Using Real-World Data For Clinical Research: Challenges And Opportunities
By Adrian F. Hernandez MD, MHS, Director of the Health Services and Outcomes Research, Duke Clinical Research Institute
Across the country, there is growing awareness of challenges facing the clinical research enterprise. Among these challenges are the expense and logistical complexities of performing clinical trials, much of which is created by the need for elaborate, stand-alone data collection systems that are distinct from the systems used for normal clinical care.
But as more and more Americans use electronic health records (EHRs), personal health records, and personal electronic devices that directly record health and fitness information, many researchers are now hoping to tap into new, rich sources of data. These real-world sources provide an alternative to stand-alone data collection systems, allowing important measurements and outcomes to be gathered as a “byproduct” of patient interactions with their healthcare providers, or measured directly by patients themselves. These new approaches, it is hoped, will allow researchers to efficiently gather data from ordinary patients in a typical care environment while avoiding the significant costs of redundant data collection.
But there’s a catch.
Ensuring the Quality and Relevance of Real-World Data
One reason that massive, redundant clinical research data collection systems emerged in the first place is that “real-world” data are inherently messy. Traditionally, before they can be used to answer research questions, data must undergo multiple rounds of “cleaning” to ensure that they are complete, accurate, consistent, and fit for use—and this can be a surprisingly complicated process. However, as interest grows in using real-world data for research, the focus should be less on cleaning data and more on ensuring that data are reliable and meaningful. For most studies, much of the data that is collected is never used. The data that matter are those that characterize the right patient, the exposure or intervention, and the outcome.
Another crucial element in harnessing real-world data is ensuring that sources of electronic health information, particularly EHRs, are interoperable, meaning that information collected via one platform or technology can be combined in standard manner. In addition, some types of analysis may also require something called an electronic or computable phenotype, which allows researchers to precisely define the various aspects that fully describe a person, a disease, or a medical condition. By combining electronic phenotypes with data specified in ways that ensure standardization and quality, researchers can ask meaningful questions that can then be applied across multiple electronic data systems.
Finally, using EHR and other electronic data for research purposes requires careful attention to privacy and security. Patients must have confidence that their data are being kept secure and confidential, and researchers must satisfy obligations for ensuring the safety of sensitive health information.
Current Efforts, Future Promise
Despite the challenges of using “real-world” data for clinical research, we have made significant progress in recent years. The FDA’s pioneering Mini-Sentinel project gathers data from EHRs and other electronic sources to provide information about the safety of medical products using an approach known as a distributed data network. Doctors and health systems participating in Mini-Sentinel do not need to transmit large amounts of data gathered at their institutions; instead, queries can be applied to data at a particular site by specialized programs, and the relevant answers returned in secure, anonymized fashion.
Other groups are now building on these successes and lessons learned. PCORnet is implementing a large-scale pragmatic trial designed to test whether lower-dose or higher-dose aspirin is more effective in preventing heart attack and stroke in persons at high risk for these conditions. This clinical trial, the ADAPTABLE aspirin-dosing study, will use the real-world data contained in the millions of EHRs in its network partners to answer a clinical question of intense importance to patients and healthcare providers alike. A second PCORnet project, a pair of observational studies investigating obesity treatment and prevention, will employ similar methods. It is unlikely that either of these studies could be performed at their current large scales and relatively low projected costs without the advantages conferred by secure, reliable access to real-world data.
Other major initiatives underway include a push for a national medical device registry system that will furnish critical information about the safety and effectiveness of medical devices. In addition, the NIH Health Care Systems Research Collaboratory supports innovative pilot projects utilizing data from EHRs and other real-world sources and hosts a growing set of tools to enable these efforts.
These are exciting times for pragmatic research. Although much remains to be done before we can realize the full potential of these new approaches to support clinical trials, we are starting to see significant progress. We hope that as these innovative trials begin to return results, we will be able to demonstrate the enormous possibilities of using real-world data to inform patients and caregivers about the choices they face every day.