MedCity Influencers

Leveraging Natural Language Processing to uncover social determinants of health insights

NLP technologies facilitate the extraction of structured information from unstructured fields to gain population-level insights and 360-degree views of every patient for organizations moving to value-based care.

social determinants of health,

The transition to value-based payment models continues to drive demand for population health data that include comprehensive, 360-degree views of every patient. In addition to diagnoses, procedures and similar medical data, effective population health management requires details on social determinants of health (SDoH), including such factors as lifestyle choices, living conditions, and health behaviors. SDoH have a significant impact on patient health – more so, in fact than clinical care, which is associated with just 20 percent of a person’s health status.

Consider, for example, how outcomes are impacted when a provider is aware that a patient has mobility issues, lives alone and lacks reliable transportation, resulting in difficulty traveling to medical appointments or picking up medications at the pharmacy. To ensure the patient receives the care required for optimal outcomes, the clinician can connect the patient with social service organizations to arrange for transportation to medical appointments and the at-home delivery of medications. SDoH details also help providers identify patients who are more likely to be re-admitted to the hospital within 30 days of discharge due to a lack of family and social support or because of a history of risky health behaviors.

SDoH data also helps organizations assess risk more accurately and stratify patient populations. Factors such as education, housing, and community safety are clinically relevant and help with the identification of at-risk individuals that could benefit from preventive health or safety-net programs. For example, demographic and economic data can help providers identify individuals who have mentioned food insecurity as an issue or are living in “food deserts” without ready-access to affordable fresh fruit, vegetables, and other healthy foods, and who are at higher risk for obesity and chronic conditions like diabetes.

Accessing critical data
Despite the clear value of SDoH for clinical decision-making and population analysis, providers often lack ready-access to critical SDoH details. Typically, healthcare groups rely on electronic health records (EHRs) and claims data to analyze the health of patient populations and to make clinical treatment decisions. Unfortunately, neither EHRs nor insurance claims alone are ideal for understanding the health of populations, nor for providing a 360-degree view of an individual’s well-being.

To obtain a comprehensive view of each patient, providers must consider all clinical data, including information stored as structured text, such as CPT and ICD-10 codes, as well as unstructured data from physician narratives, pathology and radiology reports, discharge summaries and other clinical documents. Unstructured text often contains critical information about a patient’s health status, as well as SDoH insights that could impact the care process. As much as 80 percent of clinical information stored within EHRs are in unstructured formats, which is more difficult to analyze on a large scale. As a result, providers can easily miss vital patient and population details that are essential for the delivery of optimal patient care and successful population health management.

Leveraging technology to identify SDoH details
Healthcare organizations require 360-degree patient views, including SDoH insights, in order to execute on population health strategies and achieve value-based care goals. Providers need technology that effectively extracts critical information from unstructured data in a format that supports actionable insights. Natural language processing (NLP) is one such technology that is already helping numerous organizations to maximize the value of all their clinical data.

presented by

NLP technologies facilitate the extraction of structured information from unstructured fields. For example, organizations can leverage NLP text-mining to uncover clinically relevant details from physician narratives. If the data extraction process reveals that a particular patient recently lost her spouse, the clinical team could modify the patient’s existing care plan and increase monitoring for signs of depression.

From a population health perspective, NLP can help organizations to more accurately characterize their patient populations and identify the risk of certain conditions. For example, an organization might want to assess a population’s risk of Type 2 diabetes. In addition to an analysis of structured data to reveal risk factors associated with weight, race, and age, unstructured free text fields could also be queried using NLP. This could uncover the prevalence of known SDoH risk factors, such as limited access to healthy foods, barriers to physical activity, high-stress levels, and social isolation.

Advancing organizational care goals with NLP text-mining
Ready access to population-level insights and 360-degree views of every patient are essential for organizations seeking financial and clinical success under value-based care payment models. By leveraging NLP technology, providers can uncover critical health data from unstructured text and advance their care delivery and population health goals.

Photo: vaeenma, Getty Images