Capturing Phrases for ICU-Talk, a Communication Aid for Intubated Intensive Care Patients
S. Ashraf, A. Judson, I. W. Ricketts, A. Waller, N. Alm, B. Gordon, F. MacAulay, J. K. Brodie, M. Etchels, A. Warden, A. J. Shearer · 2002 · Proceedings of the Fifth International ACM Conference on Assistive Technologies (Assets 02) · doi:10.1145/638249.638288
Summary
This paper describes the vocabulary gathering methods used for ICU-Talk, a three-year multidisciplinary project at the University of Dundee and Ninewells Hospital that developed an augmentative and alternative communication (AAC) aid specifically for intubated intensive care patients. These patients are critically ill, sedated, and unable to speak due to endotracheal tubes facilitating mechanical ventilation. When awake but still intubated — a period ranging from hours to days — patients need to communicate but are impaired by weakness, fatigue, cognitive difficulties, and poor short-term memory. Existing low-tech methods such as alphabet boards, symbol charts, and pen and paper were designed for general use, not the ICU context. The unique challenge for ICU-Talk was that patients themselves cannot contribute to vocabulary selection, unlike most AAC users who can directly input their preferred vocabulary. The project therefore developed multiple alternative methods to build the phrase database: observing nursing staff in ICU recording patients' actual communication attempts, surveying nurses across conversation topics, and conducting structured computer-based interviews with patients' next of kin to gather personalised information about family, interests, and hobbies.
Key findings
The nursing staff observations and surveys yielded 190 unique phrases, with 75% of surveyed nurses agreeing on appropriate vocabulary. Verification confirmed these matched actual patient communication attempts. Critically, only 47 of 190 phrases (about 25%) were common across patients — the remaining observed communication attempts were specific to individuals, demonstrating the essential need for personalisation. The computer-based interview tool, designed to feel like a familiar census form, gathered information through 13 questions about the patient's family, friends, hobbies, and interests, automatically generating up to 80 personalised phrases from skeleton templates. This created two vocabulary layers: a core database of general phrases applicable to all ICU patients, and a patient-specific database of personalised phrases. To address the problem of prolonged stays where fixed phrases become stale, the team also developed a tool that takes the patient's interests, searches web news articles for relevant content, and uses text summarisation with extraction-based sentence identification to automatically generate timely conversation phrases about current events. An empirical study with 20 volunteers showed participants could reliably identify and rank the most important sentences from summarised news articles.
Relevance
This paper addresses a critically underserved population in AAC research: patients who temporarily lose the ability to speak due to medical intervention rather than a chronic disability. The vocabulary gathering methodology is particularly valuable because it solves the bootstrapping problem of building a communication system for users who cannot participate in its design. For AAC practitioners, the two-tier approach of core plus personalised vocabulary, combined with automated phrase generation from external sources, offers a model for keeping communication aids relevant over extended periods. The work also highlights that effective AAC requires understanding the specific communication context — ICU patients have very different needs from typical AAC users, including the noisy environment, physical constraints of being bed-bound with monitoring equipment, and the critical importance of being able to alert staff to emergencies. The text summarisation approach for generating current-affairs phrases anticipates modern interest in using NLP to enrich AAC vocabulary dynamically.
Tags: augmentative and alternative communication · AAC · intensive care · vocabulary selection · phrase-based communication · speech synthesis · medical accessibility · intubation