CER is concerned with the conduct and synthesis of systematic research comparing different interventions and strategies to prevent, diagnose, treat and monitor health conditions. Its purpose is to inform patients, providers and decision-makers about which interventions are most effective for patients under specific circumstances [17
]. CER may employ observational or clinical trial methodologies to compare strategies for care available in typical healthcare settings, provided by typical healthcare providers, addressing a comprehensive array of health-related outcomes. In general, outcomes are favored to be patient centered and the research should be conducted in ‘real world’ diverse and representative patient populations. Translational research includes three stages: T1 is the classic bench-to-bedside (or rather human) translation; T2 constitutes efficacy research and clinical trials; and T3 is translational research into health delivery, effectiveness and implementation that explores general medical practice [23
]. Thus, CER methods have the opportunity to evolve toward the translation of research findings that emerge from traditional efficacy research, in which comparisons between alternatives (including no treatment or placebo) are conducted in controlled clinical trials within narrowly selected patient populations (i.e., those who are presupposed to benefit) and in resource-intensive research settings in which interventions are administered by investigators and their research staff. Comparative efficacy studies include research to evaluate the effects of interventions, thus measuring outcomes related to underlying disease mechanisms or clinical changes to the disease from interventions. In short, comparative efficacy studies are critical to answering questions such as ‘Can this intervention work?’ and ‘How does this intervention work?’ under ideal and carefully controlled research conditions, whereas comparative effectiveness studies may address translational questions such as ‘Which interventions when translated into practice result in care improvements and increase the likelihood of health benefits?’ Ultimately, it is the efficacy research that leads to innovations and effectiveness research that leads to practice change and improved health ().
A straightforward approach to CER would be to apply traditional efficacy methods such as the strong research design of the RCT. However, not every important CER question can be answered by large RCTs. Even pragmatic strategies that attempt to minimize investigator control of interventions, reduce exclusions and maximize generalizability have limitations due to resource availability, sufficient patients or sample size for enrollment, and lengthy time needed for outcome evaluation and follow-up. As trial synthesis and meta-analysis currently rely on evidence from randomized clinical trials, one can anticipate tension in how to apply these methods to CER with sufficient representativeness and generalizability that will require expanded intermediary study designs. New approaches to observational designs should thus be part of the methodological development to realize the promise of CER and translational science. Synergy of these approaches, such as large database development, with health information technology mandates may create unique opportunities with a national evolution to electronic medical records and interoperability. For COPD, novel opportunities for translational and effectiveness methodologies may evolve to effectively and efficiently evaluate CER questions and guide subsequent care innovation.
One example of this kind of research that can ultimately also help guide implementation and care improvement can come from conducting CER within linked registries that have been organized to monitor and guide quality of care [24
]. By conducting translational and CER research across linked registries for patients and providers in real world conditions, researchers and policymakers can explore CER across a wide representative sample of US hospitals and monitor effects of healthcare policy and care improvement efforts serially over time. Methods that employ sophisticated modeling techniques to control for confounding and selection bias have evolved and the large samples reduce opportunities for error [26
]. In addition, subgroups such as those in racial and ethnic minorities, the socioeconomically disadvantaged and those with multimorbid conditions can be assessed for differences in responses, risks and benefits.
In some circumstances, confounding and selection bias cannot be definitively excluded in observational designs and thus large randomized clinical trials will be needed to assess the comparative effectiveness of some treatment alternatives. However, even in these trials, observational studies can guide the development and design of pragmatic RCTs by providing essential preliminary data on relevant comparators, outcomes and estimations of sample size. Even when a definitive RCT answers a particular CER question, databases and registries will be beneficial for studying implementation and practice change. Such resources will additionally enable ongoing surveillance and quality monitoring across patient-focused outcomes that can further guide practice changes. Large, representative, multisite registries, preferably enhanced in the future with clinical information derived from increased use of electronic medical records, may provide the necessary data monitoring of the effects of practice change, including assessing for harm in real world settings. With sustained funding, registry-conducted CER could document care delivery and provide quality measurement that can assist ongoing care enhancement and compliment experimental design by informing future clinical trials.
Mixed methods and quasi-experimental designs should also be considered as part of the comprehensive approach to CER in COPD [26
]. Research in real world settings may not always have the luxury of random assignment or it may be difficult to devise a control group because of ethical objections to withholding an intervention or components of the intervention from one of the study groups. In addition, the researcher may not have the level of control of the delivery of the intervention in the real world setting that can be found in an experimental setting. Quasi-experimental designs can be a strong alternative to randomized clinical trials when research is conducted in real world settings. The trade-off involved with conducting research in the sometimes difficult real world setting is increased external validity and generalizability to other real world clinical settings.
Quasi-experimental designs, such as the regression discontinuity design, have been shown, when analyzed correctly, to lead to unbiased estimates of the difference between groups [26
]. In this design, assignment to groups is based on some continuous variable; individuals who exceed some threshold are assigned to the experimental group and those that do not are assigned to the control group. This type of design mimics what happens in the clinical setting, where, for example, patients are prescribed blood pressure medicine only if their blood pressure exceeds safe limits. Interrupted time series and nonequivalent control group designs are other useful approaches in community-based research, in which outcomes are followed over time [30
]. While not as strong as the regression discontinuity or interrupted time series designs, the nonequivalent control group design can be strengthened through the use of certain design elements and statistical analyses [27
]. In this design, naturally occurring groups of participants are compared. The groups could be, for example, patients treated at different clinics or treated by different healthcare teams.
Given the impracticality of testing every clinical intervention in comparative pragmatic trials, we advocate expanding methodology that includes the use of linked registries, serially performed rigorous observational effectiveness analyses and quasi-experimental designs for CER in COPD. To further ensure that the benefits promised by CER (whether by observational, RCT or quasi-experimental design) are indeed realized, we would further recommend follow-up studies to evaluate time trends in quality metrics, health outcomes to monitor for possible harms and subgroup variations, and costs analyses to be considered as components of implementation.