Opinion VIEWPOINT Embedding Cardiovascular Research Into Practice Robert M. Califf, MD Cardiovascular medicine currently lacks high-quality dence) so that each health care encounter contributes Division of Cardiology, evidence needed to inform many important decisions incrementally to informing practice. Adopting better, Department of that patients, physicians, and other health care practi- more accessible longitudinal electronic health data pre- Medicine, Duke tioners and health systems must make regarding the pre- sents new possibilities for understanding care delivery University School of 1 Medicine, Durham, vention and treatment of cardiovascular disease. At the and outcomes. North Carolina, and same time, clinical research costs are escalating while In addition, a recent report5 provided a novel Duke Translational government and industry funding is increasingly con- framework for ethical oversight of learning activities, Medicine Institute, Duke University, strained. Bridging this gap between the need for evi- one centered on the degree of risk to research partici- Durham, North dence and the resources available to generate it will re- pants rather than on arbitrary classifications of Carolina. quireharnessingcreativesolutionsthatleveragechanges research activities. This construct recognizes that already under way in health care delivery.2 In this View- many quality improvement activities conducted under Richard Platt, MD, MSc point, we discuss key limitations of the current re- oversight that apply to clinical operations share Department of Population Medicine, search system and describe an evolving national infra- numerous characteristics with clinical research that Harvard Pilgrim Health structure with the potential to increase health research requires expensive oversight. Taken together, these Care Institute, Harvard substantially. advances—a commitment to knowledge development Medical School, as part of practice, better and more accessible data, Boston, Massachusetts. The Old Paradigm and an ethical framework based on risk rather than Due in part to the poor quality of data collection and vari- outmoded divisions between quality improvement ability of practice in routine patient care, clinical re- and research—may allow many unanswered questions search has evolved as a specialized activity conducted to be addressed and will increase patients’ and clini- in parallel to clinical practice. This in turn has given rise cians’ confidence in the choice of preventive mea- to specialized research clinics, but even research done sures, diagnostic tests, and therapies. in clinical care settings was thought to require separate Converting continuous learning into actionable, in- operational and research data systems, with the latter formative protocols that are acceptable to health care served by a separate cadre of personnel. systems poses many challenges. In addition to critical is- However, issues related to ethical research con- sues of ethics oversight and patient consent in learning duct also contributed to this bicameral arrangement. In health systems, administrators and clinicians are under addition to articulating key ethical concepts, the 1979 pressure to meet efficiency standards, leaving little time Belmont Report helped crystallize the idea of research to pursue the creation of new knowledge in practice. Ad- as an activity distinct from clinical practice and gov- ditionally, current electronic health records (EHRs) are erned by different responsibilities. But while the report not optimized for research uses, and each system has its represented a watershed event in clinical research, much idiosyncratic definitions and coding standards for clini- has changed in the decades since its publication. cal data. Separate ethical frameworks and parallel data sys- Seeking to address these issues, the National Insti- tems for research and practice have substantially in- tutes of Health (NIH) initiated the Health Care Systems creased research costs and timelines and spurred mi- Research Collaboratory6 to foster innovative ap- gration of clinical trials to countries with comparatively proaches to integrating research with care delivery. In- low labor costs.3 The mutual isolation of research and tegrated health systems with EHRs were encouraged to practice also raises concerns about whether studies per- propose pragmatic clinical trials. The first round of 7 proj- formed in a separate research context can truly inform ects selected for funding share 3 salient features: all are clinicalpractice.Further,inconsistencyinregulatoryover- multicenter studies, their interventions are incorpo- sight of knowledge-generating activities is increasingly rated into routine care rather than added to it, and each apparent at the borderline of quality improvement and uses EHRs to identify eligible patients and collect rel- research: even though they often address identical top- evant data. For example, the University of Iowa launched ics, quality improvement programs often settle for sub- the Nighttime Dosing of Anti-Hypertensive Medica- optimal designs and methods to avoid being classified tions: Blood Pressure Medication Timing Study Corresponding as research. These unintended but perverse incentives (BPMedTime), a pragmatic clinical trial; another study Author: Robert M. ultimately harm patients by slowing the discovery and coordinated by the University of California, Irvine—the Califf, MD, Duke dissemination of best practices. cluster randomized trial Active Bathing to Eliminate In- Translational Medicine Institute, Duke fection (ABATE)—will investigate whether chlorhexi- University School of The New Landscape dine bathing and Methicillin-resistant Staphylococcus Medicine, 200 Trent The Institute of Medicine’s concept of a “learning health aureus decolonization can reduce infections and read- Dr, 1117 Davison Bldg, system”4 envisions virtuous cycles in which consump- missions in over 200 non–critical care units in more than DUMC 3701, Durham, NC 27710 (robert.califf tion of evidence (evidence-based practice) is comple- 50 hospitals belonging to the Hospital Corporation of @duke.edu). mented by generation of evidence (practice-based evi- America. In its first year, the Collaboratory has made sig- jama.com JAMA November 20, 2013 Volume 310, Number 19 2037 Copyright 2013 American Medical Association. All rights reserved. Downloaded From: http://jama.jamanetwork.com/ by a University of Wisconsin -Madison User on 11/29/2013 Opinion Viewpoint nificant progress in achieving consensus on ethics oversight, curat- ing research costs. However, for long-term diseases such as heart ing and sharing data across systems, and implementing protocols failure, atrial fibrillation, hypertension, and hyperlipidemia, it will be to minimize disruption of busy practices. important to integrate registry data with data from EHRs and the growing array of patient-reported outcome measures. Initial expe- Opportunities in Cardiovascular Disease riences in Sweden (TASTE)7 and the United States (SAFE-PCI)8 dem- These recent developments offer significant opportunities for car- onstratethatrandomizedtrialscanbeconductedwithinongoingreg- diovascular specialists seeking to stem the leading cause of death istries for a fraction of the cost of traditional trials. The Collaboratory and disability. However, education and the innovative leveraging of trial described earlier (BPMedTime) will pilot electronic capture of existing capabilities will be critical to these efforts. key outcome data from EHRs to address the surprisingly unan- First, specialty education programs should incorporate the con- swered question of whether daily blood pressure medicines should ceptual basis for learning health system methodology.Physicians and be taken at night or in the morning. other clinicians should understand that robust EHRs not only en- As electronic data sources mature and methods for aggregat- able more communication with patients and other health care prac- ing and analyzing them improve, public health and personalized titioners, but also facilitate aggregate analysis to inform optimal ap- medicine can increasingly exploit a common data infrastructure. The proaches to care. Collaboratory experience shows that it is possible to implement rig- Second,educationshouldexploreimprovementstoethicalover- orous, efficient, and economical trials that combine observational sight of learning activities. The NIH is supporting a substantial ef- and interventional methods, including randomization, to yield in- fort to elucidate the views of the public, patients, and clinicians re- formation directly relevant to patient care. garding learning activities that include not only “traditional” research Technological solutions are within reach, but progress de- but also surveys, quality improvement, and comparative effective- pends on strong partnerships with clinicians and delivery systems, ness studies. Matching levels of oversight to participants’ degree of and on well-designed oversight systems that enable researchers and risk could ameliorate administrative burdens and delays and focus operations experts while protecting patients and the public and in- effort on protocols where intensive oversight is beneficial. cluding them in decision making. Rapid development of this new Third, cardiovascular specialists are well poised to contribute to model in cardiovascular medicine can have significant benefits for a sophisticated fabric for learning activities. Existing registries offer public health, but fully