The volume of healthcare data continues to increase exponentially - burdening healthcare organizations with the management of disparate data, interoperability challenges between source systems and the complexities ofunstructured data. With so much data, searching out the relevant and actionable data can seem like an insurmountable task. Population health initiatives have extensively relied on the use of structured data from health insurance claims and charge-capture coding from medical billing systems.
While valuable, this data represents approximately 20 percent of all healthcare data—the rest is unstructured. Additionally, edge data from medical and Internet of Things (IoT) devices is proliferating and adding to the challenge of wading through this wealth of information to deploy timely intervention and chronic disease management. The healthcare ecosystem is ripe for the application of new technologies that will tap into unstructured data sources to improve outcomes, contain costs and improve the experience of patients and providers.
Population Health Management
Risk stratification of patient populations helps providers to continue the move from volume to value by identifying at-risk patients and avoidable cost events.Management of chronic disease, avoidance of hospital admissions and medication adherence are examples where the application of new technologies can improve population health. Medication adherence on its own is a $300 Billion problem,1and one that can be impacted by technology.Understanding why a patient is noncompliant is not easily determined from claims and charge-capture data alone.The incorporation of social determinants of health (SDOH) may provide the answer—but this information may lie within vast amounts of unstructured data. Likewise, gaps in the continuum of care may keep this information from becoming actionable.
Avoiding costly episodes of care requires the management of patients with a particular disease or combination of chronic conditions. The CDC notes that six in 10 of adults in America have a chronic disease, and that the key lifestyle risks for chronic disease are tobacco use, poor nutrition, lack of physical activity and excessive alcohol use.2Additionally, nearly 30 percent of all healthcare expenditures are associated to the top five percent of individuals with four or more chronic conditions.3
Machine Learning for Actionable Intelligence
With such large data sets, machine learning can be used to build models for patient risk scores, cost prediction and patient behavior. To tap into unstructured data, incorporate SDOH and IoT edge data, as well as reconcile with structured data, machine learning holds tremendous promise. Machine learning approaches are continuous, adaptive and aware—uncovering actionable intelligence for population health management.
However, the incorporation of SDOH data often requires the application of natural language processing (NLP) to extract indicators that are buried in the unstructured notes of patient medical records. This information is invaluable to the construction of longitudinal health records that are a critical component of continuous patient risk scores, risk stratification and population health management.
In April of 2018, CMS renamed the EHR Incentive Programs to the Promoting Interoperability Programs. The specific promotion of data standards, interoperability via consistent application programming interfaces (APIs) and patient access to health information is a significant milestone for the effective deployment of machine learning and blockchain in healthcare.
The ability to track patient data in the form of a longitudinal health record across disparate systems provides a rich data set for applied machine learning and enhanced population health management—where the outputs of machine learning are disseminated on a permissioned basis via blockchain.
ACOs and Alternative Payment Models
More than 32 million individuals in the United States are covered by accountable
care organizations (ACOs).4As a result, implementation of intelligent networks for patient data distribution are an increasingly important part of population health management.
The disparate data of traditional fee-for-service doesn’t work as value-based care accelerates. Dissemination of data across the continuum of care is a challenge that needs resolution, because the development and deployment of quality programs requires a reporting feedback loop that is entirely too fragmented in its current state.
Machine learning approaches to risk scores and their integration with care management plans are an incredible opportunity for the improvement of population health. What’s more, machine learning can boost the development of tailored care plans that prospectively address individual treatment needs and improve outcomes. Increased integration of patient and device generated data becomes an important part of the overall model—providing patients and providers with decision-making tools for treatment options.
Garnering a complete picture of the patient requires increased interoperability and permissioned access to patient data on a real-time basis. This is where blockchain and distributed ledger technology (DLT) approaches can make a substantial impact.
Blockchain or Distributed Ledger Technology (DLT)
We’ve seen a tremendous surge in dialogue about blockchain (DLT) and its potential to disrupt the healthcare landscape. With multiple ecosystem stakeholders and disparate systems—sharing data, validating its provenance and addressing privacy concerns presents many challenges. DLT is showing promise in addressing these challenges by generating auditability and traceability between stakeholders. With a single source of truth that reduces errors and the need for reconciliation, DLT can simultaneously address patient privacy concerns and population health data requirements. The use of private permissioned ledgers—where only stakeholders with permission are granted access to patient data—places patients at the center of the data exchange and enables exploration of a holistic medical and social record governed by the patient’s consent.
Integration of patients into an ecosystem that has ever-increasing needs for transparency of treatment and financial options can deliver individualized population health management. The melding of machine learning outputs with DLT smart contracts (a smart contract is computer code by which DLT network participants— including patients in our discussion herein—agree to share information with each other) can help to ensure that healthcare resources and associated spending are targeted effectively to achieve the goals of value-based care. Likewise, in the ACO model, care coordination with affiliates and the need for consolidated quality reporting can take advantage of DLT smart contracts.
Along with further integration and transparency, patients have the opportunity for self-management. The Institute for Healthcare Improvement (IHI) notes that “Patients with chronic illness need support, as well as information, to become effective managers of their own health.”5To become stewards of their own well-being, real-time access to and exchange of information are critical components of chronic disease management, especially as home-based caregivers and other community resources are increasingly involved in patient care. With a fragmented environment, patients and caregivers have the need for more control. The ability to aggregate information for themselves and share it with stakeholders is a need that DLT can address.
Not so long ago, bank statement reconciliations were needed because of a lack of real-time transparency, leading to visibility after the fact. As the shift to electronic transactions accelerated, reconciliations went by the wayside and consumers could make financial decisions with up-to-the-minute information. A similar shift in patient self-management can occur with the reliability of device-generated data.
Device-generated data is pervasive and enormously valuable for population health management. To take full advantage of this plethora of information, healthcare providers and the patients in their care need to know this information is reliable. In- home care and interventions can dramatically impact the total cost of care.
In late 2017, the FDA issued final guidance with respect to patient access to data generated by medical devices. "Providing patients with accurate and complete information about their diagnosis and treatment, including the data collected from medical devices like blood pressure or heart rhythm monitors, allows patients to be better informed about their health and more active participants in their health care decisions.”6This is strong recognition of the value of device-generated data withrespect to in-home care for high-acuity patients, chronic and complex case management and early intervention.
DLT smart contracts can automate the permissioned disclosure of device- generated data while also providing verifiability of provenance and reliability of the data. Through the shared ledger, Physicians and other patient care team members can access reliable information that includes the history of the device’s calibration and historical readings. Anomaly readings can trigger permissioned notifications to appropriate team members for immediate action— based on verified data. In-home monitoring, telehealth consultations and post- acute outreach can all benefit from the use of DLT for data sharing.
With more than 318,000 health apps and over 340 consumer-wearable devices available,7digital health and “digiceuticals” can help with behavior modification, medication adherence, and improved population health. Charlie Hartwell writes inMedium, “Imagine having a solution that transforms and heals while eliminating all those adverse side effects listed at the end of pharmaceutical commercials.”
In 2016, Deloitte published a decision-making framework for blockchain use cases.8This framework outlined the relationships between the use of smart contracts, value transfer, and the transfer and authentication of information. The combination of machine learning and blockchain will continue to advance population health management. In a healthcare ecosystem that had $3.3 Trillion in annual cost (2016),9these technologies hold tremendous potential to play a role in addressing the Triple Aim10of improved patient experience, improved population health and reduced per capita cost.