How AI is contributing to improvements in healthcare
A recent trade news article suggests that current artificial intelligence applications in healthcare may not be “sexy,” but plenty of solutions are having a positive impact on healthcare delivery and operations.
Natural-language processing, for example, is an AI technology helping healthcare providers extract key information from unstructured text to glean quantitative, actionable insights. About 80 percent of clinical information is unstructured text, making a vast amount of rich patient data difficult to access for clinical decision-making or research studies.
NLP tools, however, make unstructured data usable by facilitating queries for the identification and extraction of key concepts from large volumes of data. Findings can then be transformed into structured data for analysis, visualization or integration with structured data in data warehouses.
Health systems, payers and pharma currently leverage NLP in a wide variety of use cases to advance quality of care initiatives, streamline operations and accelerate drug discovery and development. What follows are a few specific examples of NLP and AI deployments making a positive impact on the industry.
St. Louis-based Mercy includes more than 40 acute care and specialty hospitals, plus 800 physician practices and outpatient facilities. As part of a collaborative project with a global medical device manufacturer, Mercy has analyzed cardiology measures from physician data and other clinical documentation for about 100,000 patients going back to 2011.
Because some core measurements—including ejection fraction measures and symptoms such as dyspnea, fatigue and dizziness—are not typically stored in discrete fields, Mercy relies on NLP tools to extract relevant details. The data is then analyzed to evaluate heart failure device performance, which helps the device company improve its implantable products and enables Mercy clinicians to analyze the impact of different therapies and medications, and make more informed treatment decisions.
Without NLP, Mercy researchers would need to spend many years extracting the same information manually. Based on the success of the cardiac study, Mercy recently announced a new collaboration with another global medical device organization to leverage Mercy’s real-world data and NLP capabilities to improve regulatory decision-making and health outcomes.
Atrius Health, a not-for-profit health system with 30 practice locations and 900 physicians across eastern Massachusetts, leverages NLP to identify at-risk patients and close care gaps, enabling the organization to increase efficiencies, improve patient care and capture additional revenues.
To satisfy Accountable Care Organization reporting requirements and facilitate care initiatives, Atrius Health uses NLP to access information stored in narrative form in the EHR.
For example, Atrius Health analyzes unstructured echo reports to analyze heart pump function and identify high-risk heart failure patients. In 2017, Atrius Health identified 92 otherwise undocumented congestive heart failure and chronic obstructive pulmonary disease patients. This has enabled Atrius Health to take proactive measures and close care gaps, as well as advise payers that these additional patients required risk adjustment for care-budgeting purposes. This also generated additional risk-adjusted revenue to support these individuals’ care.
In addition to advancing safety-net initiatives, Atrius Health relies on NLP to facilitate quality metric reporting and to ensure clinical documentation accuracy.
Payers also utilize NLP technologies to extract member insights from unstructured big data to improve population risk stratification.
One top-5 payer has implemented NLP within an automated workflow to ingest data from Hadoop systems, which improves their ability to stratify risk for individuals with congestive heart failure. The payer uses NLP to mine risk factors such as mentions of disease and disease severity such as body mass index, along with lifestyle characteristics, like smoking and social determinants of health factors, such as social isolation. Data is extracted and transformed to a structured format that is uploaded into Hadoop HIVE to create detailed insights for population risk stratification.
Payers can easily extend the use of NLP to support other such as HEDIS quality measurement reporting.
In the future, look for AI technologies for healthcare to become increasingly sophisticated. For example, Secure Exchange Solutions (SES) is developing a platform that leverages NLP technology to improve patient care by streamlining current bottlenecks in the prior authorization and medical review process.
SES’s solution uses NLP to extract both free text and codified data from EHRs, compares the data with current policy guidelines, then offers recommendations on acceptability of procedure, treatment and medical device authorization requests. The aim is to reduce manual processes and inefficient workflows between payers and providers and facilitate the rapid delivery of appropriate patient care.
Current and future AI technologies offer great potential for improving health outcomes and reducing operational and care delivery costs. While today’s solutions aren’t necessarily sexy, they are making an undeniable difference in healthcare.