Halt and Catch Fire: Can the Digital Revolution Empower the Move Toward Value-Based Cancer Care?
FROM THE AFFORDABLE CARE ACT1
to countless pieces in the New England Journal of Medicine
the consensus of thought leaders from academia, government, and industry recognizes the need to shift from a fee-for-service model toward a more coherent system of predicating payment on value delivery. In 2015, the national healthcare expenditures in the United States rose to $3.2 trillion, which accounted for 17.8% of the American gross domestic product.4
The push toward value-based care delivery is largely driven by the unsustainable growth rate of healthcare expenditures and the underwhelming American healthcare outcomes that result despite this extraordinary expenditure rate.5
The value conundrum is particularly challenging within the domain of cancer care, in which treatment-related costs dwarf overall healthcare spending: According to estimates from the National Cancer Institute, cancer care-related costs are projected to grow by 39% ($172.8 billion) by 2020.6
Pharmaceuticals and therapeutic innovation wield an extraordinary impact on these costs—cancer drug spending was estimated at $37.8 billion in 2016, which represents a 33% increase ($9.4 billion) for new drugs alone since 2010.7
The growth of genomic technologies (including somatic and germ line testing) will further inflate cancer care costs; the current world market for genomic testing is $9.2 billion and is expected to grow to more than $20 billion by 2022.8
The move toward developing transparency around value delivery in cancer care is undermined by 3 key factors:
- The lack of a national data set for assessing cancer outcomes data, on either a provider or institutional basis, that is avail- able for performance comparison purposes. While the Center for International Bone Marrow Transplant Research routinely provides risk-adjusted survival outcomes data to consumers, these data are limited to only those patients who undergo allogeneic transplantation.9
- Coding and billing data lack sufficient data richness to adequately risk-stratify cancer patients in a manner that allows for a transparent assessment of cancer outcomes and costs as related to clinical risk.10
- As our healthcare system increasingly works to reduce costs by commoditizing services such as laboratory testing (including genomic testing), imaging studies, and therapeutic delivery (through an increased reliance upon specialty pharmacy services and third-party pharmacy benefit managers), it becomes increasingly difficult for cancer care providers to understand their own care delivery costs because of the balkanization of health records and the proprietary nature of many data sources.
There is no shortage of academic, industry, and government sources that identify value as equaling cost/outcomes; there is far less uniformity of opinion when it comes to defining what that means for a particular patient affected by cancer. Many of the current “value” models for cancer care delivery look for the value of isolated healthcare decisions/transactions rather than the aggregate costs/outcomes of the delivery model.11,12
It is no longer adequate to simply aggregate data by histological diagnoses. In this modern era, patients are defined with increasing precision (hence the EGFR-negative, ALK-negative patient who expresses PD-L1 for whom the predicted cost of care is far more predicable); the goal then becomes one of defining the risk-banded costs of care based on a level of data richness and analytics that defies the capacities of most electronic health records (EHRs) or the analytical capacities of most healthcare providers and cancer care delivery networks. This level of iterative risk/cost model evolution needs a depth of data that is largely unprecedented in healthcare today. These analytics must have the ability to incorporate a multiplicity of data sources, reconcile multiple identifiers for a single patient, and simultaneously leverage an evolving data set of genomic risk factors.
This seemingly impossible task now represents a key focus of several efforts that attempt to master/ reconcile the breadth of relevant care delivery data in the pursuit of increasing transparent, data-rich models for assessing care. The American Society of Clinical Oncology (ASCO) has published an updated version of its value framework that has evolved to include more data sources and better integration into decision support tools to ensure that this construct can be employed more consistently in care delivery.13
The meaningfulness of decision support tools and outcomes analytics will, however, require a profoundly different information architecture to ensure that such systems are based on sufficiently rich data resources, are meaningful, and can base data assessments on an accurate risk segmentation of the population in question.
This level of analytic capacity must be based on the big data model of information technology. In their recently published book chapter, “Big Data Analytics in Healthcare: A Cloud-Based Framework for Generating Insights,” Anjum et al, envision a move toward systems that utilize scalable cloud- based data analytics architecture. They argue that to be effective, these cloud-based systems will need to ensure that genomic and clinical data are correctly identified and linked while ensuring that data from a diverse array of sources, systems, and “disparate locations” are aggregated in a robust, quality-controlled manner.14
A robust big data analytics model in the cancer care domain can yield the following potential benefits:
- Clinical trials matching
- Increasingly precise patient risk segmentation
- More robust cost/risk assessments
- A tool for more transparent value mapping of genomic/precision medicine care delivery
Toward that end, several vendors have entered the marketplace with models and tools directed at making this quantum leap toward more meaningful value-based analytics. ASCO’s big data informatics model, CancerLinQ, intends to provide both practice and research planning tools that leverage data in an innovative way that far exceeds the analytical capacities of the typical EHR system.15
Other vendors have entered the market space with propriety big data–based analytical systems, which include products and services from Flatiron Health16
and Cota Healthcare.17
These products are marketed as tools for value-based data analytics for clinical practice, research planning, and revenue cycle management.
Recently, the importance of these new analytical service tools platforms has been highlighted by the inclusion of the Cota Healthcare system as a key part of the Oncology Physician-Focused Payment Model (PFPM) submitted by Hackensack Meridian Health. The PFPM Technical Advisory Committee did ultimately recommend to the HHS secretary that the proposed oncology bundled payment model (which uses Cota’s CNA-Guided [Cota Nodal Address] Care to establish risk-cost bands within the bundles) should be accepted for testing as a pilot advanced alternative payment model (AAPM).18
This AAPM approval was followed shortly thereafter by an announcement from Memorial Sloan Kettering Cancer Center and Cota Healthcare regarding a 5-year exclusive deal in which these 2 entities would collaborate on projects focused on leveraging this suite of big data analytics to bring more effective precision medicine solutions to patients with cancer.19
The growing intensive information demands of the new precision medicine paradigm of cancer care, coupled with the drive to achieve a more meaningful alignment between cancer risk and the cost of care, is likely to increasingly push big data technologies to the forefront of cancer care. As the “black box” paradigm of per-capita reductions in the cost of care articulated in the “Triple Aim of Care”20
is challenged by new cancer care diagnostic and therapeutic technologies, big data analytic solutions can help to create a far more transparent and meaningful paradigm for how we can more intelligently move toward more value-based care for cancer patients.