Cancer Outcomes Research and the Arenas of Application

In 2004, an estimated 1.5 million persons in the United States will be diagnosed with cancer; about 560 000 will die from cancer; and more than 9.5 million will be undergoing curative treatment, coping with progressive disease, or living free of cancer after successful therapy (1). But many of these survivors still feel the aftershocks and downstream side effects arising from diagnosis and treatment, and many are fearful of recurrence. Substantial progress in reducing the suffering and death caused by cancer is being pursued by the National Cancer Institute (NCI) and cancer agencies and organizations worldwide through a variety of initiatives, programs, and projects. At the NCI, these efforts emphasize the joint importance of basic and applied scientific discovery, the development and testing of promising interventions, and the delivery of quality care to prevent, detect, and treat cancer and to improve the length and quality of life of cancer survivors (2). It is therefore vital for decision makers, at all levels, to have a firm understanding of just how effectively the fruits of discovery and development are being applied to enhance cancer care delivery in ways that reduce suffering and death. The scientific pursuit of such decision-relevant information is the central business of cancer outcomes research. The overall aim of this Monograph is to provide an empirically grounded review and evaluation of the peer-reviewed literature in cancer outcomes research. The intent is to identify both important recent contributions and the challenges that remain in bringing scientifically sound information to bear in cancer care decision making. There is, as yet, no consensus definition of outcomes research, much less cancer outcomes research. But recent statements by the U.S. Agency for Healthcare Research and Quality (AHRQ) and the NCI convey a common sense of purpose and scope of application. According to AHRQ, “... outcomes research seeks to understand the end results of particular health care practices and interventions. End results include effects that people experience and care about, such as change in the ability to function. In particular, for individuals with chronic conditions—where cure is not always possible—end results include quality of life as well as mortality” (3,4). At NCI, outcomes research “describes, interprets, and predicts the impact of various influences, especially (but not exclusively) interventions, on `final' endpoints that matter to decision makers” (5). These decision makers may include patients, families, individuals at risk for cancer, providers, private and public payers and purchasers of cancer care, regulatory agencies, health care accrediting organizations, and society at large. In cancer, final endpoints (outcomes) include such traditional and important biomedical outcomes as survival and disease-free survival, but also health-related quality of life (HRQOL), perceptions of and satisfaction with health care, and economic burden. Final outcomes are distinguished from “ intermediate” outcomes (e.g., whether the individual was screened for cancer, or quit smoking, or received appropriate adjuvant therapy following cancer surgery) and “clinical” outcomes (e.g., whether the patient's tumor shrank or disappeared, or whether the tumor recurred). Although such measures are frequently pivotal in assessing the proximate success of particular interventions, the ultimate test is, or should be, whether improvement in the clinical or intermediate outcome predicts success in improving final outcomes.1 A potentially useful way to conceptualize the idea of final outcomes is to invoke the economist's traditional model of rational choice: the decision maker acts so as to maximize “utility” subject to whatever constraints on behavior are relevant and binding. In symbols, the decision maker maximizes a function U(H, x) subject to specified constraints, where H is a vector of health-related outcomes and x is a vector of all other goods that affect utility (happiness). Within this stylized framework, H comprises what we have termed final outcomes, and, likewise, all relevant final outcomes are in H. Intermediate and clinical outcomes can then be modeled as influencing the generation of H via the production function H = f(I, C, Y), where I and C are relevant intermediate and clinical outcomes, respectively, and Y is a vector of all other factors influencing final outcomes. Further, the potential interplay between I and C could be modeled, recognizing that achieving particular intermediate outcomes influences clinical outcomes. All this underscores the distinction between final outcomes, which are the “ ends” we seek, and intermediate outcomes, which are means to those ends. Note also that what we have termed intermediate outcomes are closely related to what quality-of-care researchers call “process measures” of quality. For the three intermediate measure examples in the text, we can construct the corresponding QOC process measures: percent of eligible population screened for (a particular) cancer, percent of smokers who quit, and percent of patients eligible for adjuvant therapy following surgery who elect to receive it and do. Prominent biomedical outcomes like survival and disease-free survival have long been employed both in clinical investigations and in widely circulated reports on progress against the cancer burden at the population level. Although pockets of controversy remain about matters of definition and measurement (for example, the ongoing discussion about cause-specific versus relative survival measures), most researchers and many in the lay public would likely believe they have a basic understanding of what these particular final outcomes mean, if not also how to construct them.2 On the other hand, there remains much debate about, and wide variations in the use of patient-reported outcomes like HRQOL and measures of patient satisfaction and economic burden. Consequently, this Monograph focuses largely on the actual, and potential, contributions of such patient-reported outcomes on decision making along the cancer continuum, which includes prevention, early detection, diagnosis, treatment, life after a cancer diagnosis (survivorship), and end-of-life care. As will be seen, the bulk of the existing outcomes research literature is devoted to screening, diagnosis, and treatment applications. To be sure, there is much ongoing discussion and different viewpoints about the appropriate role of such prominent surrogate endpoints as rate of tumor progression in cancer treatment trials. But these are what we would term clinical, not final, biomedical outcomes. For outcomes research to achieve its potential to enhance care delivery, there are at least three prerequisites: 1) technically sound and decision relevant outcome measures; 2) persuasive evidence about the impact of interventions on those outcomes; 3) the capacity, determination, and ingenuity to translate findings into useful information for decision making (5,6). There are clear interconnections between outcomes research and health services research, though there is no consensus yet about the precise relationship (owing in part to varying definitions of both enterprises). The National Library of Medicine MESH heading “health services research” connotes “the integration of epidemiologic, sociological, economic, and other analytic sciences in the study of health services...” (7). To date, there is no MESH heading for outcomes research; the closest option appears to be “outcome assessment (health care),” although this appears to encompass not only what we term final outcomes but also, “... abnormal states (such as elevated blood pressure)” (7). In their detailed review of the history and role of outcomes research in oncology, Lee et al. (7) appear to settle upon a definition that does not differ significantly from that of health services research, as spelled out by the Association for Health Services Research (AHSR) in the late 1990s.3 In the introduction to a volume of papers discussing future directions in cancer outcomes research, Ramsey (10) concluded that it may be premature to impose a consensus definition “ until we know what matters to whom”4. Academy Health, the successor organization to AHSR, now defines health services research similarly to NLM, and it defines outcomes research in the same spirit as AHRQ and NCI (9). Yet another perspective on these matters of definition and function comes from the NCI itself, where its Outcomes Research Branch (ORB) (11) and Health Services and Economics Branch (HSEB) (12) operate in parallel and complementary ways within the Applied Research Program (ARP) (13) of the Division of Cancer Control and Population Sciences (14). In terms of the three prerequisites for effective outcomes research cited earlier, the ORB has focused substantially on assessing and improving outcome measures. HSEB (and ARP more broadly) has initiated a number of projects to strengthen the evidence base and analyze the impact of interventions on outcomes. ORB, with support from HSEB and many units and organizations within and beyond the NCI, has devoted substantial resources to synthesizing and translating evidence for cancer care decision making. To pursue its goal of improving the quality of cancer care by strengthening the scientific basis for public and private decision making, NCI has created a research and applications agenda that is supported by a range of outcomes research and health services research projects, data resources, and collaborations that extend well beyond the NCI itself (15). These various efforts are discussed in the Monograph's final article. In designing its agenda, NCI has been guided by the Institute of Medicine's well-known definition, “Quality of care is the degree to which health services for individuals and populations increase the likelihood of desired health outcomes and are consistent with current professional knowledge” (16). In this definition, desired means desired by individuals receiving interventions—underscoring the idea of patient-centered outcomes. NCI's comparatively broad conceptualization of outcomes research, and its recognition of the interplay between outcomes research and HSR, underscores the idea that increasing the likelihood of desired health outcomes requires sustained and well-coordinated progress in several areas. Specifically, we should measure outcomes that matter; investigate the impact of interventions on these outcomes; use the findings to improve quality of cancer care in the community; and monitor progress over time to identify successes, shortcomings, knowledge gaps, and new opportunities for research and application. As Donabedian (17) emphasized years ago, quality of care may be indexed, alternatively, by structural variables (the quantity and quality of inputs), process variables (what is done, or not done, to the patient), or outcomes (by which he meant final outcomes, as defined here). But in the end, the validity of structural or process variables as quality measures hinges on the strength of the evidence that they are predictive of outcomes that matter. As AHRQ has noted, “ By linking the care people get to the outcomes they experience, outcomes research has become the key to developing better ways to monitor and improve the quality of care” (3). With these functional definitions and policy aims in mind, we turn now to the specific framework employed in this Monograph for categorizing and characterizing the applications of cancer outcome measures, as seen in the peer-review literature over the decade of the 1990s. This framework provides a functionally useful backdrop for our efforts to present a systematic review and evaluation of a major segment of that literature and inferences about future directions for cancer outcomes research. The framework adopted here recognizes three broad categories, or arenas, for the application of cancer outcome measures5 (Table 1). The macro-meso-micro rubric has been previously used by both Sutherland and Till (18) and Osoba (19), but in each of these papers the terms are defined somewhat differently than in Table 1 here. Arenas of applications for cancer outcomes measures Arenas of applications for cancer outcomes measures Macro. Population surveillance of trends in cancer-related outcomes and progress against the cancer burden—including survival but ideally also capturing selected patient-reported outcomes. Meso. Descriptive and analytical studies to understand the impact of cancer, patterns of service use, and effects of intervention on cancer-related outcomes. Included are a diverse range of analyses: Randomized clinical trials examining intervention efficacy. Observational investigations of the effectiveness of interventions in real-world, community practice and of the burden of cancer on patients, survivors, and families. Patterns-of-care studies that not only examine variations in service utilization, but also relate these variations to differences in outcomes across population groups; and studies to monitor the quality of cancer care by tracking adherence of individuals or populations to consensus recommendations about appropriate care. Clinical modeling, evaluation, and priority-setting analyses, which typically integrate and synthesize information on outcomes (and related explanatory variables) from a variety of sources to identify the best course(s) of action for the cancer-related decision at issue. Included here are clinical decision modeling analyses to select an optimal intervention for the population (or subgroup) of interest; cost-effectiveness and cost-benefit analyses (which may, or may not, employ a decision-model); and studies to evaluate existing cancer intervention programs or to guide the establishment of new ones. Micro. Patient-clinician decision making enhanced by the use of patient-reported outcome measures, risk and outcome prediction models, or other tools to improve the quantity and quality of information available at the bedside, in the clinic and office, and in electronic (telephone or computer-assisted) communications. This Monograph presents 10 invited papers,6 that are distributed across the arenas of application as follows: one macro, seven meso, one micro, and one paper that suggests how to connect the schema in Table 1 to a more general “ health outcomes framework” that envisions a creative interplay among applications at all levels. Five of the seven meso papers examine outcomes research applications, in turn, for the four most prevalent adult cancer disease sites (breast, colorectal, lung, and prostate) and one major childhood cancer (acute lymphoblastic leukemia). Two additional meso papers provide critical assessments of the application of HRQOL measures and economic cost measures, respectively, within and across cancer disease sites. The Monograph concludes with a paper by NCI staff scientists on future directions for cancer outcomes research. of these invited papers was supported through from the National Cancer Institute and was to the review process that is for all of the National Cancer Institute Two of this and by the of the National Cancer Institute as the scientific for the is a of these invited papers within the studies to variations in important outcomes to decision makers in and at populations In particular, how these studies typically monitor trends over trends within or across units (e.g., within the United across across trends among population (e.g., in outcomes the burden of cancer (e.g., survival functional measures), or the use of services to that burden (e.g., cancer surveillance studies are not for inferences about the of variations in population outcomes. But frequently they important (e.g., and factors but in population in patterns of cancer that can be more through concludes there is a to traditional biomedical like survival, with patient-reported outcomes, like health-related quality of to improve our ability to monitor progress against the as well as caused by There is no intent here to that cancer as it is and is with cancer outcomes research. a range of outcome measures important to decision makers can and should be part of what is being and by the surveillance although diverse in and specific have in common the broad aim of understanding (and also the impact of especially interventions, on outcomes that matter. these analyses have as the of the individual cancer or at studies may to test by may that support applications they may provide important findings that into decision making about cancer care and and and a range of quality improvement clinical use, and The et al. and et al. and et al. with a common literature and general for As the it clear there differences across colorectal, lung, and childhood in the of the available relevant how best to for and how to findings most Consequently, the papers here have been to these important common paper on and a we in several with the each paper the literature the from through each paper focuses largely on applications of patient-centered outcome as health-related quality of patient and economic the of modeling and economic that examine only and other biomedical endpoints not Although the of papers differ in their particular and common the outcome measures can provide information for the impact of interventions, the on patient of that reduce side There is consensus about how to measure health-related quality of life for of the cancer here. are used across studies (and within a our to out and other of findings is this of consensus about the most appropriate HRQOL measure is a of about the for a there is wide in how the economic of cancer and cancer interventions are defined and with HRQOL many studies measure patient perceptions of the quality of cancer care or their satisfaction with of the care should be to patient for outcomes, for those treatment or where HRQOL effects are and important to decision making. There is substantial in the scientific quality of studies examining patient-centered outcomes, in clinical trials where a biomedical outcome (e.g., is the and the of data and For findings about HRQOL effects to be by decision makers, study and measure and analyze these outcomes with the same scientific as more biomedical In a of the use of HRQOL measures across cancer disease to understanding the of within specific of applications. For example, better of the between a measure like the and measures, like the and in cancer treatment trials would enhance understanding of the contributions of each the ability to whether a HRQOL change is would be by the of data on the in diverse In to the of HRQOL and the burden of cancer and cancer care the idea of economic cost has been long and in to and that there are no widely to become the for cost This has in in how cancer are defined and which the of findings across These developing and testing and data for studies use outcome measurement to and improve decision making. the literature on the use, and effectiveness of patient-centered measures like HRQOL in on the progression of disease and the impact of and improving the quality of cancer care. Although there have been and reports in the literature of successful applications, HRQOL (and patient-centered outcomes are not yet used in practice in the United The way is not or that HRQOL the should be on by improving information in the HRQOL assessment into the of care and of cancer care to reduce and improve to and use HRQOL The Monograph's invited papers with for a “health outcomes framework” to macro, meso, and applications by a of HRQOL for use at all levels, at each by additional The aim is to improve the and of HRQOL within each of whether the is a disease or a of The framework could as to cancer outcome measure of The key is that the measures are used in all arenas, the within one from that In turn, the broad use of such measures would a much better understanding of health-related quality of life across populations and In The Outcomes of Outcomes Research at by what is now the U.S. Agency for Healthcare Research and useful framework was for progress and future research directions The framework the of a with the impact of research on final outcomes as one Specifically, one from research that the knowledge and the quality of future research but does not or final outcomes; to research that does practice to research that influences care delivery to research on patient outcomes in the community is Outcomes Research from outcomes of outcomes research at by and an of from over the the concluded that outcomes research studies been most successful in health care delivery developing tools for patient-centered outcomes and economic and future research There in such as clinical decision modeling, and cost-effectiveness the noted, there that research findings or final outcomes. prominent success was the use of data and quality improvement to achieve a substantial in surgery as well as the findings in this Monograph's the importance of developing and a cancer outcomes research agenda that to the Although 1 that such are they are more likely and is, successful efforts at each will not only at the but also at that and at It is frequently that improving the quality of care requires not only to the of appropriate interventions but also to the and of services the means decision making that is to and as well as to an appropriate as In the end, the aim is to and provide the at the time and to the decision across the arenas of application. In the Monograph's final NCI staff scientists upon their and by the 10 papers that to provide about the current of the and future directions in cancer outcomes research the has is from As the Monograph's papers there has been much progress in recent years understanding the impact of interventions on final outcomes, the patient-centered outcomes emphasized here. the literature is and our knowledge more than about how such information is used (or or in about practice 2) or the delivery of cancer care The challenges in a understanding of the ways in which information at each comes to decision making, and the outcomes that matter to decision makers, to a new “ for outcomes research. to this in the Monograph's final paper with a cancer how the has been also the challenges in the of evidence that research to improvement in outcomes.