Abstract 163: Telestroke Consultations Are Executed Quickly and Efficiently

Abstract

Introduction: Telestroke systems have demonstrated the ability of technology to provide high quality acute stroke care to previously underserved populations. Although hospital systems across the U.S. are rapidly implementing these systems, little data exists to define the time commitment required by stroke specialists to complete a telestroke consultation. Objective: To describe the length of time physicians spend responding to and completing a telestroke consultation and to examine associated factors that may impact these times. Methods: This is a retrospective review of data logs generated between 7/2010 and 2/2011 using StrokeRESPOND telestroke software (InTouch Technologies, Inc.). Demographics and clinical data were abstracted from 846 consecutive consultations at eight hub hospitals and were linked to time metadata generated by clinician interaction with the software. Complete time-stamped logs were available for 203 cases. Response time was defined as patient arrival to physician log-on. Consult length was defined as time logged on to the robot and was exclusive of any telephone interaction or documentation time. Statistical analysis relating clinical variables’ effect on mean times was performed with ANOVA and paired t-test using SAS-JMP and SASv9.4 software (Cary, NC). Results: Patients had a mean age of 69.3 years and were male 41.9% (85/203). Mean consult length was 14.5 min (IQR 9.2-18.4). There was no independent association between consult length and age, diagnosis, time of arrival from symptom onset, neurological exam findings, known tPA contraindications, and absence of vascular risk factors. Mean consult length was statistically longer in tPA-recommended cases (20.0 vs 15.3 min; p= 0.04). Mean time from ED arrival to consult was 76.3 min (IQR 39.4 - 94.0) with a non-significant trend (p= 0.82) towards longer mean times (91.9 min) at night from 12:01 am to 6:00 am. Conclusions: The mean consult length of 14.5 min suggests that specialists are not using telestroke to run stroke codes remotely from start to finish and that work-up is largely completed prior to telestroke initiation. In this model, the utility of the telestroke is to have a specialist efficiently render an expert opinion on a gathered set of data (e.g. make the clinical diagnosis of stroke vs. mimic). The considerable range in response time suggests an area of improvement for telestroke. These findings will need to be validated, but this initial experience highlights the profound value of clinical metadata and the ongoing importance of thoughtful software design for accurate data abstraction with minimal data loss.