AI Algorithms identify documentation of serious illness conversations in electronic health records

THIS MONTH’S EDITOR’S CHOICE FROM PALLIATIVE MEDICINE …

Charlotta Lindvall, Alex Chan and Anne Walling explain the background to their longer article selected as Editor’s Choice in the February issue of Palliative Medicine.

Top left clockwise: Alex Chan, Charlotta Lindvall and Anne Walling.

Electronic Health Records (EHRs) contain enormous amounts of data that may be used to facilitate treatment discovery, guide quality and safety initiatives, and enhance patient satisfaction. Quantitative data, for example body temperature and blood work, can be logged and tracked. However, much of the qualitative data, such as patients’ descriptions of their symptoms or treatment goals, are manually recorded in the clinical visit notes using a free-text format. It is estimated that 70-80 per cent of patient information resides in unstructured free-text notes, including many outcomes related to palliative and end-of-life care. Manual chart reviews required to extract these data are extremely time-consuming and expensive, so many endorsed palliative care measures are not assessed, and their impact on distal and important patient outcomes have been insufficiently evaluated. A key example of this is timely documentation of patient care preferences in critically ill older adults.

In the current issue of Palliative Medicine, we show that AI (Artificial Intelligence) algorithms can be applied to assess a palliative care quality measure endorsed by the National Quality Forum. Our team, led by Dr Lindvall at the Dana-Farber Cancer Institute, built and tested AI algorithms in the form of natural language processing to assess documentation of patient care preferences in clinical notes from the EHR at Beth-Israel Deaconess Medical Center in Boston. Our natural language processing algorithms analyzed clinical notes more than 18,000 times faster than clinician coders (0.022 seconds/note vs. 402 seconds/note) with an accuracy approaching that of human coders (92 per cent). Applied to 10,250 clinical notes from1,350 critically ill patients aged 75 or over, we found that 65 per cent of patients had care preferences documented within the first 48 hours. We also detected variations in care. For example, clinicians in surgical critical care units documented patient care preferences less frequently than clinicians in medical critical care units (45 per cent vs. 75 per cent). Being able to detect differences in care at the hospital system level in just a few minutes makes these methods promising for quality improvement projects.

The Lindvall lab focuses on natural language processing to extract patient-centred information from EHRs. While we interact with AI technology every day, for example with shopping or banking, it is underutilised in clinical medicine. This is unfortunate as it offers great potential for the improvement of patient care. Natural language processing facilitates the understanding, interpretation and manipulation of human language using computers. This makes it possible to analyze massive amounts of language-based data in a consistent, efficient and unbiased manner. Our algorithms presented in the current issue of Palliative Medicine, analyzed clinical notes in a tiny fraction of the time needed for manual review, offering a practical option for rapid audit and feedback regarding care preference documentation at the system and clinician level.

READ THE FULL ARTICLE IN PALLIATIVE MEDICINE
This post relates to the longer article, ‘Deep Learning Algorithms to Identify Documentation of Serious Illness Conversations During Intensive Care Unit Admissions’ by Alex Chan, Isabel Chien, Edward Moseley, Saad Salman, Sarah Kaminer Bourland, Daniela Lamas, Anne M Walling, James A Tulskyand Charlotta Lindvall, published in Palliative Medicine 2019 Volume: 33 issue: 2, page(s): 187-196. Article first published online: November 14, 2018. Issue published: 1 February 2019.

More about the authors …

Charlotta Lindvall, MD, PhD is a physician-scientist in the Department of Psychosocial Oncology and Palliative Care, Dana-Farber Cancer Institute, Boston, USA. Contact Charlotta Lindvall by email.

Alex Chan, MS, MPH is Gerhard Casper Fellow at Health Research and Policy, Stanford University School of Medicine, Stanford, USA.

Anne Walling, MD, PhD is Associate Professor at the Division of General Internal Medicine and Health Services Research, David Geffen School of Medicine, University of California Los Angeles; and Palliative Care, VA Greater Los Angeles Healthcare System, Los Angeles, USA.

EAPC MEMBERS CAN DOWNLOAD THE FULL ARTICLE FREE OF CHARGE
If you are currently an Individual or Associate EAPC Member you have full access to the EAPC website, and the chance to download free of charge all ‘Editor’s choice’ articles and many other papers too. Just click here then enter your email address and membership password and choose from the list of journal articles.

How to join as an Individual/Associate Member, or to renew your membership

Click here to join as an Associate or Individual Member, or to renew membership.  Choose relevant national palliative care association if you wish to register free of charge as an Associate Member, now login with your email address and password and choose from the list of articles.

Note: You can apply to be an Associate Member FREE of charge provided that you are a member of your country’s national palliative care association, and that the association is an EAPC National Association member.

This entry was posted in EAPC-LINKED JOURNALS, Palliative Medicine: Editor's Choice. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.