An earlier OT article on this topic ((10/25/10 issue) described a Rapid-Learning System (RLS) for cancer care—individual patient data are gathered, aggregated, and analyzed, along with clinic trial results, and then disseminated to clinicians to inform treatment decision-making. The idea is that through an iterative process, the results of those treatment decisions would then be fed back to the databases to update and help refine the knowledge base.
Electronic health record (EHR) systems already exist in some organizations, and new, interoperative ones are being developed, from which data can be gleaned for an RLS.
This article describes the practical aspects needed for such a system to make it acceptable to practitioners and patients.
Amy Abernethy, MD, Director of the Duke Cancer Care Research Program and lead author of a Journal of Clinical Oncology paper on an RLS (J Clin Oncol 2010 Sep 20;28(27):4268-74. Epub 2010 Jun 28), explained that given the life-threatening nature of many cancers, their costs, impact on populations, and patient involvement, oncology appears to be a good specialty in which to roll out an RLS. New treatments appear rapidly and frequently, and widespread off-label use often precedes the studies that support FDA approval. Oncologists are very data-driven, and an RLS should be able to aggregate and analyze data where no formal clinical trial outcomes exist.
Oncologists understand the importance of research and “clinical trials are not anathema for us,” she said. Oncologists are used to tumor registries and the Surveillance, Epidemiology and End Results (SEER) database, among others, and evolving treatment guidelines.
ASCO President George Sledge, MD, Professor of Medicine at Indiana University Simon Cancer Center, noted that registries tend to focus more on a particular disease, whereas a Rapid-Learning System “has the potential to be something much more global.”
An RLS will not be limited by disease type but will be integrated by drugs, co-morbidities, symptomatology, drug interactions, genetics/genomics, etc. One could look at it “as registries on speed,” he said, but an RLS is really something much larger, given the constant process of data collection, aggregation, analysis, knowledge base development, and feedback to the treatment decision process.
Physician and Patient Acceptance
Gregory Masters, MD, of Medical Oncology Hematology Consultants in Newark, Delaware, and Associate Professor of Medicine at Thomas Jefferson University Medical School, said he thinks an RLS would be highly worthwhile since he wants to get as much information as possible to use in his practice. He used the example of how to better manage diarrhea with the use of irinotecan—”People would be hungry for that kind of information,” he said.
Although most oncologists would want to participate, some of those who are older may not be as comfortable with the electronic media, Dr. Masters cautioned, also expressing concern that specific guidelines to comply with privacy laws will have to be spelled out, and that practitioners may have “some baseline suspicion” about sending out data because of how it may be used.
“Is it somehow going to be used in affecting how we're reimbursed, for example?” Patients will also want assurance that their data will not be extracted by insurance companies to deny reimbursements.
Dr. Abernethy emphasized that any RLS has to be smooth and seamless for practitioners to participate, not interfering with the usual workflow—”You can't slow the doctor down, period. I've got too many things in my day, and am way too busy for somebody to throw in another thing as a roadblock.
“I live in cognitive overload all day long, and an RLS that would bubble up the stuff that I need now to take care of this person in front of me should help me be more efficient if it does not slow me down.”
Dr. Sledge agreed: “The big issue for oncologists, as for all practitioners, is time.” He calculated that one minute added to each patient visit could add 20 minutes of extra work a day, which is 100 minutes a week, and therefore a loss of three working days a year and about 60 fewer patient visits a year. So to be accepted and successful, any Rapid Learning System cannot require any net extra time, he emphasized.
Furthermore, Dr. Abernethy said, “silos” have to be dismantled, meaning that large data sets have to be shared with confidence and trust so that all researchers can take advantage of the power of the aggregated information, regardless of discipline, location, or sector.
To gain patient participation, an RLS will have to assure privacy of communication based on communication standards. People will need to feel confident that information about them is not going to be used against them in their jobs, and they will need to feel that their participation is going to help them and possibly others.
Dr. Sledge said that although “privacy concerns are absolutely legitimate and crucially important,” he thinks that 95% of patients would opt into such a system if they believed that it would make their lives safer, increase the possibility of cure, and make their time with their doctor more efficient.
Development of Systems
Many different prototype Rapid-Learning Systems will be developed, Dr. Abernethy predicted, noting that already at Duke pilot systems are making care more efficient and facilitating conversations with patients. Technical developments and RLS's will likely mature over the next 10 to 20 years, she added.
Dr. Masters said he believes it is important to involve oncologists early in development so they feel that they have control of and ownership of the system, making them more likely to move it forward.
The first level may involve quality improvement and so will not require patients' informed consent or institutional review board approval, according to Dr. Abernethy. She said that as systems evolve, she expects patients to have a say over how their data are used and in what kind of research.
Duke researchers have already learned from their prototype systems that psychosocial issues are important to their cancer patients, and they can now match patients with social workers most in tune with their needs, as opposed to a random assignment or a practitioner choosing a favorite one, Dr. Abernethy noted.
Furthermore, patients reported on computers things they would not say in person. Among breast, lung, and gastrointestinal cancer patients, sexual distress was the third most important symptom, with 30% reporting moderate to severe distress. Clinicians could look for correlations between these reports and shortness of breath and quality of life, as well as develop new interventions and monitor their impact.
Pharmaceutical companies, as well as pharmacy benefits managers, are very interested in being able to access large data sets to be able to understand the landscape in which they are marketing and delivering their drugs, giving them the ability to be much more precise in these efforts. It should be possible to fulfill these needs while safeguarding patient privacy.
Dr. Masters suggested that rather than having an RLS feed raw data to clinicians, “you would want to see a panel of experts review that data and come up with a consensus recommendation, to have that vetted in some way to make sure that it is reasonable.”
In summary, Dr. Abernethy said, an RLS should give oncologists a better way to optimize care by putting data in front of them about what should work best for a particular patient. “The intent of Rapid-Learning Systems are that [oncologists] care for this individual patient, informed by all the similar individuals who came before her, and that her care is reinvested in the system to inform care in the future.”
Dr. Sledge noted that there will be significant costs involved to develop and roll out such a system, especially for smaller practices, once funding from the current health care reform bill stops. “Having said that, though, there is no question but that this is where we're going,” he concluded.