Using Virtual Characters to Standardize Patient Hand-Offs in Surgical Training Programs - Description of a Novel System

Abstract

Following implementation of the 80-hour work week by the ACGME in July 2003 many residency programs have transitioned to a model where multiple teams of residents are responsible for a given patient's care. This evolution has increased discontinuity of care which has resulted in reports of increased medication errors, preventable adverse events, and other discrepancies in care. To this end the Joint Commission has developed expectations that hospital systems would “implement standardized approaches to “hand-off” communications” in their National Patient Safety Goals, 2008. Research so far has focused on recommendations for improvements in the process of patient hand-off; however, little has been done to systematize an educational curriculum or series of metrics in order to prepare residents for this responsibility. We have developed a computer simulation using the Virtual People Factory (VPF), a platform developed at the University of Florida. This system allows us to capture a dialogue between a real user and a virtual character; the system learns from each user interaction thereby improving the virtual character's “intelligence.” We have programmed the system to reflect a physician in the process of “checking-out” a patient to a real physician. For proof of concept we developed a scenario where a patient underwent a laparoscopic adjustable gastric banding earlier in the day and is having a postoperative course remarkable only for tachycardia. The user interacts with the system by typing in a box, similar to instant messaging, and the system responds with a preprogrammed answer. The system tracks the information exchange and an educator defines critical information that must be gleaned, these are termed “discoveries.” So far 25 users have logged into the system (figure) with 14 residents interacting for more than 2 minutes on the exchange. The critical discovery of tachycardia was identified by 28% of users. Residents that spend more time on the system are more likely to identify the critical finding of tachycardia and their clinical concern for the patient is enhanced by that finding (table). The system learns over time such that there is a near-doubling of effective questions between users 13 and 22. Conclusions: We can capture unique details about the hand-off interchange. The system can learn and approach a peak within 25 users allowing rapid wide-scale web-based deployment. A catalogue of hand-offs could be easily developed. Performance metrics based on the identification of critical discoveries could be used to assess readiness of the user to carry off a successful hand-off. View Large Image Figure Viewer