Mark Scott, Chief Marketing Officer, Apixio
Every year, the U.S. healthcare system generates 1.2 billion clinical documents—documents that contain detailed insights into patient conditions, health history, care, and social determinants of health (SDoH). If harnessed at scale, the data captured by physicians in these documents could transform the way we approach operational decision-making, care coordination, clinical treatment recommendations, and even drug discovery. But there are two critical problems standing in the way of this vision becoming reality. The first is data fragmentation. Healthcare documents are generated and stored in siloed systems that often don’t talk to one another, making it difficult to create and maintain a centralized health record for each patient. The second is searchability. Without being able to search for and extract insights from healthcare documents at scale, having a centralized patient record provides little value. 80 percent of healthcare data is unstructured, free-form text, which makes it very difficult to analyze using traditional computational techniques.
These challenges have prevented organizations from mining the wealth of available clinical data to improve healthcare costs, quality, and patient experience. However, in the past few years, technology and data acquisition methods have advanced in ways that can now support a centralized view of patient health. Leveraging HL7, FHIR, and other interoperable data exchange frameworks, enterprise data platforms can now ingest, process, and store healthcare data that is secure, normalized, quality checked, and queryable. Advanced optical character recognition (OCR) systems trained for specific healthcare use cases can now unlock data in patient charts stored as PDF or image files. Artificial intelligence techniques like statistical natural language processing (NLP), machine learning (ML), and deep learning have evolved to support specialized algorithm creation and deployment at scale, allowing data science organizations to build models that identify insights and trends within large, diverse structured and unstructured datasets.
Implementing these techniques and tools, healthcare organizations have an immense opportunity ahead of them: To merge the broad view of care activities happening for a given patient population captured by payer claims with the deep, patient-level view of conditions and care captured by clinicians at the point of care, and combine them into a comprehensive longitudinal patient record. This record would bring together claims, charts, labs, prescriptions, encounter or physician notes, health history data, demographic information, and other documents in a centralized, patient-centered view of health status and care over time. The most valuable, detailed information captured in unstructured physician notes can be transformed into machine-readable text and mined for deeper insights to augment the traditional structured data sources many organizations are already using, providing a richer, more nuanced view of medical history, risk factors, social determinants of health (SDoH), and anecdotal health information.
Layering semantic search capabilities on top of this longitudinal record, provider groups, health plans, researchers, and other authorized healthcare entities can seek out targeted intelligence to inform operational decision-making and care delivery decisions. For example, a provider group could search for patients with specific comorbidities to understand common prescribing patterns; a health plan could identify members experiencing food insecurity to enroll in a SDoH program; a research company could assess drug trial eligibility using condition data for a specific candidate pool; and a telehealth company could review clinical histories for upcoming patients to ensure continuity of care. The potential use cases for longitudinal patient records powered by an enterprise data platform are wide ranging, and will continue to grow over time as additional data types are collected and integrated into EHRs and other healthcare systems.
The changing regulatory environment will also help facilitate the development of a longitudinal patient record. As interoperability standards evolve and regulations to facilitate data exchange among healthcare entities come into effect in 2021 and beyond, plans, providers and other organizations will have a richer dataset to incorporate into aggregated patient profiles.
Now is the time for organizations to rethink their data infrastructure strategy and start evaluating data platform and technology investments to help them transform their administrative and clinical data into actionable insights. Those who do will be well-positioned to turn the historical healthcare data problem into a strategic advantage for their business.