Presentation Title

Uncertainty in the Palliative Care Context

Presenter's Name(s)

Brigitte N. Durieux, UVMFollow

Time

9:00 AM

Location

Silver Maple Ballroom - Health Sciences

Abstract

Background

Improvement of communication among clinicians, seriously ill patients, and family members is a current national priority. Systematic measurement of conversational features in the natural clinical setting is an essential step for better understanding, distributing, and incentivizing high-quality serious illness communication. Uncertaintyis one important feature of the serious illness conversation which can be detected and measured using natural language processing.

Objectives

To use natural language processing to create a tool for measuring Uncertainty,and to use the resulting measurements for the exploration of associations with other important features of the conversation.

Design

This was a cross-sectional analysis of 354 transcripts of inpatient palliative care consultations from the Palliative Care Communication Research Initiative multisite cohort study.

Setting/Subjects

Hospitalized people with advanced cancer.

Measurements

We conceptualized a corpus of uncertainty language in the palliative care context by compiling uncertainty example terms present in context-relevant current literature. We accumulated synonyms and related words to this original list, and hedged the list to words with only uncertainty-related definitions. An adjudication team comprised of the full lab decided on the strongest of these terms. Using this final list, we explored subgroups of uncertainty language, and used natural language processing to detect instances of uncertainty and proportion of uncertainty within the PCCRI transcription dataset.

Preliminary Results

Manual comparison with 5 full transcripts showed no missed word instances and confirmed the program’s numbers for word counts. The use of natural language processing to detect the presence of uncertainty within textual data has resulted in variables which are currently being analyzed against relevant PCCRI survey data.

Primary Faculty Mentor Name

Robert Gramling

Status

Undergraduate

Student College

College of Arts and Sciences

Program/Major

Biochemistry

Primary Research Category

Health Sciences

Second Program (optional)

French

This document is currently not available here.

Share

COinS
 

Uncertainty in the Palliative Care Context

Background

Improvement of communication among clinicians, seriously ill patients, and family members is a current national priority. Systematic measurement of conversational features in the natural clinical setting is an essential step for better understanding, distributing, and incentivizing high-quality serious illness communication. Uncertaintyis one important feature of the serious illness conversation which can be detected and measured using natural language processing.

Objectives

To use natural language processing to create a tool for measuring Uncertainty,and to use the resulting measurements for the exploration of associations with other important features of the conversation.

Design

This was a cross-sectional analysis of 354 transcripts of inpatient palliative care consultations from the Palliative Care Communication Research Initiative multisite cohort study.

Setting/Subjects

Hospitalized people with advanced cancer.

Measurements

We conceptualized a corpus of uncertainty language in the palliative care context by compiling uncertainty example terms present in context-relevant current literature. We accumulated synonyms and related words to this original list, and hedged the list to words with only uncertainty-related definitions. An adjudication team comprised of the full lab decided on the strongest of these terms. Using this final list, we explored subgroups of uncertainty language, and used natural language processing to detect instances of uncertainty and proportion of uncertainty within the PCCRI transcription dataset.

Preliminary Results

Manual comparison with 5 full transcripts showed no missed word instances and confirmed the program’s numbers for word counts. The use of natural language processing to detect the presence of uncertainty within textual data has resulted in variables which are currently being analyzed against relevant PCCRI survey data.