Giant EHR-based network will compare treatments

June 24, 2014

A new, voluntary government program will use electronic health records (EHR) data to compare different treatments for particular conditions, using the largest clinical database ever assembled. This program has the potential to advance medical research and improve care in physician practices.

 

A new, voluntary government program will use electronic health records (EHR) data to compare different treatments for particular conditions, using the largest clinical database ever assembled. This program has the potential to advance medical research and improve care in physician practices.

By September 2015 the program, known as the National Patient-Centered Clinical Research Network (PCORnet) is expected to give investigators access to data on 20 million to 30 million patients from 29 research networks across the country.

These include 11 clinical data research networks (CDRNs) based in healthcare systems and 18 patient-powered research networks (PPRNs) that are operated and governed by patients with particular conditions and their caregivers. The EHR data will remain within the local networks and will be made available to researchers via data mining applications.

The physicians who participate in PCORnet will not only help advance medical research, but may also be able to apply the results of the studies directly in their own practices. The Patient-Centered Outcomes Research Institute (PCORI), which is funding the networks and tying them together through PCORnet, envisions that some networks will provide the participating doctors with study results to help them improve the quality of care.

“We haven’t made it a requirement, but what we’re funding has the capacity to generate those kinds of reports that can be fed back to doctors,” Joe V. Selby, MD, MPH, executive director of PCORI, told Medical Economics.

The CDRN based at New York’s Weill Cornell Medical College intends to do just that, says Rainu Kaushal, MD, MPH, chair of the college’s department of Healthcare
Policy and Research. “We plan to develop a rigorous dissemination plan which will enable all participating health systems, including both clinicians and patients, to readily benefit in a timely manner from the research happening in our network,” she explained.

From the viewpoint of research, PCORnet’s major advantage is its potential to increase greatly the speed and efficiency of comparative effectiveness research (CER).

“PCORnet has significant potential to improve the efficiency of conducting comparative effectiveness research, not only due to the depth and breadth of the data network, but also by streamlining the contracting and regulatory approval processes,” Kaushal says.

The ability of PCORnet to answer research questions quickly, using a large and rich
database, looks promising to some observers.

“What’s powerful about EHR research is that it’s quick and inexpensive to acquire a large amount of data that’s important for certain questions,” says Steven Ornstein, MD, a professor of family medicine at the Medical University of South Carolina. 

Ornstein should know. PPRNet, a practice-based research network that he founded, has been doing observational research with EHR data since 1995.

Next: Ramifications for research

 

Ramifications for research

PCORI is a lot newer to the field, but has immensely greater reach and resources than does PPRNet.

Established by the Affordable Care Act, PCORI currently funds nearly 200 randomized clinical trials (RCTs), according to Selby. Its annual budget of $500 million looks big, but it can cost up to $150 million for a large, multi-center RCT to answer a major clinical question, such as which is the best first-line therapy for hypertension. So to PCORI, it made sense to build the EHR network. To date, it has invested over $100 million in PCORnet.

Selby says PCORI will underwrite some PCORnet trials. He hopes that the National Institutes of Health (NIH) and pharmaceutical and device companies also will fund some research. Studies will be initiated either by the funding entities or by investigators within the research networks, he adds.

CER usually compares how different treatments work in the real world. Unlike an academic clinical trial, which uses carefully selected patients and follows predetermined protocols in a controlled setting, CER follows typical patients through a normal course of care. This approach, known as a pragmatic or practical trial, can be retrospective or prospective, observational or randomized. A large portion of PCORnet research will consist of observational studies.

Because these studies can’t establish cause and effect, CER based on a retrospective analysis of EHR data is mainly important for “identifying trends that help you figure out which research questions should be asked prospectively,” says Leonard Zwelling, MD, interim chief medical officer at Legacy Community Health in Houston and a retired professor of medicine and pharmacology at the University of Texas M.D. Anderson Cancer Center.

Selby adds, “The methods of observational research have improved, and we do feel that we can get a somewhat better sense of whether these observational data are giving us a truly valid or accurate result.”

The size of the cohorts used in PCORnet studies, he says, will enable researchers to gather very precise data about the impact of therapies on subgroups of patients. He concedes, however, that those studies can’t necessarily establish whether one treatment is more effective than any other. To get closer to that answer, he says, well-designed observational studies must compare control and study groups of  similar patients who are being treated in different ways.

In addition, researchers can use PCORnet “to launch randomized trials more efficiently and more quickly,” he says. “EHRs allow you to identify the patients who would be eligible. You approach the patients with the help or the consent of their health systems. Then the EHR and possibly patient-reported data help you follow these patients over time after you’ve randomized the treatment choices.”

Kaushal says that the Weill-Cornell network plans to do both observational and “prospective interventional” trials, using  EHR data as well as patient-reported and patient-generated data, including genomic information. The network participants have all had extensive clinical trial experience, she adds.

Among the 22 organizations involved, Kaushal says, are several healthcare systems; their clinical and translational science award centers; patient engagement organizations; a practice-based research network; a genome center; a central institutional review board (IRB); regional health information organizations; and the New York state health department.

Next: Translating research into practice

 

Translating research into practice

The results of pragmatic studies are more likely to change how doctors practice than are the results of traditional clinical trials, say advocates of PCORI’s approach.

“If the study is done in a real-world delivery system and not in a highly selective population in a research clinic, other physicians and other delivery systems are going to be more likely to adopt the findings into practice,” Selby says, adding that it’s easier to generalize the results of pragmatic studies than those of traditional clinical trials.

Kaushal agrees that the outcomes of PCORnet studies will be “far more generalizable than research that might occur in a laboratory environment or outside of the clinic setting.”

Zwelling is skeptical, but doesn’t rule out the possibility that observational studies based on EHR data could influence practice.

“A lot of that depends on what you found out, the statistical power of what you found out, and your clear articulation of the question being asked and of the answer you received,” he says.

EHR limitations

From a technical standpoint, PCORnet faces several significant challenges.

First, the data networks will have to find ways to combine data from multiple EHRs in a standardized format that can be analyzed. Second, much EHR data is unstructured, and the information of interest to researchers interested isn’t always entered in structured fields or entered correctly.

Some of the PCORnet funding of the data networks is earmarked for data standardization, Selby notes. Networks are expected to put the data they gather into a standard data model to participate in PCORnet. That allows the researchers to do “distributed data analysis” without having to roll up the data into a central database.

The larger challenge, he says, is that the EHR data can’t be used in research unless it’s structured. “Data completeness is a big problem, and it’s one reason why we say there has to be a business case in this for physicians. They’ve got to step up their game. It will be much better when data entry into EHRs is more systematic.”

The “business case” he refers to is the feedback of study results to practices. Ornstein’s PPRNet has been doing that for many years in its translational research, and participating physicians have been able to use PPRNet reports to intervene with patients who have care gaps.

Here’s how it works: the doctors send de-identified data extracts from their EHRs to Ornstein and his colleagues at MUSC. After the data has been analyzed, it’s sent back to the physicians, and they can use a specialized software tool to re-identify the data on a spreadsheet.

Next: Patient consent

 

Patient Consent

Zwelling, who formerly administered clinical trials at M.D. Anderson Cancer Center,  says that PCORI would have to figure out how to obtain patient consent for its networks’ EHR studies.

Patients can’t give blanket permission for use of their clinical data in a non-specific fashion, he points out. Physicians either have to ask every patient for permission for each study, or they need to request a waiver of consent from an IRB, even if the data is de-identified.

Selby says that PCORI would deal with this issue in different ways for different studies. In very large observational studies,  involving 1 million people or more, it would be impossible to contact and obtain consent from all the participants, he says So in those cases, participating data networks would seek IRB waivers.

In other studies that require doctors to change how patients are treated, individuals will be asked for their consent-a given in the patient-powered data networks.

Doctors’ support is essential

By all accounts, PCORnet offers unparalleled opportunities for comparative effectiveness research, but its success depends on the support of physicians.

Whether doctors will ever see a “business case” for this research is open to question, since they won’t be paid for it. But the physicians that participate, either on their own or through their institutions, can make a significant contribution to medical research. Moreover, the researchers’ analysis of their EHR data, if it’s made available to the doctors, could help them provide better care to their patients, say proponents of PCORnet.

Adam Stracher, MD, a faculty member at Weill-Cornell and a practicing internist, says of his institution’s clinical data network, “The benefit to us, as it is to all physicians, is to get a rich depository of good clinical data to be able to use for outcomes research, for developing clinical guidelines, and to figure out the best ways to take care of patients with multiple diseases.”

Stracher also looks forward to getting feedback on his own patients’ outcomes from the Weill Cornell-led network.

By seeing how his patients fare compared with other doctors’ patients with similar diseases and demographics, he says, “You can get a sense of whether you’re providing the best, most valuable care you can be giving, and learning how others are taking care of patients with similar diseases.”