This "learning health system" prototype was first unveiled on March 27 at an ASCO-hosted panel discussion on big data in cancer care at the National Press Club in Washington, DC.
Sandra M. Swain, MD, FACP
A proof-of-principle prototype for CancerLinQ™ was demonstrated at the 2013 ASCO Annual Meeting. This “learning health system” prototype was first unveiled on March 27 at an ASCO-hosted panel discussion on big data in cancer care at the National Press Club in Washington, DC. At the ASCO Annual Meeting, attendees were able to view a demonstration of the CancerLinQ prototype in the exhibit hall as well as attend an educational session on the product, which ASCO calls the first demonstration of the feasibility of a health information technology-based learning health system. When the full build of CancerLinQ is available, it is expected to collect and analyze cancer care data from millions of patient visits and, using expert guidelines and other information, generate “real-time, personalized guidance and quality feedback for physicians.” Motivation for CancerLinQ was provided in part by an Institute of Medicine report that defined learning health systems as critical to the future of the nation’s healthcare system, and oncology as wellsuited for rapid health learning system development.
ASCO President Sandra M. Swain, MD, said, “We are motivated by the stunning fact that 97% of today’s cancer care does nothing to advance our collective knowledge of the disease or its treatment because only 3% of adult cancer patients participate in clinical trials. The learning health system is intended to bridge this gap.” Swain pointed out that the information on that 97% of patients not participating in trials is in paper files and electronic records that currently don’t talk to each other well, if at all. “CancerLinQ will transform cancer care by unlocking that wealth of information and enabling every patient to be a cancer knowledge donor,” she says. The CancerLinQ prototype shows that such a learning health system is feasible and is providing lessons about the challenges of building a full-scale system. The full-scale system is planned to have the core functions of real-time data collection, clinical decision support, data mining and visualization, and quality feedback.
The prototype was built in about 5 months using de-identified (anonymous) data from over 150,000 patients with breast cancer that was supplied by 26 practices in the US. The demonstration at ASCO included data from 100,809 patients that were available at the time the demonstration was prepared. Building the prototype required the collaboration of the oncology and information technology (IT) communities and included efforts to improve data standards in oncology, which is described below. Technology and legal analyses are also part of the development effort.
Since the first demonstration in March, three working groups were established to help take the lessons learned from the prototype and implement them in the development of the full system; these groups are working on clinical decision support, data quality, data governance, and natural language processing governance. Clifford A. Hudis, MD, president-elect of ASCO compared the prototype demonstration he gave at a press briefing to the Wright brothers’ first flight—although they demonstrated that man could fly, no one bought a ticket from them. Likewise, the prototype of CancerLinQ is not necessarily the exact shape and form of the final product. However, the prototype, which uses several open-source IT applications and at least one proprietary application, has demonstrated that the core functions of the health learning system are feasible. These core functions include the following:
Ultimately, Hudis said, CancerLinQ will be able to establish correlations that could inform care and establish hypotheses for prospective clinical research. Physicians set up the system once, and every time an electronic health record (EHR) is used, each patient automatically becomes a data donor. The data set could confirm one study and eliminate the need for additional studies, and also help avoid doing some studies that might be fruitless. Robert Hauser, Pharm D, PhD, senior director of Quality at ASCO, emphasized that CancerLinQ will not replace clinical trials, but will be hypothesis generating.
Although CancerLinQ is designed to be “data agnostic,” and will not have to wait for new data standards for implementation, ASCO has been developing interoperability standards for cancer care data to improve data sharing among healthcare providers, researchers, and patients. The first standard, the Breast Cancer Treatment Plan and Summary Standard and Implementation Guide, was announced at the 2013 Annual Meeting. The draft breast cancer standard was approved by open ballot through Health Level Seven International (HL7), an accredited Standards Developing Organization, and should be ready for implementation later this year. The major elements include diagnosis, patient health and comorbidity, surgical history and pathology, goals of therapy, chemotherapy regimen and dosage, duration of treatment and number of cycles, and major chemotherapy side effects.
There are many issues that remain to be resolved. Currently, CancerLinQ uses ASCO quality measures determined from ASCO Clinical Practice Guidelines, and although these guidelines do not cover the entire oncology territory, Josh Mann, associate director, Oncology Technology Solutions, Quality and Guidelines, ASCO, noted that there are no arrangements with other content providers at this time. The full CancerLinQ system will have to deidentify data, but when full genomic sequencing becomes available, theoretically it could be used to identify patients. Hudis expected that by the time the full system is up and running, this problem will have been solved. Hauser noted that those data used to measure quality already fall under HIPAA and can be part of the system. A state-by-state legal analysis is in progress, and some state laws regarding healthcare data disagree with federal regulations, so CancerLinQ likely may roll out targeting some states first where the legal situation is clearest. ASCO is also looking at international data laws and how cancer is managed worldwide.
Who will have access to CancerLinQ also remains a question. Mann said that the current thinking is that CancerLinQ will be free to ASCO members. Because EHRs and CancerLinQ will be implemented at the practice level, if one provider in a group practice is a member of ASCO but the other members aren’t, it isn’t clear how this will be managed. Physicians will be able to log in and access their own data for their own practice’s purposes. Other access is likely to be restricted. CancerLinQ developers are still defining data governance, and will have an as-yet-to-be-defined process in place to review data access requests. This may involve a panel of experts like that used to review a grant proposal. Researchers would have to apply to ASCO; ASCO could either do its study for them or give them access to data. Pharmaceutical companies would also have to prepare a research proposal for access to any data. Patients will have some tools available to them, (eg, to stay in touch with their healthcare team, report symptoms, and access treatment information).
Funding is also an issue. The prototype was funded by several million dollars from ASCO, primarily from the Conquer Cancer Foundation, and by donations from pharmaceutical companies, the Susan G. Komen Foundation, and private donors. Additional funding is needed to go forward, affecting how rapidly CancerLinQ develops.
Hauser said the CancerLinQ developers are hoping by 2014 to have something mainly geared at quality reporting that would use QOPI quality measures.Currently, practices in the US have access twice a year to compare their performance to these standards. QOPI is free to every ASCO member but requires manual chart extraction and manual entering into QOPI. The next goal for CancerLinQ is to have a version 1.0 live pilot within about 18 months that would encompass all tumor types and be available for a few dozen practices. Hauser expects that the version 1.0 CancerLinQ will be automatic in real-time, and generate reports as a QOPI “replacement” algorithm.
George W. Sledge, Jr, MD, Stanford University, who chaired the educational session on CancerLinQ, suggested the system might one day play a role in CME, automated recertification, exploring drug toxicity, and linking patients to clinical trials.
Mann said that over 400 demonstrations of the CancerLinQ prototype were given at the exhibit booth, and that at one point there was a line of 20 people waiting to see the demonstration. He said the reaction was extremely positive, and that the international practitioners seemed to be even more eager than those in the US to see the product developed. Mann hopes that the questions that come up in the next development phase will resolve rapidly. ASCO plans to publish its lessons learned—it has identified more than 60 so far—over the coming year, and will use them to inform its development of the full CancerLinQ system.