Medical care is estimated to account for only 10-20% of healthcare outcomes. As a result, healthcare executives who wish to deliver high-quality care have to consider other elements that impact patient health, including income, access to healthcare, racial discrimination, adequate medication and dietary intake.
These are social determinants of health. They offer a wealth of information about non-clinical factors that have an impact on a patient’s wellbeing. But identifying a patient’s SDOH can be challenging because details aren’t always easily accessible, especially at the time when clinicians make key treatment decisions.
SDOH data often reside in EHRs, but are essentially trapped as unstructured text within clinical notes, patient-reported data, secure e-mail exchanges, patient portal messages and other places.
In this interview with Healthcare IT News, Dr. Elizabeth Marshall, director of clinical analytics at AI-powered natural language processing vendor Linguamatics, talks about how healthcare providers can leverage AI technologies such as NLP to eliminate manual and time-consuming chart reviews to find critical patient information.
Q. Please explain why SDOH are important today, where the data reside in the EHR and why they are difficult to work with.
A. SDOH include the social, economic and environmental conditions that impact an individual’s health and quality of life, and population health outcomes. Medical care is estimated to account for approximately 20% of healthcare outcomes, which is an important reason why healthcare leaders should consider SDOH factors and their impact on patient health.
The pandemic has highlighted how certain SDOH factors, including access to care, and work, and living conditions, have a disparate impact on health outcomes. We are also seeing lower COVID-19 vaccination rates among individuals who lack ready access to the Internet and public transportation.
“An estimated 80% of clinical data is stored in an unstructured format that is difficult to search and access. As a result, clinicians are often unaware of key SDOH details that impact decision-making and patient outcomes.”
Dr. Elizabeth Marshall, Linguamatics
While SDOH provides a wealth of information about nonclinical factors impacting a patient’s overall wellbeing, identifying SDOH can be difficult for clinicians. SDOH data often resides in EHRs, but are essentially trapped as unstructured text within clinical notes, patient-reported data, patient portal messages and telehealth transcripts.
In fact, an estimated 80% of clinical data is stored in an unstructured format that is difficult to search and access. As a result, clinicians are often unaware of key SDOH details that impact decision-making and patient outcomes.
Q: What is natural language processing and how does it work?
A: NLP is a type of artificial intelligence that is concerned with the interactions between computers and human language. Using NLP, a computer can interpret the contents of documents, then extract information and insights.
NLP-based text mining transforms the free (unstructured) text in documents and databases into normalized, structured data suitable for analysis or to drive machine learning algorithms. NLP text mining can rapidly identify facts, relationships and assertions that would otherwise remain buried in the mass of textual big data.
Once extracted, this information is converted into a structured form that can be further analyzed, or presented directly using clustered HTML tables, mind maps, charts, etc.
Q: How exactly does NLP technology go through EHRs and find SDOH? And what does it do with the SDOH data?
A: NLP allows providers to unlock SDOH from EHRs to gain a more complete picture of each patient’s circumstances. A user can quickly create a query to extract key concepts and relationships from unstructured patient data to identify issues such as social isolation, transport problems and cultural factors that may impact health and outcomes.
The data can then be used with analytic tools, such as machine learning algorithms, predictive analytics and risk stratification models. In addition to using data to estimate the likelihood of future outcomes based on patterns in the historical data, healthcare organizations can use data to identify resource gaps, such as transportation issues that hinder a patient’s ability to pick up medications at the pharmacy and attend their appointments.
Based on identified SDOH issues, providers can develop new care programs, connect patients with additional resources, or offer other interventions that drive better patient outcomes.
Q: What are end results? Please connect the dots.
A: Using NLP, providers can identify patients at risk of poorer outcomes due to SDOH issues. With this insight, healthcare providers can then take proactive measures to connect patients with additional resources, which might include financial assistance for medication, educational materials to manage chronic diseases, or access to health screenings to identify underlying health conditions that could compromise health outcomes.