Utilizing pure language processing to unlock SDOH in unstructured EHR knowledge

Medical care is estimated to account for less than 10-20% of healthcare outcomes. Because of this, healthcare executives who want to ship high-quality care have to think about different parts that influence affected person well being, together with earnings, entry to healthcare, racial discrimination, sufficient remedy and dietary consumption.

These are social determinants of well being. They provide a wealth of details about non-clinical components that have an effect on a affected person’s wellbeing. However figuring out a affected person’s SDOH will be difficult as a result of particulars aren’t all the time simply accessible, particularly on the time when clinicians make key therapy selections.

SDOH knowledge typically reside in EHRs, however are basically trapped as unstructured textual content inside medical notes, patient-reported knowledge, safe e-mail exchanges, affected person portal messages and different locations.

On this interview with Healthcare IT Information, Dr. Elizabeth Marshall, director of medical analytics at AI-powered pure language processing vendor Linguamatics, talks about how healthcare suppliers can leverage AI applied sciences similar to NLP to get rid of handbook and time-consuming chart opinions to search out important affected person info.

Q. Please clarify why SDOH are vital in the present day, the place the info reside within the EHR and why they’re tough to work with.

A. SDOH embody the social, financial and environmental situations that influence a person’s well being and high quality of life, and inhabitants well being outcomes. Medical care is estimated to account for roughly 20% of healthcare outcomes, which is a vital purpose why healthcare leaders ought to think about SDOH components and their influence on affected person well being.

The pandemic has highlighted how sure SDOH components, together with entry to care, and work, and residing situations, have a disparate influence on well being outcomes. We’re additionally seeing decrease COVID-19 vaccination charges amongst people who lack prepared entry to the Web and public transportation.

“An estimated 80% of medical knowledge is saved in an unstructured format that’s tough to look and entry. Because of this, clinicians are sometimes unaware of key SDOH particulars that influence decision-making and affected person outcomes.”

Dr. Elizabeth Marshall, Linguamatics

Whereas SDOH gives a wealth of details about nonclinical components impacting a affected person’s general wellbeing, figuring out SDOH will be tough for clinicians. SDOH knowledge typically resides in EHRs, however are basically trapped as unstructured textual content inside medical notes, patient-reported knowledge, affected person portal messages and telehealth transcripts.

In truth, an estimated 80% of medical knowledge is saved in an unstructured format that’s tough to look and entry. Because of this, clinicians are sometimes unaware of key SDOH particulars that influence decision-making and affected person outcomes.

Q: What’s pure language processing and the way does it work?

A: NLP is a kind of synthetic intelligence that’s involved with the interactions between computer systems and human language. Utilizing NLP, a pc can interpret the contents of paperwork, then extract info and insights.

NLP-based textual content mining transforms the free (unstructured) textual content in paperwork and databases into normalized, structured knowledge appropriate for evaluation or to drive machine studying algorithms. NLP textual content mining can quickly determine info, relationships and assertions that might in any other case stay buried within the mass of textual large knowledge.

As soon as extracted, this info is transformed right into a structured kind that may be additional analyzed, or introduced straight utilizing clustered HTML tables, thoughts maps, charts, and so on.

Q: How precisely does NLP expertise undergo EHRs and discover SDOH? And what does it do with the SDOH knowledge?

A: NLP permits suppliers to unlock SDOH from EHRs to achieve a extra full image of every affected person’s circumstances. A consumer can rapidly create a question to extract key ideas and relationships from unstructured affected person knowledge to determine points similar to social isolation, transport issues and cultural components which will influence well being and outcomes.

The information can then be used with analytic instruments, similar to machine studying algorithms, predictive analytics and danger stratification fashions. Along with utilizing knowledge to estimate the probability of future outcomes based mostly on patterns within the historic knowledge, healthcare organizations can use knowledge to determine useful resource gaps, similar to transportation points that hinder a affected person’s skill to choose up drugs on the pharmacy and attend their appointments.

Primarily based on recognized SDOH points, suppliers can develop new care applications, join sufferers with extra sources, or supply different interventions that drive higher affected person outcomes.

Q: What are finish outcomes? Please join the dots.

A: Utilizing NLP, suppliers can determine sufferers prone to poorer outcomes as a consequence of SDOH points. With this perception, healthcare suppliers can then take proactive measures to attach sufferers with extra sources, which could embody monetary help for remedy, academic supplies to handle persistent illnesses, or entry to well being screenings to determine underlying well being situations that might compromise well being outcomes.

Twitter: @SiwickiHealthIT
Electronic mail the author: [email protected]
Healthcare IT Information is a HIMSS Media publication.

Source link