A goal of Twenty-First Century Healthcare is to establish and enhance the Learning Healthcare System (LHS). As discussed in numerous forums, journals, and social media, the LHS is viewed as critical to improving healthcare. Fundamentally, the LHS converts data about care and operations into knowledge that it translates into evidence-based clinical practice and health system change. In so doing, the LHS utilizes as vehicles health information technology, databases, the electronic healthcare record (EHR) and, importantly, a research infrastructure. The continuing narrow evidence base for clinical care combined with the need for substantial amounts of data to fill large evidentiary gaps, among other factors, have fostered the LHS concept.
To assure the utility and validity of data converted and then translated into improvements by the LHS, we need rigorous research approaches that are also efficient. Research is defined as “a systematic investigation…designed to develop or contribute to generalizable knowledge.” At present, evidentiary inputs to the LHS range from activities not generally considered research (e.g., programmatic and quality improvement evaluations) to various forms of research that are sufficiently rigorous, but for various reasons, can be difficult to employ and translate into the routine workings of the LHS.
For example, the randomized clinical trial (RCT) — considered the cornerstone, or “gold standard” methodology — provides the best data, but by its very nature, is separate from the workings of clinical care in a given healthcare system (HCS). Instead, the RCT functions in an alternate environment precisely controlled for the approach’s particulars. Further, the RCT is costly and time-consuming, and its rather narrow entry criteria may diminish its ability to be generalized to routine patients (e.g., those with comorbidities, or those outside of the entry criteria of age and severity).
The Practical Clinical Trial, which has broader entry criteria but a somewhat different methodology, shares some of the same limitations as the RCT. Observational research is limited by its inability to control for known and unknown bias. Health services research (HSR), which examines factors related to obtaining care makes use of these techniques, statistical database analysis, and other techniques.
In each of these cases, research data is then translated by a variety of means that are generally separate from the research itself. While programmatic analysis and the like may pass directly into use, results of research per se are generally published first in peer-reviewed journals and then translated into care via education or other forms of persuasion. At times, evidence moves directly into guidelines while undergoing peer review in parallel. In any event, translation is often slow and cumbersome, with many years passing before research findings find their way into patient care. (Balas EA, Boren SA. Managing clinical knowledge for healthcare improvement. In: Yearbook of Medical Informatics. Bethesda, MD: National Library of Medicine; 2000:65-70)
Seeking To Improve The Functioning Of The Learning Healthcare System
The bright line that has persisted between research and the routine clinical environment, coupled with the exclusion by research studies of many patient types seen in clinical settings, has impeded the smooth functioning of the LHS model. Additionally, in the case of HSR, studies are often conceived by those not immersed in the day-to-day practicalities of care delivery and frequently focused on isolated HCS issues. Consequently, we need new research approaches so that the best evidence is applied as seamlessly as possible.
To address this need, the U.S. Department of Veterans Affairs (VA) is piloting two new approaches: Point-of-Care Research (POC-R), a method of performing clinical trials within the daily practicalities of the HCS (with the intent of advancing these systems to LHS); and Collaborative Research to Enhance and Advance Transformation and Excellence (CREATE), an HSR program designed for more expeditious translation.
Point-of-Care Research. The first approach, POC-R, is an adaptation of the RCT in which the trial is embedded in the “real life” HCS environment and therefore utilizes system providers, healthcare personnel and resources. It is designed for, and applicable to, certain types of research questions; i.e., generally comparative effectiveness research (CER) on accepted treatments in routine care. The VA pilot is underway in the Department’s clinical trials program (the Cooperative Studies Program) and is a CER study of two alternative, approved treatments considered equivalent and offered at a decision point-of-care.
The approach works as follows: After prescreening by the EHR for eligibility criteria (of proven benefit), the option to include a patient in the study is presented to the healthcare provider. If the provider elects that option, the patient is invited to enroll in the study and informed consent obtained. The patient is then randomized into one of the study’s treatment arms with their own provider administering the study treatment in the routine clinical setting. Outcomes are assessed by automatically extracting data directly from the patient’s EHR.
The VA pilot underway takes place at the Boston and Providence VA Medical Centers where two alternative regimens — weight-based and sliding-scale — of in-hospital insulin administration are being compared. Both regimens are commonly used, approved treatments considered equivalent in usual practice. Hospital length of stay (reflecting efficiency of the method of administration) is the primary endpoint (i.e., the measure of success of the therapy) and secondary endpoints are glycemic control, and readmission to the medical center because of poor glycemic control. More than 100 subjects have been enrolled.
The randomization approach and the “real world” healthcare setting of the POC-R, combined with use of the electronic health record, bridge the evidence gap while at the same time facilitating translation. Providers engaged in the study will be familiar with the treatment regimen, which treatment choice is more effective, and the EHR functionalities underpinning the study. Consequently, when they utilize the treatment on behalf of patients who have agreed to be study subjects, the EHR will employ the data elements identified above. Also, as the study is within the healthcare system, the EHR can obtain a more realistic assessment of cost and resource use than can be achieved by the usual RCT.
It should be noted, however, that the POC-R approach has certain limitations (aside from its circumscribed use on accepted treatments). There are system and personnel requirements, such as the need to identify real time events, to identify candidates, and to randomly assign and track data. Further, the study protocols are unblinded and there is less restriction in determining subject eligibility than in the usual RCT (though this may enhance the realistic nature of the approach, as it more closely matches the precision of daily clinical practice).
Additionally, because it crosses the bright line between clinical care and research, POC-R entails a cultural shift in the doctor-patient relationship. Physicians must clearly understand expectations in the conduct of research, the extra physician time required, and oversight issues that may arise. As the trial is part of routine care, patients must also adjust to their doctor appearing not to know the correct treatment. Experience has shown thus far that patients are accepting of these “new rules” if the relationship with their own physician is maintained. Patient and provider attitudes are being assessed in a separate study. Also, letters of intent for new studies are being considered.
Collaborative Research to Enhance and Advance Transformation and Excellence. The second VA approach, CREATE, enhances the acquisition of systems-relevant data for translation by stipulating two basic requirements for peer-reviewed HSR funding proposals: an upfront, formalized agreement by health system partners to engage throughout the project, and the coordination of 3-5 groups of research projects around a similar theme.
Although the fundamental goal of HSR is to improve healthcare and its systems, individual study findings often aren’t translated into practice. Further, some research findings have limited implementation potential. As observed in an IOM report, “despite the accelerating pace of scientific discovery, the current clinical research enterprise does not sufficiently address pressing clinical questions.”
Why is this the case? One reason is the lack of, or insufficient, engagement by HCS providers. In many instances, they play a minor role (and often no role at all) in research design and execution, learning instead of research findings after the fact (once the study has been published). Relatedly, research is often published in the form of stand-alone findings which may or may not be integrated into an overall strategy for health system improvement. Generally speaking, such single, isolated studies may not sufficiently support planning efforts to be adopted in the HCS.
By requiring investigators to certify that health system partners agree to engage in study development and execution, CREATE assures that the most relevant HCS questions are raised. Further, CREATE mandates that partners commit upfront to implementing research findings within their health system as appropriate.
The other CREATE mechanism, which requires that groups of 3-5 studies be funded together (rather than individually), ensures that research findings will have greater impact on a particular system or clinical issue. To pass peer review (in the VA Merit Review system), studies must employ rigorous HSR methodology. Given the engagement of HCS providers who have participated in study design and committed to use findings, this new approach should spur implementation. For example, engagement of health system leadership at the beginning of guideline development facilitates implementation.
The first review of CREATE proposals yielded a diverse set of studies: use of information technology in quality and safety, evidence-based therapies for posttraumatic stress disorder (focusing on patient interactions), access to care for those with a substance-abuse disorder, long-term care, pain management in Patient Aligned Care Teams (the VA version of the medical home), and implementation of women’s healthcare. A second review resulted in funding for proposals that include prevention strategies, rural access to mental healthcare, MRSA prevention, population analytics, and use of Natural Language Processing.
CREATE requires that providers and system leaders overcome certain barriers related to how each collects data (i.e., speed and efficiency of acquisition for programmatic data, rigor of method for research). Already, as the approach has become more established within VA, certain (predictable) conversations have taken place between researchers and VA medical center administrators; e.g., administrators learning to understand the longer timelines needed for research compliance (as compared to programmatic evaluation), or the detailed requirements for data acquisition (e.g., IRB need). Notably, these conversations have also led to greater understanding of the research process by systems individuals and the modes of system improvement by researchers.
Both POC-R and CREATE are rigorous research approaches that seek to close the longstanding evidentiary gap in clinical practice and healthcare systems. By addressing barriers between research and clinical care, fostering efficiency, and facilitating data use, they offer more promising paths to translation. Both require research infrastructure, a major asset to healthcare systems if they undertake learning. Both enhance data use by embedding translation within the research itself: POC-R, by its in-system presence, and CREATE, by its engagement of the HCS from the beginning. Ultimately, it is hoped, with these and other initiatives, that a new synthesis can be achieved between clinical and research cultures that offers better knowledge acquisition and use, and truly takes “learning” to a new level in the Learning Healthcare System.
Note: The views presented here are solely those of the author and do not necessarily represent the views of the U.S. Department of Veterans Affairs.