Sunghwan Sohn, Ph.D.

Location

Contact

SUMMARY

Sunghwan Sohn, Ph.D., has expertise in mining large-scale electronic health records (EHRs) to unlock unstructured and hidden information through natural language processing (NLP) and machine learning techniques. Dr. Sohn develops strategies for the best use of informatics, ranging from precision medicine to population health, in order to achieve better solutions for people.

Focus areas

• Information extraction and normalization. Dr. Sohn applies NLP techniques to extract unstructured medical concepts from clinical narratives and map them to standard forms. This technique facilitates clinical research and allows seamless information exchange across healthcare institutions.
• Computational phenotyping. Dr. Sohn develops and implements artificial intelligence (AI) algorithms to automatically determine specific medical conditions such as asthma, peripheral artery disease, abdominal artery aneurysm, prosthetic joint infection, falls and delirium. He analyzes various EHRs to allow large-scale epidemiological studies.
• People's characterization toward precision population health. Dr. Sohn analyzes individuals' health trajectories from longitudinal EHRs, providing a crucial perspective on precision population health. This approach helps discern patterns and trends in people's health over time. These patterns and trends offer insights into disease progression, treatment responses and potential risk factors for specific populations such as progression and risk factors for Alzheimer's disease and related dementias. By leveraging AI-powered EHR data analytics, Dr. Sohn improves people's outcomes and enhances population health by identifying broader health determinants.
• Mitigating bias in AI models concerning social determinants of health. AI models may have harmful biases, which could increase differences among underserved populations. Dr. Sohn evaluated the performance of AI predictive models across various socioeconomic status (SES) groups and discovered that these models are less effective for lower SES groups. He aims to conduct a study to examine the extent to which inequalities in EHR quality, stemming from low SES, contribute to differential performance of AI models.
• Secondary use of EHRs to improve clinical competency and education. Dr. Sohn is engaged in development and validation of an informatics tool to assess clinicians' adherence to asthma guidelines. This tool provides personalized clinical effectiveness data that helps clinicians achieve clinical competence in asthma care and documentation.

• Active medical device surveillance. Despite the release of a final rule by the Food and Drug Administration to establish the unique device identifier (UDI) system, its impact remains limited. This limited impact is due to the unavailability of UDIs in a structured format in EHRs or administrative claims data.

  However, descriptions of medical devices and adverse outcomes are routinely recorded in the unstructured text of EHRs. Dr. Sohn's aim is to develop automated and scalable AI models using EHRs to promote safter medical device use. These AI models overcome the limitations of current passive device surveillance and enhance real-world evidence generation.

Significance to patient care

Dr. Sohn's research helps the best use of EHRs to solve clinical problems and improve public health. His work provides biomedical scientists and clinicians access to the rich yet untapped information embedded in clinical narratives. Leveraging AI-driven EHR data analytics, his work provides the healthcare community with the valuable insights needed for advancing clinical research and improving care to people.