Abstract:Outcomes 1. Describe two natural language processing tools. 2. Describe a method to reliably and accurately identify populations of interest (i.e. denominator) using electronic health record data. 3. Design keyword libraries for rule-based natural language processing. 4. Test zero-shot and few-shot text prompting, using large language models. Chronic disease is the leading cause of death and disability in the United States, accounting for more than 1.7 million deaths each year. These patients accrue hundreds of millions of clinical notes over the course of their illness care, including documentation of data critical for palliative care delivery, such as goals of care conversations, symptoms, and social support. Until recently, text-based data required manual chart abstraction, which is extremely cumbersome, expensive, and not scalable to the population level. Natural language processing (NLP) presents an opportunity to identify text-based data within electronic health records with increased speed and efficiency compared to manual chart review. In this session, we aim to show how NLP-assisted chart abstraction supports scalable extraction of text-based data for clinical trials, quality improvement, and pay-for-performance programs. We will review a valid and reliable approach to identifying study populations, steps needed to carry out NLP-assisted chart abstraction, and easy-to-use NLP tools. We will present lessons learned from NLP work conducted at three US healthcare systems: (1) At a large cancer center, NLP- assisted chart abstraction is used to measure documentation of goals of care conversations among patients with advanced cancer. The results are reported to major payers in a pay-for-performance program and generate data for research and quality improvement. (2) A large language model was used to identify symptoms discussed in audio-recorded conversations between patients and their oncologists. This study illustrates the importance of building robust text prompts. (3) A research study in the Veterans Administration illustrates the feasibility of identifying a reliable and generalizable denominator of patients with advanced cancer and implementing NLP-assisted chart abstraction for a national sample. (4) In a proof-of-concept study that is among the first to employ this technique in pediatric serious illness research, we used rule-based NLP to evaluate critical palliative and end-of-life care process measures for children with cancer who died across a nine-year span at a single center. Honing this rigorous, scalable approach enables us to assess care quality for children with serious illness in priority domains where care has not previously been measured or improved.

Desiderata for delivering NLP to accelerate healthcare AI advancement and a Mayo Clinic NLP-as-a-service implementation

Behind the scenes: A medical natural language processing project

Northwestern University resource and education development initiatives to advance collaborative artificial intelligence across the learning health system

Using Deep Learning Based Natural Language Processing Techniques for Clinical Decision-Making with EHRs

Scalability and cost of a cloud-based approach to medical NLP

Natural Language Processing for Smart Healthcare

NLP to Improve Accuracy and Quality of Dictated Medical Documents-Final Report

Natural Language Processing in medicine and ophthalmology: A review for the 21st-century clinician

An Introduction to Natural Language Processing Techniques and Framework for Clinical Implementation in Radiation Oncology

Enhancing Patient Experience by Automating and Transforming Free Text into Actionable Consumer Insights: A Natural Language Processing (NLP) Approach

Speaking the Same Language: Leveraging LLMs in Standardizing Clinical Data for AI

Foundation Models, Generative AI, and Large Language Models: Essentials for Nursing

Natural Language Processing in Biomedicine: A Unified System Architecture Overview

The Growing Impact of Natural Language Processing in Healthcare and Public Health

Implementing a Portable Clinical NLP System with a Common Data Model - a Lisp Perspective

Transforming Written and Spoken Words into Quantitative Data for Palliative Care Research Using Natural Language Processing (ME301)

Generation and evaluation of artificial mental health records for Natural Language Processing

Natural language processing data services for healthcare providers

NLP for Maternal Healthcare: Perspectives and Guiding Principles in the Age of LLMs

A Study on NLP Based Approach in AI and Text Data Mining for Automated Highlighting of New Information in Clinical Notes