CJL, Salomon JA, Tandon J. Enhancing the validity and cross-cultural comparability of survey research. Available at http://gking.harvard.edu.
65 Idler EL, Russell LB, Davis D. Survival, functional limitations,
and self-rated health in NHANES I Epidemiologic Follow-up Study, 1992. First
National Health and Nutrition Examination Survey. Am J
Epidemiol 2000;152(9): 874
-83.
66 Anastasi A. Psychological testing. 6th ed. New York (NY): Macmillan, 1988.
67 Erickson P. Trends in population-based health-related quality of life: Implications for interpretation of change. Value in Health 2001;4(6): 418 .
68 Wainer H. Computerized adaptive testing: a primer. Hillsdale (NJ): Lawrence Erlbaum Associates, 1990.
69 Ware JE, Jr, Bjorner JB, Kosinski M. Practical implications of item response theory and computerized adaptive testing: a brief summary of ongoing studies of widely used headache impact scales. Med Care 2000; 38(9 Suppl): II73 -82.[Medline]
70 Nelson E, Wasson JH, Johnson DJ. Dartmouth COOP functional health assessment charts: brief measures for clinical practice in quality of life and pharmacoeconomics in clinical trials. 2nd ed. Spilker B, editor) Philadelphia (PA): Lippincott-Raven Publishers, pp. 161 -8, 1996.
71 McEwen J, McKenna S P. Nottingham Health Profile in quality of life and pharmacoeconomics in clinical trials. 2nd ed. Spilker B, editor. Philadelphia (PA): Lippincott-Raven Publishers, pp. 281 -6, 1996.
72 Erickson P, Scott J. The On-Line Guide to Quality-of-Life Assessment (OLGA): resource for selecting quality-of-life assessments. In Walker SR, Rosser RM, editors. Quality of life assessment: key issues in the 1990s. Dordrecht (The Netherlands): Kluwer Academic Publishers, 221 -32, 1993.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. Lipscomb, M. S. Donaldson, and R. A. Hiatt Cancer Outcomes Research and the Arenas of Application J Natl Cancer Inst Monographs, October 1, 2004; 2004(33): 1 - 7. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||