Our long-term objective is to enable computer-based decision-support systems to contribute to the practice of medicine during patient-centered activities. To make a contribution, systems must communicate with their recipients, and communication is a social act. We therefore propose to investigate whether the method use to communicate information and the social role that it reflects affects the extent to which physicians accept that information and embody it in their subsequent behavior. The research builds on TraumAID, a validated decision-support system aimed at improving the delivery of quality trauma care during the initial definitive phase of patient management. We have designed and begun to implement three different methods for delivering TraumAID's patient management support to a trauma team during the course of patient care; (1) computer-graphic display of TraumAID's management plan on a monitor positioned in the trauma bay; (2) text communication with the Scribe Nurse concerning clinically-significant problems found with on-going patient care, which the nurse may in turn convey to the trauma team; and (3) direct, unmediated delivery of that information to the trauma team through state-of-the-art speech synthesis techniques. In each method, TraumAID's social role differs. In the first case, it has none. In the second, it interacts with the nurse through the medium of the computer monitor he or she is attending to. In the third, it interposes itself into the established group of trauma team members. We propose to complete the initial implementation of the three information delivery methods, drawing upon a proposed observational study of current communicative practices among members of the trauma team. We propose to then deploy the resulting enhanced TraumAID system in one of the MCP trauma bays, and carry out two experiments. The first will compare the three above-mentioned intervations in terms of their ability to get the head of the trauma team to change his/her behavior.. The second experiment will consider whether there are advantages to be gained from """"""""user-adapted"""""""" interaction - in particular, whether tailoring messages to the background and credentials of their primary recipient can increase the level of response. Here, we will consider three classes of recipient: experienced emergency medicine physicians, experienced trauma surgeons and trauma surgery residents. In parallel with these experiments, we propose to develop general techniques that can be used to improve the quality of text provided to the Scribe Nurse (written) and trauma team (spoken) by making it more succinct and more natural-sounding. This work will take advantage of state-of-the- art techniques in modelling discourse context and using it to generate contextually-appropriate prosodic contours and elliptic forms.
| Clarke, J R; Spejewski, B; Gertner, A S et al. (2000) An objective analysis of process errors in trauma resuscitations. Acad Emerg Med 7:1303-10 |
| Bierner, G (1998) TraumaTalk: content-to-speech generation for decision support at point of care. Proc AMIA Symp :698-702 |
| Gertner, A S; Webber, B L; Clarke, J R et al. (1997) On-line quality [corrected] assurance in the initial definitive management of multiple trauma: evaluating system potential. Artif Intell Med 9:261-82 |
| Carberry, S; Harvey, T; Clarke, J R (1997) Integrating communicative goals for real-time clinical decision support. Proc AMIA Annu Fall Symp :734-8 |
| Kaye, J M (1997) Modeling mechanical cardiopulmonary interactions for virtual environments. Proc AMIA Annu Fall Symp :464-8 |
| Cawsey, A J; Webber, B L; Jones, R B (1997) Natural language generation in health care. J Am Med Inform Assoc 4:473-82 |
