Tech & Innovation in Healthcare

Question: Providers in our healthcare practice make it a priority to document social determinants of health (SDOH) during our patient encounters. However, the medical coding and billing service doesn’t always report the SDOH factors.

How can we encourage the medical coders and billers to report the SDOH codes?

Kentucky Subscriber

Answer: Reporting SDOH is important to improving patients’ health and delivering optimal care. Healthcare organizations have made large strides in gathering SDOH information over the past few years, but several challenges continue to stand in the way of data capture, such as:

• Sharing of data
• Managing consent
• SDOH data collection and storage standardization

However, artificial intelligence (AI) has the potential to aid in your quest for greater SDOH reporting, and here’s one example. Researchers from Indiana University and Regenstrief Institute developed three natural language processing (NLP) algorithms to pull out financial, employment, and housing data from unstructured electronic health records (EHRs).

In a study published to JAMIA Open in July 2023, researchers applied three NLP algorithms to more than 2.4 million records from over 655,000 patients in time periods of 2019 and from Sept. 21, 2020, to March 31, 2021. The researchers had access to a training dataset of notes, which also included all clinical documentation, from two health systems in the Indianapolis area from the Indiana Network for Patient Care.

Researchers applied patient identifiers to clinical and demographic information, such as age, gender, race, ethnicity, rural/urban status, in the unstructured data. “We purposefully selected these different sources to support model generalizability,” the researchers wrote in the study.

While the AI NLP system ran, the technology read each note and created tags or indicators to alert researchers that the record contains SDOH-related data. “Our findings indicate that the use of NLP to analyze existing, routinely collected free-text reports may be a feasible approach for healthcare organizations and researchers to identify patients with documented social factors,” the researchers wrote.

In the end, the algorithms “demonstrated strong performance” in identifying adult patients experiencing:

• Housing instability
• Financial insecurity
• Unemployment

While the NLP algorithms were successful in this study, the researchers did identify challenges to using the technology as it

currently stands. One challenge the researchers acknowledged is gathering sufficient data to cover different patient populations is critical. Using data from one area of the country and hoping to achieve the same predictable performance in a patient population in a vastly different area is infeasible.